Diselenide-bridged mesoporous silica nanoplatform for baicalin delivery facilitates spinal cord injury repair via CHCHD2-mediated mitochondrial homeostasis restoration

Abstract

During the secondary phase of spinal cord injury (SCI), excessive accumulation of reactive oxygen species (ROS), mitochondrial dysfunction and amplified inflammatory cascades reinforce one another, forming a vicious cycle that drives progressive neuronal damage and limits functional recovery. To enhance lesion-site drug exposure and microenvironment-responsive delivery, we engineered a biomimetic nanoplatform (Ba@Se-MSN&BV2) by loading baicalin into diselenide-bridged mesoporous silica nanoparticles (MSNs) and cloaking them with BV2 microglial membranes, enabling lesion-targeted accumulation and ROS-triggered, on-demand release in ROS-rich lesions. In glutamate-injured HT22 cells and primary hippocampal neurons, Ba@Se-MSN&BV2 reduced ROS, partially restored mitochondrial membrane potential, suppressed apoptosis and enhanced neurite outgrowth, with protection generally exceeding that of free baicalin or Ba@Se-MSN. In lipopolysaccharide-stimulated BV2 microglia, the nanoplatform lowered IL-1β and TNF-α, increased Arg1 and IL-10 and shifted polarization toward an anti-inflammatory, reparative phenotype. In a mouse contusion SCI model, Ba@Se-MSN&BV2 preferentially accumulated at the lesion, mitigated oxidative burden, attenuated glial scarring and preserved axons and neurons. These tissue-level benefits were accompanied by sustained improvements in locomotor recovery and motor-evoked responses. Label-free quantitative proteomics highlighted CHCHD2 as a mitochondria-associated candidate that was downregulated in glutamate-injured neuronal cells and restored by Ba@Se-MSN&BV2. Knockdown/rescue assays supported a contributory role of CHCHD2 in the neuroprotective profile of the nanoplatform. Collectively, Ba@Se-MSN&BV2 mitigates oxidative–inflammatory imbalance during secondary SCI by coupling lesion-targeted homing, ROS-responsive baicalin release and CHCHD2-mediated mitochondrial stabilization, representing a lesion-matched, microenvironment-coupled nanotherapy with translational potential for post-SCI intervention.

1. Background

Spinal cord injury (SCI) is a devastating insult to the central nervous system that often causes lifelong disability, including persistent motor, sensory, and autonomic deficits together with systemic comorbidities such as chronic pain and metabolic disturbances [1,2]. After the primary insult, SCI rapidly evolves into a secondary pathobiological cascade in which overproduction of reactive oxygen species (ROS) and reactive nitrogen species (RNS), mitochondrial dysfunction, and amplified neuroinflammation establish a self-reinforcing feed-forward loop [[3], [4], [5]]. Redox imbalance induces mitochondrial depolarization and bioenergetic failure, which in turn precipitate excitotoxicity and ionic homeostatic disruption [6,7]. These secondary events drive maladaptive microglial activation and peripheral leukocyte infiltration [8,9]. Collectively, these processes converge and amplify, culminating in axonal degeneration and reactive gliosis with ensuing glial scar formation and remodeling, thereby compromising neural circuit reconstruction by disrupting structural scaffolds and creating an inhibitory microenvironment [[10], [11], [12]]. Current clinical care is centered on early decompression and spinal stabilization, supplemented by pharmacological agents, neurotrophic factor delivery, and cell-based therapies [13,14]. However, clinical gains remain modest, largely because drugs fail to accumulate at therapeutic concentrations within the lesion, delivery is rarely controlled in a spatiotemporally precise and cue-responsive manner, and long-term functional recovery is limited and often inconsistent [15]. Thus, concurrently integrating lesion targeting, pathological cue responsiveness, and neural repair within the complex SCI microenvironment remains a key unmet challenge for therapeutic advancement.

Nanoscale drug delivery offers a rational and engineerable strategy to increase lesion-site drug exposure and improve therapeutic consistency [16]. Mesoporous silica nanoparticles (MSNs), with tissue-penetrable dimensions, feature tunable pore architectures and high specific surface areas, enabling high drug loading and controllable release kinetics [17]. Abundant surface silanol groups facilitate the construction of pore-gating structures and the anchoring of biomimetic shells, enabling synergistic integration of pore regulation and surface functionalization for precise control of drug delivery and release [18]. However, within the inflammatory and oxidative microenvironment of SCI, passive diffusion solely driven by concentration gradients cannot align drug release with lesion-derived cues, and off-target leakage with premature drug loss remains prevalent, undermining durable lesion-site bioavailability [19,20]. Therefore, incorporating pathology-responsive gatekeepers into such carriers is essential to match drug release with local pathological signals. Accumulating evidence indicates that diselenide bonds are highly susceptible to oxidative stress [21,22]. Elevated ROS triggers redox-mediated bond cleavage, generating release kinetics that scale with local ROS burden, which minimizes nonspecific premature leakage while enhancing lesion-site exposure. Moreover, selenium can bolster endogenous antioxidant systems and act synergistically with intrinsic redox networks [23]. Thus, diselenide bonds serve as ROS-responsive gatekeepers and intrinsic antioxidative moieties, matching the pronounced oxidative imbalance of secondary SCI and supporting controlled in situ release with sustained and reproducible therapeutic efficacy.

Beyond pathology-triggered release, the therapeutic payload should target key nodes of secondary injury, particularly oxidative stress, neuroinflammation, and apoptosis susceptibility [24]. Baicalin is a structurally defined bioactive flavonoid with well documented antioxidant, anti-inflammatory, and anti-apoptotic properties [25,26]. Accumulating evidence from models of neurodegeneration [27] and cardiovascular injury [28] supports the neuroprotective and vasoprotective potential of baicalin. Nonetheless, baicalin has poor water solubility and low membrane permeability. It is rapidly cleared systemically and exhibits high nonspecific protein binding, leading to a low free-drug fraction, inadequate lesion exposure, and consequently limited efficacy and reproducibility [29]. Encapsulating baicalin within diselenide-gated MSNs enables spatiotemporally controlled intralesional release in lesions with high ROS burden. To translate this programmable release into consistent in vivo efficacy, lesion residence and tissue penetration must be further enhanced through nano–bio interface engineering. Cell-membrane coating preserves transmembrane receptors and adhesion ligands on nanoparticles. This self-mimicking immune signature attenuates mononuclear phagocyte clearance and strengthens interactions with inflamed endothelium and extracellular matrix, thereby promoting lesion-targeted accumulation [30,31]. Given the microglia-dominated inflammatory landscape after SCI, BV2-derived microglial membrane coating is more likely to engage lesion-associated receptors and ligand receptor interactions, thereby enhancing selective accumulation and deep tissue penetration. This stabilized intralesional exposure, in turn, supports diselenide-driven controlled release [[32], [33], [34]].

During the post-injury repair phase of SCI, preserving mitochondrial homeostasis is a key determinant of neuronal survival and axonal regeneration [[35], [36], [37]]. CHCHD2 (coiled-coil-helix-coiled-coil-helix domain containing 2) is a highly conserved protein localized to the mitochondrial intermembrane space that supports cristae integrity and mitochondrial bioenergetics [38]. It participates in energy metabolism, oxidative stress responses, and apoptosis susceptibility, and has been implicated in neuroprotection across multiple neurological disorders [39]. Accumulating evidence from neurodegenerative models, particularly Parkinson's disease, indicates that CHCHD2 loss or pathogenic mutations disrupt cristae architecture, reduce mitochondrial membrane potential, and induce persistent mitochondrial ROS generation [40,41]. These changes aggravate neuronal apoptosis, axonal degeneration, and neurological dysfunction. Conversely, restoring or enhancing CHCHD2 preserves mitochondrial function and neuronal viability [42]. Together, these findings position CHCHD2 as a key regulator of mitochondrial homeostasis in the central nervous system [43]. However, during the secondary phase of SCI, the spatiotemporal dynamics of CHCHD2, its interplay with the oxidative and inflammatory microenvironment, and its roles in spinal cord neuronal survival and axonal regrowth remain poorly defined. Accordingly, incorporating CHCHD2 as a mechanistic node within ROS-responsive nanotherapy is necessary. Defining its contribution to mitochondrial homeostasis restoration and post-SCI neurorepair will help bridge nanomaterial-mediated effects with endogenous neuroprotective pathways.

Guided by these considerations, a diselenide-bridged MSN-based nanocarrier loaded with baicalin and cloaked with a BV2-derived microglial membrane was engineered, yielding the biomimetic nanoplatform Ba@Se-MSN&BV2 (Scheme 1). The BV2 membrane provides an inflammation-tropic interface that enhances lesion-site accumulation and local bioavailability. Concurrently, diselenide bonds confer oxidative stress-responsive and pathology-coupled release, thereby facilitating synchronized suppression of secondary oxidative stress and neuroinflammatory amplification. Comprehensive multiscale in vitro and in vivo assessments were used to determine whether this strategy restores mitochondrial homeostasis, mitigates neuronal injury, and promotes axonal regeneration and locomotor recovery. Proteomic profiling combined with CHCHD2-targeted perturbation indicates that the therapeutic benefits are largely attributable to CHCHD2-mediated restoration of mitochondrial homeostasis. This mechanistic link provides a basis for coupling ROS-responsive nanotherapy with endogenous neuroprotective circuits.

Scheme 1.

Schematic illustration of Ba@Se-MSN&BV2 facilitating spinal cord recovery via alleviating oxidative stress and restoring CHCHD2-mediated mitochondrial homeostasis.

2. Results and discussion

2.1. Construction and characterization of Ba@Se-MSN&BV2

A diselenide-bridged mesoporous silica nanocarrier was engineered, with baicalin loaded into the mesopores and subsequent cloaking with BV2-derived microglial membranes, yielding the biomimetic nanoplatform Ba@Se-MSN&BV2. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) showed that Ba@Se-MSN particles were well dispersed and near-spherical, with well-ordered mesopores and intact pore walls. The average dry diameter was approximately 60 nm (Fig. 1A and B), a size favorable for nanodelivery while maintaining mesostructural integrity and compatible with tissue penetration in the central nervous system [44]. Under oxidative stimulation (100 μM H₂O₂ for 2 h), the framework underwent diselenide bond cleavage accompanied by pore wall collapse and fragmentation (Fig. S1). These changes are consistent with oxidative remodeling of the diselenide-bridged skeleton and ROS-responsive release behavior, suggesting that the carrier can couple drug liberation to oxidative cues characteristic of the secondary phase of SCI [45]. After membrane cloaking, Ba@Se-MSN&BV2 displayed a unimodal dynamic light scattering (DLS) distribution with a hydrodynamic diameter of 79.90 ± 0.58 nm and a polydispersity index (PDI) of 0.16, indicating a monodisperse colloid (Fig. 1C). Relative to Ba@Se-MSN, a modest size increase was observed without peak broadening or aggregation, indicating that membrane coating preserved colloidal stability. The zeta potential shifted toward that of native BV2 membranes with a larger absolute value (Fig. 1D), indicating a transition from a silanol-rich silica surface to a biomimetic interface dominated by membrane proteins and phospholipids, which is expected to reduce nonspecific protein adsorption and favor lesion-relevant interactions in vivo.

Fig. 1.

Synthesis, structural characterization and in vivo distribution of Ba@Se-MSN&BV2. (A) SEM images showing the morphology and mesoporous structure of Ba@Se-MSN particles. (B) TEM images revealing the mesopore arrangement and continuity of the Ba@Se-MSN pore walls. (C) Dynamic light scattering (DLS) analysis of the hydrodynamic size distribution and polydispersity index (PDI) of Ba@Se-MSN&BV2 (n = 3). (D) Zeta potential measurements comparing the surface charges of BV2 cell membranes, Ba@Se-MSN and Ba@Se-MSN&BV2 (n = 3). (E) UV–Vis absorption spectra of Se-MSN, baicalin, Ba@Se-MSN and Ba@Se-MSN&BV2. (F) In vitro cumulative release profiles of Ba@Se-MSN&BV2 in PBS (±100 μM H₂O₂) (n = 3). (G) ABTS•⁺ radical scavenging assay comparing the 2 h scavenging efficiencies of baicalin, Ba@Se-MSN and Ba@Se-MSN&BV2 (n = 3). (H) DPPH• radical scavenging assay comparing the 2 h scavenging efficiencies of baicalin, Ba@Se-MSN and Ba@Se-MSN&BV2 (n = 3). (I) ABTS•⁺ kinetic assay showing time- and concentration-dependent radical scavenging by Ba@Se-MSN&BV2 at different mass concentrations. (J) In vivo fluorescence imaging in mice comparing the biodistribution of coumarin-6 (C6)-labeled Ba@Se-MSN and Ba@Se-MSN&BV2. (K) Ex vivo fluorescence imaging of the spinal cord and major organs 6 h after administration to assess tissue distribution.

UV–Vis-based quantification using a matrix-matched calibration curve showed that Ba@Se-MSN achieved an encapsulation efficiency (EE) of 93.5% and a loading capacity (LC) of 11.12%. After membrane coating, EE remained high (91.82%), whereas LC decreased modestly to 8.85% without detectable baicalin leakage, primarily reflecting the additional membrane weight. Coomassie-stained SDS-PAGE revealed a highly similar protein profile between Ba@Se-MSN&BV2 and native BV2 membranes, with retention of key membrane proteins (Fig. S2), corroborating successful biomimetic interface construction and aligning with the observed zeta potential shift. UV–Vis spectra showed that Ba@Se-MSN&BV2 preserved baicalin's characteristic peaks at 280 and 316 nm without shifts or new bands (Fig. 1E), indicating that encapsulation and membrane cloaking did not perturb the drug's spectral profile. In vitro release proceeded slowly under physiological-mimicking conditions but was markedly accelerated in the presence of 100 μM H₂O₂, resulting in a higher cumulative release (Fig. 1F). This ROS-accelerated profile mirrored oxidative fragmentation of the framework (Fig. S1), supporting a mechanism in which diselenide bond cleavage loosens the mesostructure, promotes pore opening, and enhances baicalin diffusion. Such ROS-triggered release is expected to better align local drug availability with pathological ROS surges in ROS-rich SCI lesions than passive, diffusion-driven release alone. Given its intravenous administration, Ba@Se-MSN&BV2 was further subjected to hemocompatibility evaluation, which showed acceptable blood compatibility under the tested conditions (Fig. S3). On this basis, the in vitro antioxidant capacity and in vivo spinal lesion accumulation of Ba@Se-MSN&BV2 were systematically evaluated.

2.2. In vitro antioxidative activity and in vivo lesion accumulation

Building on these physicochemical and biocompatibility characteristics, the in vitro antioxidant capacity and in vivo spinal lesion accumulation of Ba@Se-MSN&BV2 were then systematically evaluated. In ABTS and DPPH radical-scavenging assays, baicalin, Ba@Se-MSN, and Ba@Se-MSN&BV2 all displayed dose-dependent radical-scavenging activity. At equivalent concentrations and incubation times, Ba@Se-MSN&BV2 exhibited higher scavenging efficiency and a steeper dose–response curve (Fig. 1G–I, Fig. S4–S6), consistent with preserved antioxidant function and an added benefit from nanoscale carrier and interface engineering. Such enhancement at equivalent doses suggests that packaging baicalin within a ROS-responsive, high–surface-area carrier may improve the effective utilization of its radical-scavenging capacity. Fig. S7 further shows that the empty membrane-coated nanocarrier Se-MSN&BV2 exhibits only limited free radical-scavenging activity, suggesting that the major antioxidative effect is mainly attributable to baicalin delivery, whereas the carrier itself likely provides only a modest auxiliary contribution.

To translate the in vitro antioxidant advantage into in vivo lesion accessibility, biodistribution and spinal lesion accumulation were evaluated using coumarin-6 (C6) as a fluorescent tracer. After tail-vein injection of equivalent doses of C6-labeled formulations, in vivo imaging performed at 6 h post-injection revealed preferential fluorescence accumulation at the SCI lesion. While Ba@Se-MSN-C6 showed evident enrichment, Ba@Se-MSN&BV2-C6 produced a stronger and more spatially confined signal (Fig. 1J), suggesting that BV2 membrane cloaking augments lesion targeting and residence. This pattern is in line with the inflammation-tropic homing reported for other immune cell–derived biomimetic carriers and supports the rationale for using microglial membranes to guide spinal lesion localization [46,47]. Ex vivo imaging further confirmed robust Ba@Se-MSN&BV2-C6 accumulation at the lesion and revealed predominant hepatic and renal signals, consistent with systemic clearance and metabolism of the intravenously administered nanoplatform (Fig. 1K). Together, these data indicate that BV2 membrane cloaking enhances both lesion accumulation and residence. In combination with the ROS-responsive, diselenide-gated carrier, this design preserves and reinforces the antioxidant performance of baicalin, ensuring that the payload is not only potent in vitro but also preferentially delivered to, and retained within, ROS-rich SCI lesions. However, because the present imaging analyses were conducted at the whole-lesion level, they do not directly define nanoparticle uptake by specific cell populations within the lesion microenvironment.

2.3. ROS mitigation and apoptosis inhibition in HT22 cells

To evaluate antioxidant performance under neuronal oxidative stress, glutamate-challenged mouse hippocampal HT22 cells were used as an excitotoxicity model [48,49]. Cell Counting Kit-8 (CCK-8) assays indicated no evident cytotoxicity of baicalin, Ba@Se-MSN, or Ba@Se-MSN&BV2 at concentrations up to 25 μg/mL. Accordingly, 25 μg/mL was selected for subsequent experiments (Fig. 2A). At this dose, oxidative burden was assessed using complementary probes, with DCFH-DA reporting total ROS and DHE detecting intracellular superoxide. Flow cytometry revealed a rightward shift and elevated mean fluorescence intensity (MFI) in the injury group, consistent with robust ROS accumulation. All treatments attenuated ROS signals, with Ba@Se-MSN&BV2 eliciting the most pronounced normalization toward control levels (Fig. 2B and C). Confocal microscopy corroborated the flow cytometric findings. DCFH-DA fluorescence was markedly increased after injury and was only partially reduced by baicalin or Ba@Se-MSN (Fig. 2D and E), whereas Ba@Se-MSN&BV2 produced the greatest reduction. DHE fluorescence followed the same trend (Fig. 2F and G), further supporting superior suppression of superoxide by the biomimetic nanoplatform.

Fig. 2.

In vitro assessment of antioxidant activity, apoptosis, and neuronal morphology. (A) CCK-8 assay of the effects of baicalin, Ba@Se-MSN and Ba@Se-MSN&BV2 on HT22 cell viability (n = 3). (B) Flow cytometric analysis of overall ROS in HT22 cells stained with DCFH-DA after the indicated treatments. (C) Quantification of DCFH-DA fluorescence by flow cytometry reflecting total ROS levels (n = 3). (D) Representative DCFH-DA fluorescence images of HT22 cells in each group (scale bar = 100 μm). (E) Quantification of mean DCFH-DA fluorescence intensity (n = 3). (F) Representative DHE fluorescence images of HT22 cells in each group (scale bar = 100 μm). (G) Quantification of mean DHE fluorescence intensity (n = 3). (H) Flow cytometric analysis of apoptosis in HT22 cells following the indicated treatments. (I) Quantification of apoptosis rates in HT22 cells in each group (n = 3). (J) Representative βIII-tubulin immunofluorescence images of primary hippocampal neurons derived from 1-day-old Sprague–Dawley rats after glutamate injury and subsequent treatment with baicalin, Ba@Se-MSN, or Ba@Se-MSN&BV2 (scale bar = 50 μm). (K) Quantification of total neurite length in primary hippocampal neurons (n = 30). (L) Quantification of total branch number in primary hippocampal neurons (n = 30). (M) Quantification of primary neurite length in primary hippocampal neurons (n = 30). (N) Quantification of primary neurite number in primary hippocampal neurons (n = 30). (ns, not statistically significant; ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001 vs. Injury).

To assess downstream apoptotic outcomes, Annexin V/PI dual staining followed by flow cytometry was performed. Early and late apoptotic populations increased markedly in the injury group, whereas all treatments reduced total apoptosis, with Ba@Se-MSN&BV2 yielding the lowest apoptotic fraction among the treated groups (Fig. 2H and I). These findings are consistent with ROS attenuation contributing to a survival advantage at matched doses. Taken together, Ba@Se-MSN&BV2 was more effective than baicalin and Ba@Se-MSN in reducing total ROS and superoxide and in suppressing glutamate-induced apoptosis in HT22 cells. This superiority is consistent with enhanced intracellular bioavailability conferred by the biomimetic membrane and ROS-responsive release mediated by diselenide gating [21,50]. The improved efficacy is likely related to increased cellular uptake and pathology-coupled drug liberation, providing a functional basis for subsequent analyses of neuronal structural protection.

2.4. Microglial inflammation attenuation and M1/M2 polarization remodeling

Building on the observed neuronal antioxidant and anti-apoptotic effects, the impact of Ba@Se-MSN&BV2 on microglial inflammatory activation and polarization dynamics was next evaluated. In lipopolysaccharide (LPS)-stimulated BV2 microglia, immunoblotting revealed marked increases in IL-1β, IL-6, TNF-α, and iNOS, consistent with a proinflammatory M1-like phenotype. Treatment with Ba@Se-MSN&BV2 reduced these mediators while elevating Arg1 and IL-10 (Fig. S8), indicating concurrent suppression of proinflammatory outputs and enhancement of anti-inflammatory markers. Immunofluorescence staining corroborated these findings: IL-6 signals were increased by LPS, whereas Ba@Se-MSN&BV2 decreased IL-6 and enhanced IL-10 fluorescence (Fig. S9).

Microglial polarization markers were then examined. Immunoblotting showed that LPS increased the M1 marker CD80 and reduced the M2 marker CD206, whereas Ba@Se-MSN&BV2 lowered CD80 and restored CD206 expression (Fig. S10), indicating a shift from a proinflammatory toward a reparative phenotype. Immunofluorescence analysis mirrored these changes, with elevated CD80 and reduced CD206 after LPS, and an opposite pattern following Ba@Se-MSN&BV2 treatment, accompanied by increased IL-10 staining (Fig. S11). Together with the antioxidative findings, these results indicate that Ba@Se-MSN&BV2 exerts superior immunomodulatory effects, consistent with enhanced intracellular exposure conferred by the biomimetic membrane and ROS-responsive release mediated by diselenide gating. These attributes act synergistically to mitigate the inflammatory microenvironment and bias microglial M1/M2 balance toward M2-like reparative phenotypes, thereby creating a more permissive immune milieu for spinal cord repair.

2.5. Morphological reconstruction and neurite outgrowth in primary neurons

In light of the reduced oxidative stress and the shift toward reparative microglial phenotypes, the impact of Ba@Se-MSN&BV2 on neurite structural reconstruction in primary hippocampal neurons was evaluated. CCK-8 assays detected no evident cytotoxicity at 12.5 μg/mL for any formulation; this concentration was therefore used for subsequent experiments (Fig. S12). Representative images revealed marked neurite shortening and reduced branching after injury. All treatments partially rescued these deficits, with Ba@Se-MSN&BV2 producing the most pronounced restoration of neurite architecture in the representative fields (Fig. 2J). Morphometric quantification showed that, relative to injury, Ba@Se-MSN&BV2 significantly increased total neurite length, branching complexity, and both the length and number of primary neurites (Fig. 2K–N). A scratch-based neurite outgrowth assay yielded consistent findings. Gap coverage was markedly impaired after injury, whereas all treatments improved coverage and closure, with Ba@Se-MSN&BV2 achieving the greatest recovery of gap-bridging neurite outgrowth (Fig. S13). Integrating the morphometric and scratch assay outcomes, Ba@Se-MSN&BV2 was more effective than baicalin or Ba@Se-MSN in promoting post-injury neurite regeneration and network reconstruction at matched doses, thereby linking redox modulation to structural repair in injured neurons.

2.6. In vivo control of oxidative stress

Based on the in vitro antioxidant and neuroprotective findings, in vivo ROS control was next assessed in an impactor-induced mouse spinal cord contusion model. Mice received the assigned formulations via tail-vein injection once daily for seven consecutive days starting immediately after injury, thereby covering the acute oxidative–neuroinflammatory window (Fig. 3A). Acute oxidative burden was quantified on day 3 after injury using chemiluminescence imaging with L-012 after three consecutive doses (Fig. 3B). Normal and sham mice exhibited only baseline luminescence, whereas injured mice showed a marked increase, indicating an acute oxidative burst. All treatments reduced L-012 signal intensity, with Ba@Se-MSN&BV2 eliciting the most pronounced attenuation and thus the strongest in vivo suppression of ROS at matched doses. These data are consistent with BV2 membrane–mediated lesion targeting combined with ROS-responsive baicalin release via diselenide gating, which together may amplify local antioxidant effects during the acute oxidative–neuroinflammatory phase of SCI.

Fig. 3.

Integrated evaluation of oxidative stress burden and motor function following in vivo administration. (A) Schematic diagram of the in vivo experimental protocol. (B) L-012 chemiluminescence imaging of reactive oxygen species at the spinal cord lesion site after treatment with baicalin, Ba@Se-MSN or Ba@Se-MSN&BV2. (C) Time course of Basso Mouse Scale (BMS) scores from surgery to 2 months post-injury (n = 5). (D) Representative images of hindlimb weight-bearing and standing behavior at 2 months after surgery. (E) Representative images of unweighted swimming behavior at 2 months after surgery. (F) Schematic representation of hindlimb joint trajectories during locomotion; the red curve represents the descending phase of the hindlimb trajectory from its highest position during the gait cycle. (G) Time-resolved joint angle curves illustrating the angular changes of major hindlimb joints. (ns, not statistically significant; ∗∗P < 0.01; ∗∗∗P < 0.001 vs. Injury).

2.7. Evaluation of motor and gait recovery

Given the early in vivo suppression of ROS, long-term motor recovery was next assessed in the same contusion model. Basso Mouse Scale (BMS) scores gradually increased over the 2-month follow-up and tended to stabilize at later time points, indicating progressive locomotor improvement after injury. At the endpoint, Ba@Se-MSN&BV2 achieved the highest BMS scores, Ba@Se-MSN provided a moderate benefit, whereas the injury and baicalin groups showed only limited recovery (Fig. 3C and D). Under non–weight-bearing conditions, the Louisville Swim Scale (LSS) further showed that Ba@Se-MSN&BV2 exhibited the best performance in hindlimb excursions, alternating rhythm, and postural balance (Fig. 3E, Fig. S14), indicating enhanced hindlimb drive and postural control.

For higher-resolution kinematic profiling, gait parameters were analyzed by high-speed lateral videography with joint tracking. The injury and baicalin groups showed markedly reduced hip, knee, and ankle ranges of motion, prolonged gait cycles, and disrupted inter-joint phase relationships. Ba@Se-MSN and Ba@Se-MSN&BV2 increased joint excursion amplitudes and partially normalized gait timing and coordination, with Ba@Se-MSN&BV2 showing restoration closest to control (Fig. 3F and G). Radar-plot integration of multiple kinematic measures, including ankle, knee, and hip excursion, toe and sacral heights, and mean velocity, ranked Ba@Se-MSN&BV2 highest across key motor dimensions, in line with the behavioral scales (Fig. S15).

At 2 months after injury, footprint analysis showed shortened stride length and increased stride width in the injury and baicalin groups, indicative of disrupted walking rhythm and compromised weight-bearing stability. Ba@Se-MSN partially improved both parameters but did not reach control levels, whereas Ba@Se-MSN&BV2 yielded the greatest recovery in stride length and largely normalized stride width (Fig. 4A–C). Metal-grid walking and inclined-plane tests showed a similar hierarchy of improvement: the injury and baicalin groups exhibited higher foot-fault rates and lower maximal holding angles, Ba@Se-MSN induced partial recovery, and Ba@Se-MSN&BV2 minimized fault rates and restored holding angles toward normal values (Fig. S16A-B), indicating concurrent gains in coordination and balance.

Fig. 4.

Integrated quantitative framework for assessing gait and electrophysiological function in mice. (A) Footprint analysis visualizing locomotor trajectories after spinal cord injury. (B) Quantification of stride length to evaluate stepping capacity (n = 5). (C) Quantification of step width to assess stability and left–right symmetry (n = 5). (D) Gait sequence plots depicting paw contact order and phase relationships between left and right limbs (n = 5). (E) Three-dimensional paw pressure maps illustrating load distribution and propulsion patterns. (F) Schematic diagram of stimulation and recording electrode placement for motor evoked potential (MEP) measurements. (G) Quantification of MEP amplitude recorded at the motor cortex to evaluate cortical output efficacy (n = 5). (H) Quantification of MEP amplitude recorded rostral to the lesion to assess local spinal conduction (n = 5). (I) Quantification of MEP amplitude recorded caudal to the lesion to evaluate conduction across the injured segment (n = 5). (J) Representative MEP waveforms showing latency and amplitude scales at each recording site. (ns, not statistically significant; ∗∗P < 0.01; ∗∗∗P < 0.001 vs. Injury).

CatWalk analysis at the 2-month endpoint was used to quantitatively evaluate walking quality and load-bearing capacity. Step-sequence plots showed irregular, discontinuous stepping in the injury group, partial reorganization with Ba@Se-MSN, and near-normal, symmetric left–right alternation with Ba@Se-MSN&BV2 (Fig. 4D). Three-dimensional paw-contact heatmaps and parametric quantification showed that Ba@Se-MSN&BV2 reconstructed a typical biphasic contact profile and achieved the highest proportion of normal step sequences, as well as the greatest recovery in ipsilateral contact intensity and walking speed, whereas baicalin had minimal effects (Fig. 4E, Fig. S17). Taken together, across behavioral and gait readouts, Ba@Se-MSN&BV2 was the most effective in correcting gait geometry, restoring weight-bearing, and recovering rhythmic coordination at matched doses. These functional gains are consistent with sustained intralesional actions enabled by lesion accumulation and ROS-responsive baicalin release, and they provide a rationale for subsequent histological and mechanistic interrogation.

2.8. Motor-evoked potentials reveal corticospinal reconnection

To test whether behavioral recovery was accompanied by restored corticospinal conduction, a three-site motor-evoked potential (MEP) protocol was performed at the 2-month endpoint to evaluate cortical output and translesional transmission. Electrode placement followed the schematic in Fig. 4F. Transcranial stimulation elicited cortical MEPs, and rostral and caudal spinal MEPs were recorded simultaneously above and below the lesion. Cortical output MEP amplitudes were substantially reduced in the injury and baicalin groups, partially restored by Ba@Se-MSN, and most strongly restored by Ba@Se-MSN&BV2 (Fig. 4G), indicating superior recovery of corticospinal output with the biomimetic nanoplatform. MEPs recorded rostral to the lesion showed no between-group differences (Fig. 4H), indicating preserved conduction proximal to the injury. The key divergence emerged at a caudal site approximately 5 mm distal to the lesion. Responses were almost undetectable in the injury and baicalin groups, present at moderate amplitudes with Ba@Se-MSN, and robustly enhanced with Ba@Se-MSN&BV2, approaching sham levels (Fig. 4I). Representative waveforms showed the greatest amplitude recovery and a tendency toward reduced latency in the Ba@Se-MSN&BV2 group, smaller improvements with Ba@Se-MSN, and weak or absent traces in the injury and baicalin groups (Fig. 4J). The pattern of electrophysiological recovery across cortical, rostral, and caudal sites paralleled improvements in footprint, CatWalk, and balance measures, supporting an association between Ba@Se-MSN&BV2–mediated behavioral gains and partial corticospinal reconnection with enhanced translesional transmission. These electrophysiological findings suggest that the behavioral improvements were accompanied by recovery of descending signal conduction across the lesion, rather than merely reflecting compensatory motor performance.

2.9. Tissue reconstruction and neuronal preservation

To corroborate tissue-level repair underlying the sustained behavioral and electrophysiological improvements, histological analyses were performed at the 2-month endpoint. Gross inspection and H&E staining showed that injury controls developed marked cord thinning, interruption of continuity, and large cavitations with parenchymal collapse and a thickened lesion rim, whereas Ba@Se-MSN&BV2 yielded the smallest cavity, better preserved cord continuity, and a smoother, more continuous rim, suggesting more complete parenchymal refilling (Fig. 5A and B). GFAP/NFH co-staining revealed in injury controls a dense GFAP-positive astroglial scar ring encasing fragmented, disorganized NFH-positive axons. Baicalin induced only minimal change, whereas Ba@Se-MSN partially attenuated glial scarring and improved NFH-positive axonal alignment (Fig. S18A). In contrast, Ba@Se-MSN&BV2 markedly reduced GFAP-positive scarring and restored continuous NFH-positive bundles traversing the lesion and realigning with surrounding white-matter tracts (Fig. 5C), in line with recovered caudal MEPs and indicative of enhanced translesional connectivity. NeuN staining showed a parallel gradient of neuronal preservation, with sparse and discontinuous perilesional NeuN⁺ neurons in injury controls, minimal rescue by baicalin, increased density and continuity with Ba@Se-MSN, and the highest, most continuous perilesional NeuN⁺ distribution in the Ba@Se-MSN&BV2 group (Fig. 5D; Fig. S18B). Systemic histopathology did not reveal appreciable lesions in major organs, including heart, liver, spleen, lung, kidney, and brain (Fig. S19), supporting a favorable biocompatibility profile at the tested dose. Collectively, these multilevel histological findings parallel the behavioral and electrophysiological data, supporting that Ba@Se-MSN&BV2 reduces cavitation and astroglial scarring, promotes translesional axonal continuity, and preserves neurons, thereby providing an anatomical substrate for sustained pathway integrity and long-term functional recovery. Accordingly, the locomotor improvements observed in behavioral assays appear to reflect integrated neurofunctional recovery supported by both electrophysiological restoration and structural preservation of the injured spinal cord.

Fig. 5.

Histological and immunofluorescence assessment after spinal cord injury in mice. (A) Gross brain–spinal cord specimen and schematic localization of the lesion segment. (B) Hematoxylin and eosin (HE) staining of transverse spinal cord sections showing overall histoarchitecture and lesion extent (scale bar = 200 μm). (C) Immunofluorescence staining for NFH and GFAP on transverse spinal cord sections (scale bar = 200 μm). (D) Immunofluorescence staining for the neuronal marker NeuN on transverse spinal cord sections (scale bar = 200 μm).

2.10. Identification and validation of CHCHD2 as a key target for mitochondrial homeostasis

The preceding data indicate that Ba@Se-MSN&BV2 alleviates oxidative stress and promotes neuronal structural and functional recovery. To elucidate the potential molecular basis underlying these effects, with a particular focus on oxidative stress–related signaling pathways that shape secondary SCI pathogenesis, label-free quantitative proteomics was performed in glutamate-injured HT22 cells. After normalization and batch correction, sample distributions were highly concordant, enabling robust differential analysis (Fig. S20A-B; Fig. S21). Differentially expressed proteins (DEPs) were filtered using a treatment-reversal strategy, selecting proteins that were downregulated by injury and rescued by Ba@Se-MSN&BV2, or upregulated by injury and reduced by treatment. This yielded two reversal sets comprising 364 proteins shifted upward and 297 shifted downward upon Ba@Se-MSN&BV2 exposure (Fig. 6A). Hierarchical clustering clearly separated Injury from Ba@Se-MSN&BV2 groups and showed tight within-group clustering (Fig. S20C), indicating broad correction of injury-induced expression dysregulation. Given the central role of redox imbalance in SCI, oxidative stress–related and metabolic proteins were examined in greater detail. Heatmaps showed that most glutamate-perturbed redox- and metabolism-related proteins trended back toward baseline after Ba@Se-MSN&BV2 treatment (Fig. 6B), suggesting partial normalization of redox–metabolic programs. Chord-plot enrichment analysis highlighted CHCHD2 as a hub node connected to oxidative response, hypoxia signaling, and mitochondrial homeostasis pathways (Fig. 6C). Given that mitochondrial dysfunction is a major pathological driver of secondary SCI and promotes ROS amplification and neuronal vulnerability [51], these proteomic findings suggest that Ba@Se-MSN&BV2 does not simply modulate isolated protein changes, but partially reprograms coordinated molecular networks related to oxidative stress control and mitochondrial homeostasis. Among the treatment-reversed candidates, CHCHD2 was prioritized for further validation because its close association with mitochondrial homeostasis made it a plausible mechanistic link between the proteomic changes and the neuroprotective effects of Ba@Se-MSN&BV2.

Fig. 6.

Proteomic screening and experimental framework for validating CHCHD2-mediated mechanisms. (A) Venn diagram illustrating the overlap between differentially expressed proteins and treatment-reversed subsets. (B) Heatmap showing the expression profiles of oxidative stress–related proteins. (C) Gene Ontology (GO) chord plot depicting biological processes and pathways associated with candidate proteins. (D) Western blot analysis of CHCHD2 expression in HT22 cells under the indicated conditions. (E) Densitometric quantification of CHCHD2 bands (n = 3). (F) Western blot analysis of CHCHD2 expression in the injury background following si-CHCHD2 transfection and Ba@Se-MSN&BV2 treatment. (G) Densitometric quantification of CHCHD2 (n = 3). (H) Flow-cytometric MitoSOX histograms showing the distribution of mitochondrial superoxide fluorescence. (I) Quantification of MitoSOX-positive populations by flow cytometry (n = 3). (J) Flow-cytometric analysis of JC-1 fluorescence distribution reflecting mitochondrial membrane potential states. (K) Confocal JC-1 images showing the subcellular localization of aggregate and monomer signals (scale bar = 10 μm). (L) Annexin V–FITC/PI flow-cytometric analysis of apoptotic distribution. (M) Western blot analysis of Bcl-2 and Bax expression in the injury background following si-CHCHD2 transfection and Ba@Se-MSN&BV2 treatment. (N) Quantification of the BAX/Bcl-2 ratio (n = 3). (O) Representative immunofluorescence images of primary hippocampal neurons in the injury plus si-CHCHD2 background with or without Ba@Se-MSN&BV2 treatment, showing neuronal morphological features (scale bar = 50 μm). (P) Quantification of total branch number in primary hippocampal neurons (n = 30). (Q) Quantification of primary branches length in primary hippocampal neurons (n = 30). (ns, not statistically significant; ∗, P < 0.05; ∗∗, P < 0.01; ∗∗∗, P < 0.001 vs. si-CHCHD2/Injury).

Western blot analysis in HT22 cells substantiated the proteomic findings, showing that CHCHD2 was highly expressed under basal conditions, markedly reduced after glutamate injury, and restored toward baseline following Ba@Se-MSN&BV2 treatment (Fig. 6D and E), thereby closely recapitulating the proteomic trend. To further assess the functional relevance of CHCHD2, an si-CHCHD2 knockdown model was successfully established, in which CHCHD2 expression was efficiently suppressed (Fig. S22). Notably, Ba@Se-MSN&BV2 partially rescued CHCHD2 expression even under CHCHD2-silencing conditions (Fig. 6F and G). Immunofluorescence analysis showed a concordant pattern, with CHCHD2 signals weakened by glutamate injury and further reduced after si-CHCHD2 transfection, whereas Ba@Se-MSN&BV2 treatment substantially restored CHCHD2 immunoreactivity (Fig. S23). Taken together, these results validate CHCHD2 as a treatment-responsive candidate identified by proteomic screening and support its potential involvement in the mitochondrial protective effects of Ba@Se-MSN&BV2.

2.11. CHCHD2-mediated mitochondrial homeostasis regulation and neuronal protection

To determine how CHCHD2 loss and its modulation by Ba@Se-MSN&BV2 affect mitochondrial homeostasis and neuronal integrity, mtROS and total ROS were quantified by flow cytometry using MitoSOX (mitochondrial superoxide) and DCFH-DA (total ROS). CHCHD2 silencing markedly increased mitochondrial superoxide, whereas Ba@Se-MSN&BV2 attenuated this elevation (Fig. 6H and I). DCFH-DA fluorescence showed a similar pattern, with CHCHD2 knockdown enhancing total ROS and Ba@Se-MSN&BV2 partially reversing this increase (Fig. S24). Dual-channel confocal imaging corroborated concurrent reductions in mtROS and total ROS under CHCHD2-silencing conditions after Ba@Se-MSN&BV2 treatment (Fig. S25-S26). Mitochondrial functional status was evaluated using JC-1. CHCHD2 knockdown further decreased mitochondrial membrane potential, whereas Ba@Se-MSN&BV2 partially restored JC-1 aggregation, consistent with improved ΔΨm (Fig. 6J, Fig. S27). Confocal JC-1 imaging showed the same trend, with red-to-green shifts exacerbated by CHCHD2 silencing and mitigated by Ba@Se-MSN&BV2 (Fig. 6K, Fig. S28). Apoptosis assays demonstrated that CHCHD2 knockdown increased total apoptosis, while Ba@Se-MSN&BV2 substantially reduced apoptotic cells and lowered the Bax-to-Bcl-2 ratio (Fig. 6L–N, Fig. S29). In primary hippocampal neurons, CHCHD2 silencing further impaired neurite outgrowth, whereas Ba@Se-MSN&BV2 rescued neurite length and branching complexity under CHCHD2-deficient conditions (Fig. 6O–Q, Fig. S30).

Given that CHCHD2 silencing aggravated mitochondrial dysfunction and neuronal injury, gain-of-function experiments were performed to determine whether restoration of CHCHD2 was sufficient to recapitulate the protective phenotype. In HT22 cells, transfection with a GFP-CHCHD2 expression plasmid markedly increased CHCHD2 expression, as confirmed by Western blotting (Fig. S31). Under the same injury conditions, enforced CHCHD2 expression reduced mitochondrial ROS accumulation, as indicated by MitoSOX flow cytometry (Fig. S32). In primary hippocampal neurons subjected to glutamate injury, neurite morphology was quantified in GFP-positive transfected neurons, demonstrating that CHCHD2 overexpression improved neurite architecture, with increased neurite length and branching complexity compared with injured neurons (Fig. S33). Together with the loss-of-function and rescue data, these gain-of-function findings identify CHCHD2 as a critical regulator of mitochondrial homeostasis in the injured neuronal context. CHCHD2 deficiency exacerbated mitochondrial and cellular injury, whereas restoration of CHCHD2 was sufficient to suppress mitochondrial oxidative stress and promote neuronal structural recovery. These findings further support CHCHD2 as a functionally relevant downstream mediator of the neuroprotective effects of Ba@Se-MSN&BV2.

3. Conclusion

In this study, a ROS-responsive, microglia-mimetic nanoplatform (Ba@Se-MSN&BV2) tailored to the secondary phase of SCI was constructed, in which BV2 cell membrane–mediated lesion homing together with diselenide bond–triggered, on-demand baicalin release enables targeted intervention within the oxidative and proinflammatory microenvironment. Systematic in vitro and in vivo investigations demonstrate that this platform alleviates oxidative stress and mitochondrial dysfunction, attenuates apoptosis, promotes axonal and neuronal preservation and improves motor-circuit function. Label-free quantitative proteomics combined with si-CHCHD2 knockdown and rescue experiments indicate that the neuroprotective efficacy of Ba@Se-MSN&BV2 is closely associated with CHCHD2-mediated restoration of mitochondrial homeostasis, thereby helping translate relief of oxidative burden into improved neuronal structure and function. Taken together, these findings highlight a microenvironment-responsive, mitochondria-anchored nanotherapeutic strategy for SCI and suggest that CHCHD2-mediated mitochondrial stabilization may constitute a tractable intervention axis for neuroinflammatory injury driven by disruption of redox homeostasis. From a translational perspective, this platform shows potential because it can be systemically administered, preferentially accumulate at the lesion, and provide microenvironment-responsive drug release. However, further studies are still needed to address scalable manufacturing, pharmacokinetics, long-term biosafety, and validation in more clinically relevant settings. In addition, because the present work used a single dose level and a fixed dosing schedule, future studies should systematically evaluate dose-response relationships, dosing frequency, treatment duration, and therapeutic window to optimize the pharmacological profile and translational applicability of this nanoplatform. Although Ba@Se-MSN&BV2 demonstrated clear advantages in the current study, future head-to-head comparisons with representative benchmark SCI therapies will be important to better define its relative translational value and innovation. Addressing these issues will be essential for advancing this biomimetic nanoplatform toward future clinical application.

4. Materials and methods

4.1. Materials

Major reagents, antibodies, and their sources are listed in Supplementary Materials. Female C57BL/6 mice (6–8 weeks of age; Guangzhou Ruige Biotechnology Co., Ltd., Guangzhou, China) were used throughout the study. Animals were housed under specific pathogen-free conditions (approximately 25 °C, 50–60% relative humidity) with free access to food and water.

4.2. Synthesis and characterization of materials

Se-MSNs were synthesized according to a previously reported protocol [52]. Briefly, CTAB was used as a soft template and ammonia as a base catalyst to co-condense BTESePD and TEOS in an alkaline CTAB-containing aqueous solution under continuous stirring at 80 °C until gelation. The resulting product was collected by centrifugation, washed, dried, and refluxed in an ethanol/HCl mixture to remove CTAB, yielding diselenide-bridged Se-MSNs. For baicalin loading, dried Se-MSNs were dispersed in baicalin solution (1.0 mg mL⁻¹ in 50% ethanol/PBS, pH 7.4) at a drug-to-carrier mass ratio of approximately 1:5, sonicated briefly in the dark, and stirred at 25 °C for 12–18 h. The suspension was then centrifuged, the pellet was washed with PBS to remove unbound drug, and the solid was vacuum-dried at 40 °C to obtain Ba@Se-MSN. BV2 cell membranes were isolated using hypotonic lysis and repeated freeze–thaw cycles, followed by differential centrifugation to collect membrane fractions. The crude membranes were extruded through 400 nm and 200 nm polycarbonate membranes to form BV2-CMVs. Ba@Se-MSN were mixed with BV2-CMVs at a membrane protein-to-nanoparticle mass ratio of 1:1, incubated on ice, and extruded at 4 °C through 200 nm and 100 nm membranes for multiple passes. After removal of free membrane fragments by centrifugation, the pellet was resuspended in PBS to obtain Ba@Se-MSN&BV2.

Nanoparticle morphology and mesostructure were examined by SEM and TEM. Dynamic light scattering (Litesizer 500) was used to determine the hydrodynamic diameter and polydispersity index of Ba@Se-MSN&BV2, and zeta potentials of Se-MSNs, Ba@Se-MSN, and Ba@Se-MSN&BV2 were measured using a Malvern Zetasizer Nano. A matrix-matched baicalin calibration curve was established by UV–Vis spectrophotometry (Lambda 365), and drug LC and EE were calculated accordingly. In vitro baicalin release was assessed in PBS (pH 7.4) at 37 °C under gentle shaking. At predetermined time points, supernatants were collected and replaced with equal volumes of fresh PBS. Baicalin concentrations were quantified by UV–Vis spectroscopy, and cumulative release profiles were plotted. Oxidation-responsive release mediated by the diselenide framework was evaluated in parallel in PBS with or without 100 μM H₂O₂, representing near-physiological and oxidative conditions, respectively. Hemocompatibility of Ba@Se-MSN&BV2 was evaluated by a standard hemolysis assay. Briefly, erythrocytes isolated from fresh mouse blood were washed with PBS and diluted to a 2% suspension, followed by incubation with Ba@Se-MSN&BV2 at 37 °C for 2 h. PBS and 5% Triton X-100 served as the negative and positive controls, respectively. After centrifugation, the absorbance of the supernatant was measured at 540 nm, and the hemolysis ratio was calculated accordingly.

4.3. Free-radical scavenging assays

Free-radical scavenging activity was evaluated under identical solvent systems and equivalent baicalin molar inputs across a range of mass concentrations to enable direct comparison of the in vitro scavenging capacities of baicalin, Ba@Se-MSN, Se-MSN&BV2 and Ba@Se-MSN&BV2. Samples were mixed with radical working solutions at a 1:1 vol ratio, incubated at room temperature in the dark, and absorbance was recorded at predefined time points over 0–120 min to monitor scavenging kinetics. For the ABTS assay, ABTS stock solution was prepared by mixing 7 mM ABTS with 2.45 mM potassium persulfate and allowing the mixture to react for 12–16 h in the dark to generate ABTS⁺·. The stock was then diluted with the corresponding solvent to obtain a working solution, mixed with samples at equal volume, and the absorbance was measured at 734 nm. For the DPPH assay, a 0.10 mM DPPH solution in ethanol was mixed with samples at equal volume, incubated at room temperature in the dark, and the absorbance was measured at 517 nm. Radical scavenging efficiency (%) was calculated as (A_control − A_sample)/A_control × 100%.

4.4. In vivo targeting evaluation

C6 was used as a hydrophobic fluorescent probe to assess the in vivo biodistribution and SCI targeting of Ba@Se-MSN and Ba@Se-MSN&BV2. Under the same baicalin-equivalent dosing conditions as in the therapeutic experiments, C6 was co-loaded into the formulations at 1 wt% relative to baicalin. After gentle stirring for 12 h in the dark, particles were collected by centrifugation, washed repeatedly with PBS until no detectable fluorescence remained in the supernatant to remove free dye, and resuspended in PBS to obtain C6-Ba@Se-MSN and C6-Ba@Se-MSN&BV2. SCI mice were randomly assigned to treatment groups and administered the corresponding formulations via tail-vein injection under isoflurane anesthesia. At 6 h post-injection, spinal cords and major organs were harvested and subjected to ex vivo fluorescence imaging using an AniView 600 system.

4.5. Cytotoxicity assay

The cytocompatibility of baicalin, Ba@Se-MSN, and Ba@Se-MSN&BV2 was evaluated using a Cell Counting Kit-8 (CCK-8) assay. HT22 cells and primary neurons were seeded into 96-well plates and, after stable attachment, were exposed to serial concentrations of baicalin, Ba@Se-MSN, or Ba@Se-MSN&BV2 for 24 h. The culture medium was then replaced with fresh medium containing 10% CCK-8 working solution (100 μL per well) and incubated at 37 °C for 1–2 h. Absorbance was measured at 450 nm using a microplate reader, and relative cell viability was calculated by normalizing to untreated control wells.

4.6. Intracellular ROS and superoxide measurements

Total intracellular ROS and superoxide levels were assessed using DCFH-DA and DHE fluorescent probes, respectively. The antioxidant effects of baicalin, Ba@Se-MSN, and Ba@Se-MSN&BV2 were evaluated by combining flow cytometry with confocal microscopy. After attachment, HT22 cells were exposed to glutamate (final concentration 120 mM, as determined in preliminary experiments) for 12 h to induce oxidative stress, followed by treatment with baicalin, Ba@Se-MSN, or Ba@Se-MSN&BV2 at equivalent baicalin doses for an additional 24 h. At the end of treatment, cells were incubated with DCFH-DA (10 μM) or DHE (5 μM) working solutions prepared in serum-free, phenol-red–free medium for 20–30 min at 37 °C in the dark, washed three times with PBS, and then resuspended or overlaid with phenol-red–free complete medium. Mean fluorescence intensity (MFI) was quantified by flow cytometry, and representative images were acquired by confocal microscopy to provide spatial localization and qualitative visualization of ROS and superoxide signals.

4.7. Apoptosis analysis

Annexin V-FITC/PI dual-staining flow cytometry was used to assess the anti-apoptotic effects of baicalin, Ba@Se-MSN, and Ba@Se-MSN&BV2 under oxidative injury. HT22 cells were subjected to oxidative injury as described in Section 4.6, then treated with baicalin, Ba@Se-MSN, or Ba@Se-MSN&BV2 at equivalent baicalin doses for 24 h. After treatment, culture supernatants and trypsinized adherent cells were pooled, washed with PBS, centrifuged at 300 g for 5 min, and resuspended in binding buffer. Annexin V-FITC and PI (5 μL each) were added and incubated for 10–15 min at room temperature in the dark. Binding buffer was added to 400–500 μL, and samples were immediately analyzed by flow cytometry with at least 1×10^4 events collected. Early and late apoptosis rates were calculated using Annexin V/PI quadrant gating.

4.8. Western blotting

Western blotting was performed to analyze protein expression changes under different treatments. Cells or tissues were lysed on ice for 30 min in pre-chilled RIPA buffer, followed by centrifugation at 12,000 g for 15 min at 4 °C to collect supernatants. Protein concentrations were determined by BCA assay, mixed with 5× loading buffer, and denatured at 95 °C for 5 min. Equal amounts of protein (20–30 μg per lane) were separated by SDS–PAGE and transferred onto PVDF membranes. Membranes were blocked with 5% non-fat milk for 1 h at room temperature, incubated with primary antibodies overnight at 4 °C, and then with HRP-conjugated secondary antibodies for 1 h at room temperature. Bands were visualized using ECL chemiluminescence. Densitometry was performed in ImageJ, and relative expression was normalized to internal controls.

4.9. Immunofluorescence staining

Immunofluorescence staining was used to assess subcellular localization and expression of target proteins. After treatment, cells were gently rinsed with PBS, fixed with 4% paraformaldehyde for ∼15 min at room temperature, washed three times, permeabilized with 0.1% Triton X-100 for 10 min, washed again, and blocked with 5% BSA for 1 h at room temperature. Primary antibodies were applied at recommended dilutions and incubated overnight at 4 °C. After three PBS washes, fluorophore-labeled secondary antibodies were added and incubated for 1 h at room temperature in the dark. Nuclei were counterstained with Hoechst 33,342 for 5 min. All images were acquired under identical laser power, gain, and exposure settings. Mean fluorescence intensity or integrated density was quantified in ImageJ and normalized to controls for group comparisons.

4.10. In vitro neuroprotection

Primary hippocampal neurons were prepared from 1-day-old Sprague–Dawley rats. After gentle papain digestion, cells were seeded onto poly-L-lysine–coated dishes and cultured for 48 h to allow stable attachment and neurite outgrowth. Oxidative injury was induced with glutamate, after which baicalin, Ba@Se-MSN, or Ba@Se-MSN&BV2 was added at equivalent baicalin doses and co-cultured for 24 h. Neurons were then subjected to Tuj1 immunofluorescence staining. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, blocked with 5% BSA, incubated with anti-β-III-tubulin (1:1000) overnight at 4 °C, followed by fluorescent secondary antibodies (1:1000) at room temperature in the dark. After DAPI nuclear staining, images were captured under identical parameters. Neurite length, total branch number, and related morphological parameters were quantified using ImageJ to compare neuroprotective effects across treatments.

4.11. Establishment of the spinal cord injury model

Female C57BL/6 mice (6–8 weeks old) were randomly assigned to the Normal, Sham, Injury, Baicalin, Ba@Se-MSN, and Ba@Se-MSN&BV2 groups. After deep anesthesia with intraperitoneal 3% tribromoethanol, a laminectomy was performed at T8–T10 to expose the spinal cord. A standardized contusion injury was produced at T9 using a spinal cord impactor (impact depth 1 mm, dwell time 0.5 s) [53]. Sham mice received laminectomy without impact. Postoperatively, animals received routine warming, analgesia, and antibiotic prophylaxis, and bladders were manually expressed daily until spontaneous voiding recovered.

4.12. In vivo assessment of oxidative stress

L-012 chemiluminescence was used for in vivo imaging to assess oxidative stress at the SCI site. Mice were intravenously treated via the tail vein for 3 consecutive days at equivalent baicalin doses. On day 3, freshly prepared L-012 solution was injected intraperitoneally (50 mg kg⁻¹) in the dark. After 30 min incubation under isoflurane anesthesia and normothermia, chemiluminescent signals from the injured segment were acquired using a small-animal imaging system in bioluminescence mode with identical exposure and gain settings.

4.13. Behavioral and locomotor assessments

Mice were randomly divided into six groups (Normal, Sham, Injury, Baicalin, Ba@Se-MSN, and Ba@Se-MSN&BV2, n = 5 per group) and longitudinally assessed at predefined time points after contusion. Overall hindlimb function was evaluated using the Basso Mouse Scale (BMS) and the Louisville Swim Scale (LSS). For BMS, mice were observed for 4 min in an open field (∼1 m diameter) and scored independently by two trained, blinded raters, with averaged values reported. For LSS, mice performed 3–5 free-swimming trials in a transparent tank and were similarly scored in a blinded manner [54]. At 2 months post-injury, refined gait kinematics were analyzed. Mice walked on a treadmill at constant speed while lateral high-speed videos were recorded with markers placed on the hip, knee, and ankle joints. Gait phases, left–right coordination, and dynamic parameters were derived from joint trajectories and angle time series. For the inclined plane test, the slope angle was gradually increased to determine the maximal angle maintained without slipping, and mean values from repeated trials were calculated. In the grid-walking task, each mouse completed at least three valid traversals (interval ≥2 min), with ≥20 steps per run, and hindlimb and combined foot-fault rates were computed. For footprint analysis, hind paws were dipped in non-toxic dye and mice walked along an enclosed straight corridor. Step length, step width, and left–right symmetry were quantified from the footprints. CatWalk automated gait analysis was performed under constant illumination with unified detection thresholds. Paw contact area, stance duration, weight bearing/intensity, and gait regularity were extracted to comprehensively evaluate coordination and gait recovery.

4.14. Electrophysiological recordings

Electrophysiological assessments were performed at 2 months post-injury to quantify descending pathway output and conduction across the lesion. Under stable anesthesia and normothermia, the skull was exposed and the right motor cortex was targeted for placement of a stimulating electrode. A recording electrode was inserted into the contralateral gastrocnemius muscle. Transcranial MEPs were recorded using a biosignal acquisition system (BL-420N, Chengdu Taimeng), including amplitude and latency. Short square-wave pulses were delivered and adjusted suprathreshold to obtain stable, reproducible waveforms. Signals were recorded repeatedly and averaged to reflect cortex-driven descending output. The spinal cord ∼1 cm rostral and caudal to the lesion was re-exposed through the original incision. A stimulating electrode was positioned ∼1 mm rostral to the lesion, while recording electrodes were placed ∼0.5 mm rostral and ∼5 mm caudal to the lesion. Spinal conduction-evoked potentials were recorded simultaneously. Changes in amplitude and latency between rostral near-lesion and caudal across-lesion responses were compared to evaluate proximal conduction preservation and distal functional reconnection.

4.15. Tissue processing and histology

At 2 months post-injury, mice were deeply anesthetized and perfused transcardially with cold PBS followed by 4% paraformaldehyde. Spinal cords centered on the lesion, together with major organs, were post-fixed in 4% paraformaldehyde at 4 °C for 12–24 h, dehydrated through graded ethanol, cleared, paraffin-embedded, and sectioned transversely at 6 μm. Sections were stained with hematoxylin and eosin (H&E) to evaluate cavity formation and tissue architecture. Adjacent sections were subjected to immunohistochemistry for GFAP, NFH, and NeuN to assess glial scar responses, axonal preservation/regeneration, and neuronal survival/distribution.

4.16. Proteomics analysis

HT22 cells were allocated to Control, glutamate-injury, and Ba@Se-MSN&BV2 treatment groups. After 24 h of treatment, cells were collected on ice and lysed in pre-chilled RIPA buffer. Supernatants were obtained by centrifugation at 12,000 g for 15 min at 4 °C. Protein concentrations were determined by BCA assay, and equal amounts of protein were reduced, alkylated, and digested with trypsin. Peptides were desalted using C18 columns and analyzed by high-resolution LC–MS/MS in DIA mode. Label-free identification and quantification were performed using Spectronaut. Differential proteins were defined by |fold change| ≥ 1.2 with p < 0.05. “Treatment-reversed” proteins were further defined as those significantly upregulated by injury versus control and downregulated by treatment versus injury, or vice versa, and were considered treatment-related candidate targets. Differential and reversed proteins were subjected to GO annotation and enrichment analyses to identify dominant biological processes and potential mechanisms.

4.17. siRNA transfection

HT22 cells were seeded into 24-well plates preloaded with sterile coverslips. At 60–70% confluence, siRNA transfection was performed using Lipofectamine 8000. SiRNA (final concentration ∼50 nM) and transfection reagent were separately diluted in serum-free Opti-MEM (each ∼50 μL), incubated for 5 min at room temperature, mixed to form complexes, and further incubated for 15–20 min before being added dropwise to the wells with gentle agitation. After 6 h, medium was replaced with complete medium containing 10% FBS, and cells were cultured for another 48 h. Knockdown efficiency was validated by Western blotting or immunocytochemistry. Under si-NC or si-CHCHD2 conditions, Ba@Se-MSN&BV2 was added according to grouping for rescue treatment before subsequent functional assays. Primary hippocampal neurons were transfected with siRNA using Lipofectamine 3000 according to the manufacturer's instructions. The si-CHCHD2 sequence was: sense 5′-AAGUGUGGACCCUUAUAUU-3′ and antisense 5′-AAUAUAAGGGUCCACACUU-3′.

4.18. Mitochondrial function assays

After treatment, HT22 cells were stained with MitoSOX Red and JC-1 to assess mtROS and Δψm, respectively. For mtROS measurement, cells were incubated with 5 μM MitoSOX Red at 37 °C for 15–20 min in the dark, washed three times with PBS, and imaged by confocal microscopy. Parallel samples were analyzed by flow cytometry under identical staining conditions, and mtROS was quantified as MFI in the MitoSOX channel. For Δψm measurement, cells were incubated with JC-1 (5 μg mL⁻¹) at 37 °C for 20 min in the dark. After PBS washing, red and green fluorescence signals were collected. The red/green ratio from both confocal imaging and flow cytometry was used to reflect Δψm changes. All imaging and flow cytometric analyses were performed using consistent instrument settings and compensation parameters. Gating and quantification were conducted uniformly in FlowJo.

4.19. Plasmid transfection and CHCHD2 overexpression

For gain-of-function experiments, a recombinant plasmid encoding GFP-CHCHD2 and the corresponding GFP empty vector were used for transfection. HT22 cells or primary hippocampal neurons were transfected using Lipofectamine 3000 according to the manufacturer's instructions. Briefly, cells were seeded and transfected at 60–70% confluence in serum-free Opti-MEM. After 6 h, the medium was replaced with complete culture medium, and cells were further cultured for 24–48 h before subsequent experiments. Overexpression efficiency was verified by Western blotting. In primary hippocampal neurons, GFP fluorescence was used to identify successfully transfected cells for neurite outgrowth analysis.

4.20. Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics 23.0 (IBM Corp., Armonk, NY, USA). Quantitative data are expressed as mean ± SEM. Unless otherwise specified, in vitro experiments were independently repeated at least three times with biological replicates. The normality of continuous variables was assessed using the Shapiro–Wilk test, and homogeneity of variances was evaluated with Levene's test. When parametric assumptions were satisfied, multiple-group comparisons were conducted using one-way analysis of variance (ANOVA) followed by Tukey's honestly significant difference (HSD) post hoc test. When datasets failed to meet normality or variance homogeneity and could not be corrected by transformation, the Kruskal–Wallis test with Dunn's post hoc comparisons was applied as a non-parametric alternative. All tests were two-sided, and P < 0.05 was considered statistically significant. Unless otherwise specified, significance is annotated as ∗ P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001 versus the Injury group.

Ethics approval and consent to participate

All experimental procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee of Jinan University (IACUC-20251024-01) and were conducted in accordance with institutional guidelines, relevant national regulations on laboratory animal welfare, and the ARRIVE reporting guidelines.

Consent for publication

Not applicable.

Declaration of generative AI use

During the preparation of this work, the authors did not use AI in scientific writing.

Funding

This work was supported by the Guangdong Natural Science Foundation (General Program, No. 2026A1515010293 and No. 2024A1515010268 ), the National Natural Science Foundation of China (No. 82372504), the Guangdong Provincial Medical Science and Technology Research Fund Project (No. A2023374 ), the Open Fund of the Guangdong Provincial Key Laboratory of Spine and Spinal Cord Reconstruction (No. 2023B121203001), and the Traditional Chinese Medicine Research Program of the Guangdong Provincial Administration of Traditional Chinese Medicine (No. 20251083). The authors gratefully acknowledge the K.C. Wong Education Foundation. This study was also supported by the Science and Technology Projects of Guangzhou (Nos. 2024A03J1040 and 2024A04J1310 ), the Guangdong Medical Research Foundation (No. A2024426 ), the Fundamental Research Funds for the Central Universities at Jinan University (No. 11624314 ), the Guangzhou Youth PhD “Qihang” Program (No. 2025A04J3715), and the Guangzhou Municipal University Collaborative Project (No. 2024A03J0817).

CRediT authorship contribution statement

Yongchun Xiao: Data curation, Investigation, Software, Validation, Visualization, Writing – original draft. Guang Tang: Data curation, Validation, Visualization, Writing – original draft. Zhiwan Chen: Data curation, Resources, Validation, Visualization, Writing – original draft. Hua Yang: Data curation, Visualization, Writing – original draft. Jianyu Zou: Funding acquisition, Investigation, Resources, Writing – review & editing. Ke Chen: Investigation, Resources, Supervision, Writing – review & editing. Jiong Wang: Investigation, Resources, Writing – review & editing. Juanjuan Li: Investigation, Resources, Writing – review & editing. Ping Wu: Resources, Software, Writing – review & editing. Ke Wang: Resources, Software, Writing – review & editing. Suhang Tan: Resources, Software, Writing – review & editing. Chengen Li: Software, Writing – review & editing. Yujing Gao: Data curation. Siming Yu: Conceptualization, Formal analysis, Methodology, Supervision, Writing – review & editing. Hongsheng Lin: Conceptualization, Methodology, Supervision, Writing – review & editing. Ying Bai: Conceptualization, Methodology, Supervision, Writing – review & editing. Zhisheng Ji: Conceptualization, Formal analysis, Funding acquisition, Methodology, Resources, Supervision, Writing – review & editing.

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 upon reasonable request.