An inflammation-targeted lipid nanoparticle inhibiting ferroptosis for spinal cord injury repair Authors: Yao Lv, Yujie Hu, Zixia Xu, Hangzhe Sun, Yixiao Lv, Xingtong Dong, Chengluo Ye, Guofeng Yu, Haoru Dong, Chuan Chang, Min Ge Published: 2026 ABSTRACT Erythropoietin (EPO) has been considered as a potentially effective drug candidate for the treatment of spinal cord injury (SCI). However, due to insufficient drug accumulation in the lesion site and systemic off-target effects, the clinical translation of EPO is limited. To overcome these limitations, we rationally designed a mannose-modified lipid nanoparticle (MLNP) system as a bioresponsive nanocarrier for targeted delivery of EPO mRNA to CD206-enriched inflammatory macrophages/microglia within the injured spinal cord. The engineered EPO@MLNPs exhibited well-defined physicochemical properties, high mRNA encapsulation efficiency, and enhanced stability, enabling preferential accumulation at the SCI lesion and sustained local translation of therapeutic EPO protein. In a mouse model of SCI, this materials-enabled delivery strategy effectively attenuated neuroinflammation, reduced neuronal loss, preserved serotonergic axonal integrity, and markedly improved motor functional recovery. Mechanistically, integrated transcriptomic profiling and experimental validation demonstrated that EPO@MLNP treatment suppressed ferroptosis by regulating iron metabolism and lipid peroxidation pathways. Overall, this study presents a targeted mRNA nanotherapeutic strategy that modulates both inflammatory and ferroptotic pathways, providing a promising approach for SCI treatment and other neuroinflammatory disorders. Introduction Spinal cord injury (SCI), typically caused by trauma, often leads to severe, long-term neurological deficits and permanent disability [[1], [2], [3]]. Pathologically, SCI involves an initial mechanical damage followed by a prolonged secondary phase, in which dysregulated inflammation and neuronal apoptosis play key roles in driving further tissue loss and functional impairment [4]. Notably, ferroptosis, a distinct form of cell death [5], has been recognized as an important contributor to secondary SCI pathology [6,7]. Recent studies indicate that ferroptosis in the injured spinal cord is closely associated with iron overload, lipid peroxidation, and oxidative stress, and may also interact with post-injury inflammatory responses, thereby influencing tissue preservation and functional recovery [[8], [9], [10]]. Therefore, therapeutic strategies aimed at modulating both inflammation and ferroptosis in the injury process are essential. Erythropoietin (EPO) is a 34-kDa single-chain acidic glycoprotein, primarily known as a key growth factor for the survival, proliferation, and differentiation of erythroid progenitor cells [11,12]. Beyond its hematopoietic role, EPO has emerged as a promising therapeutic candidate for SCI due to its multi-target and synergistic neuroprotective mechanisms [13]. Mechanistically, EPO can suppress excessive inflammatory responses by inhibiting pro-inflammatory cytokine (TNF-α, IL-1β, IL-6) production [14], and limit neuronal apoptosis [15]. Importantly, recent studies indicate that EPO may also interfere with ferroptosis by modulating iron metabolism, reducing lipid peroxidation, and upregulating key anti-ferroptotic regulators such as GPX4 [16]. These interconnected actions position EPO as a potential regulator of the inflammation–ferroptosis axis. It exerts protective effects throughout the secondary injury cascade, combating oxidative stress and inflammation in the early phase [17], inhibiting neuronal apoptosis and excitotoxicity in the intermediate phase, and promoting vascular regeneration and neural repair in the late phase [18], thereby providing structural support for functional recovery [13,18]. However, despite these promising preclinical effects, the translation of this potential into clinical efficacy faces significant challenges. Key evidence from recent high-quality clinical trials indicates: In patients with traumatic brain injury, two large randomized controlled trials, including an international multicenter study involving 606 patients, both showed that subcutaneous or intravenous EPO did not improve neurological outcome at 6 months. [19,20]. In extremely preterm infants, a trial involving 941 infants demonstrated that high-dose EPO did not reduce the risk of death or severe neurodevelopmental impairment at 2 years of age [21]; and in a pilot study of traumatic cervical SCI the combination of EPO and methylprednisolone also failed to show superior neurological improvement compared to methylprednisolone alone [22]. Consistently, these results show that systemic administration of EPO by subcutaneous or intravenous injection does not effectively translate into patient benefit in these complex clinical situations, despite clear preclinical mechanisms. This translational bottleneck is partly due to its nature as a large-molecule protein, which hinders effective penetration of the blood-brain/blood-spinal cord barrier [23], resulting in insufficient drug accumulation at the target site. Furthermore, systemic administration can significantly increase red blood cell count and hematocrit, directly elevating blood viscosity, slowing blood flow, and thereby increasing the risk of thrombosis [24]. Therefore, the key to future research lies in overcoming delivery barriers and developing innovative strategies that can precisely, safely, and efficiently deliver EPO to the site of central nervous system injury, which may be a necessary direction for realizing its neuroreparative potential. Notably, advanced delivery platforms such as lipid nanoparticles (LNPs) present a promising avenue [[25], [26], [27]]. LNPs have emerged as one of the most established non-viral platforms for mRNA delivery, as they can protect mRNA from degradation, facilitate cellular uptake, and enable intracellular release for therapeutic protein expression [28]. Unlike repeated systemic administration of recombinant EPO protein, mRNA-LNPs can deliver genetic instructions to target cells and induce local production of EPO within injured tissues, thereby potentially increasing therapeutic exposure at the lesion site while reducing unnecessary systemic protein exposure [29]. LNP delivery platforms have shown favorable safety profiles in the treatment of SCI [30]. Furthermore, mRNA-LNPs can modulate the injured spinal cord microenvironment, and previous studies have confirmed that intraspinal delivery of mRNA-LNPs encoding anti-inflammatory proteins can regulate IL-10, influence macrophage and lymphocyte migration, and exert neuroprotective effects [31]. Capitalizing on the delivery potential of mRNA-LNPs, this study constructed LNPs encapsulating EPO mRNA and administered them via intravenous injection in SCI mice to evaluate their impact on motor functional recovery after SCI. However, systemically administered mRNA-LNPs distribute to various tissues and are taken up by diverse cell types, resulting in limited accumulation at the injury site and insufficient targeting [32]. Therefore, simply encapsulating EPO mRNA into untargeted LNPs may not fully overcome the delivery-related limitations that contributed to the failure of systemic EPO therapy, including insufficient lesion-site exposure and off-target systemic effects [33]. An effective delivery platform for SCI should thus not only enhance accumulation at the injury site, but also promote selective uptake by key inflammatory cells within the lesion. To address this limitation, we employed a mannose-based targeting ligand, selected for its high affinity toward the inflammatory microenvironment of SCI where M2-polarized microglia are enriched [34]. This ligand was used to construct LNPs loaded with EPO mRNA. The resulting formulation enables preferential delivery of LNPs to inflammatory regions within the injured spinal cord, supporting localized and sustained expression of EPO mRNA. This enhanced targeting strategy is expected to improve the efficacy of motor function recovery in SCI mice. Here, we engineered an inflammation-targeted lipid nanoparticle (MLNP) platform for EPO mRNA delivery in SCI. Beyond therapeutic evaluation, this materials-enabled system also provides mechanistic support for how nanoscale delivery vehicles facilitate neural repair after SCI, at least partly through suppression of ferroptosis. This study supports an emerging targeted delivery strategy that reduces the systemic limitations and off-target effects of EPO, and further suggests that ferroptosis inhibition may contribute to SCI repair. Moreover, the modular and adaptable design of the modified LNP platform offers a potential technical framework for precision drug delivery in the central nervous system and other inflammation-associated diseases. In summary, this work introduces an inflammation-targeted LNP-based nanotherapeutic paradigm that bridges materials science and neurobiology, advancing translational strategies for SCI treatment (Fig. 1).Fig. 1Schematic illustration of MLNP-mediated delivery of EPO mRNA to treat SCI. MLNP loaded with EPO mRNA selectively target CD206-expressing inflammatory macrophages and microglia in the injured spinal cord. Following cellular internalization, the nanoparticles release EPO mRNA for translation into functional EPO protein. Locally produced EPO can suppress inflammatory responses during the acute phase, while concurrently inhibiting neuronal ferroptosis by regulating iron metabolism and reducing lipid peroxidation, thereby exerting dual therapeutic effects. These coordinated actions collectively promote neuronal survival, axonal preservation, and functional motor recovery following SCI. Schematic illustration of MLNP-mediated delivery of EPO mRNA to treat SCI. MLNP loaded with EPO mRNA selectively target CD206-expressing inflammatory macrophages and microglia in the injured spinal cord. Following cellular internalization, the nanoparticles release EPO mRNA for translation into functional EPO protein. Locally produced EPO can suppress inflammatory responses during the acute phase, while concurrently inhibiting neuronal ferroptosis by regulating iron metabolism and reducing lipid peroxidation, thereby exerting dual therapeutic effects. These coordinated actions collectively promote neuronal survival, axonal preservation, and functional motor recovery following SCI. Results and discussion Preparation and characterization of LNPs LNPs and mannose-modified LNPs (MLNPs) were prepared using a rapid-mixing technique. After dialysis and ultrafiltration, the size and surface properties of the particles were analyzed by dynamic light scattering (DLS). The mean diameter of both LNP and MLNP formulations was approximately 105 nm, with a polydispersity index (PDI) below 0.15 (Fig. 2A and B). The zeta potential of both nanoparticles was around −4 mV (Fig. 2C). Encapsulation efficiencies reached 94 ± 0.9 % for LNPs and 93 ± 1.1 % for MLNPs. Transmission electron microscopy (TEM) images revealed compact spherical morphologies with uniform size distributions, consistent with the DLS data (Fig. 2D and E). Both targeted MLNPs and non-targeted LNPs exhibited good blood compatibility (Fig. 2F). To further verify the colloidal stability of MLNPs, particle size and PDI were monitored over time, demonstrating that the formulation remained stable for at least three days (Fig. 2G and H). In line with previous reports [32,35], the developed MLNPs show favorable structural integrity and physicochemical properties.Fig. 2Preparation and characterization of EPO-encapsulated LNPs. (A-C) Characterizations of LNPs containing average particle size (A), PDI (B), and zeta potential (C). measured by DLS after preparation, dialysis, and ultrafiltration. (D-E) Representative transmission electron microscopy (TEM) images of LNPs. Scale bars, 100 nm. (F) Hemolytic ratio of LNPs and MLNPs after incubation with red blood cells. Statistical significance between groups is indicated in figure. (G-H) Stability analysis of LNPs. Average particle size (G) and PDI (H) of LNPs were stable up to 72 h. Data are presented as the mean ± SEM. ns = not significant, ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Preparation and characterization of EPO-encapsulated LNPs. (A-C) Characterizations of LNPs containing average particle size (A), PDI (B), and zeta potential (C). measured by DLS after preparation, dialysis, and ultrafiltration. (D-E) Representative transmission electron microscopy (TEM) images of LNPs. Scale bars, 100 nm. (F) Hemolytic ratio of LNPs and MLNPs after incubation with red blood cells. Statistical significance between groups is indicated in figure. (G-H) Stability analysis of LNPs. Average particle size (G) and PDI (H) of LNPs were stable up to 72 h. Data are presented as the mean ± SEM. ns = not significant, ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) The MLNP achieves efficient mRNA expression in the damaged region of spinal cord To assess the expression efficiency of LNPs in the damaged region after SCI, two types of LNPs loaded with luciferase mRNA were intravenously injected, followed by bioluminescence imaging 6 h later [32,36]. Bioluminescence imaging revealed that both types of LNPs accumulated in the SCI region, while MLNPs demonstrated significantly higher expression efficiency (Fig. 3A–C). Mannose serves as a ligand for CD206 [37], a marker highly expressed on inflammatory macrophages/microglia that are notably enriched in the injured spinal cord [38,39]. By targeting CD206, MLNPs enable specific recognition of these cells, thereby facilitating localized expression of the encapsulated mRNA within the injury site [40,41].Fig. 3In vivo expression of LNPs. (A) Biodistribution and transfection of LNPs at 6 h after intravenous injection. (B-C) Expression of LNPs in the spinal cord. (D) Immunostaining of EPO (red) and microglia (green) in the injured spinal cord after tail vein injection of EPO@MLNP, indicating the EPO@MLNP successfully enriched in damaged region of mouse spinal cord. Scale bars, 200 μm and 50 μm. (E-F) Western blot to verify the expression duration of EPO@MLNP. Data are presented as the mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) In vivo expression of LNPs. (A) Biodistribution and transfection of LNPs at 6 h after intravenous injection. (B-C) Expression of LNPs in the spinal cord. (D) Immunostaining of EPO (red) and microglia (green) in the injured spinal cord after tail vein injection of EPO@MLNP, indicating the EPO@MLNP successfully enriched in damaged region of mouse spinal cord. Scale bars, 200 μm and 50 μm. (E-F) Western blot to verify the expression duration of EPO@MLNP. Data are presented as the mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) EPO@MLNP was constructed by encapsulating EPO mRNA within MLNPs and administered to SCI mice via tail vein injection. EPO expression was then assessed by immunofluorescence and Western blotting. The results showed highly efficient EPO expression in the injured spinal cord region (Fig. 3D). As EPO is a secretory protein, immunofluorescence images revealed not only intracellular EPO expression but also substantial extracellular distribution of EPO protein, which is essential for its neuroprotective function [42,43]. Furthermore, temporal analysis of EPO protein levels in the injured spinal tissue demonstrated that EPO expression peaked 2 days post-administration, with residual expression still detectable after 2 weeks (Fig. 3E and F). This expression profile spans the acute phase of SCI, providing a foundation for the anti-inflammatory effects of EPO [44,45]. During this period, the injured spinal cord undergoes vascular remodeling, demyelination/remyelination, glial scar formation, and initiation of axonal repair. Given that EPO/EPOR signaling has been implicated in endothelial protection, angiogenic regulation, and neural repair, sustained local EPO expression may help establish a microenvironment favorable for later tissue remodeling [[46], [47], [48]]. In summary, this section confirms that EPO mRNA encapsulated within EPO@MLNPs is effectively translated in the injured spinal cord region. Both the spatial distribution and sustained duration of the expressed EPO protein contribute to the repair processes following SCI [14,49]. EPO@MLNP inhibits inflammation of SCI in vitro and in vivo We conducted transfection experiments in BV2 cells to evaluate the anti-inflammatory effects mediated by EPO@MLNP. Lipopolysaccharide (LPS) can induce the inflammatory response in microglia and iNOS is a key biomarker [50]. EPO@MLNP treatment substantially suppressed iNOS production at protein level in LPS-stimulated microglia, which is likely a result of the effective production of bioactive EPO via MLNP-mediated mRNA delivery (Fig. 4A and B). CD86-positive microglia were evaluated as an inflammatory indicator after SCI [51]. The results showed that EPO@MLNP significantly reduced the level of CD86-positive microglia (Fig. 4C and D). Furthermore, treatment with EPO@MLNP also decreased the extent of microglial infiltration, which is another important indicator of inflammation in SCI [52]. The pro-inflammatory cytokines including TNFα, IL-1β, and IL-6 are important mediators of inflammation after SCI [53]. To investigate the anti-inflammatory properties of the various interventions, serum was obtained from SCI mice on the third day following EPO@MLNP administration. Using ELISA kits, the concentrations of the pro-inflammatory cytokines IL-1β, IL-6, and TNF-α were quantified. Analysis revealed a significant elevation in all three cytokines in the SCI group compared to the sham group, successfully recapitulating the acute inflammatory phase post-SCI. Notably, EPO@MLNP treatment led to a marked downregulation in the expression of TNF-α, IL-6, and IL-1β (Fig. 4E–G). EPO@MLNP can locally translate and generate the EPO protein at the injury site, where EPO exerts anti-inflammatory effects and inhibits secondary damage after SCI [54].Fig. 4Anti-inflammatory effects of EPO@MLNP in vitro and in vivo. (A-B) Immunostaining and quantification of iNOS in microglia cells (BV2) stimulated with 500 ng/mL LPS for 4 h and treated with EPO@MLNP. Scale bars, 50 μm. (C-D) Immunostaining and quantification of CD86 (red) and microglia (green) in the spinal cord regions from SCI mice after the treatment of EPO@MLNP. Scale bars, 50 μm. (E-G) ELISA analysis of TNF-α, IL-1β, and IL-6 levels in mouse plasma collected on day 3 after treatment of EPO@MLNP. Data are presented as the mean ± SEM. ns = not significant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Anti-inflammatory effects of EPO@MLNP in vitro and in vivo. (A-B) Immunostaining and quantification of iNOS in microglia cells (BV2) stimulated with 500 ng/mL LPS for 4 h and treated with EPO@MLNP. Scale bars, 50 μm. (C-D) Immunostaining and quantification of CD86 (red) and microglia (green) in the spinal cord regions from SCI mice after the treatment of EPO@MLNP. Scale bars, 50 μm. (E-G) ELISA analysis of TNF-α, IL-1β, and IL-6 levels in mouse plasma collected on day 3 after treatment of EPO@MLNP. Data are presented as the mean ± SEM. ns = not significant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) The neuroprotective effects of EPO@MLNP Neuroprotection was evaluated through the expression levels of NeuN and serotonergic axons (5-HT) in spinal cord tissues [3,39]. The results revealed a significant increase in neuronal numbers in the injured spinal cord of the EPO@MLNP group compared to all other groups. Meanwhile, the number of NeuN positive cells distributed in the lesion center area increased in the EPO@MLNP group (Fig. 5A and B). In addition, the area of neuronal loss, corresponding to the region of neural injury, was measured across different groups. The results indicated that EPO@MLNP more effectively reduced the size of the spinal cord lesion (Fig. 5C). EPO@MLNP treatment resulted in a significantly higher density of 5-HT near the injury epicenter compared to all other groups. While axon density gradually declined with increasing distance from the epicenter, it remained elevated in EPO@MLNP-treated mice (Fig. 5D and E). It was observed that, compared to the injury group, the LNP group also exhibited a certain degree of neuroprotective effect. This may be attributed to the inherent inflammatory tropism of LNP itself. After SCI, disruption of the blood-spinal cord barrier and local inflammatory responses increase tissue permeability and promote the recruitment of phagocytic macrophages/microglia, which may allow non-targeted LNPs to deliver EPO mRNA to the lesion site to some extent. Therefore, the beneficial effect observed in the EPO@LNP group likely reflects the intrinsic delivery capacity of LNPs under inflammatory injury conditions rather than a completely lesion-specific targeting effect. In contrast, MLNP enhances targeting specificity to the injured area, resulting in a more pronounced protective outcome [55].Fig. 5Neuroprotective effects of EPO@MLNP in vivo. (A-B) Immunostaining of GFAP (red) and NeuN-positive neurons (cyan) and quantification of NeuN in the injured center of a mouse spinal cord injury model treated with EPO@MLNP were performed. Scale bars, 500 μm and 100 μm. (C) Quantification of injury area in mouse spinal cord. (D-E) Immunostaining and quantification of GFAP (red) and 5HT-stained axons (cyan) in mouse spinal cord injury treated with EPO@MLNP. Scale bars, 500 μm. Data are presented as the mean ± SEM. ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Neuroprotective effects of EPO@MLNP in vivo. (A-B) Immunostaining of GFAP (red) and NeuN-positive neurons (cyan) and quantification of NeuN in the injured center of a mouse spinal cord injury model treated with EPO@MLNP were performed. Scale bars, 500 μm and 100 μm. (C) Quantification of injury area in mouse spinal cord. (D-E) Immunostaining and quantification of GFAP (red) and 5HT-stained axons (cyan) in mouse spinal cord injury treated with EPO@MLNP. Scale bars, 500 μm. Data are presented as the mean ± SEM. ∗∗∗p < 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) EPO@MLNP promotes motor function recovery in mice with SCI Motor function recovery was assessed weekly until euthanasia using the BMS scoring system, footprint analysis, CatWalk XT automated gait assessment, and the inclined plane test. The EPO@MLNP group demonstrated significantly enhanced motor function recovery relative to the other groups (Fig. 6A). Footprint analysis revealed that footprints in the SCI group were closely spaced, indicating impaired hindlimb motor function. In contrast, all treatment groups showed varying degrees of stride recovery, with the EPO@MLNP group exhibiting the most significant improvement in stride length (Fig. 6B and C). The relative intensity of hind paw strikes was significantly higher in the EPO@MLNP group than in the SCI and EPO@LNP groups, and these mice also moved faster than PBS controls, collectively indicating improved motor control (Fig. 6D–G). Additionally, the EPO@MLNP group achieved a significantly greater maximum incline angle compared to the SCI and EPO@LNP groups, reflecting enhanced limb strength and coordination (Fig. 6H).Fig. 6The EPO@MLNP promotes motor function recovery of SCI mice. (A) BMS scores of SCI mice treated with EPO@MLNP assessed on days 1, 3, 7, 14, 21, and 28 after SCI. ∗ MLNP vs. LNP, # MLNP vs. SCI. (B-C) Representative footprint images and quantification of the stride length in SCI mice treated with EPO@MLNP. (D-G) Representative hind paw pressure intensity (D, F) and stride patterns (E, G) obtained by CatWalk XT analysis at 0s, 0.25s, and 0.5s intervals. (H) Inclined plane test results of SCI mice treated with EPO@MLNP. Data are presented as the mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. The EPO@MLNP promotes motor function recovery of SCI mice. (A) BMS scores of SCI mice treated with EPO@MLNP assessed on days 1, 3, 7, 14, 21, and 28 after SCI. ∗ MLNP vs. LNP, # MLNP vs. SCI. (B-C) Representative footprint images and quantification of the stride length in SCI mice treated with EPO@MLNP. (D-G) Representative hind paw pressure intensity (D, F) and stride patterns (E, G) obtained by CatWalk XT analysis at 0s, 0.25s, and 0.5s intervals. (H) Inclined plane test results of SCI mice treated with EPO@MLNP. Data are presented as the mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. EPO@MLNP attenuates ferroptosis after SCI To elucidate the mechanism of EPO, RNA-sequencing data (GSE1999) were analyzed. Following appropriate normalization (Fig. 7A and B), differential expression analysis identified a total of 268 differentially expressed genes (101 downregulated and 167 upregulated) in the injury group compared with the sham group, and 123 differentially expressed genes (60 downregulated and 63 upregulated) in the EPO-treated group relative to the injury group (Fig. 7C–E and Figs. S1–S3). Although the number of DEGs was relatively modest, this may indicate that EPO does not induce broad transcriptomic reprogramming, but rather exerts selective regulatory effects on specific injury-related pathways. In addition, the limited DEG number may also be influenced by experimental factors such as sample size, treatment intensity, or treatment duration, which could reduce the power to detect subtle transcriptional changes. Functional enrichment analysis revealed significant enrichment in signaling pathways associated with lipid metabolism (e.g., fatty acid transmembrane transporter activity, lipid transporter activity), redox regulation (e.g., NAD + ADP-ribosyltransferase activity), and iron ion metabolism (e.g., ferrous iron binding) (Fig. 7F). GSEA further demonstrated that pathways related to SCI repair, such as axon guidance, were upregulated in the EPO group, whereas ferroptosis-related pathways were downregulated (Fig. 7G). Given that lipid metabolism, NADPH homeostasis, and iron ion metabolism are established key processes in ferroptosis [56,57], these GSEA results collectively indicate that EPO exerts an inhibitory effect on ferroptosis. Finally, we examined the expression of ferroptosis-related markers in the transcriptomic data. The results showed that the expression of ferroptosis markers was upregulated in the SCI group, while EPO treatment downregulated the expression of these marker genes (Fig. 7H). However, several markers did not reach statistical significance, possibly due to biological variability and the limited sample size of the dataset. But the overall trend was consistent with the above findings.Fig. 7Bioinformatic analyses show that EPO treatment could inhibit ferroptosis. (A-B) PCA plot (A) and boxplot (B) show the data were well corrected. (C) All differentially expressed genes show in the heatmap. (D-E) Volcano plots of two group show the differentially expressed genes. (F-G) The result of functional enrichment analyses including GO (F) and GSEA (G) analysis. (H) The expression levels of ferroptosis related genes between different groups. ns = not significant, ∗p < 0.05, ∗∗p < 0.01. Bioinformatic analyses show that EPO treatment could inhibit ferroptosis. (A-B) PCA plot (A) and boxplot (B) show the data were well corrected. (C) All differentially expressed genes show in the heatmap. (D-E) Volcano plots of two group show the differentially expressed genes. (F-G) The result of functional enrichment analyses including GO (F) and GSEA (G) analysis. (H) The expression levels of ferroptosis related genes between different groups. ns = not significant, ∗p < 0.05, ∗∗p < 0.01. Based on the bioinformatic results, in vitro validation was performed by assessing Fe2+, MDA, GSH, and ferroptosis-related markers (ACSL4, GPX4) levels. RSL acts as a ferroptosis inducer [58]. Treatment with RSL elevated intracellular ferrous iron levels, increased the expression of the ferroptosis marker ACSL4, and raised MDA levels, while decreasing the levels of the ferroptosis-related negative regulators GPX4 and GSH. EPO@MLNP was able to reverse RSL-induced ferroptosis (Fig. 8). In summary, our findings demonstrate that EPO@MLNP promotes repair after SCI by inhibiting neuronal ferroptosis.Fig. 8EPO@MLNP treatment could attenuate ferroptosis in HT22 cells. HT22 cells were treated with 10 μM RSL for 24 h to induce ferroptosis, followed by the indicated treatments. (A-B) Ferrous ion staining indicated that EPO@MLNP treatment could alleviate ferrous ion concentrations. Scale bars, 20 μm and 5 μm. (C-D) RT-qPCR analysis of mRNA expression levels of ferroptosis-related markers after EPO@MLNP treatment. (E) The expression level of MDA after EPO@MLNP treatment. (F) The expression level of GSH after EPO@MLNP treatment. Data are presented as the mean ± SEM. ∗∗∗p < 0.001. EPO@MLNP treatment could attenuate ferroptosis in HT22 cells. HT22 cells were treated with 10 μM RSL for 24 h to induce ferroptosis, followed by the indicated treatments. (A-B) Ferrous ion staining indicated that EPO@MLNP treatment could alleviate ferrous ion concentrations. Scale bars, 20 μm and 5 μm. (C-D) RT-qPCR analysis of mRNA expression levels of ferroptosis-related markers after EPO@MLNP treatment. (E) The expression level of MDA after EPO@MLNP treatment. (F) The expression level of GSH after EPO@MLNP treatment. Data are presented as the mean ± SEM. ∗∗∗p < 0.001. Despite the encouraging therapeutic effects observed in the present mouse SCI model, several issues must be considered before clinical translation. Although mouse models are useful for the preliminary mechanistic study and assessment of therapeutic efficacy, they cannot fully reflect the complexity and heterogeneity in human SCI [59]. Particularly, significant differences in injury pattern, lesion extent, systemic inflammation, the therapeutic window, and the post-injury inflammatory microenvironment of human SCI may all affect the targeting efficiency and therapeutic performance of EPO@MLNP in humans [60,61]. Additionally, despite its improved targeting and reduced off-target effects compared to conventional EPO administration, clinical application requires further evaluation, such as long-term safety, immunogenicity, tolerability, and hematological risks [62]. Furthermore, the strategy still needs to be validated in a larger animal model, coupled with the systematic improvement of dose, and treatment window, coupled with scalable manufacturing and quality control [[63], [64], [65]]. Ultimately, SCI patients represent marked heterogeneity, appropriate stratification and careful selection of outcome measures may pose challenges to future clinical trial design [64]. Therefore, while our findings support the therapeutic potential of inflammation-targeted EPO mRNA delivery, further preclinical and translational research are needed before clinical application. Conclusion In conclusion, this study establishes MLNP as a rationally engineered nanomaterial platform for cell-selective mRNA delivery to inflammatory macrophages/microglia in SCI. By enabling sustained and localized EPO protein production at the lesion site, EPO mRNA–loaded MLNPs (EPO@MLNPs) transformed a systemically constrained biologic into a precision nanotherapeutic. Beyond attenuating neuroinflammation, this materials-enabled strategy revealed, for the first time, a protective role of EPO against neuronal ferroptosis through the regulation of iron homeostasis and lipid peroxidation. As a result, EPO@MLNP treatment significantly improved neuronal survival, preserved axonal integrity, and promoted robust functional motor recovery in SCI mice. Collectively, this work not only provides an effective and versatile mRNA nanodelivery strategy for neural injury but also demonstrates how rational nanomaterial design can uncover new therapeutic mechanisms and advance translational interventions for central nervous system disorders. Materials and methods Construction of the LNP LNPs were prepared using a manual mixing technique. Individual lipid components were first dissolved in ethanol at designated stock concentrations. The organic phase was formulated with the following molar composition: SM102: DSPC: cholesterol: DMG-PEG-Mannose (or DMG-PEG) = 23.8:23.8:47.6:4.8. DMG-PEG-Mannose was purchased from QIYUE BIOLOGY (XI’AN, China) with a purity of >95%. mRNA was dissolved in a 30 mM sodium acetate buffer (pH 5.0). The two solutions were combined at a total lipid-to-mRNA (EPO mRNA; APExBIO, Houston, USA) mass ratio of 20:1, with the organic phase rapidly injected into three volumes of the aqueous phase under mixing. Immediately after mixing, the solution was transferred into a dialysis cassette and dialyzed against PBS at 4 °C for 2 h to remove residual ethanol. The dialyzed product was then concentrated to the target concentration using a 50 kDa ultrafiltration device. The resulting LNPs were characterized for particle size and zeta potential by dynamic light scattering (DLS), and their morphology was examined by transmission electron microscopy (TEM). Encapsulation efficiency was determined using the RiboGreen assay (P9745, Solarbio, China). First, free RNA in the LNP suspension was quantified. The LNPs were then lysed with Triton X-100 to release all encapsulated RNA, and total RNA was measured. Encapsulation efficiency was calculated as (total RNA − free RNA)/total RNA × 100%. In vivo biodistribution of LNPs To evaluate the in vivo expression of LNPs, luciferase-encoding mRNA was encapsulated using the method described above. Each mouse received an intravenous injection of LNPs containing 4 μg of mRNA into the spinal cord. Six hours post-injection, 100 μL of fluorescein (15 mg/mL) was administered intraperitoneally, followed by in vivo bioluminescence imaging of the whole animal and excised spinal cord tissues. Cell culture The obtained HT-22 and BV2 cells were incubated in DMEM medium (Meilunbio, Liaoning, China) containing 10% bovine serum (Gibco, NY, USA), 100 U/mL penicillin and 100 μg/mL streptomycin (Beyotime, Shanghai, China). The cells were cultured in an incubator at 37 °C with 5% CO2 and were performed cell passaging every two or three days. The BV2 cells were collected at logarithmic growth phase and treated with 500 ng/mL LPS (Beyotime, Shanghai, China) for 4 h to establish the in vitro model. The HT22 cells treated with 10 μM RSL (Beyotime, Shanghai, China) for 24 h to induce the neuronal ferroptosis. Determination of MDA, GSH, and ferrous iron Malondialdehyde (MDA) levels were measured to evaluate oxidative stress using a commercial assay kit (Beyotime, Shanghai, China) according to the manufacturer's instructions. Briefly, protein extracts from cell samples and MDA standard solutions were prepared, mixed with the MDA working solution, and incubated at 100 °C for 15 min. After incubation, the mixtures were centrifuged at 1000×g for 10 min. The absorbance of the supernatant was then measured at 532 nm using a microplate reader. The absorbance value reflects the relative MDA concentration in the protein extracts. Glutathione (GSH) levels were measured using a commercial assay kit (Beyotime, Shanghai, China). Briefly, cells were digested, collected by centrifugation, and washed twice with PBS. Cell lysis was then performed via three cycles of freezing and thawing. After centrifugation at 8000×g for 10 min, the supernatant was collected. GSH detection was carried out following the manufacturer's protocol, and the GSH content in the samples was calculated based on a standard curve. The cell ferrous iron (Fe2+) level in HT-22 cells was measured using a commercial Iron Assay Kit (Meilunbio, Liaoning, China) according to the manufacturer's instructions. Briefly, after co-incubation of cells from different groups with the probe, an irreversible orange-red fluorescent product is generated upon reaction between the probe and Fe2+. RT-qPCR Total RNA was isolated from cultured cells using TRIzol reagent and quantified with a NanoDrop microspectrophotometer. cDNA was then synthesized following the manufacturer's protocol of a reverse transcription kit. For RT-qPCR, reactions were performed using 2 × RealStar Green Power Mix (GenStar), with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) serving as the internal control. The primer sequences for all target genes are provided as follows: GAPDH forward 5′-ACCCTTAAGAGGGATGCTGC-3′ and reverse 5′-CCCAATACGGCCAAATCCGT-3’; ACSL4 forward 5′-ACTGGCGATATTGGAGAAT-3′ and reverse 5′-CACATAGGACTGGTCACTT-3’; GPX4 forward 5′-CAGGAGCCAGGAAGTAAT-3′ and reverse 5′-CAGCCGTTCTTATCAATGAG-3’. Mice and spinal cord injury model Eight-week-old C57BL/6 mice (weighing 25–30 g) were obtained from Shanghai Jihui Laboratory Animal Care Co., Ltd. A SCI model was established following a previously described protocol [25]. All animal procedures were approved by the Animal Welfare Committee of Fudan University (Shanghai, China) and conducted in accordance with its guidelines. Briefly, mice were placed under general anesthesia and a laminectomy was performed to expose the T10 spinal cord segment. The cord was then crushed for 3 s using F11020-11 forceps (RWD), which were calibrated to deliver a constant closing force of 100g (forceps tips were calibrated with a force transducer before each surgery to ensure reproducibility). Before wound closure, experimental groups received designated drugs via local injection at the injury site. Postoperatively, manual bladder expression was performed twice daily until spontaneous micturition resumed. Each group consists of 6 mice. Locomotor function assessment Hindlimb motor function recovery was assessed at specific time points before and after SCI (days 1, 3, 7, 14, 21, and 28) using the Basso Mouse Scale (BMS), the inclined plane test, and the CatWalk XT automated gait analysis system. The BMS provides a score ranging from 0 (complete paralysis) to 9 (normal hindlimb motor function). Neuromuscular function was evaluated with the inclined plane test, in which mice were placed on an adjustable ramp and the maximum angle at which they could maintain their position for 10 s without slipping was recorded as an indicator of limb strength and coordination. Gait performance was further analyzed using the CatWalk XT system, an automated platform that captures detailed paw placement, stride length, step sequence, and hindlimb pressure distribution as the animal traverses a glass walkway. Infrared illumination and high-speed cameras enabled quantitative assessment of gait parameters, including stance duration, swing speed, and interlimb coordination providing objective and sensitive measures of locomotor recovery. All motor function evaluations were performed independently by two blinded assessors, and the average of their scores was taken as the final result. Western blot After treating with EPO@MLNP of different times, spinal cord tissues were collected. Proteins were extracted with RIPA lysis containing cocktail and their concentrations were detected by BCA. These protein samples were separated by SDS-PAGE electrophoresis and transferred to PVDF membranes. After blocking, the membranes were incubated overnight at 4 °C with primary antibody including anti-EPO (Abmart, TA5190, 1:1000), β-actin (Abmart, P30002, 1:2000) primary antibody. The next day, the bands were incubated with the corresponding secondary antibody and scanned with gel imaging system (ChemiDoc, BIO-RAD). Sample collection and immunofluorescence The mouse was anesthetized and secured on a surgical platform. The thoracic cavity was opened to expose the heart, after which the right atrium was incised. A scalp vein needle was inserted into the left ventricle and fixed with a needle holder, with the other end connected to a perfusion apparatus. Perfusion was initiated with physiological saline to flush out the blood, followed by formaldehyde solution. During perfusion, tail twitching and whole-body stiffness were observed. Once the liver became pale and enlarged, the spinal cord was extracted. Following fixation and dehydration, the spinal cord tissue was embedded in paraffin and sectioned. The T9–T11 spinal cord segments were carefully isolated and fixed in 4% paraformaldehyde. Following dehydration, the tissues were embedded in paraffin and sectioned at 5 μm thickness. Sections were baked at 60 °C for 1 h, deparaffinized in xylene, and rehydrated through a graded ethanol series. After blocking with 5% BSA for 30 min at 37 °C, sections were incubated overnight at 4 °C with the following primary antibodies: anti-GFAP (sc-33673, Santa Cruz, USA; 1:500), anti-5HT (20080, Immunostar, USA; 1:25000), anti-EPO (sc-5290, Santa Cruz, USA; 1:500), anti-NeuN (ab104224, Abcam, UK), and anti-IBA1 (17198, CST, USA; 1:200). Sections were then incubated with appropriate secondary antibodies for 1 h at 37 °C, followed by DAPI staining (C0065, Solarbio, China) for 5 min. Finally, the whole section was observed using a slide scanner (VS200, Olympus, Japan), and the central area of the injury was analyzed. Bioinformatics analysis To explore the mechanism of action of EPO, this study adopted from gene expression database (GEO) (https://www.ncbi.nlm.nih.gov/geo/) to obtain the public microarray dataset: GSE1999 (platforms: GPL1261). This dataset was selected to validate the conserved neuroprotective effects of EPO across CNS injury models. Although the injury model and treatment strategy differ from our SCI study, the dataset enables assessment of EPO-responsive pathways shared across neural injuries. It also allows us to examine whether EPO regulates iron metabolism and lipid peroxidation pathways, the core mechanisms of ferroptosis, across different neural injury contexts. This dataset was generated using forebrain/right hemisphere tissues from 7-day-old mice in three experimental conditions: vehicle-treated sham, vehicle-treated hypoxic-ischemic injury, and EPO-pretreated hypoxic-ischemic injury. Data preprocessing was performed in R (Version 4.3.1), including background correction, log2 transformation, and normalization. Probe sets were annotated according to the GPL1261 platform annotation file. Principal Component Analysis (PCA) and boxplot analyses were then conducted using the ggplot2 package to evaluate sample distribution and data analysis. Differential expression analysis was then performed with the limma package (thresholds: fold change >1 and adjusted P value < 0.05), followed by enrichment analyses such as the Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), and Gene Set Enrichment Analysis (GSEA) using the clusterProfiler package. Finally, ferroptosis-related markers were visualized with the ggpubr package. Statistical analysis All experiments in this paper were repeated at least three times. Data were analyzed using R language (Version 4.3.1) were expressed as mean ± standard error of the mean (Mean ± SEM). Statistical analysis of experimental data was performed using one-way ANOVA followed with Bonferroni's post-hoc test or paired t-test (∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001). Ethics approval and consent to participate All animal experimental procedures were in accordance with the institutional policies and federal regulations established by the Institutional Animal Care and Use Committee of Fudan University. The experimental protocols were specifically approved by the Ethics Committee of Fudan University, with the official registration number 2024-HSYY-690. CRediT authorship contribution statement Yao Lv: Conceptualization, Data curation, Formal analysis, Writing – original draft. Yujie Hu: Formal analysis, Validation, Writing – original draft. Zixia Xu: Data curation, Investigation, Methodology, Visualization. Hangzhe Sun: Investigation, Visualization. Yixiao Lv: Investigation, Visualization. Xingtong Dong: Investigation, Visualization. Chengluo Ye: Investigation, Visualization. Guofeng Yu: Investigation, Visualization. Haoru Dong: Conceptualization, Writing – review & editing. Chuan Chang: Conceptualization, Funding acquisition, Project administration, Writing – review & editing. Min Ge: Conceptualization, Project administration, 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.