GNPs-pIL-4 reprograms macrophage polarization and activates the OSM/GSNOR/ENG axis to improve angiogenesis in ischemic limbs

Abstract

Therapeutic angiogenesis based on gene therapies is a potential peripheral artery disease (PAD) treatment yet needs a more stable, efficient, and high-affinity delivery vector and an optimized delivery strategy. Here, we engineered three-dimensional graphene nanoparticles modified with folic acid and polyethyleneimine for macrophage-specific delivery of interleukin-4 plasmids (pIL-4), forming GNPs-pIL-4 for local intramuscular injection. GNPs-pIL-4 had uniform size, positive surface charge, and strong nucleic acid loading capacity (K_d: 25 nanomolar). In vivo, GNPs-pIL-4 reshaped the ischemic microenvironment by inducing reparative M2 macrophage polarization. Mediated by macrophages, GNPs-pIL-4 improved muscle contractility, delayed strength loss, and enhanced blood perfusion and oxygen saturation. Mechanistically, GNPs-pIL-4 specifically turned on the oncostatin M–mediated macrophage-endothelium communication and then activated the angiogenesis, with increased endothelial sprouting, migration, and tube formation. These effects attributed to the down-regulated S-nitrosoglutathione reductase expression, thereby increasing S-nitrosylation at the C209 site of endoglin. GNPs-pIL-4 represents a promising gene therapy strategy for PAD.

INTRODUCTION

Peripheral artery disease (PAD) is a chronic ischemic condition caused by arterial stenosis or occlusion, substantially impairing patient quality of life. Now, more than 230 million adults worldwide are affected by lower-limb PAD. Among them, patients with critical limb ischemia exhibit particularly poor prognosis, with 1-year mortality rates reaching 25 to 35%, amputation rates of up to 30%, and 5-year amputation-free survival as low as 27.1% (1). Although current therapies—including revascularization procedures, antiplatelet treatment, and exercise rehabilitation—can alleviate symptoms, there remain few effective strategies to restore microvascular perfusion and promote tissue regeneration (2).

Therapeutic angiogenesis has garnered increasing interest in recent years due to its potential to restore tissue perfusion (3, 4). However, the exogenous delivery of individual proangiogenic factors such as vascular endothelial growth factor (VEGF) or fibroblast growth factor (FGF), while capable of initiating angiogenesis, often results in the formation of immature and leaky vessels (5). Actually, the microenvironment within ischemic tissues plays a dual role in angiogenesis—both initiating the process and determining the structural and functional integrity of neovessels (6). This suggests that functional vascular regeneration requires not only molecular stimulation but also coordinated interactions among multiple cell types and dynamic regulation of the local microenvironment (5, 7, 8). Therefore, identifying interventions that reshape the local tissue microenvironment to enhance reperfusion and regeneration is a critical unmet need in PAD treatment (9).

Typically, local hypoxia and inflammatory cytokines [e.g., tumor necrosis factor–α (TNF-α) and interleukin-6 (IL-6)] drive macrophage polarization toward the proinflammatory M1 phenotype, initiating an immune response (10). Over time, macrophages transition toward the reparative M2 phenotype, characterized by anti-inflammatory cytokine secretion [e.g., IL-10 and transforming growth factor–β (TGF-β)], which promotes angiogenesis and tissue repair (11). Excessive activation of M1 macrophages or insufficient of M2 macrophages can disrupt immune homeostasis and contribute to the deteriorating progression of PAD. Therefore, targeting macrophage polarization from M1 to M2 represents a promising strategy to modulate the ischemic microenvironment and enhance regenerative outcomes.

IL-4 is a classical inducer of M2 polarization and plays a vital role in immune modulation and tissue repair (12, 13). However, IL-4 protein exhibits a short half-life in vivo and is prone to enzymatic degradation, necessitating frequent high-dose administration and raising concerns about off-target effects. Its broad immunoregulatory activity can also exacerbate T helper cell 2–mediated responses, contribute to allergic disorders, and potentially promote tumor immune evasion in oncologic settings (14). In contrast, gene delivery approaches enabling sustained IL-4 expression offers a more controlled and durable therapeutic alternative, provided that delivery is safe, targeted, and efficient (15).

Gene delivery systems now rely on either viral vectors or nonviral nanocarriers (16). While viral vectors are high efficiency, they carry risks of genomic integration and immunogenicity (17). Nonviral systems, such as lipid nanoparticles and cationic polymers, are generally safer but often lack cell type specificity (18). Graphene-based nanoparticles, owing to their large surface area, membrane-penetrating capability, and functional modifiability, have emerged as promising platforms for nucleic acid delivery. Previous studies have demonstrated that IL-4 plasmid-loaded graphene-based nanoparticles can target cardiac macrophages and mitigate inflammation following myocardial infarction (19). However, their therapeutic potential in angiogenesis and PAD remains underexplored.

In addition, nitric oxide (NO) plays a central role in regulating endothelial cell function and vascular remodeling (20). Protein S-nitrosylation, a key NO-dependent modification, is regulated by S-nitrosoglutathione reductase (GSNOR), which maintains intracellular redox balance. It has been reported that GSNOR is activated during myocardial ischemia and resuscitation. Inhibition of its activity or genetic deletion can restore protein S-nitrosylation levels, reduce oxidative stress and inflammation, and protect brain tissue (21). In addition, GSNOR deficiency has been shown to promote angiogenesis and perfusion recovery during myocardial ischemia (22). However, the role of GSNOR in limb ischemia remains unclear.

In this study, a graphene nanoparticle–based IL-4 plasmid delivery system (GNPs-pIL-4) to reshape the local microenvironment, enable the targeted macrophage reprogramming, and enhance ischemic tissue regeneration was successfully developed. We then systematically evaluated its effects on macrophage polarization, angiogenesis, and muscle functional recovery, in which the OSM–GSNOR–endoglin (ENG) regulatory signaling axis played an important role in orchestrating the sprouting angiogenesis and functional repair. This work provided a mechanistic framework and potential therapeutic strategy for immune-targeted nanomedicine in PAD.

RESULTS

Physicochemical characterization, nucleic acid loading capacity, and biological effects of GNPs in vitro and in vivo

To develop a safe and efficient nonviral gene delivery system, we first synthesized three-dimensional graphene nanoparticles (3DGNPs) and then performed a systematic physicochemical characterization. Scanning electron microscopy revealed a uniform spherical morphology with a porous and wrinkled three-dimensional structure (Fig. 1A). Following 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide and N-hydroxysuccinimide (EDC/NHS) activation, polyethyleneimine (PEI) and folic acid (FA) were successfully conjugated to the GNP surface, generating a positively charged and targetable nanoplatform (Fig. 1B). Dynamic light scattering analysis showed an average hydrodynamic diameter of ∼168.5 nm, a polydispersity index (PDI) of 0.168, and a zeta potential of +41.9 mV, indicating great dispersibility and stability, which was suitable for binding negatively charged nucleic acids (Fig. 1C). Fourier transform infrared spectroscopy further confirmed the successful conjugation of PEI and FA onto the GNP surface (Fig. 1D). Microscale thermophoresis (MST) analysis demonstrated strong binding between GNPs and plasmid DNA, with a dissociation constant (K_d) of ∼25 nM (Fig. 1E), while agarose gel retardation assays showed complete DNA loading at weight ratios of 2:1 or higher (Fig. 1F).

Fig. 1. Physicochemical characterization, nucleic acid loading capacity, and biological effects of GNPs.

(A) Simulated and scanning electron microscopy images of GNPs morphology. (B) Schematic of GNP synthesis and surface modification with PEI and FA. (C) Dynamic light scattering analysis of GNP hydrodynamic size and zeta potential distribution. (D) Fourier transform infrared spectra of GNPs, PEI, FA, and GNPs-FA-PEI. (E) MST assay evaluating the binding affinity between GNPs and plasmid DNA, with fitted K_d values shown. (F) Agarose gel retardation assay demonstrating DNA condensation by GNPs at different weight ratios. (G to I) Serum stability analysis: temporal changes in hydrodynamic size (G), PDI (H), and zeta potential (I). (J) In vitro release profile of GNPs/DNA complexes under a macrophage-mimicking acidic pH environment. (K) Confocal microscopy images of RAW264.7 cells after incubation with FITC-labeled GNPs (green); nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI), and membranes were stained with 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI; red). (L) TEM images showing intracellular localization of GNPs. (M) Flow cytometric analysis of FITC-labeled GNP uptake by F4/80⁺ macrophages. (N) Fluorescence microscopy images comparing transfection efficiency of GNPs/pEGFP and LNP/pEGFP in RAW264.7 cells. (O) Fluorescence microscopy images of mouse gastrocnemius muscles showing EGFP expression (green). Nuclei were stained with DAPI (blue). (P) Western blot analysis of EGFP and the internal reference glyceraldehyde phosphate dehydrogenase (GAPDH) protein expression in muscle tissues (left). Bar graph showing the relative EGFP protein expression level normalized to GAPDH (right). (Q) Cell viability of RAW264.7 cells treated with GNPs, measured by CCK-8 assay. (R and S) In vivo biodistribution of DiR-labeled GNPs following intramuscular injection: (R) whole-body imaging at 6 hours postinjection; (S) fluorescence signals in dissected organs and gastrocnemius muscles. Statistical analysis: Student’s t test was used for two-group comparisons [(N) and (O)], and one-way analysis of variance (ANOVA) with Bonferroni or Šídák correction was applied for multigroup comparisons [(P) and (Q)]. Data were shown as means ± SD. h, hours; d, days.

We next evaluated the stability of GNP/DNA complexes under simulated conditions. Serum stability tests showed that the hydrodynamic size, PDI, and zeta potential of the complexes remained stable for up to 4 days (Fig. 1, G to I). In an acidic pH environment mimicking macrophage intracellular condition, DNA exhibited a gradual release profile, reaching a plateau at ∼20 hours, after which the release rate stabilized (Fig. 1J). At the cellular level, fluorescein isothiocyanate (FITC)–labeled GNPs were efficiently internalized by RAW264.7 macrophages, showing clear colocalization with 4′,6-diamidino-2-phenylindole (DAPI)–stained nuclei and 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate–labeled membranes (Fig. 1K). Cell-specific uptake assays confirmed that GNPs were preferentially taken up by RAW264.7 macrophages, with minimal internalization in human umbilical cord endothelial cells (HUVECs) and NIH3T3 fibroblasts (fig. S1). Transmission electron microscopy further confirmed intracellular localization of GNPs within the cytoplasm of RAW264.7 macrophages (Fig. 1L). Flow cytometric analysis of dissociated gastrocnemius muscle showed that most of the cells internalizing FITC-labeled GNPs were F4/80⁺ macrophages (Fig. 1M), indicating that macrophages in muscle tissue exhibit a significantly higher uptake of GNPs compared to other cell types.

GNPs efficiently delivered a reporter plasmid (pEGFP) into RAW264.7 cells, which served as a surrogate to evaluate gene transfection capability. Compared with lipid nanoparticles (LNPs), GNPs achieved significantly higher transfection efficiency (Fig. 1N). Next, we evaluated the in vivo gene delivery effects of the GNPs. To compare with a commercially available LNP transfection reagent, both GNPs and LNPs were loaded with the pEGFP plasmid and administered via intramuscular injection into the mouse gastrocnemius muscles respectively. We found that GNPs-pEGFP exhibited higher green fluorescent protein expression efficiency in the mouse gastrocnemius muscles, which was consistently confirmed by fluorescence microscopy observation (Fig. 1O) and semiquantitative Western blotting data (Fig. 1P). To evaluate the cytotoxicity, we established a gradient of increasing concentrations of GNPs to treat RAW264.7 cells, ranging from 0.4 to 25.6 μg/ml. CCK-8 assays revealed that even at a high concentration of 6.4 μg/ml, GNPs did not significantly affect the RAW264.7 cell viability (Fig. 1Q). In vivo biodistribution studies showed that DiR-labeled GNPs were primarily retained in the gastrocnemius muscles 6 hours after intramuscular injection, with markedly reduced nonspecific accumulation in the liver compared with free dye controls (Fig. 1, R and S). These data demonstrated that GNPs had ideal physicochemical properties, strong nucleic acid loading and controlled release capacity, efficient macrophage-specific uptake, and favorable biodistribution and biosafety profiles, thereby providing a solid foundation for subsequent pIL-4 delivery and functional validation.

GNPs-pIL-4 promotes M2 macrophage polarization

To evaluate the biological effects of GNPs-pIL-4 in vivo, we established a murine hindlimb ischemia (HLI) model via femoral artery ligation and intramuscularly injected saline (NS), GNPs, pIL-4, or GNPs-pIL-4 into the gastrocnemius muscles (Fig. 2A). At day 7 posttreatment, IL-4 protein expression was significantly elevated in the HLI + GNPs-pIL-4 group compared to the Sham + NS, HLI + NS, HLI + GNPs, and HLI + pIL-4 groups, demonstrating that GNPs successfully delivered the IL-4 plasmid and induced IL-4 expression in ischemic muscles (Fig. 2B). Consistently, quantitative polymerase chain reaction (qPCR) analysis revealed that M1 marker genes Cd80 and Nos2 in the HLI + GNPs-pIL-4 group were down-regulated compared to the HLI + NS group (Fig. 2C). To determine whether the macrophages induced by GNPs-pIL-4 were of the M2 type and to preliminarily identify their specific subtype, we further examined the mRNA expression of multiple characteristic M2 markers. Data showed that, compared with the HLI + NS group, the expression of the classic M2a macrophage markers Arg1 and Mrc1 was specifically and significantly up-regulated in the HLI + GNPs-pIL-4 treatment group (Fig. 2D). In contrast, the expression of the M2b markers (IL-1β and IL-6) and M2c-related markers (MerTK and TGF-β) showed no significant changes (Fig. 2D). These data preliminarily indicated that the macrophage polarization promoted by GNPs-pIL-4 was predominantly skewed toward the M2a subtype. Multiplex cytokine assays further confirmed that GNPs-pIL-4 increased the levels of M2-associated cytokines IL-4 and IL-10 while reducing M1-associated cytokines TNF-α and CCL3 in ischemic tissues (Fig. 2, E and F). Single-cell RNA sequencing (RNA-seq) analysis revealed that macrophages in ischemic muscles from GNPs-pIL-4–treated mice exhibited elevated Mrc1 expression and reduced CD86 expression, confirming a shift toward the reparative M2 phenotype (Fig. 2G). In addition, ex vivo experiments demonstrated that GNPs-phIL-4 promoted human peripheral blood mononuclear cell (PBMC)–derived macrophages polarization toward the CD206⁺/CD80⁻ M2 phenotype (Fig. 2H).

Fig. 2. GNPs-pIL-4 promotes M2 macrophage polarization.

(A) Schematic illustration of the experimental workflow: C57BL/6J mice were subjected to femoral artery ligation to establish HLI and then received intramuscular injections of NS, GNPs, pIL-4, or GNPs-pIL-4, respectively. Gastrocnemius muscle samples were collected at day 7 postsurgery for molecular, cellular, and functional assessments. (B) A mouse model of HLI was established via femoral artery ligation and resection in C57BL/6J mice, followed by intramuscular injection of saline, GNPs, GNPs-pIL-4, or pIL-4 plasmids alone. IL-4 protein expression in gastrocnemius muscle was assessed by Western blot at day 7 postsurgery, with GAPDH as loading control (n = 4). (C) Relative mRNA expression of M1 macrophage marker genes (Cd80 and Nos2) in gastrocnemius muscle (n = 3). (D) Relative mRNA expression of M2 macrophage subtype marker genes in gastrocnemius muscle (n = 3), including markers for M2a (Arg1 and Mrc1), M2b (IL-1β and IL-6), and M2c (TGF-β and MerTK) subtypes. (E and F) Multiplex flow cytometry was performed to quantify cytokine levels in gastrocnemius tissue at day 7 postsurgery (n = 4). (G) Single-cell sequencing analysis of mouse gastrocnemius muscle revealed the expression levels of M2 macrophage marker Mrc1 and M1 marker CD86. (H) Human peripheral blood monocytes were differentiated into macrophages in vitro and subsequently treated with GNPs-phIL-4; M1/M2 polarization ratios were analyzed by flow cytometry. Statistical analysis: One-way ANOVA with Bonferroni or Šídák correction was applied for multigroup comparisons [(B) to (F)]. Data were shown as means ± SD. (G) was based on single-cell transcriptomic analysis. Representative images were shown in (H). GM-CSF, granulocyte-macrophage colony-stimulating factor.

GNPs-pIL-4 promotes perfusion recovery and tissue regeneration in murine ischemic hindlimb

Perfused vessels were labeled by intravenous injection of FITC-dextran, and then the ischemic areas in these mice were detected by two-photon intravital imaging at day 7 posttreatment. The data showed that compared with the HLI + NS group, the HLI + GNPs-pIL-4 group exhibited the strongest and much more continuous FITC fluorescence signals in the ischemic muscles, whereas no improvement was observed in the HLI + GNPs or HLI + pIL-4 groups (Fig. 3A). These findings indicated that only the combined GNPs-pIL-4 treatment effectively increased the number of perfused vessels in ischemic muscles, while GNPs or pIL-4 alone treatment had no significant benefit. Tissue oxygenation was assessed using ultrasound-photoacoustic imaging. Compared with the HLI + NS group, the HLI + GNPs-pIL-4 group exhibited significantly enhanced oxygenation signals in the ischemic muscles, whereas no improvement was observed in the HLI + GNPs or HLI + pIL-4 group (Fig. 3B). In addition, capillary density and vascular maturity were assessed by CD31 and α–smooth muscle actin (α-SMA) immunohistochemical staining, respectively. Compared with the HLI + NS group, the HLI + GNPs-pIL-4 treatment markedly increased the number of CD31-positive capillaries in the ischemic muscles, whereas no improvement was observed in the HLI + GNP–treated or HLI + pIL-4–treated group (Fig. 3C). α-SMA staining revealed that these newly formed capillaries in the GNPs-pIL-4–treated group were frequently surrounded by α-SMA–positive mural cells (Fig. 3D), indicating enhanced vascular maturation and stability. Furthermore, hindlimb perfusion recovery was monitored respectively by use of laser speckle imaging. The data showed that blood perfusion was markedly improved in the HLI + GNPs-pIL-4 group compared with the HLI + NS group, whereas no significant improvement was observed in the HLI + GNPs or HLI + pIL-4 group (Fig. 3E). A hallmark of newly regenerated fibers is the presence of centrally located nuclei (23). To evaluate the impact of GNPs-pIL-4 on postischemic skeletal muscle regeneration, we performed a detailed morphometric analysis of myofibers with hematoxylin and eosin (H&E) staining and wheat germ agglutinin (WGA) fluorescence staining (Fig. 3, F and G). Compared to the HLI + NS group, the HLI + GNPs-pIL-4 treatment group showed a significant increase in the percentage of centrally nucleated myofibers, confirming its superior efficacy in promoting myogenic repair over the HLI + GNPs, HLI + pIL-4, or HLI + NS group (Fig. 3H). Quantification of myofiber count per high-power field (HPF) revealed a significant reduction in the GNPs-pIL-4 treatment group compared to the HLI + NS group (Fig. 3I). This decrease, observed by day 7 posttreatment, likely reflected a progression in the repair process from a phase rich in proliferating, newly formed small fibers to a stage characterized by active myofiber fusion, hypertrophy, and maturation, leading to a lower count of individual fibers per field. Consistently, the average myofiber cross-sectional area (CSA) in the HLI + GNPs-pIL-4 group was significantly larger than that in the HLI + NS group (Fig. 3J), indicating promoted myofiber growth. Further analysis of the CSA distribution provided deeper insight. The Sham + NS group exhibited a distribution skewed toward larger fiber areas, while the HLI + NS group showed a pronounced shift toward smaller areas, indicative of impaired regeneration. In contrast, the HLI + GNPs-pIL-4 treatment group demonstrated a more balanced and up/right-shifted distribution across area bins (Fig. 3, K and L). This pattern indicated that GNPs-pIL-4 treatment not only actively stimulates the regeneration of myofibers but also effectively guides them toward a more mature and hypertrophic phenotype. These systematic morphometric data demonstrated that GNPs-pIL-4 treatment effectively promoted substantial skeletal muscle regeneration and maturation after ischemic injury. The muscle fatigue curve reflects the functional status of ischemic muscle. We found that the HLI + GNPs-pIL-4 group exhibited greater resistance to fatigue (a slower decline in the fatigue curve) and higher minimal contraction force retention compared with the HLI + NS group, whereas no improvement was observed in the HLI + GNPs or HLI + pIL-4 group (Fig. 3M).

Fig. 3. GNPs-pIL-4 promotes tissue regeneration and vascular recovery in ischemic hindlimbs.

(A) Two-photon in vivo imaging of perfused vessels (FITC-dextran injection) in ischemic limbs; vessels within a volume of 500 μm by 500 μm by 100 μm were quantified (n = 3). (B) Ultrasonographic evaluation of tissue oxygen saturation on day 7 postsurgery. (C) Representative images of CD31 immunohistochemical staining (brown) in ischemic gastrocnemius muscles. The number of CD31-positive vessels per HPF was quantified. (D) Representative immunofluorescence images of α-SMA (red) and DAPI (blue) costaining in ischemic muscles. The bar graph (right) quantified the relative α-SMA fluorescence intensity. (E) Laser speckle contrast imaging of hindlimb perfusion on day 7 post-HLI (n = 5). (F) H&E-stained cross sections of gastrocnemius muscles from the Sham + NS, HLI + NS, HLI + GNPs-pIL-4, HLI + GNPs, and HLI + pIL-4 groups. (G) Fluorescent images of muscle sections stained with WGA (magenta, myofiber borders) and DAPI (blue, nuclei). (H) Quantitative analysis of the percentage of centrally nucleated myofibers from the WGA/DAPI-stained images. (I) Quantitative analysis of the total number of myofibers per HPF from the WGA-stained images. (J) Analysis of the average myofiber CSA calculated from the WGA-stained images. (K) Violin plots depicting the distribution of myofiber CSA for each group. (L) Frequency distribution histogram of myofiber CSA, binned by area. (M) A muscle tension measurement system was used to record the fatigue curves and calculate the minimum force retention of toe flexor muscles under continuous stimulation (n = 4). Statistical analysis: One-way ANOVA was used for multiple group comparisons in (A), (C), (D), (E), (H), (I), (J), (K), and (M), followed by Bonferroni correction or Šídák multiple comparisons test where applicable. Data were shown as means ± SD. (A), (C), and (D) to (G) showed the representative images.

To verify the long-term benefits of GNPs-pIL-4 treatment and compare the performance of its delivery vehicle, GNPs, with the traditional LNP system, we conducted extended observations for 14 days and performed a parallel comparative study. Specifically, regarding long-term efficacy, longitudinal laser speckle imaging revealed that mice in the GNPs-pIL-4 treatment group maintained superior perfusion recovery compared to the HLI + NS and HLI + LNP-pIL-4 groups from postoperative day 3 through day 14 (fig. S2, A and B). In vivo two-photon microscopy on day 14 postsurgery demonstrated a denser and better-connected network of perfused microvessels within the muscles of the GNPs-pIL-4 group (fig. S2, C and D). Concurrently, a significant increase in the density of α-SMA–positive vessels was observed in the muscle sections of this group (fig. S2, E and F), indicating that the treatment promoted vascular maturation. Correspondingly, GNPs-pIL-4 treatment still significantly improved muscle fatigue resistance and minimal contraction force retention on day 14 (fig. S2, G and H). Regarding the vehicle comparison, compared to LNP-pIL-4, GNPs-pIL-4 induced greater blood flow recovery (fig. S2, A and B), the formation of a greater number of perfused vessels with better structure (fig. S2, C and F), and superior functional muscle recovery on day 14 (fig. S2, G and H). These integrated data demonstrated that GNPs-pIL-4 treatment provided some sustained, multifaceted benefits for up to 14 days and that the GNP vehicle itself contributed substantially to achieving superior therapeutic outcomes compared to the traditional LNP system.

Macrophages mediate the GNPs-pIL-4–induced murine muscle regeneration and reperfusion after ischemia

To confirm GNPs-pIL-4–induced tissue repair by targeting macrophage in vivo, clodronate liposomes (CLs) were administered intraperitoneally to selectively deplete murine macrophages, in which empty liposomes (ELs) served as the control. In the ischemic hindlimb model, GNPs-pIL-4 treatment significantly delayed the fatigue progression under continuous stimulation and enhanced the minimal contraction force retention of the flexor digitorum muscles, suggesting the improved muscle function (HLI + EL + GNPs-pIL-4 versus HLI + EL + NS). However, these effects were markedly attenuated following macrophage depletion (HLI + CL + GNPs-pIL-4 versus HLI + EL + GNPs-pIL-4) (Fig. 4A). H&E staining further demonstrated the enhanced skeletal muscle regeneration (more centrally nucleated regenerating fibers) with GNPs-pIL-4 treatment (HLI + EL + GNPs-pIL-4 versus HLI + EL + NS), while this regenerative effect was substantially diminished in the CL-pretreated group (HLI +CL + GNPs-pIL-4 versus HLI + EL + GNPs-pIL-4) (Fig. 4B), indicating a critical regulatory role of macrophages in GNPs-pIL-4–induced muscle repair.

Fig. 4. Macrophages are essential mediators of GNPs-pIL-4–induced murine muscle regeneration and reperfusion after ischemia.

C57BL/6J mice were intraperitoneally injected with CLs or ELs to deplete macrophages for 7 consecutive days (administered every other day; initial dose: 200 μl/25 g body weight, maintenance dose: 100 μl/25 g). On day 7, mice underwent femoral artery ligation to induce HLI, followed by immediate intramuscular injection of GNPs-pIL-4 or control treatments. (A) On postoperative day 7, toe flexor muscle fatigue curves and minimum force retention were recorded using a muscle tension system (n = 4). (B) H&E staining was performed to assess muscle morphology and myofiber regeneration at day 7. (C) Laser speckle contrast imaging was conducted pre-HLI, 30 min post-HLI, and on day 7 to monitor perfusion recovery; perfusion was quantified at day 7 (n = 3). (D) Oxygen saturation in ischemic limbs was assessed using ultrasonography at day 7. (E) Perfused vasculature was visualized by two-photon in vivo imaging following intravenous injection of FITC-dextran; and vessel counts were quantified in ischemic limbs (n = 3). (F) CD31 immunohistochemical staining was performed to label and quantify vascular density in transverse gastrocnemius sections (n = 3). Statistical analysis: One-way ANOVA followed by Bonferroni or Šídák multiple comparisons tests was used for statistical evaluation in (A), (C), (E), and (F). Data were shown as means ± SD. (B) and (D) show representative images.

In terms of perfusion, GNPs-pIL-4 markedly improved the blood flow recovery ratio in the ischemic limb by day 7 postsurgery (HLI + EL + GNPs-pIL-4 versus HLI + EL + NS) (Fig. 4C). Consistent with this finding, oxygen saturation imaging revealed superior tissue oxygenation in the HLI + EL + GNPs-pIL-4 group compared with the HLI + EL + NS group (Fig. 4D). Two-photon intravital imaging and CD31 immunohistochemistry also demonstrated that GNPs-pIL-4 significantly increased the number of perfused vessels and capillary density (HLI + EL + GNPs-pIL-4 versus HLI + EL + NS). However, all these proangiogenic effects were largely abolished upon macrophage depletion (HLI + CL + GNPs-pIL-4 versus HLI + EL + GNPs-pIL-4) (Fig. 4, E and F). Therefore, macrophages played an indispensable role in mediating the proregenerative and properfusion effects of GNPs-pIL-4 in murine ischemic tissues.

GNPs-pIL-4 activates macrophage secretory OSM signaling to further promote endothelial cell sprouting, migration, and tube formation

To explore the molecular mechanisms underlying the GNPs-pIL-4–induced perfusion recovery, we analyzed endothelial cells data from single-cell RNA-seq of gastrocnemius muscle with or without GNPs-pIL-4 treatment after ischemia. Gene set enrichment analysis (GSEA) revealed that endothelial cells from the HLI + GNPs-pIL-4 group exhibited significant activation of sprouting angiogenesis–related pathways compared with the HLI group (Fig. 5A and fig. S4A). The enrichment plot of sprouting angiogenesis showed a normalized enrichment score of 1.94 (Fig. 5B and fig. S4, B and C), with many related genes significantly up-regulated by GNPs-pIL-4 treatment (HLI + GNPs-pIL-4 versus HLI + NS) (Fig. 5C and fig. S4, D and E).

Fig. 5. GNPs-pIL-4 activates macrophage secretory OSM signaling to promote endothelial cell sprouting, migration, and tube formation.

(A to H) Single-cell suspensions were prepared from the gastrocnemius muscles of mice subjected to HLI, with mature myofibers removed before single-cell RNA-seq and analysis. (A) GSEA plot showing up-regulated pathways in endothelial cells from the HLI + GNPs-pIL-4 versus HLI groups. [(B) and (C)] GSEA enrichment curves and volcano plot of sprouting angiogenesis–related gene expression. (D) Quantification of the cell-cell interaction number and interaction strength across groups. (E) CellChat analysis of the interactions between macrophages and other cell types. (F) Comparison of signaling pathway information flow intensity between groups. (G) Expression of OSM and its receptors [leukemia inhibitory factor receptor (LIFR), oncostatin M receptor (OSMR), and IL-6 signal transducer (IL6ST)] in endothelial cells and macrophages. (H) Violin plot of OSM expression levels. (I) UMAP visualization of Osm expression in gastrocnemius muscles. (J) OSM protein expression in mouse gastrocnemius muscles following GNPs-pIL-4 treatment in the HLI model, with or without macrophage depletion using CL (n = 3). (K) Phase-contrast microscopy images and quantification of sprouting area in murine femoral artery explants cultured with or without recombinant OSM (20 ng/ml) (n = 3). (L) Scratch assay monitoring EOMA endothelial cell migration with or without OSM (20 ng/ml) treatment (n = 3). (M) Tube formation assay of EOMA endothelial cells on Matrigel with or without OSM (20 ng/ml) treatment (n = 3). (N to P) Effects of conditioned medium from GNPs-pIL-4–treated macrophages (MφCM) on explant sprouting (N), endothelial cell migration (O), and tube formation (P), evaluated with or without anti-OSM neutralizing antibody (n = 3). Statistical analysis: Student’s t test was used for two-group comparisons [(K) to (M)], and one-way ANOVA with Bonferroni or Šídák correction was applied for multigroup comparisons [(J) and (N) to (P)]. Data were shown as means ± SD. (K) to (P) showed the representative images. NES, normalized enrichment score.

To investigate how GNPs-pIL-4–stimulated macrophages modulate endothelial cells, we performed cell-cell interaction analysis and found that the total signal communication and interaction between macrophages and endothelial cells were markedly enhanced by GNPs-pIL-4 treatment (HLI + GNPs-pIL-4 versus HLI) (Fig. 5, D and E). Among the intercellular signals, OSM signaling was uniquely activated in the HLI group compared with the Sham group and further up-regulated by GNPs-pIL-4 treatment (Fig. 5F). There are three OSM receptors including leukemia inhibitory factor receptor, oncostatin M receptor (OSMR), and IL-6 signal transducer in endothelial cells, which all together received the OSM signaling. Receptor-ligand analysis identified the overall OSM postreceptor signal enhancement in endothelial cells of the HLI group than the Sham group and much more in the GNPs-pIL-4–treated group, although increase mode was likely different (Fig. 5G). In addition, OSM expression itself was also significantly up-regulated in macrophages upon GNPs-pIL-4 treatment (HLI + GNPs-pIL-4 versus HLI) (Fig. 5H). Uniform manifold approximation and projection (UMAP) visualization showed the highest localized enrichment of OSM (red) in macrophages of the GNPs-pIL-4 treatment group (Fig. 5I), suggesting that the effects of macrophages promoting endothelial sprouting are likely associated with the OSM signal activation.

To determine the role of macrophage-derived OSM in angiogenesis, we first confirmed by Western blot that GNPs-pIL-4 significantly increased OSM protein expression in murine ischemic muscle tissues (HLI + EL + GNPs-pIL-4 versus HLI + EL + NS), which was markedly attenuated following macrophage depletion (HLI + CL + GNPs-pIL-4 versus HLI + EL + GNPs-pIL-4) (Fig. 5J). In addition, recombinant OSM significantly promoted sprouting angiogenesis (OSM versus control) (Fig. 5K). Scratch assays showed that OSM enhanced endothelial cell migration compared with the control group (Fig. 5L), and Matrigel-based tube formation assays demonstrated the increased capillary-like network formation by OSM (OSM versus control) (Fig. 5M and movie S1). Furthermore, we collected the macrophage-conditioned medium (MφCM) from GNPs-pIL-4–treated or untreated RAW 264.7 cells, together with or without OSM neutralization antibody, and used it to stimulate endothelial cells. Data showed that the conditioned media from GNPs-pIL-4–treated macrophages significantly promoted endothelial sprouting, migration, and tube formation (GNPs-pIL-4 + MφCM versus MφCM), all of which were substantially diminished upon neutralization of OSM with blocking antibodies (GNPs-pIL-4 + MφCM versus GNPs-pIL-4 + MφCM + AntiOSM) (Fig. 5, N and P, and movie S2). Consistently, qPCR analysis revealed that GNPs-pIL-4–treated MφCM up-regulated key genes that enriched in sprouting angiogenesis pathways, including Kdr, Fgf1, Pdgfb, and Angpt1 (GNPs-pIL-4 + MφCM versus MφCM), which were diminished by OSM neutralization antibody (GNPs-pIL-4 + MφCM versus GNPs-pIL-4 + MφCM + AntiOSM) (fig. S5, A to D). Therefore, GNPs-pIL-4 promoted OSM secretion from macrophages, which, in turn, programed endothelial cells to sprouting angiogenesis.

OSM knockdown alleviates GNPs-pIL-4–induced muscle regeneration and angiogenesis in ischemic tissues of mice

To determine whether OSM mediates the GNPs-pIL-4–induced tissue repair in vivo, we used adeno-associated virus (AAV)–shOSM to knock down Osm expression in the gastrocnemius muscle of mice under ischemia. Infected 2 weeks later, the ischemic gastrocnemius muscles of mice were collected for qPCR analysis. The data showed that Osm expression was significantly reduced in the AAV-shOSM group compared to the AAV-shNC group, showing that AAV-shOSM effectively suppressed the expression of Osm (fig. S6). After establishing a murine HLI model and immediately administering intramuscular GNPs-pIL-4, we observed that Osm knockdown (AAV-shOSM group) significantly attenuated the therapeutic effects of GNPs-pIL-4 on muscle functional recovery at day 7, compared to the AAV-shNC group. Specifically, mice in the AAV-shOSM + HLI + GNPs-pIL-4 group exhibited reduced fatigue resistance and lower minimal contraction force retention compared to the AAV-shNC + HLI + GNPs-pIL-4 group, indicating impaired functional improvement following GNPs-pIL-4 treatment after OSM knockdown (Fig. 6A). H&E staining further revealed a marked reduction in the number of regenerating myofibers in OSM knockdown mice despite GNPs-pIL-4 therapy (AAV-shOSM versus AAV-shNC) after ischemia, suggesting the compromised muscle regeneration (Fig. 6B). Laser speckle perfusion imaging showed that the properfusion effect of GNPs-pIL-4 was substantially diminished in the OSM knockdown group (AAV-shOSM + HLI + GNPs-pIL-4 versus AAV-shNC + HLI + GNPs-pIL-4) (Fig. 6C). Tissue oxygenation assessed by ultrasound imaging was also markedly reduced by AAV-shOSM compared with the AAV-shNC group after GNPs-pIL-4 treatment under ischemia (Fig. 6D), underscoring the OSM’s role in modulating GNPs-pIL-4–induced vascular recovery.

Fig. 6. OSM knockdown attenuates GNPs-pIL-4–induced muscle regeneration and angiogenesis in ischemic tissue of mice.

AAV–short hairpin RNA was used to knockdown OSM expression in the gastrocnemius muscle of mice (shOSM or control shNC). Then, HLI was induced by femoral artery ligation, followed by immediate intramuscular injection of GNPs-pIL-4 or saline control. (A) On day 7 postsurgery, fatigue curves and minimum force retention of toe flexor muscles were measured using a muscle tension testing system (n = 4). (B) H&E staining was performed to evaluate skeletal muscle regeneration in gastrocnemius tissue. (C) Laser speckle contrast imaging was conducted pre-HLI, 30 min post-HLI, and on day 7 to assess perfusion recovery, with quantification of perfused vessels at day 7 (n = 3). (D) Ultrasonography was used to evaluate oxygen saturation in ischemic muscle on day 7. (E) Following intravenous injection of FITC-dextran, two-photon in vivo imaging was performed to visualize perfused vasculature and quantify vessel numbers (n = 3). (F) CD31 immunohistochemistry was used to assess vascular density in gastrocnemius sections (n = 3). (G) qPCR was conducted to evaluate expression levels of angiogenesis-related genes (Kdr, Fgf1, Pdgfb, and Angpt1) in the gastrocnemius muscle of the different experimental groups. Statistical analysis: One-way ANOVA followed by Bonferroni or Šídák multiple comparisons tests was used for group comparisons in (A), (C), and (E) to (G). Data were shown as means ± SD. (B) and (D) show representative images.

Two-photon intravital imaging demonstrated that the number and integrity of perfused vessels induced by GNPs-pIL-4 under ischemia were significantly decreased in the AAV-shOSM group than it in the AAV-shNC group (Fig. 6E). Similarly, CD31 immunohistochemistry revealed a substantial reduction in neovascular density by AAV-shOSM compared with it in the AAV-shNC group after GNPs-pIL-4 treatment under ischemia, indicating impaired angiogenesis (Fig. 6F). In addition, qPCR analysis showed that GNPs-pIL-4–induced up-regulation of key angiogenic genes, including kinase insert domain receptor (Kdr), fibroblast growth factor 1 (Fgf1), platelet-derived growth factor subunit B (Pdgfb), and angiopoietin 1 (Angpt1), was significantly suppressed in the OSM knockdown group (AAV-shOSM + HLI + GNPs-pIL-4 versus AAV-shNC + HLI + GNPs-pIL-4) (Fig. 6G). These data collectively demonstrated that OSM knockdown markedly impaired the regenerative and proangiogenic effects of GNPs-pIL-4 in vivo, confirming OSM as a pivotal mediator in GNPs-pIL-4–induced ischemic tissue repair.

GNPs-pIL-4 promotes perfusion and muscle recovery by down-regulating GSNOR protein expression in ischemic tissues

To explore the mechanisms by which GNPs-pIL-4 promotes endothelial sprouting and angiogenesis, we analyzed the top 10 pathways ranked by GSEA enrichment scores in the treated endothelial cells with or without GNPs-IL-4 from the single-cell RNA-seq dataset under murine HLI. Compared with the HLI group, the HLI + GNPs-pIL-4 group showed the greatest enrichment of the “Protein Nitrosylation” pathway (Fig. 7A). A volcano plot of this pathway highlighted significant down-regulation of gene Adh5 by GNPs-pIL-4 treatment (Fig. 7B). Notably, Adh5 encodes GSNOR, a key negative regulator of protein S-nitrosylation by degrading S-nitrosoglutathione (GSNO). Then, we validated Adh5 expression in both our murine single-cell dataset and a human PAD transcriptomic dataset (GSE120642) (24). Data showed that Adh5 expression was down-regulated in the HLI group and further reduced following GNPs-pIL-4 treatment (HLI + GNPs-pIL-4), compared to the Sham group (Fig. 7C). Similarly, in patients with PAD, GSNOR expression progressively declined from healthy adults to intermittent claudication (IC) and chronic limb ischemia (CLI) stages (Fig. 7D).

Fig. 7. GNPs-pIL-4 promotes perfusion and muscle recovery by down-regulating GSNOR protein expression in ischemic tissues.

(A and B) GSEA pathway enrichment analysis (A) and volcano plot (B) of differentially expressed genes (DEG) in endothelial cells from the GNPs-pIL-4–treated versus untreated HLI mice. (C) Adh5 gene expression in endothelial cells of different experimental groups. (D) Adh5 expression in skeletal muscles from the healthy adults, patients with IC, and patients with CLI. (E) GSNOR protein expression in endothelial cells treated with MφCM, GNPs-pIL-4–stimulated MφCM, or the same with anti-OSM neutralizing antibody. (F) GSNOR protein expression in gastrocnemius muscles from the HLI mice treated with NS or GNPs-pIL-4. (G to K) EOMA endothelial cells treated with or without OSM (20 ng/ml, 24 hours). (G) Adh5 mRNA expression levels. (H) Volcano plot of differentially expressed genes (RNA-seq) in the OSM-treated versus control EOMA cells. (I) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis (bubble plot) of differentially expressed genes. (J) Volcano plot of differentially expressed genes in EOMA cells. (K) Predicted STAT3 transcription factor binding sites in the Adh5 promoter region. (L) Ex vivo sprouting assay of femoral artery explants from GSNOR⁺/⁺ and GSNOR⁻/⁻ mice cultured in Matrigel (n = 5). (M) Two-photon in vivo imaging of the perfused vessels in the Sham and HLI mice (n = 3). (N) Oxygen saturation in ischemic limbs was evaluated by ultrasonography. (O) Laser speckle contrast imaging of blood flow recovery in ischemic limbs of the GSNOR⁺/⁺ and GSNOR⁻/⁻ mice (day 0 and day 7) (n = 6). (P) Muscle tension recordings (fatigue curves and residual force) of the toe flexor muscles (n = 4). Statistical analysis: [(G), (L), and (O)] Two-tailed unpaired Student t test. [(E), (F), (M), and (P)] One-way ANOVA followed by Bonferroni or Šídák multiple comparisons test. Data were shown as means ± SD. HIF-1, hypoxia-inducible factor 1; TH17, T helper cell 17; ECM, extracellular matrix.

To test whether GNPs-pIL-4–stimulated macrophages regulate endothelial GSNOR expression, we treated endothelial cells with MφCM activated with or without GNPs-pIL-4. Western blot analysis showed that MφCM activated with GNPs-pIL-4 (GNPs-pIL-4 + MφCM) significantly reduced GSNOR protein levels in endothelial cells compared with MφCM without the GNPs-pIL-4 treatment group (MφCM), and this effect was abolished by anti-OSM neutralizing antibody (GNPs-pIL-4 + MφCM + AntiOSM versus GNPs-pIL-4 + MφCM) (Fig. 7E), indicating that macrophage-derived OSM mediates the suppression of endothelial GSNOR by GNPs-pIL-4. Consistently, GSNOR expression was also markedly decreased in ischemic muscle tissues of the HLI + GNPs-pIL-4 group, compared to the HLI group (Fig. 7F).

To obtain direct evidence that OSM is sufficient to down-regulate GSNOR in endothelial cells, we firstly stimulated mouse endothelial cells (EOMA) directly with recombinant OSM protein for 24 hours and examined the mRNA expression level of Adh5 by reverse transcription qPCR. OSM treatment significantly reduced the Adh5 mRNA expression (Fig. 7G). Subsequently, bulk RNA-seq of these samples confirmed that OSM treatment markedly down-regulated Adh5 expression at the transcriptomic level as well (Fig. 7H). In terms of the potential downstream signaling mechanisms, we analyzed the transcriptomic profile of OSM-stimulated endothelial cells. Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis revealed that the differentially expressed genes were highly significantly enriched in the “JAK-STAT signaling pathway” (Fig. 7I), indicating activation of this canonical axis by OSM in endothelial cells. A volcano plot of the differentially expressed genes (Fig. 7J) confirmed extensive transcriptomic remodeling, with significant induction of the OSM receptor gene Osmr itself. Furthermore, bioinformatic analysis predicted potential STAT3 (signal transducer and activator of transcription 3) transcription factor binding sites within the Adh5 promoter region (Fig. 7K), suggesting that STAT3, a key downstream effector of the OSM-OSMR pathway, may directly bind to and repress Adh5 transcription.

We further assessed the role of GSNOR in sprouting angiogenesis in Gsnor^+/+ and Gsnor^−/− mice. GSNOR knockout significantly enhanced sprouting angiogenesis, with a greater sprouting area compared to Gsnor^+/+ (Fig. 7L). In vivo two-photon imaging showed increased numbers of perfused vessels in Gsnor^−/− mice compared with Gsnor^+/+ either in the Sham group or in the HLI group (Fig. 7M). Similarly, both ultrasound-photoacoustic oxygen saturation imaging demonstrated improved tissue oxygenation and perfusion in GSNOR-deficient mice under both Sham and HLI conditions. Quantitative analysis of oxygen saturation revealed comparable levels between GSNOR^−/− and wild-type (WT) mice in the Sham groups, whereas GSNOR^−/− mice exhibited significantly better recovery following HLI (Fig. 7N). Consistently, laser speckle imaging showed that perfusion recovery was markedly enhanced in GSNOR^−/− compared with GSNOR^+/+ mice under HLI conditions, while no significant difference was observed in the Sham groups (Fig. 7O). In the Sham group, GSNOR^−/− alone did not affect the fatigue resistance and minimal contraction force retention under sustained stimulation. However, muscle function tests demonstrated improved fatigue resistance and increased minimal contraction force retention in Gsnor^−/− mice compared with Gsnor^+/+ mice under sustained stimulation in HLI groups, suggesting enhanced regenerative capacity by GSNOR knockout (Fig. 7P). Therefore, GNPs-pIL-4 promoted angiogenesis, blood flow recovery, and functional tissue regeneration by down-regulating GSNOR expression.

Endothelial GSNOR deficiency enhances S-nitrosylation of ENG and promotes sprouting angiogenesis

To investigate how GSNOR regulates protein S-nitrosylation under ischemic conditions, we used the iodo tandem mass tag (iodoTMT)–switch assay to measure global S-nitrosylated proteins in the gastrocnemius muscle of WT and Gsnor^−/− mice under Sham and HLI conditions, with or without GNPs-pIL-4 treatment. The data showed that S-nitrosylation levels were significantly elevated in Gsnor^−/− mice under both Sham and HLI conditions (Gsnor KO + Sham versus WT + Sham; Gsnor KO + HLI versus WT + HLI), with a further increase under ischemia (Gsnor KO + HLI versus Gsnor KO + Sham). In WT mice, GNPs-pIL-4 treatment also increased protein S-nitrosylation in ischemic muscles (WT + HLI + GNPs-pIL-4 versus WT + HLI) (Fig. 8A, quantified data in fig. S7A).

Fig. 8. Endothelial GSNOR deficiency enhances S-nitrosylation of ENG and promotes sprouting angiogenesis.

(A) S-nitrosylated proteins (SNO-proteins) in gastrocnemius tissues of GSNOR^+/+ and GSNOR^−/− mice under sham or HLI conditions were detected using the iodoTMT-switch assay. (B) Endothelial cells were treated with MφCM derived from GNPs-pIL-4–stimulated macrophages and MφCM treated with or without recombinant OSM, and levels of S-nitrosylated proteins were assessed by Western blot; GAPDH served as the loading control. (C) Venn diagram illustrating the overlap among the endothelial nitrosoproteome and angiogenesis-related genes in both humans and mice. (D) Validation of the S-nitrosylation of key proteins (ENG and CDC42) in endothelial cells treated by MφCM derived from GNPs-pIL-4–stimulated macrophages and MφCM treated with or without recombinant OSM. (E) Homology analysis of the S-nitrosylation site C209 of ENG in humans and mice and the schematic diagram of the C209A point mutant. (F) After knockdown of endogenous ENG via siRNA, endothelial cells were transfected with WT or C209A-mutated ENG constructs, respectively, and subsequently ENG protein expression was evaluated by Western blot and quantified (n = 3). Scratch wound assay (G) and tube formation assay (H) were used to evaluate the effects of various treatments, including control vector, OSM stimulation, ENG knockdown + OSM, ENG knockdown with WT ENG rescue + OSM, and ENG knockdown with C209A-mutant ENG rescue + OSM, on the migration and angiogenic capacity of EOMA endothelial cells (n = 3). (I) Ex vivo aortic sprouting assay was performed to assess the effect of GNPs-pIL-4 + MφCM, with or without AAV-shENG or AAV-shNC, on sprouting angiogenesis; representative images and quantification of sprouting area are shown (n = 3). Statistical analysis: One-way ANOVA followed by Šídák multiple comparisons test was used for statistical analysis in (F), (H), (I), and (J). Data were shown as means ± SD. OE, overexpression.

To assess whether macrophage-derived OSM contributes to the regulation of S-nitrosylation in endothelial cells, we treated endothelial cells with normal media, MφCM, MφCM derived from GNPs-pIL-4–stimulated macrophages and MφCM treated with recombinant OSM. Western blot data showed that both GNPs-pIL-4–derived MφCM and MφCM treated with recombinant OSM markedly increased total protein S-nitrosylation levels in endothelial cells (Fig. 8B, quantified data in fig. S7B). Then, we referred to previously published endothelial SNO-proteome datasets and cross-referenced them with angiogenesis-related genes from both human and mouse databases (25). Among the candidate targets, ENG and cell division cycle 42 (CDC42) emerged as functionally relevant proteins (Fig. 8C). Further validation showed that GNPs-pIL-4–derived MφCM and MφCM treated with recombinant OSM significantly increased the S-nitrosylation level of ENG but had no effect on CDC42 S-nitrosylation [Fig. 8D, quantified data in fig. S7 (C and D)].

According to the analysis of SNO-proteome datasets, the cysteine at position C209 of ENG is a potential site of S-nitrosylation. Sequence conservation analysis revealed that cysteine 209 is a conserved putative S-nitrosylation site on ENG (Fig. 8E). On the basis of this, we generated a C209A mutant to assess its functional importance. To determine the impact of ENG S-nitrosylation on vascular behavior, we silenced endogenous ENG using small interfering RNA (siRNA) and then reconstituted these cells with either WT or C209A mutant ENG. Western blot confirmed the comparable expression of both constructs (Fig. 8F). Recombinant OSM significantly promotes endothelial cell migration and tube formation (vector + OSM versus vector) (Fig. 8, G and H). Knockdown of ENG in endothelial cells markedly suppressed the promigratory and proangiogenic effects of OSM (ENG-KD + OSM versus vector + OSM) (Fig. 8, G and H). Furthermore, reexpression of WT ENG in ENG-deficient cells rescued the promotive effects of OSM, whereas expression of the ENG-C209A mutant failed to rescue these effects (ENG-KD → ENG-rescue + OSM versus ENG-KD → ENG-C209A-mutant-rescue + OSM) (Fig. 8, G and H, and movie S3).

The ex vivo femoral artery explant sprouting assay demonstrated that GNPs-pIL-4–stimulated MφCM promoted microvessel sprouting from arterial explants (GNPs-pIL-4 + MφCM versus MφCM). AAV-shNC had no effect on explant growth or sprouting capacity (AAV-shNC + GNPs-pIL-4 + MφCM versus GNPs-pIL-4 + MφCM). Knockdown of ENG using AAV-shENG significantly attenuated the prosprouting effect of GNPs-pIL-4 + MφCM (AAV-shENG + GNPs-pIL-4 + MφCM versus AAV-shNC + GNPs-pIL-4 + MφCM) (Fig. 8I).

Therefore, GSNOR deficiency enhanced ENG S-nitrosylation at cysteine 209, thereby activating endothelial sprouting angiogenesis. In addition, an OSM-GSNOR-snoENG signaling axis regulated sprouting angiogenesis in ischemic tissues.

GNPs-pIL-4 reprograms macrophage polarization and activates the OSM/GSNOR/ENG axis to improve angiogenesis and tissue repair in ischemic limbs

In HLI, macrophages polarized toward the proinflammatory M1 phenotype, leading to increased inflammation, impaired sprouting angiogenesis, reduced perfusion, and compromised muscle function. GNPs-pIL-4 therapy delivered IL-4 to promote M2 macrophage polarization, which enhanced the secretion of OSM. OSM, in turn, activated endothelial cells by down-regulating GSNOR and increasing S-nitrosylation of ENG, thereby promoting angiogenesis, improving blood perfusion, enhancing muscle regeneration, and facilitating tissue repair (Fig. 9).

Fig. 9. GNPs-pIL-4 reprograms macrophage polarization and activates the OSM/GSNOR/ENG axis to improve angiogenesis and tissue repair in ischemic limbs.

In HLI, macrophages polarized toward the proinflammatory M1 phenotype, leading to increased inflammation, impaired angiogenesis, reduced perfusion, and compromised muscle function. GNPs-pIL-4 therapy delivered IL-4 to promote M2a-like macrophage polarization, which enhanced the secretion of OSM. OSM, in turn, activated endothelial cells by down-regulating GSNOR and increasing S-nitrosylation of ENG, thereby promoting angiogenesis, improving blood perfusion, enhancing muscle regeneration, and facilitating tissue repair.

DISCUSSION

PAD is a prevalent yet often underrecognized chronic vascular condition that not only markedly impairs lower limb perfusion but also increases the risk of myocardial infarction and stroke. Because of the low rate of early diagnosis, the 5-year mortality rate among patients with severe limb ischemia exceeds 50%, and many ultimately face the risk of amputation (1). Although current standard treatments—such as antiplatelet therapy, stent implantation, and bypass surgery—can partially improve perfusion, they are often insufficient in promoting neovascularization (2). In this study, we developed an IL-4 gene delivery system based on three-dimensional GNPs. This system promoted therapeutic angiogenesis through mechanisms: GNPs-pIL-4 induces macrophage polarization toward the M2 phenotype and activates the OSM signaling pathway to down-regulate GSNOR expression in endothelial cells, thereby enhances S-nitrosylation of the protein ENG, accelerates vascular remodeling, increases the number of perfused vessels and tissue oxygenation, and ultimately restores the ischemic muscle function. Animal studies demonstrated that this strategy has both potent therapeutic efficacy and good biocompatibility.

The postischemic tissue microenvironment is a dynamic system composed of various cell types, cytokines, and signaling pathways. In the early stages of acute ischemia, local tissue undergoes acidosis due to hypoxia and metabolic dysregulation, triggering inflammatory responses (26). Immunocytes infiltrate the damaged site and release proinflammatory cytokines such as TNF-α and IL-1β, which exacerbate tissue injury. Concurrent oxidative stress induces cell apoptosis, further expanding the damage. As ischemia progresses, hypoxia-inducible factor 1α is activated, up-regulating angiogenic genes such as VEGF and initiating endogenous angiogenesis and fibrosis (27). Macrophages play a dual role in the PAD microenvironment. In the early phase, M1-type macrophages dominate and release inflammatory cytokines (e.g., TNF-α and IL-6), exacerbating tissue injury (10). In the later stages, they transition to an M2 phenotype, secreting factors such as IL-10 and TGF-β to promote tissue repair and angiogenesis (11, 26). Imbalance between the M1 and M2 macrophages often contributes to the progression of PAD. Targeting the ischemic microenvironment, we locally administered GNPs-pIL-4 in a mouse HLI model, successfully inducing M2 macrophage polarization. This led to significant up-regulation of anti-inflammatory cytokines including IL-4, IL-10, and CCL2, alleviating the proinflammatory M1-dominant state and fostering a local environment favorable for neovascularization and tissue repair. These effects were significantly diminished upon macrophage depletion, confirming their pivotal role in postischemic repair. Consistent with our findings, other studies have shown that miR-93 enhances M2 polarization by down-regulating interferon regulatory factor-9, thereby improving perfusion and angiogenesis in PAD models (28). In diabetic mice, reconstituted high-density lipoprotein nanoparticle therapy has been reported to suppress chronic inflammation and promote M2 polarization, improving collateral remodeling (29). Conversely, dynamin-related protein 1 deficiency or VEGF165b overexpression promotes M1 activation and impairs vascular regeneration. Together, these findings support macrophage phenotype modulation as a promising therapeutic strategy for PAD.

There is a continuum and diversity of macrophage functional states in vivo. In addition, there are various M2 macrophage subtypes. Our gene expression analysis of macrophage markers provided greater precision in defining the phenotype induced by GNPs-pIL-4 and clarified its subtype specification. Compared to the HLI + NS group, we observed a significant and selective up-regulation of classic M2a-associated markers (Arg1 and Mrc1) but not markers characteristic of other M2 subtypes such as M2b (e.g., IL-1β and IL-6) or M2c (e.g., TGF-β and MerTK) (Fig. 2D). This indicated that our delivery strategy may specifically and effectively drive macrophage polarization toward an M2a-like phenotype. It was consistent with the reported role of IL-4 signaling as the primary driver for M2a activation.

Previous studies have shown that IL-4 plasmid delivery targeting cardiac macrophages can attenuate inflammation following myocardial infarction and improve cardiac remodeling and function (19). These findings align with our data, suggesting that IL-4 gene therapy has broad applicability across ischemic conditions and could be further optimized for clinical use. Now, the clinical extensive application of IL-4 is limited by its short half-life, poor stability, and potential systemic side effects. In the present study, we constructed a stable nanocarrier system with high binding force for relatively targeted and sustained delivery of IL-4 plasmids to macrophages, with few cytotoxicity. The GNPs we designed, as a nonviral gene delivery system, exhibited several intrinsic advantages. Unlike conventional two-dimensional graphene sheets, GNPs had a three-dimensional isotropic architecture, which provided an exceptionally high surface area for efficient loading of various nucleic acid molecules. Their unique isotropic structure and sharp edges facilitated cellular uptake, either through endocytosis or direct membrane penetration. In addition, GNPs were highly amenable to surface functionalization: They could be modified with cationic polymers such as PEI to enhance gene condensation and promote endosomal escape or conjugated with targeting ligands such as FA to improve cell-specific delivery. These features collectively improved intracellular delivery efficiency while conferring lower immunogenicity compared to viral vectors. These material properties endowed GNPs with the unique capability to locally modulate immune cells and promote tissue regeneration. Furthermore, GNPs can be integrated with enzyme-responsive hydrogels or pH- and redox-sensitive linkers to enable stimulus-responsive gene release within pathological microenvironments. These features highlight the translational potential of GNPs in ischemia-related regenerative medicine.

Beyond PAD, the pathological mechanisms and intervention strategy involved in this study are also highly relevant to other ischemia- and inflammation-related conditions. Chronic wounds represent a major unresolved clinical challenge and are commonly observed in diabetic foot ulcers, venous leg ulcers, arterial occlusive diseases, and in elderly patients (30). The underlying causes are multifactorial, including impaired local blood circulation, persistent infection, retained foreign material, metabolic decline, and systemic comorbidities. Among these, ischemia and chronic inflammation are recognized as central barriers to wound healing (31). It has been well documented that a clean wound with sufficient blood supply generally heals, whereas inadequate perfusion often leads to nonhealing chronic wounds (32). Thus, strategies that promote angiogenesis and restore effective perfusion in ischemic tissues hold particular promise for chronic wound management. In this context, our GNP-based nonviral gene delivery platform not only enhanced angiogenesis and tissue perfusion but also improved the inflammatory microenvironment via macrophage polarization, thereby providing a potential therapeutic strategy for chronic wounds. Future studies are warranted to validate its efficacy in models of diabetic foot ulcers or other chronic wounds, which may further facilitate clinical translation.

Furthermore, our single-cell transcriptomic analysis revealed that GNPs-pIL-4 treatment significantly enhanced the number and strength of intercellular communications in ischemic tissues, with enrichment of pathways associated with endothelial sprouting and neovascularization. Two-photon intravital imaging further confirmed a substantial increase in perfused vessels and improved vascular continuity by GNPs-pIL-4 treatment, supporting our hypothesis that through the new constructed nanomaterial, microenvironmental modulation promotes the formation of mature and functional vasculature after ischemia.

Conventional proangiogenic therapies often focus on single growth factors such as VEGF, FGF, or HGF. While these approaches have shown promise in preclinical studies and early-phase clinical trials, large-scale clinical trials have frequently failed to demonstrate significant therapeutic benefit. For example, the RAVE study involving 105 patients reported no improvement in peak walking time or ankle-brachial index after a single intramuscular injection of adenoviral VEGF121, and high-dose groups even exhibited peripheral edema (33). Similarly, the TAMARIS trial involving 525 patients showed no significant benefit in amputation or mortality rates following FGF-1 plasmid therapy (34). These outcomes highlight the limitations of single-factor strategies. In contrast, our immune microenvironment–modulating approach may offer a more robust biological rationale and improved translational potential.

In terms of mechanisms, we have reported that OSM mediated the regenerative pathway in ischemia (35). Consistently, we found in the present study that the OSM signaling pathway was activated during the ischemic recovery process and further enhanced following GNPs-pIL-4 administration. Gene expression profiling indicated that OSM is specifically expressed in macrophages, while its receptors in endothelial cells are widely up-regulated by the GNPs-pIL-4 treatment. Data from in vitro experiments showed that neutralizing OSM in conditional media derived from GNPs-pIL-4–treated macrophages markedly reduced its proangiogenic effects on endothelial cells. Similarly, in vivo knockdown of OSM in ischemic tissues significantly attenuated the therapeutic benefits of GNPs-pIL-4. These findings collectively indicated that OSM is a critical effector molecule through which M2 macrophages promote vascular regeneration and blood flow recovery. This study extended our understanding about the OSM’s effect on endothelial cell function, including cell migration, sprouting, and tube formation. Given the broad expression of OSM receptors in various cell types, OSM appeared to serve as a central mediator of cross-cellular coordination in the ischemic microenvironment. Previous work by our group has shown that OSM can induce a proangiogenic phenotype in fibroblasts, contributing to vascular regeneration, consistent with observations in this study. However, some other studies reported the effects of OSM in disease models associated with pathological angiogenesis. For example, while OSM promoted endothelial sprouting in vitro, it reduced neovascularization and vascular occlusion in an oxygen-induced retinopathy model in mice. These findings suggested that the function of OSM is highly dependent on the pathological context and the specific cellular targets involved.

Moreover, NO, a well-established regulator of vascular function, plays critical roles in vasodilation, endothelial cell proliferation, and migration (36). As a key mediator of S-nitrosylation, NO modifies protein function via covalent attachment of NO to cysteine residues. GSNOR acts as a negative regulator by removing SNO groups from proteins. Previous studies have shown that GSNOR deficiency enhances skeletal muscle contractility and fatigue resistance through RyR1 S-nitrosylation (37). In the present study, GNPs-pIL-4 treatment down-regulated GSNOR expression in ischemic tissues, which led to a significant increase in the S-nitrosylation of the endothelial protein ENG. This posttranslational modification promoted angiogenesis and improved perfusion. In GSNOR-deficient mice, vascular regeneration and functional recovery after ischemia were markedly enhanced. However, in nonischemic conditions, GSNOR deletion did not significantly affect vascular density or perfusion, indicating that GSNOR’s regulatory role is particularly critical during stress-induced vascular remodeling. Our study provided evidence about the endothelial protein ENG being a previously unidentified regulatory target of S-nitrosylation. However, GSNOR knockout has also been reported to impair angiogenic capacity in mesenchymal stem cells, highlighting the potential cell type–specific regulatory effects of GSNOR.

Mechanistic experiments revealed that GNPs-pIL-4–conditioned macrophage media suppressed GSNOR expression in endothelial cells, an effect that was reversed upon OSM neutralization. These data demonstrated that OSM served as a crucial mediator linking M2 macrophages to endothelial GSNOR regulation. By down-regulating GSNOR, OSM enhanced ENG S-nitrosylation and thereby promoted angiogenesis. This provided previously unidentified insights into the role of GSNOR in vascular regeneration and revealed a previously unidentified regulatory axis involving macrophage-derived OSM, GSNOR suppression, and ENG modification as a central mechanism in ischemic tissue repair.

Despite the promising therapeutic efficacy and biocompatibility, several challenges remained in translating GNPs-pIL-4 therapy into clinical practice. First, the femoral artery ligation model used in this study simulated the acute HLI and unlikely to fully recapitulate the chronic, low-perfusion, inflammatory microenvironment observed in patients with PAD. In addition, mice have robust collateral circulation, enabling rapid compensatory perfusion that may not fully reflect human pathophysiology, thereby limiting the translational relevance. Furthermore, common comorbidities such as diabetes and hyperlipidemia were not incorporated in this model, restricting the generalizability of the findings. Therefore, future studies should validate the safety and therapeutic stability of GNPs-pIL-4 in more complex, chronic ischemia models that better mimic the multifactorial nature of human PAD.

In summary, this study presents a IL-4 gene delivery platform based on GNPs capable of modulating macrophage polarization, remodeling the ischemic microenvironment, and markedly improving HLI. In addition, we identified the OSM-GSNOR-ENG axis as a central mechanism promoting angiogenesis and tissue recovery in the process, offering previously unidentified mechanistic insights, therapeutic targets, and translational potential for PAD treatment.

MATERIALS AND METHODS

Experimental design

This study aimed to investigate whether macrophage-targeted GNPs-pIL-4 could promote angiogenesis and tissue repair after ischemic injury. GNPs were synthesized and functionalized with PEI and FA to enable plasmid loading and macrophage targeting, followed by comprehensive characterization of their morphology, particle size, zeta potential, nucleic acid binding capacity, and biosafety in vitro. A murine HLI model was then established by femoral artery ligation in C57BL/6J mice, which were randomly assigned to receive intramuscular injections of NS, GNPs, pIL-4, or GNPs-pIL-4. The therapeutic efficacy was evaluated by measuring IL-4 expression, macrophage polarization, blood perfusion, oxygenation, vascular density, and muscle regeneration through molecular, histological, and functional analyses. To elucidate the underlying mechanism, macrophage depletion, OSM knockdown (AAV-shOSM), and GSNOR knockout mice were used to verify whether the therapeutic effects of GNPs-pIL-4 depended on macrophage-mediated activation of the OSM-GSNOR-ENG axis. In addition, single-cell RNA-seq, endothelial migration and tube formation assays, and protein S-nitrosylation analysis were conducted to further uncover the intercellular signaling and molecular mechanisms by which GNPs-pIL-4 enhances angiogenesis and tissue regeneration.

Animals and treatment

All animal protocols were approved by the Animal Care and Use Committee of Peking University (permit: LA2021071). C57BL/6J mice were purchased from the Department of Laboratory Animal Science, Peking University Health Science Center. The mice were housed in a specific pathogen–free animal facility under controlled conditions of 20° to 26°C temperature and 40 to 70% relative humidity, with a 12-hour light/12-hour dark cycle. Direct exposure to strong light and noise disturbances was avoided.

Cell culture

RAW264.7 mouse macrophages, EOMA mouse endothelial cells, HUVEC human umbilical vein endothelial cells, and NIH3T3 mouse fibroblasts were used in this study. All cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37°C in a humidified atmosphere containing 5% CO₂.

For human macrophage differentiation, PBMCs were isolated from human peripheral blood and cultured in RPMI 1640 medium containing 10% FBS. Macrophage differentiation was induced by incubation with macrophage colony-stimulating factor (100 ng/ml) for 7 days.

MφCM preparation

To obtain MφCM, RAW264.7 macrophages were seeded at a density of 1 × 10⁶ cells per well in six-well plates and cultured overnight. Cells were then treated with GNPs-pIL-4 or control vehicle for 48 hours in serum-free DMEM. After incubation, the culture supernatant was collected, centrifuged at 1000g for 10 min at 4°C to remove cell debris, and filtered through a 0.22-μm filter. The cleared supernatant was used as MφCM for subsequent endothelial cell stimulation, femoral artery ring sprouting assays, and Western blot analysis. For OSM neutralization experiments, anti-OSM neutralizing antibody was added to MφCM at a working concentration 1 hour before endothelial cell stimulation.

RNA interference and cell transfection

For ENG knockdown, EOMA cells were transfected with siRNA targeting endogenous ENG using a commercial liposome transfection reagent according to the manufacturer’s instructions. A scramble sequence was used as a negative control. At 48 hours after transfection, cells were collected for Western blot validation and functional assays. For rescue experiments, cells were transfected with plasmid encoding WT ENG or ENG-C209A mutant, and stable expression was confirmed by Western blot.

Preparation of GNPs

The 3DGNPs serve as the precursor of the GNPs used as delivery carriers. The synthesis method of 3DGNP has been described in detail in our previous work (38). Briefly, 3DGNP was synthesized using a catalyst-free plasma-enhanced chemical vapor deposition technique, with methane (CH₄) as the carbon source and a mixed argon/hydrogen (Ar/H₂) gas as the etching atmosphere. The CH₄ flow rate was set to 10 standard cubic centermeters per minute (sccm), and the gas ratio of CH₄, Ar, and H₂ was maintained at 1:0.5:4. A graphite disk of ∼3 mm thick was used as the substrate for carbon deposition and was additionally connected to a radio frequency power source to enhance ion energy. The deposition process was conducted at 800°C, with the temperature controlled and maintained by a heating plate. After synthesis, the system was allowed to cool naturally under an inert gas atmosphere. The resulting 3DGNPs were mechanically scraped from the substrate, ground using a sand mill, and filtered through a 200-nm pore-sized membrane to obtain nanoscale three-dimensional graphene.

GNPs were synthesized through a modified Hummers’ method. Briefly, 2 g of graphene powder (particle size, ∼30 μm) and 1 g of sodium nitrate were added into a 250-ml three-necked flask and precooled to 0°C. Under vigorous stirring, 50 ml of concentrated sulfuric acid was added dropwise while maintaining the temperature below 5°C for 30 min. Then, 0.3 g of potassium permanganate was slowly introduced, followed by stirring for another 30 min at a temperature not exceeding 10°C. Subsequently, 7 g of potassium permanganate was divided into three portions and added over the course of 1 hour, with the reaction temperature maintained below 20°C throughout this low-temperature oxidation stage.

After removing the ice bath, the reaction mixture was gradually heated to 35° ± 3°C in a water bath and stirred for 2 hours until a brown suspension was obtained, representing the intermediate-temperature reaction phase. To initiate the high-temperature reaction, 90 ml of deionized water was slowly added, which caused an immediate rise in temperature to ∼90°C accompanied by vigorous gas evolution. The diluted suspension was kept at this temperature for 15 min.

Next, a mixture of 7 ml of hydrogen peroxide (30%) and 55 ml of ultrapure water (preheated to 45°C) was added, resulting in a bright yellow dispersion of oxidized graphene. The resulting mixture was immediately filtered while warm to obtain a yellow-brown filter cake. This was washed three times with 150 ml of 3% hydrochloric acid (at 45°C) and then redispersed in 600 ml of water. The suspension was centrifuged at 4000 rpm for 20 min at room temperature to collect a gel-like oxidized graphene product, which was then dried in a vacuum oven at 40°C for 24 hours.

The dried oxidized graphene was redispersed in ultrapure water to yield a brown suspension, followed by dialysis to remove residual acids and salts. The purified product was diluted to a series of concentrations ranging from 0.01 to 1 mg/ml. The suspensions were subjected to ultrasonication at 200 W (80% amplitude) for 30 min to exfoliate the graphene oxide layers. The resulting dispersions were centrifuged at 3000 rpm to remove unexfoliated material.

A 10-ml aliquot of the graphene dispersion (0.1 mg/ml) was mixed with 20 mg of EDC and 10 mg of NHS, followed by sonication at 200 W (40%) for 10 min to activate carboxyl groups. Subsequently, 10 ml of PEI solution (1 mg/ml) was added dropwise under stirring. After sonication for 30 min, the mixture was stirred at room temperature for 24 hours. The product was centrifuged at 12,000 rpm for 15 min, the supernatant was discarded, and the pellet was redispersed in 10 ml of ultrapure water followed by 10 min of sonication.

Last, 10 ml of FA-PEG (folic acid–polyethylene glycol) solution (1 mg/ml), preactivated with EDC/NHS, was slowly introduced. The mixture was sonicated for 30 min (200 W, 40%) and stirred at room temperature for another 24 hours. After centrifugation at 12,000 rpm for 15 min, the supernatant was discarded, and the pellet was redispersed in 10 ml of ultrapure water and sonicated again for 10 min. The resulting dispersion contained the final GNPs.

Particle size analysis of GNPs

A total of 100 μl of the diluted GNP suspension was carefully transferred into a clean, bubble-free cuvette for dynamic light scattering measurement. The outer surface of the cuvette was wiped clean before placement into the Zetasizer Nano ZSP particle size and zeta potential analyzer (Malvern Instruments, UK). Measurements were performed using the Zetasizer software under the “Size Measurement” mode, with the temperature set at 25°C and an equilibration time of 2 min. Each sample was measured in triplicate, and the average value was reported.

Zeta potential measurement of GNPs

Zeta potential was assessed using a folded capillary electrophoresis cell (DTS1070). A 1-ml aliquot of the GNP suspension was slowly injected into the cell to avoid air bubble formation. The measurement was conducted under the “Zeta Potential Measurement” mode, with a temperature setting of 25°C and ultrapure water as the dispersing medium. Each sample was measured three times to ensure reproducibility.

FITC labeling and cellular uptake assay of GNPs

FITC powder was first dissolved in dimethyl sulfoxide to prepare a 10 mM stock solution. To perform fluorescent labeling, 100 μl of FITC solution was added to 900 μl of GNP suspension (0.1 mg/ml). The mixture was incubated in the dark for 24 hours, followed by centrifugation at 12,000g. The resulting GNPs-FITC pellet was resuspended in 1 ml of ultrapure water and sonicated at 200 W (40% amplitude) for 10 min to ensure homogeneous dispersion.

For the cellular uptake assay, RAW264.7, HUVEC, and NIH3T3 cells were seeded into 24-well plates and cultured for 24 hours. After medium replacement, each well received 2 μl of the GNPs-FITC suspension. The cells were incubated in the dark for 6 hours, followed by two phosphate-buffered saline (PBS) washes to remove uninternalized nanoparticles. The uptake of GNPs-FITC by each cell type was visualized and analyzed using a multifunctional cell imaging system (BioTek Cytation 7).

HLI mouse model and GNPs-pIL-4 treatment

To prepare the GNPs-pIL-4 complex, the amount of plasmid DNA was calculated on the basis of a dosage of 10 μg of pIL-4 plasmid per mouse. For a single mouse, 10 μg of pIL-4 plasmid was dissolved in 30 μl of PBS, gently mixed, and incubated at room temperature for 5 min. Then, 20 μl of GNPs was slowly added dropwise into the plasmid solution, and the mixture was gently pipetted to ensure homogeneity, avoiding vortexing or vigorous agitation. The mixture was incubated at room temperature for 15 to 20 min to allow the formation of stable GNPs-pIL-4 complexes.

Male C57BL/6J mice aged 7 weeks were used for modeling and were acclimated for 7 days before surgery. Mice were fasted for 6 hours before surgery with free access to water. On the day of surgery, mice were weighed and anesthetized by intraperitoneal injection of a 1.25% tribromoethanol (Avertin) working solution at a dose of 20 ml/kg (equivalent to 250 mg/kg). The depth of anesthesia was assessed by the absence of pedal withdrawal reflex. Body temperature was maintained at 37°C using a heating pad throughout the procedure. Ophthalmic ointment was applied to prevent corneal dryness. After confirming adequate anesthesia, hair on both hindlimbs was removed using depilatory cream, and the surgical area was disinfected with povidone iodine. A longitudinal incision (∼1 cm) was made along the inguinal ligament to expose the femoral artery by carefully separating the subcutaneous tissue. The proximal and distal ends of the femoral artery were doubly ligated using surgical sutures with an approximate spacing of 0.5 cm between the ligatures. The artery was then transected between the two ligation points to ensure complete interruption of blood flow and to eliminate collateral circulation. The incision was closed with sutures, followed by a second round of povidone iodine disinfection. In the sham group, mice underwent the same surgical exposure of the femoral artery without ligation. The artery was left exposed for ∼5 min before skin closure to complete the sham procedure.

The gastrocnemius muscle of the mouse hindlimb was palpated to avoid major blood vessels and nerves. Using an insulin syringe, 50 μl of GNPs-pIL-4 complex was drawn and injected perpendicularly into the mid-belly of the muscle. The injection was performed slowly, and the injection site was gently pressed for 1 min postinjection to prevent leakage. For control groups, the following treatments were administered: Sham group: 50 μl of sterile saline; GNP group: 50 μl of GNP suspension (composed of 20 μl of GNPs and 30 μl of PBS); pIL-4 group: 50 μl of plasmid solution containing 10 μg of pIL-4 plasmid DNA.

Hindlimb perfusion assessment in mice

Laser speckle imaging was performed using a full-field laser perfusion imager (FLPI2, Moor Instruments, Devon, UK). The imaging probe was adjusted to a height of 15 to 20 cm above the mouse fixation platform. In the preview interface, the imaging focus was adjusted until a clear image was obtained. Regions of interest were defined over both hindlimbs, and the ischemic (surgical) and nonischemic (contralateral) limbs were labeled accordingly. In the settings interface, the acquisition mode was set to high-resolution with a capture duration of 10 s per frame. Imaging was initiated, and recordings were saved.

Perfusion intensity was quantified using the analysis interface, and the relative perfusion ratio was calculated as the ratio of perfusion in the ischemic limb to that in the contralateral limb (ischemic/contralateral). Avertin anesthesia was used for postsurgical measurements, whereas isoflurane was used for nonsurgical time points.

Ultrasound-photoacoustic oxygen saturation imaging of mouse hindlimbs

Mice were first anesthetized in an isoflurane induction chamber using 2% isoflurane with oxygen flow (1 liter/min). Once the righting reflex disappeared, anesthesia was maintained with 1% isoflurane via a nose cone. Hindlimb hair was gently removed using a depilatory cream, and the skin was cleaned thoroughly. The mice were then positioned on the temperature-controlled stage of a multimodal ultrasound–photoacoustic imaging system (IVIS Spectrum, VisualSonics, Toronto, Canada) with limbs extended and aligned horizontally to avoid occlusion. An acoustic coupling gel was applied to the hindlimbs to form a ∼1.5-cm-thick layer over the imaging region. The ultrasound probe was positioned at the ankle joint, and its height was adjusted to center the hindlimb within the imaging frame. The imaging chamber door was then closed.

A three-dimensional scanning mode was selected with a step size of 2 mm. Using the joystick, the probe was advanced to the knee joint, which was set as the scan endpoint, and image acquisition was initiated. After data collection, the average oxygen saturation of the entire hindlimb was quantified, and a representative image was obtained by selecting a transverse section with an area of 30 mm² at the mid-hindlimb level.

Perfused vessel imaging of the mouse gastrocnemius muscle

Mice were first anesthetized in an isoflurane induction chamber with 2% isoflurane delivered in an oxygen flow of 1 liter/min. Once the righting reflex disappeared, mice were transferred to a temperature-controlled surgical platform, and anesthesia was maintained with 1% isoflurane via a nose cone. Hair was shaved from both hindlimbs, and the skin was depilated and cleaned before being disinfected with povidone iodine. Mice were positioned in the supine position with the right hindlimb extended naturally. A 1-cm longitudinal skin incision was made in the inguinal region to expose the femoral vein by blunt dissection of the subcutaneous tissue. An insulin syringe was used to puncture the femoral vein, and FITC-dextran (Sigma-Aldrich, USA) was injected slowly at a dose of 80 mg/kg. The incision was loosely closed following injection.

The mouse was then repositioned in the right lateral recumbent position, with the left hindlimb elevated and extended horizontally to expose the gastrocnemius muscle. The skin overlying the gastrocnemius was incised, and physiological saline was applied to maintain tissue hydration. The exposed area was positioned under a water-immersion objective. The excitation wavelength was set to 800 nm, and the detection channel was adjusted to 500 to 550 nm to visualize the superficial vasculature in real time. Z-stack imaging was performed with a step size of 1 μm and a total depth of 100 μm. Following image acquisition, three-dimensional reconstruction was conducted. The incision was then sutured, and the mouse was placed on a 37°C heating pad for recovery.

Paraffin embedding and sectioning of mouse gastrocnemius muscle

Mice were euthanized by cervical dislocation, and the skin of the hindlimb was incised using surgical scissors to expose the gastrocnemius muscle. The muscle was gently separated using forceps, carefully excised in its entirety, and briefly rinsed in precooled PBS to remove residual blood and debris. The tissue was then transferred into 4% paraformaldehyde and fixed at room temperature for 24 hours. After fixation, samples were placed in tissue processing cassettes and dehydrated using a vacuum tissue processor (VIP-5-Jr-J2, Sakura Finetek, Tokyo, Japan) with the following schedule: 70% ethanol for 4 hours at 25°C, 80% ethanol for 2 hours at 25°C, 90% ethanol for 2 hours at 25°C, absolute ethanol I for 1 hour at 25°C, absolute ethanol II for 1 hour at 25°C, xylene I for 30 min at 25°C, xylene II for 30 min at 25°C, paraffin I for 45 min at 58°C, paraffin II for 45 min at 60°C, paraffin III for 45 min at 60°C, and paraffin IV for 45 min at 60°C. Tissues were embedded in paraffin molds with careful orientation before solidification. After cooling, blocks were trimmed to retain a 2-mm paraffin margin.

The paraffin blocks were then precooled at −20°C for 30 min to enhance hardness. Sectioning was performed using a fully automated rotary microtome (Leica 2M2255, Leica, Mannheim, Germany) at a thickness of 5 μm. The blade angle was adjusted to be parallel to the surface of the paraffin block. Serial sections were floated on a 40°C water bath for expansion and mounted onto glass slides. Slides were then dried in a 60°C oven for 2 hours to ensure firm adhesion of the tissue.

H&E staining of paraffin sections

Paraffin-embedded tissue sections were deparaffinized by sequential immersion in xylene I for 8 min, xylene II for 8 min, absolute ethanol I for 4 min, absolute ethanol II for 4 min, 95% ethanol for 2 min, 80% ethanol for 2 min, and distilled water for 1 min. Sections were then stained with Mayer’s hematoxylin for 5 min, followed by gentle rinsing under running distilled water for 1 min. Eosin staining was performed for 3 s, followed by another 1-min rinse under running distilled water.

Dehydration and clearing were carried out through sequential washes in 70% ethanol (30 s), 80% ethanol (10 s), 95% ethanol (2 min), absolute ethanol I (2 min), absolute ethanol II (4 min), xylene I (4 min), and xylene II (8 min). Slides were mounted using neutral resin and allowed to air dry in a ventilated, dark environment. After drying, digital slide scanning was performed using a slide scanner (Pannoramic DESK, 3DHISTECH, Budapest, Hungary).

Immunohistochemical staining of paraffin sections

Paraffin-embedded sections were deparaffinized through sequential immersion in xylene I (8 min), xylene II (8 min), absolute ethanol I (4 min), absolute ethanol II (4 min), 95% ethanol (2 min), 80% ethanol (2 min), and distilled water (1 min). For antigen retrieval, the sections were immersed in a staining jar filled with citrate buffer solution, ensuring complete submersion of the tissue. The jar was sealed with plastic wrap to prevent evaporation and placed in a microwave for 20 min at medium power. After heating, the jar was left to cool naturally to room temperature.

To block endogenous peroxidase activity, sections were incubated in 3% hydrogen peroxide at room temperature for 10 min, followed by three washes with PBS, each for 5 min. Nonspecific binding was blocked with 5% bovine serum albumin (BSA) for 30 min at room temperature. Subsequently, anti-CD31 primary antibody (1:100 dilution, Ab281583, Abcam, UK) was applied, and the sections were incubated overnight at 4°C in a humidified chamber. The next day, sections were washed three times with PBS (5 min each), followed by incubation with horseradish peroxidase (HRP)–conjugated secondary antibody for 30 min at room temperature, and another three PBS washes.

DAB working solution was prepared according to the manufacturer’s instructions and applied to the sections under a microscope. The staining reaction was monitored in real time and terminated by PBS rinsing once clear brown coloration appeared. Nuclei were counterstained with hematoxylin for 1 min. Dehydration and clearing were performed through 70% ethanol (30 s), 80% ethanol (10 s), 95% ethanol (2 min), absolute ethanol I (2 min), absolute ethanol II (4 min), xylene I (4 min), and xylene II (8 min). Slides were mounted with neutral resin, air-dried in a well-ventilated, dark environment, and imaged using a digital slide scanner.

WGA fluorescence staining

Paraffin sections were deparaffinized, rehydrated, and rinsed with PBS. Sections were incubated with iFluor 555–conjugated WGA solution (1:200 dilution) at room temperature for 1 hour in the dark. After three washes with PBS, nuclei were counterstained with DAPI. Images were captured using a fluorescence microscope, and myofiber morphology and CSA were analyzed.

Morphometric analysis of skeletal muscle

For skeletal muscle morphometric analysis, paraffin sections of gastrocnemius muscle were subjected to H&E staining and WGA fluorescence staining. Images were captured using a digital slide scanner or fluorescence microscope under standardized illumination conditions. Quantitative analysis was performed using ImageJ software in accordance with established methodologies from previous studies (39, 40). Centrally nucleated myofibers: A myofiber was defined as centrally nucleated if one or more nuclei were located within the cytoplasmic region, rather than at the peripheral edge of the myofiber. The percentage of CNFs was calculated as the number of CNFs divided by the total number of myofibers per field. Myofiber CSA: The CSA of individual myofibers was measured manually or semiautomatically after outlining the myofiber borders. Only myofibers with intact morphology and complete borders were included. Myofiber size distribution: The frequency distribution histogram and violin plot of myofiber CSA were generated on the basis of the CSA values of all measured myofibers to illustrate the shift in myofiber size distribution. Myofiber number per HPF: Total myofibers were counted in five randomly selected nonoverlapping HPFs per section, and the average number was calculated. All morphometric parameters were quantified in five nonoverlapping fields per section, with three mice per group.

Western blot analysis of muscle tissue protein in mice

Mice were euthanized by cervical dislocation, and the gastrocnemius muscles of the hindlimbs were rapidly dissected. The excised muscle tissues were briefly rinsed in prechilled PBS to remove residual blood and then wrapped in aluminum foil, snap-frozen in liquid nitrogen, and stored at −80°C until further processing. For protein extraction, 100 mg of frozen muscle tissue was transferred to a 2-ml homogenization tube containing 1 ml of radioimmunoprecipitation assay (RIPA) lysis buffer and two small grinding beads. The samples were balanced and placed in a cryogenic high-throughput tissue homogenizer (SCIENTZ-48L, SCIENTZ, Zhejiang, China) and processed at 60 kHz, 4°C for 1 min, followed by a 5-min pause; the cycle was repeated three times. After homogenization, samples were sonicated three times (80 W, 5 s each) and incubated on ice for 30 min. Lysates were centrifuged at 12,000g for 15 min at 4°C, and the supernatants were collected and stored at −80°C.

Protein concentrations were determined using the BCA assay. BSA standards were prepared at concentrations of 2000, 1000, 500, 250, 125, 62.5, 31.75, and 0 μg/ml, and 20 μl of each standard was added to a 96-well plate. Tissue lysates were diluted 1:10, and 20 μl of each sample was added in parallel. Then, 200 μl of BCA working solution was added to each well. After incubation at 37°C in the dark for 30 min, absorbance at 562 nm was measured using a microplate reader, and protein concentrations were calculated from the standard curve.

Sample volumes were adjusted on the basis of protein concentration to load 20 μg of total protein per lane. SDS loading buffer (5×) was added to achieve a final concentration of 1×, and samples were denatured at 95°C for 5 min using a metal bath. A 10% SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gel was prepared using a one-step gel preparation kit (PG212, Yamei, China) and allowed to polymerize at room temperature for 15 min. Samples were loaded, and electrophoresis was performed at 90 V for 30 min through the stacking gel and at 120 V through the resolving gel until the bromophenol blue dye front reached the bottom.

After electrophoresis, the gel and polyvinylidene difluoride (PVDF) membrane were equilibrated in transfer buffer and assembled in the order of filter paper-gel-PVDF-filter paper. Protein transfer was carried out at a constant current of 220 mA for 90 min. The membrane was blocked in 5% nonfat milk prepared in TBST at room temperature for 1 hour, followed by overnight incubation with primary antibodies at 4°C. The membrane was washed three times with TBST (5 min each), incubated with HRP-conjugated secondary antibodies at room temperature for 1 hour, and washed again with TBST three times.

Enhanced chemiluminescence reagents A and B were mixed and applied evenly to the membrane surface. After 1 min of incubation in the dark, chemiluminescent signals were visualized using an imaging system (ChemiDoc Touch, Bio-Rad, Texas, USA).

Tissue mRNA extraction and reverse transcription

Frozen muscle tissue was retrieved, and 100 mg was weighed and transferred into a 2-ml homogenization tube containing 1 ml of TRIzol reagent and two small grinding beads. The tube was balanced and placed into a cryogenic high-throughput tissue homogenizer (SCIENTZ-48L, SCIENTZ, Zhejiang, China) and processed at 60 kHz and 4°C for 1 min with 5-min intervals, repeated three times. After homogenization, the lysate was allowed to stand at room temperature for 5 min, followed by the addition of 200 μl of chloroform. The sample was vortexed vigorously for 15 s until milky white and then incubated at room temperature for 15 min. Phase separation was performed by centrifugation at 12,000g for 15 min at 4°C. The upper aqueous phase was carefully transferred to a clean 1.5-ml microcentrifuge tube. An equal volume of isopropanol was added, and the tube was gently inverted 10 times to mix and then incubated at room temperature for 10 min. The sample was centrifuged again at 12,000g for 15 min at 4°C, and the supernatant was discarded, leaving a visible white RNA pellet at the bottom of the tube.

The pellet was washed with 1 ml of prechilled 75% ethanol [prepared using diethyl pyrocarbonate (DEPC)–treated water and absolute ethanol], gently resuspended by tapping the tube wall, and centrifuged at 12,000g for 5 min at 4°C. The ethanol was removed, and the pellet was air-dried by inversion for 5 min. Last, the RNA pellet was dissolved in 20 μl of DEPC-treated water and incubated at 55°C for 5 min to facilitate complete dissolution.

A 2-μl aliquot of RNA was used to determine concentration and purity using a microvolume ultraviolet-visible spectrophotometer (NanoDrop One/OneC, Thermo Fisher Scientific, Waltham, MA, USA). Reverse transcription was performed using a commercial kit (G492, ABM, Canada). A 20-μl reaction system was prepared according to the kit instructions, and the reverse transcription program was run on a PCR thermocycler (SensoQuest labcycler, SensoQuest, Göttingen, Germany).

Real-time qPCR

qPCR reactions were prepared according to the manufacturer’s instructions for the qPCR premix (AQ602-24, Allsheng, China), and amplification was performed using a real-time fluorescence quantitative PCR system (AriaMx, Agilent, California, USA).

Macrophage depletion

CLs (40337ES08, Yeasen, China) and control blank liposomes (40338ES05, Yeasen, China) were removed from the refrigerator and allowed to equilibrate to room temperature. The vials were gently inverted 8 to 10 times to ensure thorough mixing. A 26-gauge needle was attached to a 1-ml syringe, and 200 μl of CLs was drawn into the syringe. The mouse was restrained by grasping sufficient skin behind the ears and the tail with the left hand, while the head and limbs were immobilized. The mouse was held in a slightly head-down position to allow abdominal organs, originally concentrated in the lower right quadrant, to shift away from the injection site. The syringe was inverted six times to remix the liposomes before injection. The needle was inserted at a 30° angle into the lower right quadrant of the abdomen. A volume of 200 μl of either CLs or ELs was injected intraperitoneally.

Macrophage depletion can be assessed 24 hours after injection. For long-term depletion, injections were repeated every 3 days, with the dose halved for subsequent administrations.

Femoral artery ring sprouting assay

Mice were euthanized by cervical dislocation and immersed in 75% ethanol for 5 min for surface sterilization. After drying, a groin incision was made to expose the femoral artery, and a 3- to 5-mm segment of the vessel was carefully dissected. The isolated artery was immediately placed in ice-cold PBS, and surrounding fat and nerve tissues were gently removed on ice. Under a stereomicroscope, the artery was trimmed into 1-mm ring segments and longitudinally opened to expose the luminal surface.

Using a precooled pipette tip, 30 μl of growth factor–reduced Matrigel (0827045, MoBase Biotech, Xiamen, China) was dispensed into each well of a 24-well plate. The artery rings were transferred using microforceps and embedded in the Matrigel with the luminal side facing downward. The plate was placed in a 37°C incubator for 1 hour to allow the Matrigel to solidify, after which 400 μl of culture medium containing different treatment conditions was added to each well. The initial morphology of each ring was recorded under a microscope.

The cultures were maintained at 37°C in a humidified atmosphere with 5% CO₂, with media replaced every 2 days. Micrographs were taken daily to monitor and document endothelial sprouting from the vessel rings.

Single-cell RNA-seq and analysis

Fourteen days after HLI surgery and GNPs-pIL-4 treatment, mice were euthanized, and gastrocnemius muscle tissues were harvested. A total of 0.6 g of muscle tissue was collected and immediately immersed in ice-cold tissue preservation solution. Samples were transferred on ice to the processing station, where single-cell suspensions were prepared using a tissue dissociation system (gentleMACS Octo, Miltenyi Biotec GmbH, North Rhine-Westphalia, Germany) in accordance with the protocol of the muscle tissue dissociation kit (130-098-305, Miltenyi Biotec, Germany). Single-cell transcriptome libraries were subsequently constructed using the 10x Genomics 3′ single-cell RNA-seq platform.

Raw sequencing data were processed using Cell Ranger (cellranger-7.1.0) to generate matrix, barcodes, and features files. These files were loaded into R and used to create Seurat objects via the Seurat package. Quality control was performed to retain cells with nFeature_RNA > 200, nCount_RNA < 7500, and mitochondrial gene content (percent.mt) < 5%. The filtered data were normalized, and highly variable genes were identified. Data from different groups were integrated using the FindIntegrationAnchors and IntegrateData functions to correct for batch effects. Principal components analysis, UMAP, and t-distributed stochastic neighbor embedding were performed for dimensionality reduction, followed by clustering at multiple resolution levels. Differentially expressed genes in each cluster were identified using the FindAllMarkers function, and cell types were annotated on the basis of known marker genes.

Cell-cell communication analysis between different cell types was conducted using the CellChat v2 package in R. Gene expression data from single-cell RNA-seq were used as input, with reference to the CellChatDB.mouse database. Signaling pathways involving fewer than 10 cells were excluded from the analysis.

Matrigel tube formation assay

Low growth factor Matrigel was transferred from −80°C to a 4°C refrigerator 24 hours in advance to thaw and was placed on ice 30 min before the experiment. Matrigel was diluted 1:1 with prechilled serum-free DMEM to a final concentration of 4 to 6 mg/ml and gently mixed using a pipette tip. A total of 50 μl of diluted Matrigel was added to each well of a precooled 96-well plate, taking care to avoid bubble formation. The plate was incubated at 37°C for 30 min to allow gelation, forming a uniform Matrigel layer.

EOMA cells at passages 3 to 5 were serum-starved in medium containing 1% FBS for 12 hours. After trypsinization, cells were centrifuged at 200g for 5 min and resuspended in media corresponding to different treatment groups. The cell concentration was adjusted to 1 × 10⁵ cells/ml, and 100 μl of cell suspension was gently added to each well. The plate was incubated at 37°C in a humidified 5% CO₂ atmosphere, and tube formation was monitored in real time using the Cytation 7 imaging system (images captured every 10 min).

Scratch assay

Scratch culture inserts (80366, Ibidi, Martinsried, Germany) were sterilized under ultraviolet light for 30 min and placed into a 24-well plate. EOMA cells were trypsinized and resuspended at a density of 5 × 10⁵ cells/ml. A total of 70 μl of the cell suspension was added to each chamber, and the plate was gently shaken to ensure even cell distribution. Cells were cultured at 37°C in a 5% CO₂ incubator for 24 hours to form a confluent monolayer.

Under sterile conditions, the Ibidi inserts were removed, and the wells were gently washed three times with prewarmed PBS to remove detached cells. The medium was replaced with starvation medium containing 1% FBS to minimize proliferation-related effects. Cells were cultured continuously at 37°C with 5% CO₂, and real-time migration was monitored using the Cytation 7 imaging system.

Detection of protein S-nitrosylation

After isolation of the mouse gastrocnemius muscle, 100 mg of fresh tissue was weighed and transferred to a 2-ml homogenization tube containing 1 ml of RIPA lysis buffer and two small grinding beads. The sample was balanced and processed in a cryogenic high-throughput tissue homogenizer (SCIENTZ-48L, SCIENTZ, Zhejiang, China) under 60 kHz at 4°C for 1 min with a 5-min interval, repeated three times. The homogenate was then subjected to ultrasonic disruption at 80 W for 5 s, repeated three times, followed by 30 min of ice bath incubation. Lysates were centrifuged at 12,000g for 15 min at 4°C, and the supernatant was collected. To remove nonspecific binding proteins, 10 μl of Protein A/G magnetic beads were added, and the sample was rotated at 4°C for 30 min. After centrifugation, the supernatant was retained.

Protein concentration was determined by the BCA assay and adjusted to 1 mg/ml. For each 100 μl of protein sample, 2 μl of 1 M methyl methanethiosulfonate was added, followed by vortexing for 1 min and incubation at room temperature for 30 min. Then, 600 μl of prechilled acetone was added, and samples were incubated at −20°C for 1 hour to precipitate proteins. The samples were centrifuged at 10,000g for 10 min, the supernatant was discarded, and the pellet was air-dried for 10 min.

The protein pellet was resuspended in 100 μl of HENS buffer and divided equally into two tubes (50 μl per tube). Each tube received 1 μl of 20 mM iodoTMT reagent, followed by vortexing and the addition of 2 μl of 1 M sodium ascorbate. Samples were incubated at room temperature in the dark for 1 to 2 hours. Unreacted reagents were removed by a second round of acetone precipitation, and the final protein pellet was resuspended in HENS buffer.

Laemmli loading buffer (5×) was added to each sample, and samples were heated at 95°C for 5 min. A volume of 25 μl of each sample was loaded per lane for SDS-PAGE, and protein S-nitrosylation was detected using an anti-TMT primary antibody and an HRP-conjugated secondary antibody.

Statistical analysis

In all figures, n represented the number of biological replicates (independent mice for in vivo experiments, independent cell batches for in vitro experiments). Each experiment was performed at least three independent times. All experimental data were presented as mean ± SD. Statistical analysis was performed using GraphPad Prism 9.5 (GraphPad, CA, USA). A Student’s t test was used to compare differences between two groups that satisfied the assumptions of normality, homogeneity of variance, and independence. One-way analysis of variance (ANOVA) was applied for multiple group comparisons under the same assumptions. Post hoc pairwise comparisons were corrected using Tukey’s HSD, Bonferroni, or Šídák methods. A P value < 0.05 was considered statistically significant.

Supplementary Materials

The PDF file includes:

Figs. S1 to S7

Legends for movies S1 to S3

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S3