Antioxidant lipid nanoparticles enhance mRNA stability for regeneration therapy and gene editing Authors: Benhao Li, Mengyao Zhao, Bowei Yang, Taihua Yang, Yongling Wang, Yaxin Hu, Miao Zhang, Jiahao Ge, Xianglong Tang, Ze Zhang, Hengqing Cui, Xiaoyuan Chen, Yi Yan Yang, Ling Peng, Qianqian Ni Published: 2026 ABSTRACT The success of mRNA vaccines has motivated the development of mRNA therapeutics to treat a wide range of diseases, as well as applications in regenerative medicine. However, the inherent instability of mRNA under conditions of oxidative stress renders these applications challenging. Here we report a general approach to shield mRNA from oxidative degradation by incorporating antioxidant lipid within lipid nanoparticles (LNPs). Specifically, we design a library of antioxidant ionizable lipids and identify a lead LNP formulation containing a 4-hydroxyphenyl-modified antioxidant lipid. Mechanistically, we show that antioxidant LNPs directly scavenge reactive species and preserve mRNA integrity under extracellular and intracellular oxidative stress. Across multiple organ and tissue injury models, we demonstrate that AO12LNPs sustain high-level and durable protein expression and restore higher transcript integrity than conventional clinically used LNPs, leading to improved regenerative outcomes and more efficient genome editing. Our findings highlight the potential of AOLNPs as a next-generation mRNA delivery platform, capable of overcoming challenging oxidative conditions in regenerative medicine. Introduction Tissue or organ injury caused by acute or chronic disease has become a global public health burden1–3. Although organ transplantation and stem cell therapy has undergone considerable development, limited donors, complicated tissue microenvironment and high-cost burden call for the urgent development of novel therapeutic strategies4,5. The past decades have witnessed rapid and exciting progress in RNA therapeutics, especially the mRNA vaccines to combat the COVID-19 pandemic6,7. This success has largely relied on the development of lipid nanoparticles (LNPs) technology to protect mRNA from rapid degradation and enable its effective intracellular delivery and the generation of the desired antigen proteins8,9. Such advances in mRNA technology have propelled progress in enzyme replacement, protein or antibody therapy and genome editing10,11. One recent and striking example is the application of paracrine vascular endothelial growth factor A (VEGFA) mRNA delivered by LNPs in heart failure treatment12. This approach achieved the desired therapeutic threshold through the prolonged half-life of expressed VEGFA protein and precise in vivo targeting, thereby reducing the chronic toxicity associated with long-term exposure to VEGFA protein administered directly to patients. Such progress paves the way towards the development of new mRNA drugs for regeneration therapy. The inherent fragility of mRNA does, however, pose a significant challenge to regenerative treatment aimed at addressing acute and chronic organ injury, where oxidative stress is a hallmark8,13,14. Indeed, the presence of large amounts of oxidized species, such as reactive species, at injured organs or tissues can damage the integrity of mRNA and induce translation errors and failures15,16, leading to unsatisfactory protein expression. Although LNPs effectively protect mRNA from degradation caused by exo- and endonucleases9,17–20, they are less efficient against reactive species, which can lead to mRNA degradation15,21 or the formation of lipid-mRNA adducts under oxidative stress22. This ultimately renders mRNA payloads untranslatable and results in the loss or failure of protein expression. An increased stability of mRNA-LNP formulations against reactive species is essential to ensure effective translation of the delivered mRNA and the desired therapeutic outcomes with regards treating organ injuries22–24. Current chemical modification and sequence optimization measures are designed to protect mRNA from endonuclease degradation and minimize innate immunogenicity8,25, therefore, developing a cost-effective approach to mitigate the impact of reactive substances or oxidative degradation on mRNA would significantly advance mRNA technology, facilitating its therapeutic application across various indications, especially in the “harsh” inflammatory microenvironment of injured organs. The clinically approved four-component LNP systems for mRNA delivery are composed of ionizable cationic lipid, polyethylene glycol (PEG)-lipid, phospholipid, and cholesterol6,9,18,26,27. In particular, the ionizable lipid plays a crucial role both in stabilizing mRNA through electrostatic complexation28–30 and in mediating endosomal escape via its pH-responsive ionization characteristics9,31–33. We reasoned that using ionizable lipids with antioxidant functionalities could be a promising approach to improve the stability of mRNA-LNP complexes against oxidative degradation induced by reactive species. We therefore elaborated, in this study, a novel and effective strategy that incorporating antioxidant lipids into LNPs protects multiple RNA modalities, including linear mRNA, circular RNA, and CRISPR-Cas9/sgRNA, from oxidative degradation, thereby enhancing RNA stability and protein production across injured organs such as liver, lung, and psoriatic skin. Specifically, we developed a library of antioxidant lipids encompassing double/triple bond, heterocyclic, thiol, or phenol-derived antioxidant entities and incorporated them into LNPs to establish antioxidant LNPs (AOLNPs) for mRNA delivery. We then systemically compared the effect of the different antioxidant moieties on mRNA expression levels. Notably, 4-hydroxyphenyl modified lipid (AO12)-substituted LNPs (AO12LNPs) enabled an 89% antioxidation score when compared to the commercially available SM-102-based LNPs (cLNPs, antioxidation score of 57%) under oxidative stress conditions. Mechanistic studies have revealed that phenol-derived antioxidant lipids exhibit strong redox activity capable of efficiently quenching diverse reactive species and suppressing ionizable lipid oxidation (Fig. 1a). Across oxidative injury models, AO12LNPs consistently sustain high protein expression of delivered mRNA, yielding 2.4–3.1-fold higher luminescence signals in inflammatory liver and lung tissues relative to cLNPs. AO12LNPs successfully delivered therapeutic hepatic growth factor (HGF) mRNA to injured tissues, restored hepatic and pulmonary repair pathways under severe oxidative microenvironments, enabling 2.8-fold higher HGF expression, reducing ALT/AST levels by 44–92% in chronic liver fibrosis, and reducing lung collagen-positive area to 3.1% (vs. 6.3% with cLNPs). In a murine psoriasis model, AO12LNPs enhanced CRISPR-Cas9 mRNA/sgRNA stability and editing efficiency under inflammatory oxidative stress, resulting in more robust NLRP3 disruption, greater suppression of psoriatic cytokines, and improved formation and recovery of epidermal barrier (Fig. 1b). These findings position AOLNPs as a versatile platform addressing a fundamental limitation of current LNPs-based RNA delivery, its vulnerability to oxidative stress and degradation, while demonstrating the significant potential for clinical translation in regenerative medicine and oxidative stress-related diseases.Fig. 1Conceptualization of antioxidant lipid nanoparticles (AOLNPs) enhances mRNA stability for regeneration therapy and gene editing.a Schematic comparison of commercial LNPs (cLNPs) and AOLNPs. The AOLNPs platform incorporates antioxidant lipids to protect RNA molecules from oxidative degradation, including linear mRNA, circular RNA, and CRISPR-Cas9/sgRNA system from oxidative degradation, thereby enhancing protein production across multiple disease models. b The improved RNA performance in inflamed lung injury (acute lung injury and chronic pulmonary fibrosis), liver injury (acute liver injury and chronic liver fibrosis), and psoriasis. Figure 1 created in BioRender. Ni, Q. (2026) https://BioRender.com/k49p414. Conceptualization of antioxidant lipid nanoparticles (AOLNPs) enhances mRNA stability for regeneration therapy and gene editing. a Schematic comparison of commercial LNPs (cLNPs) and AOLNPs. The AOLNPs platform incorporates antioxidant lipids to protect RNA molecules from oxidative degradation, including linear mRNA, circular RNA, and CRISPR-Cas9/sgRNA system from oxidative degradation, thereby enhancing protein production across multiple disease models. b The improved RNA performance in inflamed lung injury (acute lung injury and chronic pulmonary fibrosis), liver injury (acute liver injury and chronic liver fibrosis), and psoriasis. Figure 1 created in BioRender. Ni, Q. (2026) https://BioRender.com/k49p414. Results Antioxidant lipid substituted LNPs for mRNA delivery We rationally designed and synthesized a library of antioxidant (AO) ionizable lipids, with a total of 16 antioxidant moieties introduced to the commercially available ionizable cationic lipids SM-102 (used in Spikevax™ in Moderna) accordingly. The AO lipid library comprised five categories based on antioxidant capacity and included alkyl (AO1,2), unsaturated alkene and alkyne (AO3-6), heterocyclic (AO7-9), thiol/disulfide (AO10-11), and phenol (AO12-16) derived lipids, respectively (Fig. 2a), with alkyl-based lipids (AO1,2) bearing no apparent antioxidant capacity as reference control. All synthesized lipids were purified using column chromatography on silica gel, and fully characterized using NMR, and high-resolution mass spectrometry (HRMS) (Supplementary Figs. 1–47). Taking AO12 as an example, characteristic signals at 7.05–6.74 ppm in the 1H-NMR spectrum for 4-hydroxyphenyl group and at 173.10 ppm in the 13C-NMR spectrum for ester bond carbon as well as the precise molecular weight at 858.7190, indicated the successful synthesis of AO12. With the view to protecting mRNA from oxidative degradation, we developed AONLNPs by incorporating AON lipids into the LNP formulations containing SM-102 as ionizable lipids. Modulating the incorporation ratio of AO lipids ensured the right balance between mRNA delivery efficiency and antioxidation capacity allowing effective and optimal protein production (Fig. 2b). The mRNA-loaded AONLNPs were prepared by rapid vortex to mix mRNA aqueous solution with an ethanol solution containing a mixture of ionizable cationic lipid (SM-102 and/or AON lipids), cholesterol, 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC), DMG-PEG2000. Serving as a control formulation, the four lipid compounds (SM-102, cholesterol, DSPC, and DMG-PEG2000) were formulated at a molar ratio of 50:38.5:10:1.5, to generate the original commercial SM-102 LNPs (cLNPs), which are composition-wise and proportion-wise similar to FDA-approved mRNA-LNPs products (Moderna COVID-19 vaccine). Choice of optimum incorporation ratio involved inclusion of AON lipids at 5 mol% to 40 mol% within the ionizable cationic lipid component (SM-102), while maintaining other lipid components at the same molar ratio (Supplementary Table 1). After loading the firefly luciferase (FLuc)-encoding mRNA, we characterized each LNPs for its hydrodynamic diameter and polydispersity index (PDI). The cLNPs and the 64 AONLNPs possessed stable and uniform hydrodynamic diameters (approximately 150 nm) with low PDI (<0.25), almost neutral surface potential (approximately −5 mV), and high mRNA encapsulation efficiency (>80%) (Supplementary Figs. 48–52).Fig. 2Multi-step screening of AOLNP for luciferase mRNA delivery in vitro.a Chemical structures of antioxidant ionizable lipids (AON). The antioxidative moieties are classified as five categories bearing alkyl, double/triple bond, heterocyclic, thiol/disulfide, or phenol-derived entities. b In vitro transfection efficiency was assessed for AONLNPs with different antioxidant ionizable lipids, component ratios, and reactive species. c Heat map of luciferase expression efficiency following treatment of HEK-293T cells with FLuc-mRNA loaded AONLNPs and cLNPs (Moderna formulation) with different molar ratios of AON/SM-102 from 5 to 40% via high-throughput screening platform after 24 h incubation (n = 6 biologically independent samples/group, mean ± s.d.). d Heat map of relative luciferase expression changes of FLuc mRNA loaded cLNPs and AONLNPs (molar ratio of AON to SM-102: 5:100) to reactive species (O2-·, ·OH, H2O2, ClO-, and ONOO-) at different concentrations (0.00–0.50 mM) (n = 6 biologically independent samples/group, mean ± s.d.). e Radar chart of luciferase expression efficiency following treatment with FLuc-mRNA loaded AONLNPs and cLNPs under various reactive substances (0.05 mM) for the guidance and evaluation of mRNA stability of cLNPs and AONLNPs. f Antioxidant capacity of AONLNPs and cLNPs are calculated as the weighted arithmetic mean of the individual reactive species measurements shown in (e). g Scheme diagram illustrating the treatment of RAW 264.7 macrophages with PMA to trigger the production of endogenous reactive species (upper) and heat map of relative luciferase expression changes of FLuc-mRNA loaded AONLNPs and cLNPs in RAW 264.7 macrophages (below). h Screen for the optimal antioxidant ionizable lipid AO12 calculated from expression efficiency (data from c), antioxidant capacity in vitro (data from f), and antioxidant capacity in RAW 264.7 macrophage cells (data from g). The P value was determined by an unpaired two-sided t-test. P values are indicated in the graphs. b, g created in BioRender. Ni, Q. (2026) https://BioRender.com/xxo8uug. Multi-step screening of AOLNP for luciferase mRNA delivery in vitro. a Chemical structures of antioxidant ionizable lipids (AON). The antioxidative moieties are classified as five categories bearing alkyl, double/triple bond, heterocyclic, thiol/disulfide, or phenol-derived entities. b In vitro transfection efficiency was assessed for AONLNPs with different antioxidant ionizable lipids, component ratios, and reactive species. c Heat map of luciferase expression efficiency following treatment of HEK-293T cells with FLuc-mRNA loaded AONLNPs and cLNPs (Moderna formulation) with different molar ratios of AON/SM-102 from 5 to 40% via high-throughput screening platform after 24 h incubation (n = 6 biologically independent samples/group, mean ± s.d.). d Heat map of relative luciferase expression changes of FLuc mRNA loaded cLNPs and AONLNPs (molar ratio of AON to SM-102: 5:100) to reactive species (O2-·, ·OH, H2O2, ClO-, and ONOO-) at different concentrations (0.00–0.50 mM) (n = 6 biologically independent samples/group, mean ± s.d.). e Radar chart of luciferase expression efficiency following treatment with FLuc-mRNA loaded AONLNPs and cLNPs under various reactive substances (0.05 mM) for the guidance and evaluation of mRNA stability of cLNPs and AONLNPs. f Antioxidant capacity of AONLNPs and cLNPs are calculated as the weighted arithmetic mean of the individual reactive species measurements shown in (e). g Scheme diagram illustrating the treatment of RAW 264.7 macrophages with PMA to trigger the production of endogenous reactive species (upper) and heat map of relative luciferase expression changes of FLuc-mRNA loaded AONLNPs and cLNPs in RAW 264.7 macrophages (below). h Screen for the optimal antioxidant ionizable lipid AO12 calculated from expression efficiency (data from c), antioxidant capacity in vitro (data from f), and antioxidant capacity in RAW 264.7 macrophage cells (data from g). The P value was determined by an unpaired two-sided t-test. P values are indicated in the graphs. b, g created in BioRender. Ni, Q. (2026) https://BioRender.com/xxo8uug. Assessment of mRNA transfection efficiency of AONLNPs on HEK-293T cells using luminescence assay revealed the enhanced FLuc expression afforded by the treatment of 5 mol% AO1,2,4,5,6,7,9,10,11,12,13,14,15 doped AONLNPs, with approximately 1.12–2.25-fold higher than cLNPs (Fig. 2c and Supplementary Fig. 53). Incorporation ratios of AON lipids from 5 mol% to 40 mol% then led to two different mRNA expression trends dependent on AON lipid incorporated: the FLuc expression decreased with alkyl (AO1,2), unsaturated alkene and alkyne (AO3-6), heterocyclic (AO7-9) and thiol (AO10-11) derived AOLNPs at incorporation ratios above 5 mol% but continued to increase with phenol-derived AONLNPs, especially 4-hydroxyphenyl, 3-hydroxyphenyl and 4-hydroxy-3-methoxyphenyl-derived AO12,13,14LNPs, even at a high incorporation dose of 40 mol% for AO13,14LNPs when compared to cLNPs (red dash line in Supplementary Fig. 53). Based on these findings, 5 mol% AON lipids doping concentration was selected to further assess the antioxidation capacity of all AO1-16LNPs. AO12LNPs shield mRNA from oxidative degradation To systematically evaluate the protective capacity of AONLNPs against oxidative degradation, we exposed mRNA-loaded nanoparticles to physiologically relevant reactive species, including superoxide anion (O2-·), hydroxyl radical (·OH), hydrogen peroxide (H2O2), hypochlorite (ClO-), and peroxynitrite (ONOO-), which were generated using well-established chemical systems to mimic the oxidative microenvironment of injured tissue. FLuc-mRNA-loaded cLNPs and AONLNPs were treated with the above-mentioned reactive species at a range of concentrations (0.00–0.50 mM) for 2 h. After removing the excess reactive species via dialysis, HEK-293T cells were treated with the reactive-species-exposed-cLNPs or -AONLNPs at 37 °C for 24 h to compare how oxidative stress impacts the translation efficiency of mRNA LNPs. Exposure of cLNPs to reactive species (0.05 mM) decreased the FLuc expression to 75% (O2-), 64% (·OH), 65% (H2O2), 52% (ClO-) and 29% (ONOO-) that obtained with the non-reactive species exposed cLNPs (Supplementary Table 2). Higher concentrations of reactive species up to 0.50 mM led to further reductions in FLuc expression level to approximately 50% (O2-·, ·OH, H2O2), 27% (ClO-), and 6.8% (ONOO-) (Fig. 2d, e). The average antioxidant scores (AO scores) of cLNPs were 57%, 49%, 41% and 37% at reactive species concentrations of 0.05, 0.10, 0.25, and 0.50 mM, respectively. These results demonstrated the negative impact that exposure to oxidative stress has on the translation capacity of mRNA loaded within cLNPs. Then, we examined the shielding capacity of the AO1-16 lipids incorporated at 5 mol% into SM-102 LNPs formulation against oxidative degradation of loaded mRNA. Luminescence assay data (Fig. 2e) showed that Fluc-mRNA delivered by AO1-16LNPs resisted oxidative degradation by a broad range of reactive species types at low to moderate concentration (0.05–0.50 mM). Notably, AO4,5,7,8,12,13,15LNPs attenuated mRNA degradation by O2-·, allowing protein production to levels over 90% that obtained with AONLNPs not exposed to reactive species. AO3,4,7,12,13LNPs, AO4,5,6,8,9,10,12,13,14LNPs, and AO5,7,8,11,12,13,15LNPs enabled over 75% mRNA translation efficiency in ·OH, H2O2, and ClO- treated groups, respectively, while AO5,8,12,13,16LNPs maintained over 50% mRNA expression despite ONOO- exposure. As shown in Fig. 2f, AO4,5,7,8,12,13,15LNPs possessed high AO scores over 70, with similar trends at moderate reactive species concentrations illustrating their robust antioxidant capability (yellow circle in Fig. 2h). Noteworthy, among all AONLNPs, phenol-derived AOLNPs greatly augmented oxidative stability of mRNA at a diverse range of reactive species concentrations, with AO12,13LNPs providing the most protection to FLuc-mRNA (Supplementary Figs. 54–58). The antioxidative effect of AONLNPs would be expected to extend the life-span of mRNA in a high oxidative microenvironment. To investigate this, we assessed the protective effects of AONLNPs in murine RAW 264.7 macrophages pretreated with phorbol 12-myristate 13-acetate (PMA) to generate endogenous reactive oxygen species (ROS) (Fig. 2g). After 24 h incubation with FLuc-mRNA loaded cLNPs or AONLNPs, PMA-stimulated high reactive species-containing RAW 264.7 cells incubated with any of the 16 AONLNPs exhibited higher FLuc expression levels compared to those incubated with cLNPs, with AO12,14,15,16LNPs offering the highest (2-fold) FLuc expression, demonstrating their potent antioxidant capacity at the cellular level (red circle in Fig. 2h). Among the 16 AONLNPs comprehensively evaluated for their antioxidative mRNA delivery and subsequent protein expression, we selected AO12LNPs as the optimal formulation for its superior stability against oxidative degradation, high mRNA delivery efficiency and elevated protein expression levels (Fig. 2h). To determine whether direct incorporation of antioxidant small molecules could substitute for structural lipid modification, we doped the unconjugated antioxidant precursors (Raw-AON) into conventional LNPs at 5 mol%. These formulations exhibited markedly reduced mRNA expression and limited protection under oxidative stress (Supplementary Fig. 59). Covalent integration of antioxidant moieties into the ionizable lipid backbone anchors antioxidant functionality in close spatial proximity to the encapsulated mRNA, thereby facilitating efficient scavenging of reactive species surrounding the transcript. These results indicated that structural integration, rather than simple doping, is required to simultaneously maintain transfection efficiency and confer effective oxidative protection. To investigate the generalizability of the antioxidant lipid, AO12 (5 mol%) was incorporated into two clinically validated LNP platforms based on DLin-MC3-DMA and ALC-0315 (Supplementary Tables 3–4). All formulations exhibited comparable particle size, near-neutral surface charge, and high mRNA encapsulation efficiency (>90%). AO12 incorporation enhanced transfection efficiency in HEK-293T cells and significantly improved retention of luciferase expression under PMA-induced oxidative stress (85–87% versus 53–59% for the corresponding control formulations), supporting the versatility of AO12-mediated antioxidant integration across multiple clinically established LNPs systems (Supplementary Figs. 60–61). Meanwhile, three independent batches of AO12LNPs were prepared using a microfluidic mixing system. All batches exhibited comparable particle size, polydispersity index, zeta potential, mRNA encapsulation efficiency, and transfection efficiency (Supplementary Fig. 62). Meanwhile, the batches displayed consistent antioxidant protection under PMA-induced oxidative stress in RAW 264.7 macrophages, indicating good batch-to-batch reproducibility. Taken together, our data confirm that AONLNPs shield mRNA from oxidative degradation, with 4-hydroxyphenyl derived AO12LNPs presenting the most effective mRNA protection against diverse reactive species. The subsequent antioxidant studies primarily focused on AO12LNPs, which were identified as the optimal formulation in this study. Mechanism investigations of AO12LNPs in preserving mRNA Antioxidants often scavenge reactive species via electron transfer34. To investigate the mechanism underlying the varied antioxidation capacities, we performed an electrochemical study involving cyclic voltammetry to test the redox behavior of the antioxidant lipids precursors (Raw-AON), aiming to elucidate the electron transfer mechanisms of the antioxidant lipids and the rationale behind the shielding from oxidative degradation offered by phenol-derived AO lipids (Supplementary Fig. 63). Electrochemical analysis revealed that Raw-AO10,12,13,14,15,16 exhibited superior antioxidant capacity compared to other candidates (Fig. 3a). Their lower oxidation potentials and higher current values indicate an enhanced reducing ability and more efficient electron transfer, both of which are key features for effective scavenging of reactive species.Fig. 3Mechanism of antioxidant capacity of AONLNPs.a Electrochemical properties (potential and current) of a library of Raw-AON (N: 1-16) by cyclic voltammetry (Potential data are given in V vs Ag/Ag+). b Correlation between antioxidant capacity of AONLNPs in RAW 264.7 macrophage cells and potential of Raw-AON (1-16). c Denaturing agarose gel images of FLuc mRNA loaded cLNPs or AO12LNPs after incubation with various reactive species (n = 3 biologically independent samples). d Relative intensity of different RNA bands in (c) (n = 3 biologically independent samples/group, mean ± s.d.). e The capillary electrophoresis (CE) electropherogram of FLuc-mRNA extracted from cLNPs or AO12LNPs after incubation with ·OH or ONOO⁻, showing that AO12LNPs better maintain the integrity of Fluc-mRNA. f PMA-stimulated RAW 264.7 cells were transfected with FLuc-mRNA encapsulated in either cLNPs or AO12LNPs. Total intracellular mRNA was isolated at the indicated time points post-transfection, and FLuc mRNA levels were quantified by reverse transcription-quantitative polymerase chain reaction (RT-qPCR) (n = 5 biologically independent samples/group, mean ± s.d.). g Reactive species elimination experiments of mRNA-loaded cLNPs and AO12LNPs in RAW 264.7 macrophages, assessed using confocal laser scanning microscopy (CLSM) after various treatments. Reactive species were detected using DCFH-DA. Blue channel for nuclei with Hoechst 33342: λex: 405 nm, λem: 419–476 nm; Green channel of DCF: λex: 488 nm, λem: 492−588 nm. Scale bar: 20 μm. h Relative intensity of DCF in (g) (n = 20 biologically independent samples, mean ± s.d.). i Schematic illustration of PMA-induced oxidative stress in RAW 264.7 macrophages and subsequent transfection with FLuc or RLuc mRNA delivered by AO12LNPs or cLNPs. j Dual luciferase expression changes of FLuc mRNA or RLuc mRNA loaded AO12LNPs and cLNPs in RAW 264.7 macrophages (n = 6 biologically independent samples/group, mean ± s.d.). The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S. not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. i created in BioRender. Ni, Q. (2026) https://BioRender.com/kvdhz1p. Mechanism of antioxidant capacity of AONLNPs. a Electrochemical properties (potential and current) of a library of Raw-AON (N: 1-16) by cyclic voltammetry (Potential data are given in V vs Ag/Ag+). b Correlation between antioxidant capacity of AONLNPs in RAW 264.7 macrophage cells and potential of Raw-AON (1-16). c Denaturing agarose gel images of FLuc mRNA loaded cLNPs or AO12LNPs after incubation with various reactive species (n = 3 biologically independent samples). d Relative intensity of different RNA bands in (c) (n = 3 biologically independent samples/group, mean ± s.d.). e The capillary electrophoresis (CE) electropherogram of FLuc-mRNA extracted from cLNPs or AO12LNPs after incubation with ·OH or ONOO⁻, showing that AO12LNPs better maintain the integrity of Fluc-mRNA. f PMA-stimulated RAW 264.7 cells were transfected with FLuc-mRNA encapsulated in either cLNPs or AO12LNPs. Total intracellular mRNA was isolated at the indicated time points post-transfection, and FLuc mRNA levels were quantified by reverse transcription-quantitative polymerase chain reaction (RT-qPCR) (n = 5 biologically independent samples/group, mean ± s.d.). g Reactive species elimination experiments of mRNA-loaded cLNPs and AO12LNPs in RAW 264.7 macrophages, assessed using confocal laser scanning microscopy (CLSM) after various treatments. Reactive species were detected using DCFH-DA. Blue channel for nuclei with Hoechst 33342: λex: 405 nm, λem: 419–476 nm; Green channel of DCF: λex: 488 nm, λem: 492−588 nm. Scale bar: 20 μm. h Relative intensity of DCF in (g) (n = 20 biologically independent samples, mean ± s.d.). i Schematic illustration of PMA-induced oxidative stress in RAW 264.7 macrophages and subsequent transfection with FLuc or RLuc mRNA delivered by AO12LNPs or cLNPs. j Dual luciferase expression changes of FLuc mRNA or RLuc mRNA loaded AO12LNPs and cLNPs in RAW 264.7 macrophages (n = 6 biologically independent samples/group, mean ± s.d.). The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S. not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. i created in BioRender. Ni, Q. (2026) https://BioRender.com/kvdhz1p. We then investigated the correlation between the antioxidation capacity of AONLNPs in PMA-stimulated RAW 264.7 macrophages and the redox potential of the corresponding Raw-AON (1-16). Incorporating AON lipids bearing higher antioxidant capacity (lower potential), such as Raw-AO10,12,13,14,15,16, as an additional component in AONLNPs resulted in an increase in luciferase expression under high-oxidative stress conditions (Fig. 3b). The redox potential of Raw-AON thus represents a predictive indicator of AONLNP performance in maintaining mRNA expression under oxidative stress. Afterwards, we inspected the preservation of mRNA by AO12LNPs. We exposed FLuc-mRNA loaded AO12LNPs or cLNPs to various reactive species, followed by dialysis to remove excess reactants and isopropanol precipitation to isolate FLuc-mRNA for analysis by denaturing agarose gel imaging. In the absence of reactive species treatment, both AO12LNPs and cLNPs maintained comparable levels of mRNA integrity. Upon exposure to various reactive species, AO12LNPs offered markedly enhanced mRNA preservation compared to cLNPs (Fig. 3c, d). Then, capillary electrophoresis was conducted on FLuc-mRNA extracted from cLNPs and AO12LNPs following reactive species exposure. After ·OH or ONOO⁻ treatment, the full-length mRNA fraction declined to 41% and 25% in cLNPs, respectively, whereas AO12LNPs retained 58% and 47%. (Fig. 3e), indicating enhanced preservation of mRNA integrity in AO12LNPs under oxidative stress. Meanwhile, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis of a model RNA oligonucleotide demonstrated oxidative fragmentation in cLNPs samples, whereas fragmentation was markedly reduced in AO12LNPs (Supplementary Fig. 64). Meanwhile, NMR analysis showed that 3-(4-hydroxyphenyl) propionic acid (Raw-AO12) interacts with reactive species, as evidenced by clear upfield shifts of aromatic protons in the 4-hydroxyphenyl group (Supplementary Fig. 65). This spectral signature is characteristic of phenolic antioxidants undergoing electron-donating or radical-quenching reactions, confirming that the 4-hydroxyphenyl moiety actively scavenges reactive species. These results confirmed that AO12LNPs suppress oxidative RNA degradation at the molecular level. We further evaluated whether AO12LNPs preserve mRNA integrity under oxidative stress. PMA-stimulated RAW 264.7 macrophages treated with AO12LNPs exhibited substantially higher intracellular mRNA levels compared with those treated with cLNPs, indicating superior preservation of mRNA cargo in an oxidative environment (Fig. 3f). We employed 2,7-dichlorodihydrofluorescein diacetate (DCFH-DA), which can be effectively transformed to its fluorescent form (DCF) upon exposure to reactive species. In PMA-stimulated RAW 264.7 macrophages, AO12LNPs effectively reduced the proportion of DCF-positive cells (32% vs. 95% for cLNPs) (Fig. 3g, h), confirming the efficient shielding from reactive species damage offered by AO12 to mRNA. We next evaluated the functional consequence of this protection using a dual-luciferase reporter strategy (Fig. 3i). Three LNPs formulations were prepared, including FLuc mRNA loaded AO12LNPs, FLuc mRNA loaded cLNPs, and Renilla luciferase (RLuc) mRNA loaded cLNPs. Under oxidative conditions, the expression of FLuc and RLuc delivered by cLNPs decreased to 52% and 57%, respectively, confirming substantial mRNA degradation in the oxidative environment (Fig. 3j). While AO12LNPs substantially preserved mRNA functionality, increasing FLuc expression to 86% and enhancing the expression of RLuc mRNA delivered by cLNPs to 72%. These improvements demonstrated that AO12LNPs not only directly protect their own encapsulated mRNA cargo, but also can consume the reactive species in the microenvironment, thereby mediately protect the adjacent mRNA in cLNPs without AO lipid doping. Safety and biodistribution profiles of AO12LNPs Having identified AO12LNPs as the optimal formulation, we next evaluated their cytotoxicity in HEK-293T cells. Cell viability was maintained at over 88% following incubation with AO12LNPs at a concentration of 10 μg/mL, similar to that observed with cLNPs (Supplementary Fig. 66), highlighting the negligible cytotoxicity of AO12LNPs. We next used confocal laser scanning microscopy (CLSM) to visualize the intracellular trafficking of Cy5-labeled mRNA loaded AO12LNPs and cLNPs in HEK-293T cells following 3 to 12 h incubation (Supplementary Fig. 67). AO12LNPs showed comparable endosomal escape efficiency to cLNPs. Further hemolysis assay demonstrated that while neither LNP formulation showed any membrane perturbative activity at neutral pH, both exhibited membrane-disruptive features at acidic pH (Supplementary Fig. 68). These findings are in line with the above-observed non-cytotoxicity and endosomal escape capacity. Further biosafety evaluation of AO12LNPs involved the intravenous (i.v.) administration of AO12LNPs to healthy mice. AO12LNPs exhibited a favorable safety profile, with no significant changes in body weight, organ histopathology (H&E staining), or blood biochemical parameters compared with controls (Supplementary Figs. 69–71). Serum ALT/AST levels and pro-inflammatory cytokines (IL-1β, IL-6, and TNF-α) also remained comparable to PBS groups, indicating no detectable hepatotoxicity or systemic immunogenicity (Supplementary Figs. 72, 73). Next, we investigated whether the incorporation of AO lipids alters the morphology, protein corona, and biodistribution profile of LNPs. Cryo-electron microscopy (Cryo-EM) revealed that both cLNPs and AO12LNPs exhibited comparable spherical morphologies and internal structural organization (Supplementary Fig. 74). Then, analysis of the protein corona following incubation with mouse plasma, using SDS-PAGE and quantitative proteomics, showed similar plasma protein binding profiles between the two formulations (Supplementary Figs. 75, 76), indicating that AO12 lipid incorporation does not alter surface protein adsorption. We then evaluated the in vivo biodistribution of both cLNPs and AO12LNPs. Near-infrared dye (Cy-lipid)35 labeled AO12LNPs rapidly accumulated in the liver in a manner comparable to cLNPs (Supplementary Figs. 77, 78). Meanwhile, DiR-labeled cLNPs and AO12LNPs displayed similar liver-dominant biodistribution with no significant differences in tissue accumulation profiles (Supplementary Fig. 79). To identify the distribution of mRNA loaded AO12LNPs and cLNPs within hepatic cell populations, we conducted co-immunostaining of EGFP protein expressed from delivered reporter mRNA and of cell type-specific markers for hepatocytes (CK18), endothelial cells (CD31), Kupffer cells (F4/80) and leukocytes (CD45) (Supplementary Fig. 80). Both AO12LNPs and cLNPs displayed a tropism for the same liver cell types, showing no significant differences in their distribution patterns. Specifically, AO12LNPs exhibited a marked preference for hepatocytes, consistent with previous reports for SM-102 based cLNPs in general36,37. Flow cytometry analysis performed 24 h after intravenous administration of EGFP mRNA-loaded LNPs further confirmed that both formulations predominantly transfected hepatocytes, endothelial cells and Kupffer cells, with no significant differences in cellular distribution (Supplementary Figs. 81, 82). AO12LNPs enhance and prolong protein production in injured tissues Encouraged by the promising feature of AO12LNPs, we next evaluated whether it can act as a universal antioxidant platform for enhancing protein production across differently injured organs and tissues. First, in an acute liver injury mouse model induced by acetaminophen (APAP, 550 mg kg⁻¹), we injected FLuc mRNA loaded AO12LNPs and cLNPs (0.6 mg kg⁻¹) intravenously (Fig. 4a). FLuc mRNA delivered by cLNPs exhibited a pronounced reduction of protein expression (33%) in the injured liver while AO12LNPs preserved 89% of protein expression in the oxidant-inflammatory hepatic microenvironment (Fig. 4b, c), although both formulations yielded comparable luciferase expression in healthy mice. Similarly, when delivering FLuc-encoding circular RNA (circRNA), AO12LNPs preserved 70% of protein output compared to only 29% provided by cLNPs (Fig. 4d, e, and Supplementary Fig. 83), confirming that AO12LNPs effectively shield both linear and circRNA from oxidative degradation. We next assessed its applicability in non-hepatic tissues. To achieve lung targeting, DOTAP was doped in both AO12LNPs and cLNPs, forming lung-AO12LNPs and lung-cLNPs (Supplementary Table 5)38. In a lipopolysaccharide (LPS)-induced acute lung injury model, intravenous administration of FLuc-mRNA loaded lung-AO12LNPs maintained 77% of the luciferase expression compared to that observed in healthy lungs, whereas lung-cLNPs retained only 49% protein expression (Fig. 4f–h, Supplementary Fig. 84). Moreover, lung-AO12LNPs maintained 45% of luciferase expression (19% for lung-cLNPs) in a bleomycin-induced lung fibrosis model featuring sustained oxidative stress (Supplementary Fig. 85).Fig. 4Comparison of the antioxidant capacity of FLuc-mRNA loaded cLNPs or AO12LNPs on luciferase expression in injured liver or lung tissues.a Schematic of experiments analyzing the antioxidant capacity of FLuc-mRNA or FLuc circRNA loaded AO12LNPs and cLNPs in mouse models of APAP-induced acute liver injury. Female mice were exposed to the Whitten effect and an overnight fast before a single injection of APAP. b Bioluminescence images of mice treated with/without APAP followed by treatment with linear Fluc-mRNA delivered using AO12LNPs and cLNPs. c Quantitative analysis of bioluminescence signal in (b) (n = 3 biologically independent samples/group, mean ± s.d.). d Bioluminescence images of ex vivo organs from mice treated with Fluc circRNA delivered via AO12LNPs or cLNPs, with or without APAP. e Quantitative analysis of bioluminescence signal in (d) (n = 3 biologically independent samples/group, mean ± s.d.). f Schematic of experiments evaluating the antioxidant capacity of AO12LNPs and cLNPs in LPS-induced acute lung injury model. g Bioluminescence images of ex vivo organs from mice treated with linear Fluc mRNA delivered via AO12LNPs or cLNPs, with or without LPS. h Quantitative analysis of bioluminescence signal in (g) (n = 3 biologically independent samples/group, mean ± s.d.). i Schematic of experiments analyzing the antioxidant capacity of AO12LNPs and cLNPs in CCl4-induced chronic liver fibrosis model. j Bioluminescence images of mice with/without liver fibrosis followed by treatment with Fluc mRNA delivered using AO12LNPs and cLNPs at 6 h post-treatment. k Quantitative analysis of bioluminescence signal in (j) (n = 4 biologically independent samples/group, mean ± s.d.). l Quantitative analysis of bioluminescence signal from 6 h, day 1 up to day 7 after a single injection of AO12LNPs and cLNPs. m Bioluminescence images of mice with liver fibrosis at day 7 post-treatment with AO12LNPs or cLNPs. n Quantitative analysis of bioluminescence signal in (m) (n = 4 biologically independent samples/group, mean ± s.d.). The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. a, f, i created in BioRender. Ni, Q. (2026) https://BioRender.com/51y8hlf. Comparison of the antioxidant capacity of FLuc-mRNA loaded cLNPs or AO12LNPs on luciferase expression in injured liver or lung tissues. a Schematic of experiments analyzing the antioxidant capacity of FLuc-mRNA or FLuc circRNA loaded AO12LNPs and cLNPs in mouse models of APAP-induced acute liver injury. Female mice were exposed to the Whitten effect and an overnight fast before a single injection of APAP. b Bioluminescence images of mice treated with/without APAP followed by treatment with linear Fluc-mRNA delivered using AO12LNPs and cLNPs. c Quantitative analysis of bioluminescence signal in (b) (n = 3 biologically independent samples/group, mean ± s.d.). d Bioluminescence images of ex vivo organs from mice treated with Fluc circRNA delivered via AO12LNPs or cLNPs, with or without APAP. e Quantitative analysis of bioluminescence signal in (d) (n = 3 biologically independent samples/group, mean ± s.d.). f Schematic of experiments evaluating the antioxidant capacity of AO12LNPs and cLNPs in LPS-induced acute lung injury model. g Bioluminescence images of ex vivo organs from mice treated with linear Fluc mRNA delivered via AO12LNPs or cLNPs, with or without LPS. h Quantitative analysis of bioluminescence signal in (g) (n = 3 biologically independent samples/group, mean ± s.d.). i Schematic of experiments analyzing the antioxidant capacity of AO12LNPs and cLNPs in CCl4-induced chronic liver fibrosis model. j Bioluminescence images of mice with/without liver fibrosis followed by treatment with Fluc mRNA delivered using AO12LNPs and cLNPs at 6 h post-treatment. k Quantitative analysis of bioluminescence signal in (j) (n = 4 biologically independent samples/group, mean ± s.d.). l Quantitative analysis of bioluminescence signal from 6 h, day 1 up to day 7 after a single injection of AO12LNPs and cLNPs. m Bioluminescence images of mice with liver fibrosis at day 7 post-treatment with AO12LNPs or cLNPs. n Quantitative analysis of bioluminescence signal in (m) (n = 4 biologically independent samples/group, mean ± s.d.). The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. a, f, i created in BioRender. Ni, Q. (2026) https://BioRender.com/51y8hlf. We further evaluated AO12LNPs for delivering mRNA therapeutics in chronically injured organs and tissues. To simulate the clinical progression to chronic liver injury, we established CCl4-induced liver fibrosis in C57BL/6 mice through repeated injections of CCl4 twice a week over 8 weeks. We then treated these mice with a single dose of FLuc-mRNA loaded AO12LNPs or cLNPs (0.6 mg kg−1 mRNA), and examined organ distribution and duration of the produced protein within 1-week time course (at 6 h, and on days 1, 2, 3, 4, 7) (Fig. 4i). Compared with healthy mice, liver fibrosis led to the suppression of mRNA expression particularly in the FLuc-mRNA loaded cLNPs group in which bioluminescence decreased to 19%, whereas the AO12LNPs group exhibited stronger protein translation ability with 60% bioluminescence intensity (Fig. 4j, k), noting that the FLuc-mRNA loaded AO12LNPs group exhibited comparable mRNA expression levels to the cLNPs group in the liver of healthy mice. As compared with cLNPs, the FLuc-mRNA loaded AO12LNPs allowed the continuous translation of mRNA in the high oxidant microenvironment, providing a 15-fold bioluminescence intensity at 4 days post-injection (Fig. 4l). Most importantly, mice in the AO12LNPs group showed a sustained luciferase expression over 7 days post-injection (Supplementary Fig. 86), with protein levels persisting at approximately 3.5 times higher than the background at 7 days post-injection at which point luciferase expression was nearly undetectable in the cLNPs group (Fig. 4m, n). Collectively, these results strongly support the protective effect offered by AO12LNPs as an antioxidant mRNA delivery system, effectively resisting oxidant attack and sustaining high protein expression levels over the long term. AO12LNPs deliver HGF mRNA for effective liver regenerative therapy Motivated by the antioxidant capabilities as well as the high and prolonged protein production observed with reporter mRNA delivered by AO12LNPs, we next investigated the therapeutic efficacy of hepatic growth factor (HGF)-encoding mRNA therapeutics delivered by AO12LNPs in treating acute and chronic liver injury. Liver injury in murine models is characterized by hepatocyte steatosis, inflammation, and subsequent hepatic necrosis, which can be detected with elevated levels of alanine aminotransferase (ALT) and aspartate transaminase (AST) in serum36,39,40. HGF has been shown to mitigate dyslipidemia associated with steatohepatitis and ameliorate chronic liver injury by enhancing the production of antioxidant proteins and suppressing NADPH oxidase activity, thus reducing reactive species levels and preventing hepatocyte apoptosis and subsequent liver damage36,41–43. We therefore administered HGF-encoding mRNA loaded AO12LNPs via a single i.v. injection to mice with APAP-induced acute liver injury, using cLNPs as the control (Fig. 5a, Supplementary Fig. 87). We assessed the recovery of liver function by measuring liver damage biomarkers (ALT and AST). Mice treated with the HGF-encoding mRNA loaded AO12LNPs exhibited markedly lower serum ALT and AST levels that continued to decrease over time, demonstrating the potential for stronger and long-term HGF encoding mRNA therapy (Fig. 5b, c). H&E staining and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining on the liver of mice sacrificed at 24 h post-injection permitted an assessment of hepatic apoptosis and necrosis (Fig. 5d). Beneficial therapeutic effects of HGF-encoding mRNA delivered by AO12LNPs were consistent with the imaging and serum analysis results showing the more rapid disappearance of necrotic areas and TUNEL-positive cells with the AO12LNPs compared to cLNPs (Fig. 5e). These results demonstrated the faster restoration of liver function with HGF-encoding mRNA delivery by AO12LNPs platform, significantly outperforming cLNPs. To compare mRNA-based delivery with direct protein administration, we evaluated therapeutic efficacy in the APAP-induced liver injury model. AO12LNPs mediated delivery of HGF mRNA resulted in significantly higher intrahepatic HGF protein levels compared with recombinant HGF protein administration (Supplementary Fig. 88). Consistently, the AO12LNPs treatment produced a greater reduction in serum ALT levels, indicating superior therapeutic efficacy. These findings highlight the advantage of mRNA therapeutics, which enable sustained protein production in target tissues.Fig. 5Comparison of HGF-mRNA loaded AO12LNPs or cLNPs on liver function recovery in acute and chronic liver injury.a Schematic workflow of the APAP-induced acute liver injury timeline: Mice were exposed to the Whitten effect and an overnight fast before a single injection of APAP. A single dose of HGF-mRNA loaded AO12LNPs or cLNPs was administered 24 h following APAP injection. b, c Mice were analyzed at 8 h and 24 h post-injection of HGF-mRNA loaded AO12LNPs or cLNPs to assess serum ALT and AST levels (n = 4 biologically independent samples/group, mean ± s.d.). d Representative hematoxylin and eosin (H&E) staining images (left) and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining images (right) of liver tissue sections from mice in the APAP, cLNPs, and AO12LNPs groups, scale bars: 200 μm (n = 3 biologically independent samples/group, mean ± s.d.). e Observation of TUNEL+ necrotic and apoptotic cells in liver tissue from (d). f Schematic representations of CCl4-induced chronic liver fibrosis injury model and treatment schedules. Injury scheme timeline: 6-8 weeks C57BL/6 mice was used to establish CCl4-induced liver fibrosis injury model. Mice were injected with HGF-mRNA loaded AO12LNPs or cLNPs after 8 weeks of CCl4 injection, followed by another two rounds of injections with 4 days intervals. Western blot images (g) and quantitative analysis (h) of HGF expression following treatment with HGF-mRNA loaded AO12LNPs or cLNPs (n = 3 biologically independent samples/group, mean ± s.d.). i Mice were analyzed 10 days after the first administration of HGF mRNA-loaded AO12LNPs or cLNPs to evaluate serum AST and ALT levels (n = 7 biologically independent samples/group, mean ± s.d.). j Representative H&E staining images, immunohistochemical (IHC) images of Desmin and Sirius red staining. k Quantification of the percentage area of collagen coverage in Desmin and Sirius red-stained images (n = 3 biologically independent samples/group, mean ± s.d.). l Relative mRNA expression of biomarkers for fibrogenesis (n = 5 biologically independent samples/group, mean ± s.d.). Box bounds represent the 25th and 75th percentiles, the central line indicates the median (50th percentile), and whiskers extend from minima to maxima. The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Panels a and f created in BioRender. Ni, Q. (2026) https://BioRender.com/fzy883r. Comparison of HGF-mRNA loaded AO12LNPs or cLNPs on liver function recovery in acute and chronic liver injury. a Schematic workflow of the APAP-induced acute liver injury timeline: Mice were exposed to the Whitten effect and an overnight fast before a single injection of APAP. A single dose of HGF-mRNA loaded AO12LNPs or cLNPs was administered 24 h following APAP injection. b, c Mice were analyzed at 8 h and 24 h post-injection of HGF-mRNA loaded AO12LNPs or cLNPs to assess serum ALT and AST levels (n = 4 biologically independent samples/group, mean ± s.d.). d Representative hematoxylin and eosin (H&E) staining images (left) and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining images (right) of liver tissue sections from mice in the APAP, cLNPs, and AO12LNPs groups, scale bars: 200 μm (n = 3 biologically independent samples/group, mean ± s.d.). e Observation of TUNEL+ necrotic and apoptotic cells in liver tissue from (d). f Schematic representations of CCl4-induced chronic liver fibrosis injury model and treatment schedules. Injury scheme timeline: 6-8 weeks C57BL/6 mice was used to establish CCl4-induced liver fibrosis injury model. Mice were injected with HGF-mRNA loaded AO12LNPs or cLNPs after 8 weeks of CCl4 injection, followed by another two rounds of injections with 4 days intervals. Western blot images (g) and quantitative analysis (h) of HGF expression following treatment with HGF-mRNA loaded AO12LNPs or cLNPs (n = 3 biologically independent samples/group, mean ± s.d.). i Mice were analyzed 10 days after the first administration of HGF mRNA-loaded AO12LNPs or cLNPs to evaluate serum AST and ALT levels (n = 7 biologically independent samples/group, mean ± s.d.). j Representative H&E staining images, immunohistochemical (IHC) images of Desmin and Sirius red staining. k Quantification of the percentage area of collagen coverage in Desmin and Sirius red-stained images (n = 3 biologically independent samples/group, mean ± s.d.). l Relative mRNA expression of biomarkers for fibrogenesis (n = 5 biologically independent samples/group, mean ± s.d.). Box bounds represent the 25th and 75th percentiles, the central line indicates the median (50th percentile), and whiskers extend from minima to maxima. The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Panels a and f created in BioRender. Ni, Q. (2026) https://BioRender.com/fzy883r. Encouraged by the above results, we further evaluated the therapeutic efficacy of HGF-encoding mRNA delivered by AO12LNPs to rescue liver function in chronic liver fibrosis, with cLNPs as the comparative delivery system. After 8 weeks of CCl4 injection, mice with liver fibrosis were randomly assigned to receive three doses of mRNA loaded AO12LNPs, cLNPs, or PBS at 4-day intervals, with ongoing CCl4 injection until the end of mRNA treatment (Fig. 5f). The AO12LNPs permitted a 2.8-fold and 1.3-fold higher expression of HGF protein in mouse livers compared to liver fibrosis and cLNPs groups, respectively (Fig. 5g, h). This highlights the antioxidant capacity of AO12LNPs to achieve greater mRNA expression within the high-oxidative microenvironment associated with the disease. Moreover, analyses of liver damage and hepatocyte necrosis, as determined by serum levels of AST and ALT as well as liver histopathology, demonstrated a significant reduction two days after the final injection of HGF mRNA loaded AO12LNPs (Fig. 5i). Compared with the liver fibrosis group, both the AO12LNPs and cLNPs groups exhibited lower levels of liver injury markers and collagen deposition, indicating the ability of HGF protein to reduce the severity of liver damage. AO12LNPs containing HGF mRNA permitted a 44% and 92% reduction in AST and ALT levels, respectively, compared to the liver fibrosis group, while cLNPs permitted reductions of only 26% and 31%, respectively. Consequently, mice in the HGF mRNA-loaded AO12LNPs group exhibited a 1.7-fold and 3.0-fold greater decrease in AST and ALT levels compared to those in the cLNPs group, demonstrating the most significant benefits in liver fibrosis reversal. AO12LNPs also permitted a 1.9-fold and 1.6-fold decrease in collagen formation revealed by Sirius red and Desmin staining, respectively, compared to the cLNPs group, which is consistent with the H&E staining results (Fig. 5j, k, and Supplementary Fig. 89). Furthermore, mice treated with HGF mRNA-loaded AO12LNPs exhibited a more pronounced downregulation in the expression of pro-fibrogenic factors (α-SMA: 12-fold) and inflammatory cytokines (Tnf-α: 2.3-fold) compared to those in the cLNPs group (Fig. 5l). Altogether, these results clearly highlight that AO12LNPs, as an antioxidant mRNA delivery system, achieved high expression levels of therapeutic HGF protein, reduced collagen formation and provided superior fibrosis relief. We then performed single-cell RNA sequencing (scRNA-seq) analysis to gain further insights into the cellular and molecular changes induced by the antioxidant HGF-encoding mRNA delivery platform AO12LNPs in the treatment of chronic liver fibrosis and liver function recovery (Fig. 6a). 10 days after the first injection of HGF-mRNA loaded AO12LNPs, we isolated liver cells using percoll-density gradient centrifugation for their analysis using 10× Genomics scRNA-seq, with cLNPs delivery as the control. A total of 14354 and 19661 cells were sorted from liver tissues in mice treated with HGF-mRNA loaded AO12LNPs and cLNPs respectively, and retained for subsequent analysis. Unsupervised clustering analysis identified 16 major cell lineages based on the expression of canonical markers in mouse liver. These lineages included hepatocytes (hepa), cholangiocytes (chol), endothelial cell (endo), quiescent hepatic stellate cells (qHSCs), activated hepatic stellate cells (aHSCs), inactivated hepatic stellate cells (iHSCs), dendritic cells (DCs), T cells, macrophages (Mac), B cells, Kupffer cells, neutrophils (Neu), NK cells, basophils (Baso), proliferative NK cells, and proliferative T cells (Fig. 6b, c and Supplementary Fig. 90). The HGF-mRNA loaded AO12LNPs treatment group elicited distinct differences in gene expression compared to the HGF-mRNA loaded cLNPs group, particularly in genes associated with fibrosis, oxidative stress, and antioxidative processes (Fig. 6d). Notably, the HGF-mRNA loaded AO12LNPs treatment group exhibited notable downregulation of liver fibrosis-related genes, including those involved in extracellular matrix (ECM) organization (e.g., Cola1, Col1a2, Col3a1, Col6a3, Acta2, Timp2 etc.), collagen fiber assembly (e.g., Bgn), and profibrogenic mediators (e.g., Tgfβ, Fgf1, Tgfα etc.)44–46. Furthermore, oxidative stress-related genes, such as Nox1, Cyp2e1 and Duox247–49 were also downregulated in the HGF-mRNA loaded-AO12LNPs group. In contrast, genes associated with antioxidative processes, including Mt1, Mt2, and Sesn250 were notably overexpressed (Fig. 6e). To further explore cell-type specific transcriptional responses, we examined gene expression changes in aHSCs, iHSCs and qHSCs. AO12LNPs group significantly reduced the expression of fibrosis-related genes involved in extracellular matrix organization and collagen deposition, while simultaneously suppressing oxidative stress-associated genes and enhancing antioxidant gene expression (Supplementary Fig. 91). Then, we further analyzed inflammatory gene expression across different liver cell populations (Supplementary Fig. 92). Compared with cLNPs treatment, AO12LNPs group reduced inflammatory gene signatures across multiple cell types, including hepatocytes, hepatic stellate cells, dendritic cells, and cholangiocytes. To investigate the mechanism underlying the therapeutic effects of HGF mRNA delivery, we performed ligand-receptor interaction analysis to examine hepatocyte-immune cell communication networks. Fibrosis-associated pathways, including Collagen, SPP1, FN1, and Laminin signaling, were markedly attenuated following AO12LNPs treatment (Supplementary Fig. 93). In parallel, inflammation-related signaling pathways, including CXCL, ADGRE, GALECTIN, and ICAM, were also reduced, suggesting that HGF expression reshapes intercellular signaling networks that regulate fibrogenesis and inflammation (Supplementary Fig. 94). Sub-clustering analysis of hepatocytes further revealed a reduced proportion of fibrosis- and oxidative stress-associated hepatocyte states in the AO12LNPs treated group compared with the cLNPs group (Supplementary Fig. 95). These results clearly demonstrated that HGF-mRNA loaded AO12LNPs treatment effectively down-regulates the expression of fibrosis- and oxidative stress-related genes and promotes the expression of antioxidative genes during the treatment of chronic liver fibrosis. This therapeutic effect alleviates liver fibrosis and reduces oxidative stress levels in a highly oxidative pathological microenvironment.Fig. 6Single-cell RNA sequencing confirms that the antioxidant HGF-mRNA delivery platform AO12LNPs treatment accelerates liver function recovery in chronic liver fibrosis.a Schematic overview of designed scRNA-seq experiment and workflow diagram. 6-8 weeks C57BL/6 mice were subjected to intraperitoneal injection and repeated CCl4 injection to establish CCl4-induced liver fibrosis injury model. Mice were injected with HGF-mRNA loaded AO12LNPs or cLNPs after 8 weeks of CCl4 injection, followed by another two rounds of injections with 4 days intervals. Animals were sacrificed at selected time points and livers were extracted for analysing. b Uniform manifold approximation and projection (UMAP) plot visualizing 16 major cell clusters from AO12LNPs or cLNPs groups. c Violin plots showing the expression of cluster marker genes. d Heatmap showing the expression levels of representative markers involved in fibrosis, oxidative stress, and antioxidative processes. e Relative expression levels of the genes depicted in (d). f Comparison of partial differential expressed genes (DEGs) between HGF-mRNA loaded AO12LNPs or cLNPs based treatments. The HGF mRNA loaded cLNPs group was set as the standard. g Gene ontology (GO) enrichment analysis between two different treatment groups. h GO terms gene set enrichment analysis (GSEA) to ranked expression genes by their log2FC between two groups. Fibrosis associated GO terms was shown (|NES | > 1, adjusted P < 0.005) (above); oxidation associated GO terms was shown (|NES | > 1, adjusted P < 0.005) (below). i Pseudotime trajectory analysis of hepatic stellate cells (HSCs) subtypes, indicating the development of three subtypes, rooting from quiescent HSCs (qHSCs). j Proportion changes of two groups for different cell subtypes during the pseudotime analysis. k Boxplots illustrating the scores for fibrosis, oxidative stress, and antioxidative processes in the AO12LNPs and cLNPs groups. Box bounds represent the 25th and 75th percentiles, the central line indicates the median (50th percentile), and whiskers extend from minima to maxima. Panel a created in BioRender. Ni, Q. (2026) https://BioRender.com/vdaqnbe. Single-cell RNA sequencing confirms that the antioxidant HGF-mRNA delivery platform AO12LNPs treatment accelerates liver function recovery in chronic liver fibrosis. a Schematic overview of designed scRNA-seq experiment and workflow diagram. 6-8 weeks C57BL/6 mice were subjected to intraperitoneal injection and repeated CCl4 injection to establish CCl4-induced liver fibrosis injury model. Mice were injected with HGF-mRNA loaded AO12LNPs or cLNPs after 8 weeks of CCl4 injection, followed by another two rounds of injections with 4 days intervals. Animals were sacrificed at selected time points and livers were extracted for analysing. b Uniform manifold approximation and projection (UMAP) plot visualizing 16 major cell clusters from AO12LNPs or cLNPs groups. c Violin plots showing the expression of cluster marker genes. d Heatmap showing the expression levels of representative markers involved in fibrosis, oxidative stress, and antioxidative processes. e Relative expression levels of the genes depicted in (d). f Comparison of partial differential expressed genes (DEGs) between HGF-mRNA loaded AO12LNPs or cLNPs based treatments. The HGF mRNA loaded cLNPs group was set as the standard. g Gene ontology (GO) enrichment analysis between two different treatment groups. h GO terms gene set enrichment analysis (GSEA) to ranked expression genes by their log2FC between two groups. Fibrosis associated GO terms was shown (|NES | > 1, adjusted P < 0.005) (above); oxidation associated GO terms was shown (|NES | > 1, adjusted P < 0.005) (below). i Pseudotime trajectory analysis of hepatic stellate cells (HSCs) subtypes, indicating the development of three subtypes, rooting from quiescent HSCs (qHSCs). j Proportion changes of two groups for different cell subtypes during the pseudotime analysis. k Boxplots illustrating the scores for fibrosis, oxidative stress, and antioxidative processes in the AO12LNPs and cLNPs groups. Box bounds represent the 25th and 75th percentiles, the central line indicates the median (50th percentile), and whiskers extend from minima to maxima. Panel a created in BioRender. Ni, Q. (2026) https://BioRender.com/vdaqnbe. Further gene ontology (GO) enrichment analysis and Gene Set Enrichment Analysis (GSEA) revealed that differentially expressed genes (DEGs) upon treatment with HGF mRNA-loaded AO12LNPs were particularly enriched in categories such as extracellular matrix, cell redox homeostasis, extracellular matrix structural constituent, collagen fibril organization, all closely associated with ECM organization, collagen formation, and oxidative processes (Fig. 6f–h and Supplementary Figs. 96, 97). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis between the HGF mRNA loaded AO12LNPs and cLNPs groups gave consistent results (Supplementary Fig. 98). As hepatic stellate cells (HSCs) are the primary effector cells in hepatic fibrogenesis, we performed pseudotime trajectory analysis of HSC types to track the transcriptomic changes following HGF mRNA loaded AO12LNPs and cLNPs treatments. We observed a higher proportion of qHSCs in the HGF mRNA loaded AO12LNPs group compared to the HGF mRNA loaded cLNPs group (56% vs 44%), and a lower proportion of aHSCs (42% vs 58%). These results may suggest the potential of HGF mRNA-loaded-AO12LNPs to preserve HSC quiescence and inhibit activation (Fig. 6i, j and Supplementary Fig. 99). Based on the categorization of DEGs into subtypes, the HGF mRNA loaded AO12LNPs treatment group showed decreased fibrosis and oxidation scores and elevated antioxidation score compared to the HGF mRNA-loaded-cLNPs group (Fig. 6k). Taken together, the above results indicated that HGF mRNA-loaded-AO12LNPs treatment holds promising therapeutic potential by suppressing fibrosis- and oxidative stress-related genes, upregulating antioxidative genes, and promoting the quiescence of HSCs. All these ultimately contribute to the alleviation of chronic liver fibrosis and support the recovery of liver function. AO12LNPs deliver HGF mRNA for effective regeneration therapy in lung We next studied AO12LNPs for mRNA delivery in treating lung injury. Murine model of lung injury was induced by intratracheal instillation of bleomycin (BLM) for 7 days, then mice were treated with HGF-mRNA loaded AO12LNPs and cLNPs, respectively, via intravenous administration. 14 days later, lung tissues were collected for further analysis (Fig. 7a). Co-immunostaining of lung tissue sections revealed a substantially higher abundance of HGF⁺ cells in the AO12LNPs group, confirming efficient in situ production of the therapeutic HGF protein (Fig. 7b, c and Supplementary Fig. 100). Macroscopically, lungs from healthy control mice displayed a smooth surface, pink coloration, soft texture, and good elasticity without visible hemorrhage or necrosis. In contrast, lungs from the BLM and cLNPs groups appeared dull in color, markedly firm, and exhibited reduced elasticity. Notably, the AO12LNPs group showed improved gross morphology, with lungs exhibiting a healthier pinkish appearance, moderately firm texture, and only slightly reduced elasticity (Fig. 7d). Histopathological evaluation by H&E staining revealed well-preserved alveolar architecture in healthy mice, whereas the BLM model group displayed severe structural distortion, including alveolar collapse, thickened septa, and fibrotic foci. Both cLNPs and AO12LNPs based mRNA treatments ameliorated these pathological changes, with AO12LNPs showing markedly improved therapeutic efficacy (Fig. 7e and Supplementary Fig. 101). This was corroborated by a significant reduction in Ashcroft scores to 4.3 and 1.8 in the cLNPs and AO12LNP groups, respectively, indicating a superior antifibrotic effect mediated by AO12LNPs (Fig. 7f). Masson’s trichrome staining further revealed notable reductions in collagen deposition in both treatment groups, with collagen-positive areas decreasing to 6.3% (cLNPs) and 3.1% (AO12LNPs) (Fig. 7g, h and Supplementary Fig. 102). Consistently, the content of hydroxyproline (HYP), an indicator of collagen accumulation, was also effectively reduced by AO12LNP treatment (Fig. 7i). Oxidative stress assessment demonstrated that AO12LNPs treatment significantly reduced lipid peroxidation (MDA levels), while concurrently increasing endogenous antioxidant factors including glutathione (GSH), superoxide dismutase (SOD), and catalase (CAT) (Fig. 7j–m). Collectively, these findings demonstrated that the superior mRNA delivery efficiency and antioxidative protection afforded by AO12LNPs translate into robust functional and regenerative benefits in chronic lung injury, establishing AO12LNPs as a highly effective platform for mRNA-based pulmonary therapy.Fig. 7Comparison of therapeutic efficacy of HGF mRNA loaded AO12LNPs or cLNPs on lung function recovery in bleomycin-induced lung fibrosis.a Schematic workflow of the therapeutic experiment. b Representative co-immunostaining images of lung sections from mice following the specified treatment. Blue: Hoechst-stained nucleus. Red: HGF expressed from HGF-mRNA loaded cLNPs or AO12LNPs, scale bars: 50 μm. c The proportion of HGF+ cells in (b) (n = 4 biologically independent samples/group, mean ± s.d.). d Representative images of the lung from mice following the specified treatments. e Representative H&E-stained lung sections, scale bars: 100 μm. f Ashcroft score determined by H&E-stained lung sections (n = 4 biologically independent samples/group, mean ± s.d.). g Representative Masson’s trichrome-stained lung sections, scale bars: 100 μm. h Quantification of collagen coverage using Masson trichrome-stained lung sections (n = 4 biologically independent samples/group, mean ± s.d.). Mice were analyzed 14 days after the first administration of HGF-mRNA loaded AO12LNPs or cLNPs to quantify Hydroxyproline (HYP) (i), lipid peroxidation (MDA) (j), total GSH (k), total superoxide dismutase (SOD) (l), and catalase (CAT) (m) in lung tissues (n = 5 biologically independent samples/group, mean ± s.d.). For box plots in c, i–m, boxes indicate 25th–75th percentiles, central lines represent medians (50th percentile), and whiskers span minima to maxima. The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. a Created in BioRender. Ni, Q. (2026) https://BioRender.com/hclm0k3. Comparison of therapeutic efficacy of HGF mRNA loaded AO12LNPs or cLNPs on lung function recovery in bleomycin-induced lung fibrosis. a Schematic workflow of the therapeutic experiment. b Representative co-immunostaining images of lung sections from mice following the specified treatment. Blue: Hoechst-stained nucleus. Red: HGF expressed from HGF-mRNA loaded cLNPs or AO12LNPs, scale bars: 50 μm. c The proportion of HGF+ cells in (b) (n = 4 biologically independent samples/group, mean ± s.d.). d Representative images of the lung from mice following the specified treatments. e Representative H&E-stained lung sections, scale bars: 100 μm. f Ashcroft score determined by H&E-stained lung sections (n = 4 biologically independent samples/group, mean ± s.d.). g Representative Masson’s trichrome-stained lung sections, scale bars: 100 μm. h Quantification of collagen coverage using Masson trichrome-stained lung sections (n = 4 biologically independent samples/group, mean ± s.d.). Mice were analyzed 14 days after the first administration of HGF-mRNA loaded AO12LNPs or cLNPs to quantify Hydroxyproline (HYP) (i), lipid peroxidation (MDA) (j), total GSH (k), total superoxide dismutase (SOD) (l), and catalase (CAT) (m) in lung tissues (n = 5 biologically independent samples/group, mean ± s.d.). For box plots in c, i–m, boxes indicate 25th–75th percentiles, central lines represent medians (50th percentile), and whiskers span minima to maxima. The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. a Created in BioRender. Ni, Q. (2026) https://BioRender.com/hclm0k3. AO12LNPs deliver CRISPR-Cas9 mRNA/sgRNA for psoriasis therapy Psoriasis is a chronic, recurrent inflammatory skin disorder characterized by the excessive release of pro-inflammatory cytokines3,51,52. The NOD-like receptor family pyrin domain-containing protein 3 (NLRP3), a cytosolic immune sensor responsive to cellular stress signals, is typically activated upon inflammation and assembles into the NLRP3 inflammasome53. NLRP3-driven inflammasome activation plays a pivotal role in the onset and progression of psoriasis54. Here, we developed a CRISPR-Cas9/single-guide RNA (sgRNA) system targeting the NLRP3 gene and encapsulated both Cas9 mRNA and NLRP3-specific sgRNA into antioxidant AO12LNPs for psoriasis treatment. We first validated the gene editing efficiency in vitro. RAW264.7 macrophages were stimulated with LPS to induce NLRP3 expression, followed by treatment with Cas9 mRNA/sgRNA-loaded cLNPs or AO12LNPs. Western blot analysis confirmed that AO12LNPs achieved higher NLRP3 gene disruption efficiency compared with cLNPs (Supplementary Fig. 103). Subsequently, a murine psoriasis model was induced by topical application of imiquimod (IMQ) to the dorsal skin for seven consecutive days (Fig. 8a). Repeated IMQ exposure resulted in the classical pathological features of psoriasis, including erythema, scaling, induration, inflammatory cell infiltration, and increased skinfold thickness. The psoriasis area and severity index (PASI) was assessed every two days. Compared with the untreated psoriasis group, subcutaneous injection of NLRP3-targeting Cas9/sgRNA loaded in either cLNPs or AO12LNPs substantially mitigated psoriatic symptoms, with AO12LNPs treatment achieving more pronounced improvements in erythema, scaling, and induration (Fig. 8b, c and Supplementary Fig. 104). In addition, mice receiving cLNPs or AO12LNPs showed no significant changes in body weight (Supplementary Fig. 105), indicating minimal systemic toxicity.Fig. 8Comparison of therapeutic efficacy of Cas9 mRNA/sgRNA loaded AO12LNPs or cLNPs targeting NLRP3 for the treatment of imiquimod-induced psoriasis.a Schematic workflow of the therapeutic experiment. b Representative images of the skin from mice following the specified treatments. c Psoriasis area and severity index (PASI) scores based on three pathological symptom factors, including erythema, scaling, and induration, measured every other day for 2 weeks. The scores were rated on a 0- to- 4 scale as follows: 0, none; 1, slight; 2, moderate; 3, obvious; 4, very obvious (n = 6 biologically independent samples/group, mean ± s.d.). Moisture values (d), transepidermal water loss values (e), heat loss by evaporation cooling (f), relative humidity of shin (g), and water evaporation concentration (h) during the treatment (n = 6 biologically independent samples/group, mean ± s.d.). i H&E staining of the skin sections, scale bars: 100 μm. j Thicknesses of epidermis determined by H&E-stained microphotography (n = 4 biologically independent samples/group, mean ± s.d.). k Representative images of the spleens and weights on day 14 (n = 6 biologically independent samples/group, mean ± s.d.). Enzyme-linked immunosorbent assay (ELISA) of il-1β (l), il-6 (m), and TNF-α (n) production in the skin tissues of mice after treatment (n = 5 biologically independent samples/group, mean ± s.d.). For box plots in d–h, j, l–n, boxes indicate 25th–75th percentiles, central lines represent medians (50th percentile), and whiskers span minima to maxima. The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. a Created in BioRender. Ni, Q. (2026) https://BioRender.com/nyzdcgq. Comparison of therapeutic efficacy of Cas9 mRNA/sgRNA loaded AO12LNPs or cLNPs targeting NLRP3 for the treatment of imiquimod-induced psoriasis. a Schematic workflow of the therapeutic experiment. b Representative images of the skin from mice following the specified treatments. c Psoriasis area and severity index (PASI) scores based on three pathological symptom factors, including erythema, scaling, and induration, measured every other day for 2 weeks. The scores were rated on a 0- to- 4 scale as follows: 0, none; 1, slight; 2, moderate; 3, obvious; 4, very obvious (n = 6 biologically independent samples/group, mean ± s.d.). Moisture values (d), transepidermal water loss values (e), heat loss by evaporation cooling (f), relative humidity of shin (g), and water evaporation concentration (h) during the treatment (n = 6 biologically independent samples/group, mean ± s.d.). i H&E staining of the skin sections, scale bars: 100 μm. j Thicknesses of epidermis determined by H&E-stained microphotography (n = 4 biologically independent samples/group, mean ± s.d.). k Representative images of the spleens and weights on day 14 (n = 6 biologically independent samples/group, mean ± s.d.). Enzyme-linked immunosorbent assay (ELISA) of il-1β (l), il-6 (m), and TNF-α (n) production in the skin tissues of mice after treatment (n = 5 biologically independent samples/group, mean ± s.d.). For box plots in d–h, j, l–n, boxes indicate 25th–75th percentiles, central lines represent medians (50th percentile), and whiskers span minima to maxima. The P value was determined by an unpaired two-sided t-test. No adjustments for multiple comparisons were performed. P values are indicated in the graphs. N.S., not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. a Created in BioRender. Ni, Q. (2026) https://BioRender.com/nyzdcgq. We further evaluated multiple biophysical indicators of skin barrier function during treatment, including skin moisture content, transepidermal water loss, evaporative heat loss, local skin humidity, and water evaporation concentration. Across all indicators, AO12LNPs treated mice exhibited significantly greater restoration of skin barrier integrity compared with those receiving cLNPs group (Fig. 8d–h and Supplementary Figs. 167, 107). Histological evaluation of skin lesions via H&E staining (Fig. 8i) revealed smooth, intact epidermal morphology with no signs of inflammation in the healthy control group. In contrast, IMQ-treated mice displayed severe pathological alterations, including pronounced acanthosis, hyperkeratosis, and extensive inflammatory infiltration. Both cLNP and AO12LNPs treatment attenuated these pathological changes, with AO12LNPs producing more substantial improvements, reflected by smoother epidermal structure, reduced hyperkeratosis, and more notable reduction in epidermal thickening (Fig. 8j and Supplementary Fig. 108). The IMQ-induced psoriasis mouse model was characterized by a significant increase in spleen weight and size, a known systemic consequence of the massive release of inflammatory cytokines. Subcutaneous delivery of NLRP3-targeting CRISPR/Cas9 systems encapsulated in either cLNPs or AO12LNPs effectively reversed this splenomegaly, reducing both spleen size and weight. Notably, AO12LNPs-treated group displaying spleen morphology and weight most similar to those of healthy mice (Fig. 8k). Consistently, AO12LNPs achieved the strongest suppression of pro-inflammatory cytokines (IL-1β, IL-6, and TNF-α) in skin tissue (Fig. 8l–n). Co-immunostaining further confirmed efficient gene editing in local skin tissues, as evidenced by markedly reduced NLRP3-positive cells in the AO12LNPs-treated group (Supplementary Fig. 109). Meanwhile, indel frequency analysis revealed an on-target editing efficiency of 23.67% at the NLRP3 locus treated with AO12LNPs loaded with Cas9 mRNA and NLRP3 sgRNA (Supplementary Fig. 110a, b). And targeted deep sequencing of five predicted off-target loci detected no measurable off-target editing (Supplementary Fig. 110c and Supplementary Tables 6–7). Taken together, these results demonstrated that AO12LNPs substantially enhance CRISPR/Cas9-mediated NLRP3 gene editing, leading to effective suppression of psoriatic inflammation, restoration of skin barrier function, and normalization of systemic inflammatory responses. Discussion Considerable advancements in mRNA-based therapeutics have been made since the success of COVID-19 vaccines using LNPs for mRNA delivery9,19. Though the commonly used LNPs succeed in improving endosomal escape, redirecting their biodistribution, or modulateing immunogenicity6,18,35,55, they are not capable of resisting to oxidative degradation and sequestering reactive species, such as reactive species present in inflamed or injured tissues21,28. In this study, we address this largely neglected yet critical issue of mRNA degradation under oxidative stress by incorporating antioxidant lipids into LNPs to enhance the mRNA stability in treating diseases associated with enriched reactive species and highly oxidative conditions. By integrating antioxidant lipids into the commercially available and SM-102-based formulations (cLNPs), we establish AOLNPs as a generalizable strategy to stabilize diverse RNA modalities under oxidative stress. Among the library of antioxidants examined, phenol-derived AO12 most effectively quenched reactive species, suppressed ionizable-lipid oxidation and aldehyde formation, and preserved RNA integrity, evidenced by an antioxidant score 89% compared to cLNPs (antioxidant score of 57%). These mechanistic features translated into sustained RNA functionality across multiple inflammatory models, in which AO12LNPs consistently maintained substantially higher protein production than cLNPs for linear mRNA, circular RNA, and CRISPR-Cas9 systems. Importantly, these protective effects occurred without altering biodistribution or biosafety, underscoring the compatibility of antioxidant lipids with existing LNPs architectures. By mitigating oxidative degradation, AO12LNPs preserved mRNA integrity and supported prolonged, functional protein production in oxidative microenvironments. This capability produced pronounced therapeutic benefit across distinct diseases. In acute and chronic liver injury, AO12LNP-delivered HGF mRNA enhanced hepatoprotection, reduced fibrosis, and elicited stronger transcriptomic reversal of fibrogenic and oxidative-stress pathways relative to cLNPs. In pulmonary fibrosis, AO12LNPs restored alveolar architecture, suppressed lipid peroxidation, elevated endogenous antioxidant defenses, and reduced collagen accumulation. In a psoriasis model with skin inflammation, CRISPR-Cas9/sgRNA delivered by AO12LNPs achieved more efficient NLRP3 disruption, stronger suppression of inflammatory cytokines, and more complete recovery of epidermal barrier function. This study establishes antioxidant LNPs as a broadly effective strategy to preserve RNA integrity under oxidative stress, thereby enabling sustained protein production and enhanced therapeutic and genome-editing outcomes across diverse inflammatory diseases. Although there are a few recent works56,57 performing anti-inflammatory LNPs to improve the mRNA therapeutic efficacy, they emphasize the mitigation of cellular damage or immune activation rather than systematically investigating the direct protective effects on mRNA structural integrity. This key gap is addressed here through quantitative analysis of mRNA degradation and systematic validation of transcript integrity across multiple RNA modalities. This antioxidant-engineering framework offers a versatile foundation for developing next-generation mRNA therapeutics for regeneration therapy and gene editing. Despite these advances, there are limitations. While phenol-derived lipids, such as AO12, exhibit the most robust protective effects, the design space of antioxidant lipids is still largely underexplored. Systematic optimization of phenolic structures, or integration with other stabilizing chemistries, could potentially enhance resistance to oxidative degradation even further. Second, our therapeutic evaluation was restricted to liver, lung, and skin, whereas other oxidative stress rich organs (e.g., kidney, myocardium, ischemic tissues) remain to be investigated. Third, although the short-term biosafety profiles were favorable, further investigation into the long-term behavior and potential accumulation of AOLNPs is necessary. Moreover, the scalability of AO lipid synthesis and AOLNP manufacturing need to be validated for translation for advancing mRNA therapeutics into real clinical implementation to treat diseases associated with high oxidative stress. Methods Mice and ethics Balb/c mice and C57BL/6 mice (female, 6–8 weeks old) were purchased from Shanghai SLAC Laboratory animal CO. Ltd. Animal protocols were approved by the Institutional Animal Care and Use Committee of Fudan University (202504015S). The animals were housed in a pathogen-free, individually ventilated animal facility under controlled conditions: a temperature of 20–23 °C, relative humidity of 40–60%, and a 12-h light/dark cycle. The facility’s ambient temperature was 25 °C with a relative humidity of 55–65%. Materials and apparatus 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC), cholesterol, poly(ethylene) golucol (PEG)-lipid (DMG-PEG2000) and 1,2-Dioleoyl-3-trimethylammonium propane (DOTAP) were purchased from Avanti Polar lipids. Ionizable lipids SM-102 was purchased from Wuhan Sun-Shine bio-technology Co. Ltd. Butyric acid, 4-pentenoic acid, crotonic acid, 4-pentynoic acid, 2-furanpropanoic acid, 2-thiophenepropionic acid, 3-mercaptopropionic acid, α-lipoic acid, 3-(4-hydroxyphenyl)propionic acid, 3-hydroxyphenylacetic acid, benzenepropanoic acid, 3-(3,5-di-tert-butyl-4-hydroxyphenyl)propionic acid, 3-(3,4-dihydroxyphenyl)propionic acid, toluene, dichloromethane, methanol, p-toluenesulfonic acid, diacetyldichlorofluorescein (DCFHDA) and phorbol 12-myristate 13-acetate (PMA) were purchased from Sigma Aldrich. Acetaminophen (APAP), cyclopropylacetic acid, 3-cyanopropionic acid, 3-indolepropionic acid were purchased from Energy Chemical. Carbon tetrachloride (CCl4) was purchased from Sinopharm Group Co. Ltd. LysoTracker® Green DND-26 and Hoechst 33342 were purchased from Invitrogen. Cy-lipid was synthesized and purified as reported previously35. Silica gel plates and column chromatography were purchased from the Qingdao Ocean Chemicals. Firefly luciferase (FLuc) encoding mRNA, firefly luciferase-encoding circRNA and enhanced green fluorescent protein (eGFP) encoding mRNA were purchased from ShanghaiCynbio. Renilla Luc mRNA was purchased from TriLink BioTechnologies. Cyanine5 (Cy5)-labeled FLuc mRNA (Cy5-mRNA) was purchased from APExBio. The mRNA platform investigated was 5-methoxyuridine (5moU)-modified. Hepatocyte-growth-factor (HGF)-encoding mRNA was synthesized by in vitro transcription (IVT) as reported previously36. HGF affinity purified polyclonal Ab protein (Cat. No: AF-294-SP) was purchased from R&D systems. NLRP3 single guide RNA (NLRP3-sgRNA) was purchased from Integrated DNA Technologies. NLRP3 Rabbit PolyAb (Cat. No: 30109-1-AP) was purchased from Proteintech Singapore Pte Ltd. Dulbecco’s Modified Eagle medium (DMEM), phosphate-buffered-solutions (PBS, 10×, pH = 7.4) were obtained from MesGan Biotechnology. Quanti-iT RiboGreen RNA reagent was purchased from Thermo Fisher Scientific. Nuclear magnetic resonance (NMR) spectra were recorded using a Bruker 400 MHz spectrometer. High-resolution mass spectra (HRMS) were obtained from Agilent 6546 Q-TOF LCMS in positive ESI mode. Dynamic light scattering (DLS) and zeta potential measurements were performed with a Litesizer 500, Anton Paar. Intracellular trafficking of Cy5-labeled mRNA loaded in cLNPs and AONLNPs was analyzed using a confocal microscope (Carl Zeiss, LSM 710). Bioluminescence imaging of mice was conducted with the PerkinElmer IVIS Spectrum system. NIR-II imaging was carried out using an InGaAs detector (Princeton Instruments, NIRvana 640). Synthesis and characterization of antioxidant ionizable lipids AON (N = 1–16) SM-102 (0.05 mmol, 35.5 mg), butyric acid (0.1 mmol, 8.8 mg), and p-toluenesulfonic acid (0.05 mmol, 9.5 mg) were mixed in toluene (5 mL) and heated at 110 oC for 4 h. After cooling, the solution was extracted with dichloromethane for three times. The organic extracts were dried over anhydrous sodium sulfate and evaporated to dryness. The antioxidant ionizable lipid AO1 was purified by silica gel column chromatography using dichloromethane-methanol (10:1 (v/v)). The synthesis of other antioxidant ionizable lipids AON (N = 2–16) followed the same method as described above, with the carboxylic acid derivatives replaced by the corresponding alternative. All chemicals were characterized by 1H-NMR, 13C-NMR, and high-resolution mass spectrometry (Supplementary Figs. 1–47). Preparation of mRNA loaded cLNPs and AONLNPs The mRNA loaded cLNPs and AONLNPs formulations were prepared using rapid vortex mixing as previously reported58. Briefly, cLNPs were prepared by rapidly mixing the lipid mixtures in ethanol with 10 mM citrate buffer (pH 4.0), each loaded with a single type of RNA, either linear FLuc mRNA, circular RNA, or HGF-encoding mRNA, at a volume ratio of 3:1 (citrate buffer solution to ethanol). The resulting mixture was then vortexed for 5 min to ensure thorough mixing. The cLNPs used in this study contain an ionizable cationic lipid (SM-102), DSPC, cholesterol, and DMG-PEG2000 in a molar ratio of 50:10:38.5:1.5. The mass ratio of mRNA to total lipids was maintained at 1:20. After incubating for 10 min at room temperature, the prepared solution was dialyzed against 1×PBS for at least 12 h to remove the acidic buffer and ethanol. Following dialysis, the solution was transferred into an RNase-free tube for subsequent experiments. The mRNA loaded cLNPs solution was stored in refrigerator at 4 °C. To prepare a series of mRNA loaded AONLNPs formulations by adjusting the molar ratio of AON to SM-102 from 0 to 40%, the AON lipids was first dissolved in ethanol and subsequently mixed with other lipid components. The mixture was then formulated with mRNA in buffer as described above. All formulations were named as the antioxidant lipid (AON). To prepare a series of lung-targeting LNPs, 30 mol% of the cationic lipid DOTAP was incorporated into the lipid composition, based on previously established protocols38. The detailed lipid constituents are provided in Supplementary Table 5. For microfluidic mixing, the aqueous and ethanol phases were mixed at a volume ratio of 3:1 using a microfluidic chip device (Micro&Nano). The nanoparticles size, polydispersity index (PDI), and zeta potential of cLNPs and AONLNPs were measured using a Litesizer 500, Anton Paar. Encapsulation efficiency of FLuc-mRNA loaded cLNPs and AONLNPs were measured by Quant-iT RiboGreen RNA Kit (Thermo Fisher, Cat. No. R11490) following protocols. In vitro transfection of FLuc-mRNA loaded cLNPs and AONLNPs HEK-293T cells (American Type Culture Collection (ATCC), Cat. No. CRL-3216) were seeded in 96-well plates (1 × 104 cells per well) and cultured overnight. The FLuc-encoding mRNA loaded cLNPs or AONLNPs solution was prepared in Dulbecco’s Modified Eagle Medium (DMEM). These solutions were incubated with HEK-293T cells at 37 oC for 24 h (250 ng mRNA per well). The bioluminescence assay to assess luciferase expression was conducted using the One-Lumi™ Firefly Luciferase Assay Reagent (Beyotime, RG056M). The reagent was added to each well, and the luminescence was measured with a Thermo Scientific microplate reader, following the protocol provided by the manufacturer. To evaluate the chemostability (antioxidation capacity) of AONLNPs and cLNPs, a series of FLuc-mRNA-loaded AONLNPs or cLNPs in RNase-free tubes were treated with reactive species (O2-·, ·OH, H2O2, ClO-, and ONOO-) at varying concentrations (0.00–0.50 mM) for 2 h. Then, the mixed solution was subjected to dialysis against 1× PBS for at least 12 h to eliminate excess reactive species. After dialysis, the solution was transferred into DMEM and incubated with HEK-293T cells for 24 h. The bioluminescence assay was then performed as outlined in the above text. To assess the antioxidant capacity of AONLNPs and cLNPs against intracellularly produced reactive species, murine RAW264.7 macrophage cells (ATCC, Cat. No. TIB-71) in 96-well plates (1 × 104 cells per well) was stimulated with phorbol 12-myristate 13-acetate (PMA) (0.2 μg/mL) for 1 h to induce oxidative stress. Following this, FLuc-mRNA-loaded cLNPs or AONLNPs solutions were added to the cells and incubated for 24 h. Bioluminescence measurements were subsequently carried out as described above. Under oxidative stress conditions, luciferase expression for each formulation was normalized to its corresponding non-stressed control to assess relative retention of mRNA expression. Cyclic voltammetry experiments Compounds Raw-AON (N = 1–16) were individually dissolved in DMSO to prepare 10 mM solutions for the following test. Cyclic voltammetry (CV) experiments were performed on a CHI660E electrochemistry workstation. A three-electrode system was employed, consisting of a platinum wire working electrode, a glassy carbon counter electrode, and an Ag/Ag+ double-junction reference electrode immersed in acetonitrile. Tetrabutylammonium hexafluorophosphate (Bu4PF6) served as the supporting electrolyte in DMSO. All solutions were deoxygenated by nitrogen bubbling prior to measurements. Scans were conducted at a rate of 100 mV/s. Stability of FLuc-mRNA loaded cLNPs and AONLNPs In RNase-free tubes, the FLuc-mRNA loaded cLNPs or AO12LNPs were subjected to treatment with reactive species, including superoxide (O2-·), hydroxyl radical (·OH), hydrogen peroxide (H2O2), and hypochlorite (ClO-) at concentrations of 0.05 mM, as well as peroxynitrite (ONOO-) at a concentration of 0.02 mM, for a duration of 2 h. Then, the mixed solution was subjected to dialysis against 1× PBS for at least 12 h to eliminate excess reactive species. After the dialysis, FLuc-mRNA was extracted from the mRNA-loaded cLNPs or AO12LNPs formulation by isopropanol precipitation. 100 µL of FLuc-mRNA loaded cLNPs or AO12LNPs were diluted in 900 µL of a solution containing 60 mM ammonium acetate dissolved in isopropanol. A brief vortexing was conducted, followed by centrifugation at 14,000 g for 15 min at 4 oC. The supernatant was removed, and the resulting pellet was washed with 1 mL isopropanol. After a further centrifugation (4 °C), the pellet was dried under vacuum and then resuspended in 100 µL RNase-free water at room temperature for subsequent denaturing agarose gel imaging. Quantification of total intracellular mRNA by RT-qPCR RAW264.7 macrophage cells in 12-well plates (2 × 105 cells per well) was stimulated with PMA (0.2 μg/mL) for 1 h to induce oxidative stress. Following this, FLuc-mRNA-loaded cLNPs or AONLNPs solutions were added to the cells and incubated at different time points. FLuc-mRNA in RAW264.7 macrophage cells were collected and isolated using MolPure ®Flash Cell/Tissue Total RNA Kit (YEASEN, LOT:19221ES60). Following reverse transcription of the mRNA into cDNA using the BioRT Master HiSensi cDNA First Strand Synthesis Kit (Cat: BSB40M1), the resulting cDNA was quantified by RT-qPCR (BioRad) with Taq Pro Universal SYBR qPCR Master Mix (LOT: 7e1550f4). All primer sequences used in these experiments: forward primer for Fluc mRNA: 5’-TGATGTACCGGTTCGAGGAG-3’ and reverse primer for Fluc mRNA: 5’-TTGTCGATCAGGGTGCTCTT-3’; forward primer for GAPDH mRNA: 5’-TCCATGACAACTTTGGCATTG-3’ and reverse primer for GAPDH mRNA: 5’-CAGTCTTCTGGGTGGCAGTGA-3’. Dual-luciferase assay for in vitro transfection For the Firefly/Renilla luciferase assay, murine RAW264.7 macrophages were seeded in 96-well plates (1 × 10⁴ cells per well) and stimulated with 0.2 μg/mL PMA for 1 h to induce oxidative stress. The cells were then treated for 24 h with either a combination of FLuc mRNA-loaded cLNPs and RLuc mRNA-loaded cLNPs, or a combination of FLuc mRNA-loaded AO12LNPs and RLuc mRNA-loaded cLNPs. Following incubation, bioluminescence was measured using the Dual-Lumi™ II Luciferase Reporter Gene Assay Kit (Cat: RG027). Acute and chronic liver injury mouse models For the acute liver injury model, 6–8 weeks old female Balb/c mice weighting between 18 and 20 g were administered with a single intraperitoneal injection of acetaminophen (APAP) at a dose of 550 mg/kg, dissolved in phosphate-buffered saline (PBS)36. Prior to APAP administration, the mice were subjected to a 12-h fast, following the Whitten effect, to synchronize estrous cycles. For CCl4-induced chronic liver fibrosis model, 6–8 weeks old female C57BL/6 mice were intraperitoneal injected with CCl4/olive oil (1/9, v/v, 60 μL per mouse) twice a week for a total of 8 weeks40. The levels of acute and chronic liver injury were monitored by measuring serum AST and ALT levels following standard protocols (Beyotime, Cat. No. P2715M, P2711M). Acute and chronic lung injury mouse models For the acute lung injury model, 6–8 weeks old female Balb/c mice weighting between 18 and 20 g were administered with a single intratracheal injection of lipopolysaccharide (LPS) at a dose of 5 mg/kg59. For chronic lung fibrosis model, 6–8 weeks old female C57BL/6 mice were single intratracheal instillation of bleomycin (1.5 mg/kg in 50 μL sterile saline)60. In vivo FLuc mRNA delivery FLuc-mRNA loaded cLNPs and AO12LNPs were i.v. administrated to healthy mice or mice with acute and chronic liver or lung injury (0.6 mg/kg for FLuc-mRNA). For models of acute liver injury, at 6 h after injection, bioluminescence imaging was performed using IVIS Spectrum CT Biophotonic Imager (PerkinElmer). For models of CCl4-induced chronic liver fibrosis, at 6 h, 1 day, 2 days, 3 days, 4 days, and 7 days after injection, bioluminescence imaging was performed using IVIS Spectrum CT Biophotonic Imager (PerkinElmer). In vivo therapeutic mRNA delivery for liver regenerative therapy APAP-induced acute liver injury models were prepared as described above. At 24 h, the mice were i.v. treated with PBS, HGF-mRNA loaded cLNPs and AO12LNPs (0.5 mg/kg for HGF-mRNA), respectively. For CCl4-induced chronic liver fibrosis treatment, after 8 weeks of CCl4 induction, mice received a single injection of HGF-mRNA loaded cLNPs or AO12LNPs (0.5 mg/kg), followed by another two rounds of injections with 4 days intervals. After treatment, the mice were euthanized to collect their blood and livers. Serum was separated from collected blood. Liver damage biomarkers including ALT, AST were assayed. And liver tissues were fixed with 10% formalin and embedded with paraffin, followed by H&E staining. In vivo therapeutic mRNA delivery for lung regenerative therapy Female C57BL/6 mice (6–8 weeks old) were randomly allocated into four experimental groups: a PBS control group, a bleomycin-induced model group, a group treated with HGF-mRNA loaded lung-targeting cLNPs, and a group treated with HGF-mRNA loaded lung-targeting AO12LNPs. A pulmonary fibrosis model was established by a single intratracheal instillation of bleomycin (1.5 mg/kg in 50 μL of sterile saline). Seven days post-bleomycin administration, HGF-mRNA loaded lung-targeting cLNPs or AO12LNPs were i.v. administered. Upon completion of the 14-day experimental period, lung tissues from all groups were harvested for subsequent histopathological evaluation (H&E staining, Masson’s trichrome staining) and immunohistochemical (IHC) analysis. The hydroxyproline content in lung tissues was quantified using a commercial assay kit from Beijing Solarbio Science & Technology Co., Ltd. Levels of MDA, GSH, SOD, and CAT in lung tissues were measured with assay kits purchased from Shanghai Beyotime Biotechnology Co., Ltd. In vitro gene editing in RAW264.7 macrophages cells RAW264.7 macrophages cells in 6-well plates (1 × 106 cells per well) were stimulated with lipopolysaccharide (LPS) (1 μg/mL) for 1 h to induce oxidative stress. Following this, cLNPs or AO12LNPs (2.5 μg RNA per well, 1:1, Cas9 mRNA/NLRP3-sgRNA, wt/wt) solutions were added to the cells and incubated for 24 h. NLRP3 expression was assessed in RAW 264.7 cells under both basal and LPS-stimulated conditions via Western blot analysis. Following SDS-PAGE separation of cell lysates, proteins were transferred to a membrane and immunoblotted with the 30109-1-AP anti-NLRP3 antibody, which was applied at a 1:6000 dilution and incubated for 1.5 h at room temperature The sequence of NLRP3-sgRNA used in this experiment was obtained from IDT with the sequence: UCUGCUCAUAACGACGAGGCGUUUUAGAGCUAGAAAUAGCAAGUUAAAAUAAGGCUAGUCCGUUAUCAACUUGAAAAAGUGGCACCGAGUCGGUGCUUUU. As for quantification of on-target editing efficiency, RAW264.7 macrophages were stimulated with LPS and treated with AO12LNPs loaded with Cas9 mRNA and NLRP3 sgRNA. Genomic DNA was isolated 72 h after treatment, and the genomic region surrounding the NLRP3 target site was PCR-amplified. PCR products were subjected to Sanger sequencing, and indel frequencies were quantified using EditCo platform. As for off-target analysis by deep sequencing, potential off-target sites for the NLRP3 sgRNA were predicted using the CHOPOFF platform, and the five candidate loci were selected for analysis (Supplementary Table 6). The predicted off-target loci were PCR-amplified and subjected to next-generation sequencing (NGS) on illumina Hiseq Xten/ illumina Miseq/ illumina Novaseq / MGI2000 System. Gene editing (Cas9 mRNA/NLRP3-sgRNA) for psoriasis therapy 6-8 weeks old female Balb/c mice were randomly divided into four groups, including normal health group, imiquimod modeling group, NLRP3-targeting Cas9/sgRNA loaded cLNPs group, and NLRP3-targeting Cas9/sgRNA loaded AO12LNPs group. Psoriatic model mice were induced by daily topical application of 100 mg of 5% imiquimod cream to a shaved dorsal skin area (2 cm × 2 cm) for seven consecutive days53. Four hours after imiquimod treatment, Cas9/sgRNA-loaded cLNPs or AO12LNPs, containing 25 μg Cas9 and 25 μg NLRP3-targeting sgRNA, were administered via six-point subcutaneous injections distributed across the upper and lower regions of the inflamed area at 3-day intervals. Disease severity was quantified using a modified PASI scoring system based on erythema, scaling, and induration, with each parameter graded on a 0–4 scale (0 = none; 1 = slight; 2 = moderate; 3 = obvious; 4 = very obvious). Transepidermal water loss was measured using a Tewameter® TM Hex (SN: 22204251). Skin hydration, evaporative heat dissipation, local humidity, and water evaporation density were assessed using a Corneometer® dual MPA 580 (SN: 23071308). Upon completion of the treatment regimen, skin and spleen tissues were harvested for histopathological evaluation (H&E staining), ELISA assays, and immunohistochemical (IHC) analyses. Blood chemistry analysis In the experiment comparing recombinant HGF protein and HGF mRNA-loaded AO12LNPs for liver function recovery in the APAP-induced acute liver injury model, female Balb/c mice (6–8 weeks old, three mice per group) were adopted. Mouse blood was obtained via tail vein puncture under sterile conditions. Blood samples were collected and allowed to clot in serum separator tubes at ambient temperature. Following clot formation, the tubes were centrifuged at 1000 x g for 10 min. The resulting serum was then carefully collected and stored at −80 °C until further analysis. Serum AST (Beyotime, Cat. No. P2715M) and ALT (Beyotime, Cat. No. P2711M) levels were determined using corresponding kits with a Thermo Scientific microplate reader, following the protocol provided by the manufacturer. Histopathological and immunohistochemical staining (IHC) Left lateral lobe liver samples or skin were fixed with 4% paraformaldehyde, embedded in paraffin, and sectioned. After deparaffinization and dehydration, liver tissue sections were stained with H&E, Desmin, and Sirius Red. These stained sections were then examined to assess pathological changes in the liver tissues. Biosafety evaluation Healthy mice (famale, 6–8 weeks old) were i.v. injected with PBS, cLNPs, or AO12LNPs (0.6 mg/kg for FLuc-mRNA) every 4 days for a total of three injections. Blood testing was conducted 24 h after the final injection. For hematological analysis, five mice per group were used to detect blood routine indicators including white blood cells, monocytes, lymphocytes, hemoglobin, red blood cells (RBCs), and granulocytes using a YAN-305A hematology analyzer (Shanghai Yuyan Instruments Co., Ltd., China). For hepatic biochemistry detection, three mice per group were applied to determine serum AST (Beyotime, Cat. No. P2715M) and ALT (Beyotime, Cat. No. P2711M) levels following standard protocols. Mouse body weight was monitored over a period of 30 days. Major organs including heart, liver, spleen, lung, and kidney were also harvest after dose 3 to evaluate the pathological changes through H&E staining. Liver issue dissociation and preparation of single-cell suspensions The obtained liver tissues were placed in a Petri dish on wet ice, pre-rinsed with 1× PBS (free of RNase, Ca²⁺, and Mg²⁺ ions) to remove blood, surface debris. The tissues were then chopped into small pieces (approximately 0.5 mm2) and rinsed again with 1× PBS. These fragments were transferred into a dissociation buffer containing 0.35% collagenase IV, 2 mg/ml papain, and 120 U/ml DNase I, and incubated in a water bath shaker at 37 °C with agitation for 20 min. The suspension was filtered through a 70–30 µm cell strainer and centrifuged at 300 × g for 5 min at 4 °C. RBCs were removed by adding 1 ml of 1× erythrocyte lysis buffer (MACS 130-094-183, diluted from 10× stock) and incubated at room temperature or on wet ice for 2–10 min. The suspension was then centrifuged at 300 × g for 5 min at 4 °C, and the cell precipitate was collected. To eliminate dead cells, 100 µl of Dead Cell Removal MicroBeads (MACS 130-090-101) was added, mixed thoroughly, and incubated for 15 min at room temperature. After incubation, Binding Buffer was added, and the cell mixture was passed through MS Columns (MACS 130-042-201) to remove dead cells and other reagents. The cell precipitate was centrifuged at 300 × g for 5 min at 4 °C, washed twice with 1× PBS containing 0.04% BSA, and centrifuged under the same conditions. The cells were resuspended in 100 µl of 1× PBS with 0.04% BSA, and viability was assessed using trypan blue staining, ensuring >85% cell viability. 10x Genomics scRNA-seq analysis Single-cell suspensions were prepared and loaded onto the 10x Chromium chip following the manufacturer’s protocol for the 10x Genomics Chromium Single-Cell 3’ kit (V3). cDNA synthesis and amplification, as well as library preparation, were carried out according to established procedures. The scRNA-seq sequencing data were performed using the illumina NovaseqTM 6000 system (LC Bio Technology, China). The raw sequencing data were processed into FASTQ format using bcl2fastq software (version 5.0.1). The CellRanger-generated expression profile matrix was imported into Seurat (version 4.1.0) to filter out low-quality cells from the scRNA-seq dataset. The filtered data was then normalized, downscaled, and subjected to clustering analysis. The cells were visualized in a two-dimensional space using UMAP (Uniform Manifold Approximation and Projection). Gene expression values for each cell were calculated using the LogNormalize method from Seurat’s normalization function. Differentially expressed genes (DEGs) between the HGF mRNA-loaded AO12LNPs and cLNPs treatment groups were identified using the FindMarkers() function in the Seurat package. The DEG analysis was performed with the following criteria: P < 0.01, |log₂FC | ≥ 0.26 (corresponding to ≥1.2-fold change), and gene expression detected in at least 10% of cells. GO and KEGG enrichment analyses were conducted on the differentially expressed genes identified for each cluster, relative to other clusters, using hypergeometric testing. We conducted cell interaction analysis using CellChat (v 1.6.1) with default parameters. Hepatocytes were subsetted and re-analyzed using Seurat. Principal component analysis (PCA) was performed based on the top 2000 highly variable genes, and the first 20 principal components were used for neighborhood graph construction and clustering with the FindNeighbors and FindClusters functions. The results were visualized using UMAP. Pseudotime trajectory analysis was performed using Monocle (version 2.22.0). Differential gene expression analysis was conducted using the differentialGeneTest function, and cells were ordered along differentiation trajectories based on the expression patterns of differentially expressed genes using the default Monocle workflow. Bioinformatics analysis was performed using the OmicStudio platform, available at https://www.omicstudio.cn/tool. Statistics and reproducibility Data was plotted as mean ± standard deviation (s.d.) of biological replicates, unless otherwise indicated. Detailed statistical methods are described in the text and figure legends. Statistical testing was performed using GraphPad Prism 6.01. P < 0.05 were statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Source Data