Agonal cell resuscitation strategy to promote tissue repair

Abstract

Overactive lytic regulated cell death (RCD) causes excessive inflammation and impairs tissue repair. Current strategies rely on small-molecule inhibitors targeting the initiation of lytic RCD but these show off-target effects and reduced pro-repair factor release. During lytic RCD, a transitional agonal stage exists between initiation and terminal death, during which cells either proceed to terminal death or undergo resuscitation. Here, we found that agonal cells actively upregulated uptake of extracellular vesicles (EVs), and internalised EVs fused with the plasma membrane via SNARE complexes to enhance membrane repair. We demonstrated that enhancing the membrane repair capacity of agonal cells through artificially prepared EV-mimetic nano-platelet vesicles (NPVs) effectively promoted their resuscitation. Moreover, resuscitated cells secreted substantial amounts of prostaglandin E2 and N1-Acetylspermidine to further promote tissue repair. Our therapeutic strategy for lytic RCD-related delayed tissue repair is based on EV-mediated membrane repair and aims to establish a pro-regenerative niche using NPVs that can drive agonal cell resuscitation.

During lytic regulated cell death, a transitional agonal stage exists between initiation and terminal death. Here, the authors found that agonal cells upregulated uptake and internalization of extracellular vesicles fused with the cell membrane via SNARE complexes to enhance membrane repair.

Introduction

Inflammation is an essential and complex biological response that promotes tissue repair by clearing pathogens and necrotic tissue, while excessive or persistent inflammatory responses result in tissue damage and impaired repair¹. Lytic regulated cell death (RCD), such as necroptosis and pyroptosis, exacerbates local inflammatory responses and hinders tissue regeneration by releasing large amounts of damage-associated molecular patterns (DAMPs)².

Traditionally, it was believed that the activation of lytic RCD inevitably results in cell death³. Consequently, current strategies for inhibiting lytic RCD primarily focus on suppressing key molecular machinery in cell death pathways to prevent its initiation^4,5; however, this intervention strategy has limitations. First, small-molecule inhibitors have off-target effects. For example, receptor-interacting protein kinase 3 (RIPK3) inhibitors, such as GSK840/GSK843/GSK872, can also induce apoptosis⁵. Second, inhibiting the initiation of lytic RCD exhibits a therapeutic paradox. Although lytic RCD releases DAMPs to trigger inflammation, it also releases various pro-repair factors, including macrophage-derived 11,12-epoxyeicosatrienoic acid via gasdermin D (GSDMD) pores⁶ and muscle cell-secreted tenascin-C via phosphorylated mixed lineage kinase domain-like protein (p-MLKL) pores⁷. Therefore, blocking the initiation of lytic RCD using small-molecule inhibitors could also prevent the release of such pro-repair factors.

Recent studies have revealed that cells undergoing lytic RCD do not strictly follow an irreversible trajectory^3,8–12. Upon removal of death-inducing agents, cells with p-MLKL or GSDMD p30 may undergo resuscitation, retaining their capacity to survive and proliferate^9,12. This indicates the existence of a transitional phase between the activation of lytic RCD (the appearance of p-MLKL or GSDMD p30) and terminal death (severe loss of membrane integrity), which we term the “agonal stage”. During this stage, cells may either experience a progressive loss of membrane integrity leading to terminal death, or undergo resuscitation. As cells have an inherent membrane repair ability (e.g., through lysosomal exocytosis and ESCRT-III-mediated membrane scission)¹³, we speculate that the fate of agonal cells may depend on the balance between membrane damage and repair. Therefore, augmenting the membrane repair capacity of agonal cells to drive resuscitation may represent a promising strategy for intervening in lytic RCD. This strategy circumvents the off-target effects and therapeutic paradox of small-molecule inhibitor therapies by avoiding interference with lytic RCD initiation. However, the biological features of agonal cells and the effective strategies to enhance their membrane repair capacity remain to be systematically elucidated.

Diabetic wounds, ischaemic flaps, and spinal cord injury are typical disorders characterised by excessive lytic RCD and impaired tissue repair^14–17. The vicious cycle between lytic RCD and inflammation is a key factor contributing to the poor tissue repair observed in these diseases. Therefore, promoting the resuscitation of agonal cells to disrupt this vicious cycle may represent an effective therapeutic strategy. In this proof-of-concept study, we focused on necroptosis in tissue repair and found that the ability of agonal cells to take up extracellular vesicles (EVs) was significantly greater than that of normal cells. During this process, EVs fuse with the plasma membrane (PM) via vesicle-soluble N-ethylmaleimide-sensitive factor attachment protein receptors (v-SNARE), significantly improving the membrane repair capacity of agonal cells. Furthermore, compared with natural EVs (red blood cell-derived EVs [RBC-EVs], platelet-derived EVs [PMPs], and apoptotic vesicles [ApoVs]), artificially prepared nano-platelet vesicles (NPVs)¹⁸ exhibit an abundance of v-SNARE and superior membrane repair efficiency. Agonal cells can be effectively resuscitated by enhancing their membrane repair capacity through NPV intervention. Additionally, upon necroptosis-driven activation of the Erk/MAPK pathway, resuscitated cells secrete large amounts of prostaglandin E2 (PGE2) and N1-Acetylspermidine (N1-AcSPD) to promote tissue repair. NPV-based therapeutic strategies effectively promote the resuscitation of agonal cells, reduce necroptosis, alleviate inflammation at injury sites, and accelerate healing in diabetic wounds, ischaemic flaps, and spinal cord injury. Consequently, the agonal cell resuscitation strategy exhibits potential for treating various diseases associated with lytic RCD (Supplementary Fig. 1a).

Results and discussion

Agonal cells enhance their ability to take up EVs

As NIH3T3 cells do not endogenously express RIPK3, which mediates the phosphorylation of MLKL that activates necroptosis¹⁹, the progression of necroptosis in these cells can be regulated by controlling the expression levels of exogenous RIPK3. We overexpressed RIPK3 and selected a single clone (clone 1) with a lower RIPK3 upregulation (RIPK3-OE NIH3T3) for subsequent experiments (Supplementary Fig. 1b). During necroptosis and pyroptosis, membrane attack by p-MLKL or GSDMD induces phosphatidylserine (PS) externalisation prior to a complete loss of membrane integrity (positivity for impermeable nucleic acid dyes such as Sytox Green)^9,12. This enables the identification of agonal cells through Annexin V⁺ Sytox Green⁻ staining. We used TNF-α, Lcl-161 and zVAD-FMK (TSZ), a classic necroptosis inducer, to induce necroptosis. Necroptosis is induced as follows: TNF-α engages TNFR1 to trigger death signalling, Lcl-161 antagonises the activity of inhibitor of apoptosis proteins, and zVAD-FMK blocks caspase-dependent apoptotic pathways, thereby shunting cells toward necroptosis instead of apoptosis⁹. As expected, treatment of RIPK3-OE NIH3T3 cells with TSZ demonstrated an agonal stage (Annexin V⁺ Sytox Green⁻) prior to terminal death (Annexin V⁺ Sytox Green⁺) (Supplementary Fig. 1c). Additionally, Western blot analysis confirmed the presence of p-MLKL with no detectable cleaved PARP (a marker of apoptosis) in the agonal cell obtained by fluorescence-activated cell sorting (FACS) (Supplementary Fig. 1d–f).

To investigate the characteristic changes in agonal cells, we performed RNA sequencing (RNA-seq) analysis on agonal cells obtained by FACS (Fig. 1a and Supplementary Fig. 1d, e). Principal component analysis (PCA) demonstrated significant differences between agonal and control cells (Supplementary Fig. 1g). We selected genes that were upregulated in agonal cells and had a fold change of ≥1.5 for KEGG enrichment analysis, and the results showed significant enrichment in endocytosis and SNARE interactions in vesicular transport pathway (Fig. 1b). Meanwhile, consistent with previous reports that death stimuli activate the integrated stress response²⁰, we found that downregulated genes were significantly enriched in the ribosome pathway (Supplementary Fig. 1h). Additionally, agonal cells upregulate multiple genes encoding the phosphatidylserine receptor (PSR) (Fig. 1c). EVs expose PS on their surface, a feature that facilitates their capture and uptake by cells via PSR^21–23. These results suggest that activation of the cell death pathway may stimulate cells to enhance their capacity for the uptake and transport of EVs. We also analysed single-cell RNA sequencing (scRNA-seq) data from public databases for six necroptosis-relevant diseases. Numerous studies have demonstrated that necroptosis occurs during the progression of chronic wounds, Alzheimer’s disease, acute liver failure, acute kidney injury, fatal coronavirus disease 2019 and acute myocardial infarction. Inhibiting necroptosis has been shown to confer therapeutic benefits in these conditions^5,24–31. The results demonstrated that the normalised enrichment scores (NES) of endocytosis and SNARE interactions in vesicular transport gene sets were significantly positively correlated with the NES of the programmed necrotic cell death gene set across six cell types (Fig. 1d). Then, we classified the indicated cells into three subgroups (normal, stressed, and agonal) based on the NES of the programmed necrotic cell death gene set from lowest to highest (Supplementary Fig. 2a, e, i, m, q, u). In the six disease models, the proportion of the agonal subgroup was significantly higher in the disease group (Supplementary Fig. 2b, f, j, n, r, v). Moreover, the NES of endocytosis and SNARE interactions in vesicular transport gene sets were significantly higher in the agonal subgroup than in the normal subgroup, accompanied by a notable upregulation of multiple PSR genes (Supplementary Fig. 2c, d, g, h, k, l, o, p, s, t, w, x). Collectively, both bulk RNA-seq and scRNA-seq consistently indicated that agonal cells enhanced the uptake and transport of EVs by upregulating the molecular machinery involved in endocytosis and vesicular transport as well as PSRs.

Fig. 1. Agonal cells upregulate its ability to uptake EVs.

a Schematic illustration of RNA-seq performed on cells at distinct stages of necroptosis. b KEGG enrichment analysis of upregulated genes (fold change ≥ 1.5; p < 0.05) in agonal cells derived from RIPK3-OE NIH3T3. c Heatmap showing differential expression of phosphatidylserine receptor (PSR) genes (n = 3 independent samples). d Pearson correlation analysis between the normalised enrichment scores (NES) of the indicated gene set across diverse cell types. e Schematic diagram of the experimental workflow for assessing PS⁺ liposome uptake capacity of RIPK3-OE NIH3T3 by flow cytometry. f Quantification of DiI signals in ANX V⁻ Sytox Green⁻ and ANX V⁺ Sytox Green⁻ cells (n = 3 independent samples; mean ± s.d.). g Quantification of DiI signals in agonal cells (cyto D, cytochalasin D; n = 3 independent samples; mean ± s.d.). h Immunofluorescence staining of phosphorylated mixed lineage kinase domain-like protein(p-MLKL) in wounds (nuclei, blue; p-MLKL, green; Scale bars, 50 µm). i Quantification of agonal cell and terminal dead cell in wound tissue (NW 24, 48 and 72 h, n = 6 independent mice; The remaining group, n = 5 independent mice; mean ± s.d.). j Schematic representation of isolating EVs from tissue microenvironments. k Micrograms of proteins from EVs normalised to the grams of tissue (n = 6 independent mice; mean ± s.d.). l Relative abundance of EVs subtypes in the wound tissue microenvironment. m Micrograms of proteins from different EVs subtypes normalised to the grams of wound tissue at day 4 and 8 post-injury (Day8 NW, n = 5 independent mice; The remaining group, n = 6 independent mice; mean ± s.d.). INH inhibitory neurons, PTE proximal tubular epithelial cells, AT2 Type Ⅱ alveolar epithelial cells, RBC-EVs red blood cell-derived EVs, PMPs platelet-derived EVs, ApoVs apoptotic vesicles, TC-EVs tissue cell derived EVs, NW normal wound, DW diabetic wound. P values were determined using one-sided Fisher’s Exact test with Benjamini-Hochberg multiple-comparison test (b), a two-tailed unpaired Student t test (d), two-way ANOVA with Šidák’s multiple-comparison test (i), two-way ANOVA with Tukey’s multiple-comparison test (k) or two-tailed one-way ANOVA with a Tukey post-hoc test (f, g, m).

Given that exocytosis and endocytosis, which are functionally distinct vesicular trafficking pathways, share a substantial set of molecular machinery, we further investigated the ability of agonal cells to uptake EVs using in vitro assays. To evaluate the cellular uptake of EVs, we synthesised PS⁺ liposomes that mimicked the lipid composition of EVs³². We used 1 μM DiI to label liposomes. Under identical experimental conditions, the fluorescence from non-specific aggregates formed during the DiI staining process was negligible compared to that from liposomes (Supplementary Fig. 3a, b). The results demonstrated that ANX V⁺ Sytox Green⁻ cells exhibited a significantly higher uptake of PS⁺ liposomes than ANX V^- Sytox Green^- cells (Fig. 1e, f). Furthermore, treatment with cytochalasin D, an endocytosis inhibitor, significantly reduced the uptake of PS⁺ liposomes by agonal cells (Fig. 1g), suggesting that agonal cells primarily uptake PS⁺ EVs via endocytosis.

We investigated the relationship between EVs and necroptosis in vivo. We selected diabetic wounds (DW), which are a typical tissue repair disorder associated with lytic RCD, as the experimental model¹⁴. Results revealed that over time (days 4–8 post-injury), p-MLKL consistently maintained high expression levels in the DW group (Fig. 1h, Supplementary Fig. 3c, d). Correspondingly, DW maintained a consistently higher proportion of agonal and terminal dead cells throughout the 8-day observation period (Fig. 1i and Supplementary Fig. 3e). To explore the dynamic changes in EVs in the abovementioned tissue microenvironment, we isolated EVs at different time points post-injury using established protocols (Fig. 1j)³³. The successful application of this extraction method was verified through transmission electron microscopy (TEM), western blot, and nanoparticle tracking analysis (NTA) characterisation (Supplementary Fig. 4a–c). EVs content in the tissue microenvironment was significantly elevated in normal wounds (NW) at day 4 post-injury but decreased significantly by day 8. In contrast, EVs content remained low in DW throughout the observation period (Fig. 1k). Additionally, we employed inhibitors of endocytosis (prochlorperazine, PCZ)³⁴, microvesicle secretion (pantetheine)³⁵, and exosome secretion (GW4869)^36,37 to further investigate the cause of the reduced EVs content in DW (Supplementary Fig. 4d). The results indicated comparable EVs uptake capacity between NW and DW on day 4 (Supplementary Fig. 4e), which aligns with the substantial number of agonal cells present in both groups at this time point (Figs. 1h, i and Supplementary Fig. 3c–e), while NW exhibited significantly stronger EVs secretion capacity than DW. By day 8, DW exhibited significantly greater EVs uptake capacity than NW, consistent with their divergent necroptosis patterns at this time point, while secretion capacity showed no significant differences between the two groups. These findings suggest that persistent necroptosis in diabetic wounds may be associated with inadequate EV content in the injured tissue microenvironment.

We employed nanoflow cytometry to characterise the composition of EVs. EVs were categorised based on their cellular origin as platelet-derived CD41⁺ EVs (PMPs), white blood cell-derived CD45⁺ EVs (WBC-EVs), and red blood cell-derived CD235a⁺ EVs (RBC-EVs)^38,39. EVs not marked by these markers were classified as tissue cell-derived EVs (TC-EVs). Results showed that on day 4 post-injury, the proportion of blood cell-derived EVs in both NW and DW was significantly higher than that in normal skin tissue. By day 8, the proportion of blood cell-derived EVs in NW had decreased significantly, while it remained unchanged in DW (Fig. 1l and Supplementary Fig. 4f, g). In terms of total content, on day 4 post-injury, NW showed significantly higher levels of RBC-EVs and PMPs than DW, with a more pronounced difference observed in PMPs content (Fig. 1m). This trend aligns with changes in necroptosis in the wound tissue (Fig. 1h, i and Supplementary Fig. 3c–e). In summary, these results demonstrate that agonal cells exhibit enhanced EVs uptake capacity, and that the relative deficiency of PMPs in DW may contribute to the abnormal enhancement of necroptosis in these wounds.

EVs mediate membrane repair in agonal cells via SNARE-dependent membrane fusion

p-MLKL disrupts membrane integrity, a key mechanism of necroptosis-induced cell death. Cells possess multiple inherent mechanisms to repair damaged PM, including lysosomal exocytosis, endosome exocytosis and ESCRT-III-dependent membrane resealing^13,40. We hypothesised that agonal cells take up more EVs, particularly PMPs, to enhance membrane repair and thus counteract necroptosis. To achieve a rapid induction of p-MLKL-dependent membrane disruption, NIH3T3 was transfected with previously reported hMLKL^1–140−2 × Fv, which rapidly oligomerizes and attacks the PM upon dimerizer (B/B) addition⁴¹. As reported in previous literature, the kinetics of membrane repair were assessed using a calcein fluorescence leakage assay⁴⁰. Calcein is a polar, green fluorescent molecule impermeable to the PM. Thus, only when the integrity of PM is compromised does intracellular calcein leak out through membrane pores, resulting in reduced fluorescence intensity⁴². Furthermore, the degree of PM damage is positively correlated with the leakage rate of calcein (Supplementary Fig. 5a, b). Therefore, when subjected to stimulation with the same degree of membrane damage, a slower decline in calcein intensity reflects a more robust capacity for membrane repair⁴⁰. Moreover, we observed that when cells were subjected to membrane damage induced by 150 nM B/B, impermeable nucleic acid dyes (Sytox orange) started to enter the cells only after calcein had completely leaked out (Supplementary Fig. 5c, d). This indicates that cells remained in the agonal stage before calcein was fully released. To evaluate the effect of different types of EVs on membrane repair, we tested RBC-EVs, PMPs, and ApoVs (apoptotic vesicles derived from mouse mesenchymal stem cells). TEM, Western blot, and NTA collectively validated the reliability of these vesicle preparation protocols (Supplementary Fig. 6a–c). RBC-EVs and PMPs were used as representative blood-derived EVs, whereas ApoVs represented EVs originating from local tissue cells during injury. Results showed that all three types of EVs (especially PMPs) significantly slowed the decline in calcein fluorescence, suggesting an enhanced cellular membrane repair capacity (Fig. 2a).

Fig. 2. EVs mediate membrane repair in agonal cells via SNARE-dependent membrane fusion.

a Upper: Images of calcein-labelled hMLKL^1-140 −2 × Fv-OE NIH3T3 injured by 150 nM B/B after treatment with different EVs. Down: Plot showing kinetics of calcein leakage (Ctrl, n = 26 independent cells; RBC-EV, n = 34 independent cells; PMP, n = 35 independent cells; ApoV, n = 30 independent cells; scale bar, 20 µm; mean ± s.e.m.). b Schematic diagram of the preparation of NPVs. c Quantification of PS⁺ NPV (n = 3 independent samples; mean ± s.d.). d Upper: Images of cells injured by 150 nM B/B. Down: Plot showing kinetics of calcein leakage (PMP, n = 31 independent cells; Ctrl, n = 30 independent cells; NPV, n = 28 independent cells; scale bar, 20 µm; mean ± s.e.m.). e Representative images and quantification of gold nanoparticles colocalized with plasma membrane (Scale bar, 1 μm; n = 15 independent cells; mean ± s.e.m.; PM, plasma membrane). Orange arrows are used to indicate the gold particles. f Upper: TIRFM images of hMLKL^1-140 −2 × Fv-OE NIH 3T3 containing DiI-NPVs following addition of vehicle (normal) or 100 nM B/B (agonal) (scale bars, 10 µm). Zoomed image series in the bottom panels shows NPVs moving axially (normal) or undergoing exocytosis (agonal). Down: DiI fluorescence kinetics in the zoomed images (Peak intensity: the brightest pixel; Total intensity: sum of all pixels). g Representative images of the colocalization degree of NPVs with lysosomes in Lamp1-zsGreen/RIPK3-OE NIH3T3 (scale bars, 20 µm). h Quantification of VAMP2-FITC and VAMP7-FITC in different groups (n = 3 independent samples; mean ± s.d.). i Left: Images of cells injured by 100 nM B/B. Right: Plot showing kinetics of calcein leakage (siStx1a + siStx4a, n = 39 independent cells; siStx1a + siStx4a + NPVs, n = 36 independent cells; siSCR, n = 37 independent cells; siSCR+NPVs, n = 28 independent cells; siStx1a, and siStx1a + NPVs, n = 30 independent cells; siStx4a, and siStx4a + NPVs, n = 29 independent cells; scale bars, 20 µm; mean ± s.e.m.). j Schematic representation of NPVs-mediated plasma membrane repair. P values were determined using a two-tailed one-way ANOVA with a Tukey post-hoc test (c, h), or a two-tailed nonparametric Mann–Whitney U test (e).

CD42b and CD41 can sense shear stress and induce PS exposure in the platelet membranes^43,44. Our research group had previously developed artificially prepared NPVs¹⁸. NPVs were obtained through the extrusion of intact platelets, resulting in nano-sized vesicles with an average diameter of 215.4 nm (Fig. 2b and Supplementary Fig. 6d) and the retention of platelet-specific proteins, namely CD42b and CD41 (Supplementary Fig. 6e). Under shear stress generated by high-speed centrifugation at 25,800 g, the proportion of PS⁺ NPVs was significantly increased (Fig. 2c and Supplementary Fig. 6f). This result confirmed that PS⁺ artificial NPVs were successfully prepared. Using the CCK-8 assay, we determined the maximum usable dose of NPVs on NIH 3T3 cells (Supplementary Fig. 6g). We then investigated the membrane repair capacity of NPVs and found that NPVs exhibited superior membrane repair efficiency compared with PMPs (Fig. 2d).

We used high-resolution immunoelectron microscopy (HRIEM) to examine the distribution of NPVs (CD41 labelled)³⁸ in agonal cells. HRIEM images revealed slight co-localisation of immunogold particles with the PM in normal cells, whereas agonal cells showed significantly more immunogold particles co-localised with the PM, indicating that more NPVs underwent fusion with the PM in agonal cells than in normal cells (Fig. 2e). Moreover, we observed that in agonal cells, some NPVs were located in endosomes with compromised membrane integrity, and others existed as free entities in the cytosol, unencapsulated by lysosomes or endosomes (Supplementary Fig. 7a). These findings suggest that NPVs in agonal cells may exhibit biological behaviours similar to lysosomal exocytosis (lysosomal trafficking to and fusion with the PM to promote membrane repair)⁴⁵. To further validate this hypothesis, we employed total internal reflection fluorescence microscopy (TIRFM) to dynamically monitor the behaviour of DiI-labelled NPVs near the PM during MLKL-induced cell membrane damage⁴⁶. Under identical experimental conditions, nonspecific aggregates generated during the staining process exhibited negligible fluorescence relative to the labelled NPVs (Supplementary Fig. 7b, c). Upon addition of 100 nM B/B, real-time monitoring of DiI fluorescence demonstrated that as NPVs approached the PM, both the peak and total fluorescence increased. Upon membrane fusion, the peak intensity decreased due to DiI diffusion, while the total intensity remained elevated because DiI had not yet diffused out of the observation area and was still fully excited. Conversely, a consistent trend in changes between peak and total intensities was observed in normal cells, suggesting no fusion events (Fig. 2f). Notably, the fraction of NPVs that fused with PM was significantly elevated in B/B-induced agonal cells relative to normal cells (Supplementary Fig. 7d). Under identical imaging conditions, the DiI control group showed an undetectable fluorescence signal (Supplementary Fig. 7e). Taken together, these results suggest that NPVs in agonal cells undergo exocytosis, thereby integrating their membranes into the PM.

We further explored the reason for this behaviour. Under normal conditions, NPVs are primarily internalised via endocytosis¹⁸ and degraded in lysosomes. This prompted us to investigate the mechanism by which NPVs evade the endolysosomal system in agonal cells. p-MLKL not only targets the PM but also disrupts lysosomal membranes^47,48. Based on these observations, we hypothesised that p-MLKL-mediated damage to the endolysosomal system facilitates the escape of NPVs from lysosomes. To monitor changes in endolysosomal membrane permeability, we used a previously reported monomeric Azami Green–galectin 3 (mAG-Gal3) fluorescence reporter system^49,50. When endolysosomal membrane permeability increases, mAG-Gal3 accumulates in a punctate pattern. Fluorescence imaging revealed that TSZ-induced necroptosis triggered a marked augmentation in the formation of mAG-Gal3 puncta (Supplementary Fig. 7f–h). Correspondingly, after 3 h of TSZ treatment, we observed a significant increase in lysosomal pH in the TSZ group (Supplementary Fig. 7i, j), suggesting that p-MLKL disrupts the lysosomal membrane, leading to proton leakage. This process may facilitate the lysosomal escape of NPVs. Furthermore, we used Lamp1-zsGreen/RIPK3-OE NIH3T3 cells to observe the co-localisation of DiI-labelled NPVs with lysosomes. We found that 3 h after TSZ-induced necroptosis, the Manders’ overlap coefficient of NPVs was significantly decreased (Fig. 2g and Supplementary Fig. 7k). These results demonstrate that NPVs within the endolysosomal system can successfully escape lysosomal degradation owing to the disruption of endolysosomal membrane integrity in agonal cells.

Intracellular vesicles, including lysosomes, endosomes, and synaptic vesicles, primarily mediate fusion with the PM through vesicle-associated membrane protein 2 (VAMP2)/syntaxin 1 (STX1)/synaptosome-associated protein 25 (SNAP25) or VAMP7/STX4/SNAP23 complexes^51,52. Therefore, we investigated whether NPVs and PMPs, as exogenous vesicles, also contain VAMP2 and VAMP7. Given that only the N-terminal outwards-facing VAMP2 or VAMP7 can mediate membrane fusion, we used antibodies that specifically target the N-terminal domains of VAMP2 and VAMP7 for detection⁵². The results showed that, without adding Tween20, both NPVs and PMPs exhibited detectable levels of N-terminal outward-facing VAMP2 and VAMP7. Furthermore, the mean fluorescence intensity (MFI) of VAMP2 and VAMP7 on NPVs was approximately double that on PMPs. Upon addition of Tween20 to enhance vesicle permeability⁵³, the MFI of VAMP2 and VAMP7 on both vesicle types significantly increased, with PMPs showing a slightly higher MFI than NPVs. This suggests that the total amount of VAMP2 and VAMP7 on the PMPs membranes was slightly higher than that on the NPVs, and the greater amount of N-terminal outward-facing VAMP2 and VAMP7 on the surface of NPVs than that of PMPs may be related to the conformational flip of VAMPs during the preparation of artificial vesicles (Fig. 2h). These findings implied that NPVs fused with the PM via SNARE interactions.

We knocked down STX1A and STX4 in cells using siRNAs (Supplementary Fig. 7l, m) to inhibit SNARE-mediated membrane fusion. After inducing membrane damage in hMLKL^1-140−2xFv-OE NIH3T3 cells with 100 nM B/B, we observed that knocking down either STX1A or STX4 alone partially inhibited the membrane repair efficiency of NPVs (Fig. 2i). Additionally, the simultaneous knockdown of STX1A and STX4 almost completely abolished the membrane repair efficiency of NPVs. Moreover, we compared the area under the curve (AUC) to preliminarily assess the relative contribution of STX1A/STX4-dependent vesicles from different sources to membrane repair capacity. The AUC between the fluorescence curves of the siRNA scrambled control (siSCR) and siStx1a + siStx4a groups can be regarded as an indicator of the membrane repair capacity contributed by endogenous vesicles. Similarly, the AUC between the siSCR and siSCR+NPVs groups represents the membrane repair capacity provided by NPVs. The results showed that NPVs (76.7%) contributed more than endogenous vesicles (23.3%) in the STX1A/STX4-dependent membrane repair process. These results confirm that NPVs enhance cellular membrane repair capacity mainly through SNARE-mediated membrane fusion with the PM. These findings reveal that the disruption of endolysosomal membrane integrity by p-MLKL allows NPVs taken up by agonal cells to escape the endolysosomal system. Subsequently, the v-SNAREs on the surface of these vesicles interact with the target SNAREs on the PM, facilitating vesicle-PM fusion, thereby repairing the damaged membrane (Fig. 2j).

NPVs promote agonal cell resuscitating and resuscitated cells secrete pro-repair signalling molecules

Building on our previous findings that EVs significantly enhance the cellular membrane repair capacity, we hypothesised that the upregulation of EV uptake by agonal cells may represent an intrinsic protective mechanism. We then explored whether NPV intervention could resuscitate agonal cells. First, we performed FACS to isolate agonal cells (Annexin V⁺ Sytox Green⁻) (Fig. 3a). The isolated agonal cells were cultured for 16 h with or without NPV supplementation. In the control group, most agonal cells progressed to terminally dead cells (Annexin V⁺ Sytox Green⁺). Correspondingly, in the presence of NPVs, 15% of the agonal cells were successfully resuscitated, as evidenced by their transition to an Annexin V⁻ Sytox Green⁻ phenotype (Fig. 3b). To further validate this phenomenon, we cultured the two cell groups for 7 days. The results demonstrated that the control group exhibited no proliferation, whereas the NPV-treated group displayed significant expansion (Fig. 3c). Western blot also demonstrated that the resuscitated cells exhibited undetectable levels of p-MLKL (Fig. 3d). Furthermore, RIPK3-OE NIH3T3 cells resuscitated by NPV treatment displayed slightly enhanced resistance to TSZ-induced necroptosis compared with untreated parental cells (Fig. 3e). This may be attributed to the enhanced membrane repair capacity imparted by NPVs. As is well known, during lytic RCD, significant changes occur in cellular calcium homeostasis and mitochondrial energy metabolism^12,54. Therefore, we further examined the ability of agonal cells to maintain calcium homeostasis and preserve mitochondrial energy metabolism. We employed GCaMP6 to monitor changes in cytosolic calcium levels throughout the progression of necroptosis⁵⁵. The results indicated that agonal cells exhibited a sharp increase in cytosolic calcium due to PM perforation by p-MLKL. As the process advanced to terminal death, GCaMP6 fluorescence was extinguished, resulting in a significantly lower signal than that in normal cells. In contrast, agonal cells resuscitated by NPVs showed no significant difference in calcium concentrations compared with normal cells (Fig. 3f, Supplementary Fig. 8a), demonstrating that agonal cells retain the capacity to maintain calcium homeostasis. Moreover, we monitored changes in mitochondrial energy metabolism during necroptosis using the Mito-AT1.03 fluorescent protein⁵⁶. Our results showed that agonal cells maintained near-normal mitochondrial ATP levels, whereas these levels were significantly decreased in terminal dead cells. In contrast, resuscitated cells exhibited markedly elevated mitochondrial ATP concentrations (Fig. 3g, Supplementary Fig. 8b). We further performed RNA-seq on normal cells (Ctrl) and resuscitated cells (Res) derived from NIH3T3 and found that multiple genes associated with oxidative phosphorylation were significantly upregulated in resuscitated cells (Supplementary Fig. 8c, d). Based on the above results, we proposed that the agonal stage is a unique and reversible RCD stage, and that NPVs are capable of effectively resuscitating agonal cells.

Fig. 3. NPVs promote agonal cell resuscitating and resuscitated cells secrete pro-repair signalling molecules.

a Sorting strategy for ANX V⁺ Sytox Green⁻ RIPK3-OE NIH3T3 cells. b Flow cytometry analysis and quantification of resuscitated cells. Resuscitated cells are circled in red (n = 3 independent samples; mean ± s.d.). c Clonogenic survival of resuscitated cells from (b) (n = 3 independent samples; mean ± s.d.). d Immunoblotting of pMLKL and total MLKL in cells from (b). e IncuCyte-based quantification of Sytox Green⁺ cells from (b) (Parental cells + TSZ, and Resuscitated cell, n = 4 independent samples; Parental cells, n = 3 independent samples; Resuscitated cell + TSZ, n = 5 independent samples; mean ± s.d.). f Flow cytometry analysis and quantification of GCaMP6 signals in GCaMP6/RIPK3-OE-NIH3T3 (n = 3 independent samples; mean ± s.d.). g Flow cytometry analysis and quantification of Mito-AT1.03 signals in Mito-AT1.03/RIPK3-OE-NIH3T3 (n = 3 independent samples; mean ± s.d.). h Quantification of resuscitated cells (n = 3 independent samples; mean ± s.d.). NPVs were administered at indicated time following TSZ addition. i Quantification of resuscitated cell. Following transfected with the indicated siRNA, RIPK3-OE NIH3T3 were sorted as in (a) (siSCR, scramble; n = 3 independent samples; mean ± s.d.). j Quantification of resuscitated MEFs (n = 3 independent samples; mean ± s.d.). k, l Quantification of PI⁺ cells by multifunctional microplate reader (n = 3 independent samples; mean ± s.d.). m Schematic for the generation of supernatant from MEF (MC, derived from control MEF; MP, derived from NPV-treated MEF; MD, derived from dead MEF; MR, derived from resuscitated MEF). n Upper: schematic of the wound healing assay. Down: Representative images and quantification of wound closure area at 24 h (MD, n = 3 independent samples; The remaining group, n = 4 independent samples; mean ± s.d.). o Upper: schematic of gene expression analysis. Down: qPCR for indicated gene (n = 3 independent samples; mean ± s.d.). P values were determined using a two-tailed one-way ANOVA with a Tukey post-hoc test (f–i, n, o), two-way ANOVA with Šidák’s multiple-comparison test (e), or a two-tailed unpaired Student t test (b, c, j–l).

In addition, we added NPVs at 0, 3, 6 and 9 h after TSZ treatment to determine the temporal window for resuscitating agonal cells. The results showed that NPVs maintained some efficacy in resuscitating agonal cells when administered within 6 h post-induction (Fig. 3h and Supplementary Fig. 8e). Importantly, when STX1A and STX4 were knocked down in agonal cells using siRNA, the pro-resuscitation effect of NPVs was significantly impaired (Fig. 3i and Supplementary Fig. 8f). In addition to NIH3T3 cells, NPVs could resuscitate agonal mouse embryonic fibroblasts (MEFs) (Fig. 3j and Supplementary Fig. 8g). Additionally, NPVs significantly enhanced the resistance of MEFs and bone marrow-derived macrophages (BMDMs) to necroptosis and pyroptosis in the continuous presence of death-inducing stimuli (Figs. 3k, l).

Having established that NPVs can resuscitate agonal cells, we investigated the potential role of these resuscitated cells in a tissue-injury microenvironment. Lytic RCD leads to cell lysis and release of DAMPs, which trigger excessive inflammatory responses and impair tissue repair². This prompted us to explore the potential role of secretory products from resuscitated cells in tissue repair. We collected four types of supernatants derived from MEFs: MC (supernatants derived from untreated control cells), MP (supernatants derived from normal cells treated with NPVs), MD (supernatants derived from dead cells), and MR (supernatants derived from resuscitated cells). These supernatants were tested for their effects on fibroblasts and macrophages, which are the two key cell types involved in tissue repair (Fig. 3m)¹. First, we evaluated the effect of MR on fibroblasts using an in vitro wound healing assay. When the four supernatants were added to wounds created in monolayers of primary mouse fibroblasts, the results showed that MR significantly enhanced the migration and proliferation of primary mouse fibroblasts (Fig. 3n). Similarly, MR significantly enhanced fibroblast viability, whereas MD markedly suppressed it (Supplementary Fig. 9a). Additionally, MD induced a substantial number of fibroblast deaths (Supplementary Fig. 9b). Next, we incubated naïve BMDMs with the four supernatants for 16 h before assessing the expression of pro-repair genes. Macrophages treated with MD exhibited massive cell death, precluding RNA collection (Supplementary Fig. 9a–c). In contrast, MR significantly promoted the expression of four genes associated with tissue repair (Vegfα, Thbs1, Tgm2 and Arg1)⁵⁷ in macrophages compared with those treated with MC (Fig. 3o). Our findings demonstrate that during necroptosis, NPVs can enhance the cellular membrane repair capacity to resuscitate agonal cells, and these resuscitated cells can further secrete pro-repair signalling molecules.

Resuscitated cells exhibit pro-repair ability by secreting prostaglandin E2 and N1-AcSPD

To identify the specific pro-repair components in the MR, we performed ultrafiltration using a 3-kDa molecular weight cut-off cellulose membrane to separate protein ligands (>3 kDa) and small-molecule metabolites (<3 kDa) (Fig. 4a)⁶. Silver staining confirmed the absence of proteins in the <3 kDa fraction (Fig. 4b). We then tested the effects of the >3 kDa and <3 kDa fractions on fibroblasts and macrophages. We found that only the <3 kDa fraction of MR promoted in vitro wound healing and upregulated the expression of pro-repair genes in macrophages (Fig. 4c, d). Correspondingly, only the <3 kDa fraction of MR enhanced fibroblast viability, whereas the >3 kDa fraction of MD induced substantial fibroblast death and inhibited in vitro wound healing (Supplementary Fig. 9d–f). These results indicate that small-molecule metabolites, rather than protein ligands, are the key factors responsible for the pro-repair effects of MR.

Fig. 4. Resuscitated cells exhibit pro-repair ability by secreting prostaglandin E2 and N1-acetylspermidine.

a Schematic of supernatant ultrafiltration. b Silver staining of supernatant (n = 3 independent samples). c Primary mouse tail tip fibroblasts in coculture with <3 kDa and > 3 kDa fractions of MR. Representative images and quantification of wound closure area at 24 h (n = 4 independent samples; mean ± s.d.). d Macrophages in coculture with <3 kDa and >3 kDa fractions of MR for 16 h. qPCR for indicated gene (n = 3 independent samples; mean ± s.d.). e PCA of metabolites in MC, MD, MP and MR (n = 6 independent samples). f Venn diagram of TOP50 Up-regulated metabolites (VIP > 1, p < 0.05) between different groups (VIP, variable importance in projection). g Volcano plot of metabolite levels in MR vs. MP group. h ELISA results of PGE2 concentrations in supernatants (PGE2, Prostaglandin E2; n = 6 independent samples; mean ± s.d.). i PCA for genes in normal cells (Ctrl), normal cells treated with NPVs (C + NPVs) and agonal cells resuscitated by NPVs treatment (Res) derived from MEFs (n = 3 independent samples). j PGE2 and N1-acetylspermidine (N1-AcSPD) synthesis pathway. COX Cyclooxygenase, Ptgs Prostaglandin-Endoperoxide Synthase, PGH2 Prostaglandin H2, PGES prostaglandin E synthase, ODC Ornithine decarboxylase, SRM Spermidine synthase, SMS spermine synthase, PAO Polyamine oxidase, SAT1 Spermidine/spermine N1-acetyltransferase. k Heatmap showing differential expression genes encoding the key enzymes required for the synthesis of PGE2 and N1-AcSPD (n = 3 independent samples). l Immunoblotting and quantification of COX2 and SAT1 in MEFs (n = 3 independent samples; mean ± s.d.). The samples derive from the same experiment and the blots were processed in parallel. m Immunoblotting and quantification of pErk1/2 in MEFs (n = 3 independent samples; mean ± s.d.). n Macrophages in coculture with 10 μM PGE2 or 10 μM N1-AcSPD for 16 h. qPCR for indicated genes (n = 3 independent samples; mean ± s.d.). o Representative images and quantification of wound closure area (n = 4 independent samples; mean ± s.d.). P values were determined using a two-tailed one-way ANOVA with a Tukey post-hoc test (c, d, h, l–o), or two-tailed unpaired Student t test (f, g).

To identify the specific metabolites responsible for the pro-repair effects of MR, we performed metabolomic analysis of the four supernatants. PCA revealed that the metabolite profiles of MD and MR were significantly different from those of MC, whereas the metabolite profile of MP showed no significant difference from that of MC (Fig. 4e). We then performed an intersection analysis of the top 50 metabolites enriched in MR compared to MC, MR compared to MD, and MR compared to MP. This analysis identified 13 metabolites that were significantly upregulated in all three comparisons (Fig. 4f, g and Supplementary Fig. 9g, h). Among these, PGE2 and N1-AcSPD were previously reported to promote cell proliferation or macrophage polarisation toward a pro-repair phenotype^57–59. PGE2 enrichment in the MR was further validated by ELISA (Fig. 4h). We subsequently performed RNA-seq to investigate the mechanism underlying the elevated levels of PGE2 and N1-AcSPD in MR. Consistent with the metabolomic analysis, agonal cells resuscitated by NPVs (Res) significantly upregulated the expression of key enzymes required for the synthesis of PGE2 and N1-AcSPD compared with normal cells (Ctrl) and normal cells treated with NPVs (C+NPVs) (Fig. 4i–k). Western blot analysis further confirmed that Res contained higher levels of cyclooxygenase-2 (COX-2) and Spermidine/spermine N1-acetyltransferase 1 (SAT1) (Fig. 4l). Numerous studies have shown that activation of the Erk/MAPK signalling pathway can upregulate the expression of COX-2 and SAT1^60–62. Accordingly, we observed significant upregulation of genes associated with the Erk/MAPK signalling pathway in Res (Supplementary Fig. 9i). Western blot analyses also demonstrated that NPV treatment alone did not induce activation of the Erk/MAPK signalling pathway, and that only resuscitated cells that had undergone the initiation of necroptosis exhibited activation of this pathway (Fig. 4m). This result aligned with previous reports. Necroptosis activation induces p-MLKL to form pores in the PM⁹, mitochondria⁶³ and lysosomes⁴⁷, resulting in elevated cytosolic calcium level, which in turn activates the Erk/MAPK pathway^64,65.

Based on these findings, we investigated the roles of PGE2 and N1-AcSPD in tissue repair. Our findings revealed that PGE2 significantly promoted the expression of pro-repair genes in macrophages, whereas N1-AcSPD had no such effect (Fig. 4n). In the in vitro wound healing assay, both PGE2 and N1-AcSPD enhanced the proliferation and migration capabilities of fibroblasts, with N1-AcSPD showing a more significant effect (Fig. 4o). Consistently, N1-AcSPD enhanced fibroblast viability more effectively than PGE2 and neither molecule caused significant cell death (Supplementary Fig. 9j, k). Cells can take up N1-AcSPD and convert it into polyamines (including putrescine, spermidine, and spermine)^58,66. It is well established that polyamines promote cell growth and proliferation⁶⁷, as spermidine enables eIF5A hypusination (^hypeIF5A) (Supplementary Fig. 9l), and active ^hypeIF5A regulates the translation of proteins critical for the cell cycle and cellular energy metabolism to drive this effect^68,69. Therefore, we investigated whether N1-AcSPD treatment would increase the ^hypeIF5A level in fibroblasts. Western blot results confirmed that N1-AcSPD indeed increased the level of ^hypeIF5A in fibroblasts (Supplementary Fig. 9m), indicating that N1-AcSPD can enhance the proliferation and migration capacities of fibroblasts by elevating ^hypeIF5A levels.

As the activation of cell death programmes exposes cells to multiple stressors, we investigated whether resuscitated cells underwent senescence or malignant transformation. Gene set enrichment analysis (GSEA) revealed no significant difference in the cellular senescence gene set between the Res and Ctrl groups (Supplementary Fig. 10a). Additionally, the senescence-associated β-galactosidase assay showed no notable increase in enzymatic activity in resuscitated cells (Supplementary Fig. 10b), and Western blot analysis indicated no significant difference in the levels of p21 between the two groups (Supplementary Fig. 10c). Moreover, the expression levels of proteins associated with myofibroblast transition showed no significant difference between resuscitated and normal cells (Supplementary Fig. 10d), indicating that resuscitated cells do not promote tissue fibrosis. RNA-seq revealed no significant upregulation of multiple oncogenes in resuscitated cells (Supplementary Fig. 10e)⁷⁰. In vitro colony formation assays demonstrated that neither one nor three times resuscitated cells exhibited abnormally enhanced proliferative capacity (Supplementary Fig. 10f). Furthermore, subcutaneous tumour formation assays in nude mice showed that both one and three times resuscitated cells failed to form tumours at 16 days post-transplantation (Supplementary Fig. 10g). These results collectively indicate that repeated resuscitation does not induce oncogenic transformation.

NPVs mediated agonal cell resuscitation for promoting tissue repair

Given that tissue injury induces lytic RCD not only at the initial stage of injury, but also continuously throughout the subsequent inflammatory phase (Fig. 1h, i and Supplementary Fig. 3c–e)⁷¹. The inflammatory phase typically lasts 3–5 days, and may be further prolonged in severe injury cases¹. In theory, NPVs-based administration during this phase can promote the resuscitation of agonal cells and accelerate tissue healing. Therefore, we further evaluated the therapeutic efficacy of an NPV-based agonal cell resuscitation strategy in DWs models, ischaemic random-pattern skin flap (FLAP) models, and spinal cord clip compression injury (SCI) models.

In the DW model, we first evaluated whether locally applied NPVs could effectively target agonal cells in the wound. On the day of wound induction, mice received intraperitoneal injections of z-VAD-FMK to minimise interference from early apoptotic cells^72,73. The results demonstrated that the fluorescence intensity of NPVs was significantly higher in agonal cells than that in normal cells (Supplementary Fig. 11a, 11b), indicating effective targeting of agonal cells by NPVs. Normal mouse wounds (group 1) and untreated DW (group 2) served as controls, whereas NPVs were administered locally to the DW (group 3) (Fig. 5a). The results showed that the wound healing rate was significantly higher in group 3 than in group 2 (Fig. 5b, c). Haematoxylin and eosin staining (H&E) revealed that groups 3 and 1 exhibited significantly improved re-epithelialisation and regeneration of skin appendages compared with group 2 (Fig. 5d, e). Masson’s trichrome staining demonstrated that groups 3 and 1 had significantly greater collagen deposition than group 2 (Fig. 5d, f). Immunofluorescence analysis of α-SMA and Ki67 revealed significantly enhanced angiogenic capacity and cellular proliferation in groups 1 and 3 compared with group 2 (Fig. 5g–i). Immunofluorescence staining for p-MLKL revealed no significant differences among the three groups on day 4. However, by day 8, the p-MLKL-positive area was significantly reduced in groups 1 and 3, but it remained unchanged in group 2 (Fig. 5j, k). Additionally, we assessed the levels of tissue inflammation by detecting the expression of inducible nitric oxide synthase (iNOS) and arginase-1 (Arg1) in the wound tissue. The results revealed that iNOS expression was markedly upregulated in group 2 compared with that in groups 1 and 3 on day 8, whereas Arg1 expression followed the inverse pattern (Fig. 5l, m and Supplementary Fig. 11c, d).

Fig. 5. Agonal cell resuscitation by NPVs promotes tissue repair.

a Schematic representation of the diabetic wounds assay. b Representative images of wounds. (Scale bar, 2 mm). c Percentage of wound area at each time point relative to the original wound area (n = 6 independent mice; mean ± s.d.). d Haematoxylin and eosin and Masson trichrome staining of wounds (scale bars, 400 µm). Quantification of scar width (e) and collagen deposition area (f) (n = 6 independent mice; mean ± s.d.). g Immunofluorescence staining of α-SMA and Ki67 of wounds (scale bar, 50 µm). Quantification of blood vessels (h) and Ki67⁺ cells (i) at the wound sites (n = 6 independent mice; mean ± s.d.). Immunofluorescence staining and quantification of p-MLKL (j, k), Arg1, and iNOS (l, m) of wounds (scale bar, 50 µm; n = 6 independent mice; mean ± s.d.). n Schematic representation of the in vivo study design. o KEGG enrichment analysis of upregulated genes (fold change ≥ 1.5; p < 0.05) in diabetic wound (DW) treated with NPVs. p GSEA was applied to compare the Wnt signalling pathway gene sets between DW and DW treated with NPVs (NPVs). q Immunoblotting and quantification of active β-catenin in wounds (n = 6 independent mice; mean ± s.d.). r LC–MS/MS identification and quantification of PGE2 and N1-AcSPD in wound (n = 4 independent mice; mean ± s.d.). s Schematic representation of the study design. t Immunoblotting and quantification of active β-catenin in wounds (n = 6 independent mice; mean ± s.d.). u Schematic representation of study design. v Left: Representative images of wounds (Scale bar, 2 mm). Right: Percentage of wound area (n = 6 independent mice; mean ± s.d.). The samples derive from the same experiment and the blots in q and t were processed in parallel. P values were determined using a two-tailed one-way ANOVA with a Tukey post-hoc test (e, f, h, i, k, l, t), two-way ANOVA with Šidák’s multiple-comparison test (c, v), a two-tailed unpaired Student t test (q, r), one-sided Fisher’s Exact test (o), or Permutation test with False Discovery Rate multiple-comparison test (p).

Next, we collected tissue samples 6 days after wound induction and performed RNA-seq (Fig. 5n). The results showed that NPV intervention effectively reduced the expression of genes related to regulated necrosis, IL-6, and IFN-γ (proinflammatory cytokine) signalling pathways in DW. Concomitantly, the expression of genes associated with collagen synthesis, stem cell proliferation, and keratinocyte proliferation pathways was significantly enhanced (Supplementary Fig. 11e–g). Furthermore, the NPV-treated group exhibited upregulation of genes associated with the Wnt signalling pathway (Fig. 5o, p). Accumulating evidence demonstrates that the Wnt signalling pathway plays a critical role in regulating cell growth and differentiation, and its activation can effectively facilitate tissue repair^74–76. Western blot analysis further confirmed that NPV treatment significantly enhanced Wnt/β-catenin pathway activity in DW (Fig. 5q). PGE2 and polyamines can enhance the Wnt/β-catenin pathway through multiple mechanisms^77–80. Correspondingly, we found that NPV treatment significantly increased the levels of PGE2 and N1-AcSPD in wound tissues (Fig. 5r). Therefore, we speculate that NPVs promote tissue repair by driving the resuscitation of agonal cells; subsequently, the resuscitated cells secrete large amounts of PGE2 and N1-AcSPD to activate the Wnt signalling pathway in tissue cells. To verify this, we pre-injected diabetic mice with AAVDJ-shStx1a and AAVDJ-shStx4a to abrogate the membrane repair-dependent pro-resuscitation effect of NPVs (Supplementary Fig. 11h). The results showed that suppression of the membrane repair capacity of NPVs markedly inhibited Wnt/β-catenin signalling and delayed wound healing (Fig. 5s–v). Based on the above findings, we propose that NPVs primarily facilitate wound healing by enhancing the membrane repair capacity of agonal cells, promoting their resuscitation. Additionally, we compared the therapeutic efficacy of NPVs with that of the classical RIPK1 inhibitor Nec-1s. The results showed that NPVs exhibited superior wound healing effects compared with Nec-1s (Supplementary Fig. 11i, j). These findings align with our argument that Nec-1s suppresses both necroptosis and the subsequent secretion of pro-repair mediators by resuscitated cells, thereby accounting for its inferior efficacy relative to NPVs.

In addition to evaluating the therapeutic effects of NPVs on DW, we assessed their efficacy in treating ischaemic FLAP. After establishing a FLAP model in the dorsal region of ICR mice, we locally administered NPVs via intradermal injections and evaluated flap survival on day 7 post-surgery (Supplementary Fig. 12a). The results demonstrated that the survival area of the flap in the control group was only 44.9%, whereas the flap survival area in NPV-treated mice was significantly increased to 87.7% (Supplementary Fig. 12b). Laser Speckle Contrast Imaging (LSCI) revealed that the blood perfusion intensity was significantly higher in the NPVs group than in the control group (Supplementary Fig. 12c). Masson’s trichrome staining revealed a significant increase in collagen deposition in the NPVs group (Supplementary Fig. 12d, e). Immunofluorescence staining for α-SMA and Ki67 revealed a marked increase in the number of neovascularized and proliferating cells in the NPVs group (Supplementary Fig. 12f, g). Immunofluorescence staining for p-MLKL demonstrated a significant reduction in necroptosis in the NPVs group on postoperative day 7 (Supplementary Fig. 12h, i). Furthermore, immunofluorescence staining for Arg1 and iNOS showed that Arg1 expression was significantly upregulated, whereas iNOS expression was significantly downregulated in the NPVs group (Supplementary Fig. 12h, i). Moreover, we investigated the therapeutic potential of NPVs in SCI. Our previous studies demonstrated that NPVs can effectively target inflammatory injury sites¹⁸. The results showed that the NPVs group achieved significantly higher Basso Mouse Scale scores than the SCI group (Supplementary Fig. 12j, k). H&E staining further revealed that on day 28 post-injury, the NPVs group exhibited substantially smaller lesion areas than the SCI group (Supplementary Fig. 12l). Correspondingly, the NPVs group showed significantly less p-MLKL expression (Supplementary Fig. 12m).

The biological safety of NPVs in vivo was also evaluated. The results showed no significant changes in serum levels of CRP and IFN-γ, nor any marked increases in total WBC or neutrophil (NEU) counts after multiple NPV treatments (Supplementary Fig. 13a). Furthermore, H&E staining showed no structural abnormalities, and no significant differences in inflammatory cell infiltration at topical application sites on the skin and in vital organs, including the heart, liver, spleen, lung, and kidney, compared with the control group (Supplementary Fig. 13b).

In summary, this study demonstrates that during necroptosis, agonal cells actively take up EVs to repair their damaged PM. Unlike intracellular vesicles (e.g., lysosomes) that rely on endogenous cellular phospholipids for membrane repair, EVs do not deplete cellular phospholipid reserves, thereby significantly enhancing cellular membrane repair capacity. Among various vesicles, artificial NPVs exhibit superior membrane repair efficiency. Correspondingly, NPV treatment effectively augments cellular membrane repair capacity primarily through SNARE-mediated membrane fusion, thereby driving agonal cell resuscitation. Furthermore, following necroptosis-driven activation of the Erk/MAPK signalling pathway, resuscitated cells release substantial amounts of PGE2 and N1-AcSPD, which in turn upregulate tissue Wnt/β-catenin signalling to ultimately facilitate tissue repair. Based on these findings, we propose the “agonal cell resuscitation strategy” as a therapeutic intervention targeting lytic RCD to facilitate tissue repair. By augmenting the cellular membrane repair capacity instead of blocking lytic RCD initiation, our strategy circumvents the inherent limitations of conventional small-molecule inhibitors, such as off-target toxicity and the therapeutic paradox. The NPVs used in this strategy are generated via straightforward physical manipulation of autologous platelets, thereby avoiding immunogenic risks. Therefore, this strategy demonstrates considerable potential for clinical application in the treatment of diseases associated with lytic RCD.

Methods

Plasmids

RIPK3 and Lamp1-zsGreen plasmids were synthesised by HANBIO (Shanghai, China). hMLKL^1-140 −2Fv and mAG-Gal3 plasmids were synthesised by HZREPOBIO (Hangzhou, China). The mito-AT1.03 plasmid was purchased from Beyotime (D2606) and subsequently cloned into the pCMV-puro vector by Tsingke Biotechnology (Beijing, China). The GCaMP6 plasmid was purchased from Miaoling Biotechnology (P84659), followed by in-house cloning into the pLVX-puro vector. The aforementioned plasmids are available from the corresponding author for academic research use only.

Cell culture

NIH3T3 (Cat.No. SCSP-515), 293T (Cat. No. SCSP-502) were purchased from the Cell Bank of the Type Culture Collection of the Chinese Academy of Sciences, Shanghai Institute of Cell Biology, Chinese Academy of Sciences. Primary mouse mesenchymal stem cells (MSC) were purchased from Oricellbio (MUBMX-01001, Guangzhou, China). The RIPK3-OE NIH3T3 and hMLKL^1-140 −2Fv-OE NIH3T3 cell lines were generated via lentivirus transduction. Subsequently, monoclonal cells with suitable expression levels were selected for subsequent experiments. Briefly, 293T cells were transfected with the indicated plasmid by using PolyJet™ (SignaGen, SL100688) for 72 h. Target cells were infected with filtered virus containing supernatants from packaging cells supplemented with 10 μg/mL polybrene. Infected cells were spun at 1600 × g for 60 min before being put back in the incubator. Stable transductants were selected at 72 h post-infection by puromycin (5 μg/ml, 48 h, MCE, CL13900). Cells were treated with 500 nM STS (MCE, HY-15141) for 5 h to induce apoptosis. The GCaMP6/RIPK3-OE NIH3T3, Mito-AT1.03/RIPK3-OE NIH3T3, Lamp1-zsGreen/RIPK3-OE NIH 3T3 and mAG-Gal3/RIPK3-OE NIH3T3 were generated by further transducing the RIPK3-OE NIH3T3 with the corresponding lentivirus and subjected to fluorescence-activated cell sorting (FACS) for selection.

Bone marrow cells were harvested from femora and tibiae of C57BL/6 mice and differentiated in the presence of recombinant mouse macrophage colony-stimulating factor (M-CSF; 20 ng/mL, R&D Systems) in complete MEMα (Procell) containing 10 mM glucose, 2 mM L-glutamine, 100 U/mL penicillin/streptomycin, and 10% foetal bovine serum (FBS) for 5 days for preparation of BMDMs.

Mouse tail tip fibroblasts (TF) were isolated by excising the tail tip (from six- to eight-week-old male C57BL/6 mice), mincing the tissue, and digested with 2.5 mg/mL collagenase I (dissolved in DMEM with 10% FBS, Merck, 1148089) at 37 °C for 1 h, followed by treatment with 10 μg/mL DNase I (Roche, 10104159001) for 15 min. Following digestion, cells were centrifuged to remove the enzyme-containing medium. The pellet was resuspended in DMEM with 10% FBS and cultured further for approximately 5–7 days until fibroblast confluence reached approximately 70%–80%. MEF, NIH 3T3, 293T and TF were cultured in DMEM (Pricell) with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin.

Flow cytometric quantification of cell death

To assess necroptosis, RIPK3-OE NIH3T3 were induced by adding 30 ng/ml TNF-α (SinoBiological, 50349-MNAE), 10 μM Lcl-161 (MCE, HY-15518) and 100 μM z-VAD-fmk (APExBIO, A1902) (TSZ) for the indicated times. Then, cells were trypsinized, resuspended in Annexin V binding buffer and stained with Annexin V-APC (1:200, Elabscience, E-CK-A117), 50 nM Sytox Green (Invitrogen, S7020) for 5 min. Cells with Annexin V⁺/⁻ or Sytox Green⁺/⁻ were analysed by flow cytometry using CytoFLEX LX (Beckman Coulter). The percentages of differently labelled cells were calculated by FlowJo software (Tree Star).

We employed a 6-mm full-thickness skin defect model to investigate the proportions of agonal and terminally dead cells at different time points post-injury. To minimise interference from early apoptotic cells, mice received intraperitoneal administration of 20 mg/kg z-VAD-FMK (APExBIO, A1902) on the day of wound induction and every two days thereafter⁷². At designated time points, wound tissues were collected and minced, followed by digestion with Dispase II (2U/ml, yeason, 40104ES60) and DNase I (40 U/ml, Roche) at 37 °C for 2 h. The tissue fragments were removed by filtration through a 70 μm strainer. Cells were collected by centrifugation at 300 × g for 5 min, resuspended in red blood cell lysis buffer, and incubated at 37 °C for 3 min to lyse erythrocytes. After centrifugation and one wash with PBS, the cells were resuspended in Annexin V binding buffer and stained with Annexin V–APC (1:200, Elabscience) and 50 nM Sytox Green (Invitrogen) for 5 min.

Cell death plate reader assay

The experimental steps for detecting cell death via fluorescence intensity of PI dye using a microplate reader were similar to those previously reported in the literature, as briefly described below⁸¹. First, 10,000 cells were seeded per well in a black-bottom 96-well plate, with or without the addition of NPV (75 μg/mL). After 16 h, the medium was replaced with HBSS (Procell) containing 10 mM HEPES (Procell), 10% FBS, and 1 μM PI (MCE, HY-D0815), and NPV (75 μg/mL) was added or omitted according to groups. For MEF cells, 30 ng/ml TNF-α, 10 μM Lcl-161 and 100 μM z-VAD-fmk were added to induce necroptosis, and fluorescence intensity of PI was measured using a microplate reader after 8 h. For BMDM cells, 1 μg/mL LPS (MCE, HY-D1056) was first added for 4 h, followed by 20 μM Nigericin (HY-127019) for 1.5 h to induce pyroptosis, after which PI fluorescence intensity was detected. For the positive control group, 0.05% Triton X-100 (Beyotime, ST797) was added to permeabilize the cell membrane. The microplate reader (BioTek) was set with excitation wavelength at 535 nm ± 10 nm and emission wavelength at 617 nm ± 20 nm.

Label liposomes or NPVs with fluorescent dyes

A total of 1 × 10¹¹ liposome particles (200 nm in diameter) were suspended in 4 mL PBS, followed by incubation with DiI (1 μM, Beyotime, C1036) at 37 °C for 15 min. The sample was subsequently washed twice by centrifugation (118,000 × g, 1 h), and the pellet was resuspended in 1 mL PBS. The DiI control group underwent the same procedure except that no liposomes were included in the solution. Both preparations were then added to cells at the same dilution ratio for subsequent experiments. For NPVs labelling, 1 × 10⁶ cells/ml platelets were resuspended in PBS containing 500 nM PGE1 (MCE, HY-B0131), incubated with 1 μM DiI at 37 °C for 15 min, and then washed twice by centrifugation (800 × g, 30 min). The pellet was resuspended in PBS containing 500 nM PGE1 and sequentially extruded six times through 0.4 μm and 0.2 μm filters (BIOLAND), respectively. The suspension was subsequently centrifuged at 25,800 × g for 1 h, and the final pellet was resuspended in PBS without PGE1. The DiI control was processed identically, but without platelets in the solution.

Flow cytometric quantification of cell uptake of liposomes or NPVs

First, necroptosis was induced in RIPK3-OE NIH 3T3 by adding TSZ for 3 h. DiI (1 μM, Beyotime, C1036)-labelled liposomes (1.5 × 10⁹particles/ml) were then added and co-incubated with the cells for 1 h. Subsequently, the cells were digested and stained with Annexin V-APC and Sytox Green. The MFI of DiI was measured in Annexin V-APC⁻ Sytox Green⁻ cells and Annexin V-APC⁺ Sytox Green⁻ cells using flow cytometry (CytoFLEX LX, Beckman Coulter).

To investigate the mechanisms of liposome uptake by cells, RIPK3-OE NIH3T3 were first exposed to 5 μM cytochalasin D (MCE, HY-N6682) for 30 min, followed by the addition of TSZ and DiI labelled liposomes, and the changes in the MFI of DiI were measured in Annexin V-APC⁺ Sytox Green⁻ cells by flow cytometry. For cells treated at 4 °C, the cells were first incubated at 4 °C for 30 min, then transferred to 37 °C and cultured in medium supplemented with TSZ for 3 h, and finally transferred to 4 °C and cultured with liposome for 1 h.

We employed a 6-mm full-thickness skin defect model to evaluate the in vivo targeting capability of NPVs toward agonal cells. On the day of wound induction, mice received an intraperitoneal injection of z-VAD-FMK (20 mg/kg) to minimise interference from early apoptotic cells. On day 1 post-injury, 150 μL DiI-NPV (75 μg/mL) were administered through combined topical application and intradermal injection. Wound tissues were collected 24 h later, minced, and digested with Dispase II (2 U/ml, Yeason) and DNase I (40 U/ml, Roche) to isolate cells for flow cytometry analysis.

RNA-seq

The total RNA of agonal cells and control (with three samples for each group and 10⁶ cells for each sample) was extracted using TRIzol lysis reagent (Invitrogen, 15596018CN) according to the manufacturer’s instructions (agonal cells were obtained by Annexin V⁺ Sytox Green⁻ cell sorting, CytoFLEX SRT, Beckman Coulter). To investigate why resuscitated cells secrete enhanced levels of pro-repair factors, we established three experimental groups. The control group consisted of untreated MEFs. The C+NPVs group was prepared by incubating normal MEFs with 75 μg/ml NPVs for 16 h. For the Res group, MEFs were treated with TSZ for 2.5 h, then digested and sorted by FACS to obtain agonal cells. These agonal cells were subsequently incubated with 75 μg/ml NPVs for 16 h. After medium replacement to remove non-adherent, terminally dead cells, the remaining adherent resuscitated cells were collected as the Res group. Sequencing libraries were generated using VAHTS Universal V.6 RNA-seq Library Prep Kit for Illumina (NR604-01/02) following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. Genes with P-values of <0.05 and absolute fold changes of ≥1.5 were identified as differentially expressed genes. The FPKM score was calculated on the basis of the number of transcribed fragments. KEGG and GO enrichment analysis was performed using the DAVID online database (https://davidbioinformatics.nih.gov).

Analysis of scRNA-seq data from six human diseases associated with necroptosis

The six scRNA-seq datasets of necroptosis-related diseases are publicly available, and the detailed access methods are specified in the “Data availability” section. Meanwhile, these data were reanalysed by Seurat (Version 4.1.0) in R (version 4.0.5). Except the dataset of chronic wound, the primary annotated cell types of the selected cells were identified based on the metadata.

The primary annotated cell types of chronic wound were identified based on the marker gene listed in Supplementary Table. 1. Subsequently, Elbow plots were used to select appropriate principal components, and the selected cell (INH, hepatocyte, PTE, AT2, cardiomyocytes and fibroblast) were subsequently reintegrated based on these components, respectively. To view the overall distribution pattern of selected cells, the Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction algorithm was employed. Then, we scored gene sets using the ssGSEA method from the R package scGSVA (https://github.com/guokai8/scGSVA). We performed Pearson correlation analysis on the NES of the corresponding gene sets, with a confidence interval of 0.95. Based on the NES of programmed necrotic cell death gene sets, we categorised the selected cells into three subclusters: normal, stressed and agonal. The mean NES of indicated gene sets were illustrated by the VlnPlot function in Seurat. Statistical analyses between the two groups were carried out by using the Kruskal–Wallis test.

Isolation of EVs from tissues microenvironment

To inhibit vesicle uptake, prochlorperazine (1 mg/kg, Sigma, P9178) was administered intraperitoneally every two days. To suppress EVs secretion, GW4869 (5 mg/kg, MCE, HY-19363) and pantethine (300 mg/kg, MCE, HY-B1028) were injected intraperitoneally on the same schedule. A full-thickness skin defect with a diameter of 6 mm was created on the back of ICR mice using a punch biopsy tool. The wound tissue and approximately 2 mm of surrounding skin tissue were collected. After weighing and recording the normal skin tissues and wound tissues, the tissues were minced and subjected to digestion with 2 mg/ml collagenase D (Roche, 11088858001) and 40 U/ml DNase I at 37 °C for 30 min. Following digestion, the mixture was filtered through a 70μm sterile filter. The filtrate was then centrifuged at 300 × g for 10 min at 4 °C, and the supernatant was further centrifuged at 2000 × g for 20 min. Cells and cell debris were removed by these two steps. Then, the pellet was discarded, and the supernatant was ultrafiltered through 100 kDa filters (Millipore) at 3500 × g and 4 °C for 30 min. Finally, the retentate was diluted with PBS and centrifuged at 118,000 × g for 2.5 h at 4 °C. The resulting pellet were EVs and was resuspended in PBS for further use. The protein content of EVs was determined using a BCA assay kit (Beyotime, P0009) according to the manufacturer’s instructions.

Nanoflow cytometry

The content of indicated proteins on EVs was detected using the CytoFLEX s (Beckman Coulter). Through the VSSC channel of the cytoFlex s flow cytometer, we can detect EVs with a diameter of 65 nm or larger⁸². All vesicles were first labelled with the membrane dye 500 nM STAR 580-DOPE (abberior, ST580). To detect the origin of EVs, they were separately incubated with CD41-FITC (1:100, biolegend, 133903), CD45-FITC (1:100, biolegend, 147709) and CD235a-FITC (1:100, biolegend, 116707) on ice for 60 min. To determine the total content of VAMP2 and VAMP7 in vesicles, vesicles suspended in PBS were co-incubated with 0.05% Tween 20 (FD0020, FUDE) at 4 °C for 20 min to enhance membrane permeability. Subsequently, Human TruStain FcX™ (1:20 dilution, BioLegend, Cat. No. 422301) was added to the vesicles for blocking at 4 °C for 10 min. After blocking, the primary antibodies for VAMP2(1:200, CST, 13508) and VAMP7(1:500, ProteinTech, 22268-1-AP) were added, and the mixture was incubated at 4 °C for 1 h. Following incubation, the samples were centrifuged at 118,000 × g for 1 h; the supernatant was discarded, and the pellet was resuspended in PBS. FITC-conjugated secondary antibody (1:100 dilution, ProteinTech) and the membrane dye 500 nM STAR 580-DOPE were then added to the vesicles, followed by incubation at 4 °C for another 1 hour. Finally, the samples were centrifuged again at 118,000 × g for 1 h, the supernatant was discarded, and the pellet was resuspended in PBS. Nanoflow cytometric analysis was performed using a CytoFLEX S flow cytometer (Beckman Coulter). If only the content of N-terminus-exposed VAMP2 and VAMP7 in vesicles is to be determined, the step of adding 0.05% Tween 20 to enhance membrane permeability is omitted, with all other steps performed as described above.

PMPs, RBC-EVs, ApoVs and PS⁺ liposome generation

The preparation of PMP was similar to previously reported methods⁸³. Briefly, inactivated platelets obtained by centrifugation were resuspended in PBS containing 2 mM CaCl₂ (Solarbio, C7250). Then, 2 μM calcimycin (MCE, HY-N6687) was added, and the mixture was incubated on a shaker at 37 °C for 1 h. Afterwards, the sample was centrifuged at 2000 × g at 4 °C for 20 min, the pellet was discarded, and the supernatant was collected. Then, the supernatant was ultrafiltered through 100 kDa filters (Millipore) at 3500 × g and 4 °C for 30 min. The retentate was diluted with PBS and centrifuged at 118,000 × g at 4 °C for 2.5 h. The resulting pellets were PMPs, which were resuspended in PBS and stored for further use. The preparation process for RBC-EVs was similar to that of PMPs³⁹, with the exception that the concentration of calcimycin was adjusted to 10 μM and the incubation time was extended to 16 h.

The preparation of ApoVs was achieved by multiple centrifugations of the culture medium of apoptotic cells⁸⁴. Briefly, 4 days prior to inducing apoptosis in mouse MSCs, the serum was replaced with EV-free FBS. When the mouse MSCs reached 80% confluence, the medium was switched to serum-free DMEM, and 250 nM staurosporine (MCE, HY-15141) was added. The cells were incubated for 16 h. Afterwards, the medium was collected and centrifuged at 300 × g at 4 °C for 10 min, and the supernatant was taken. This supernatant was centrifuged again at 2000 × g for 20 min at 4 °C. Then, the supernatant was ultrafiltered through 100 kDa filters at 3500 × g and 4 °C for 30 min. The resulting retentate was diluted with PBS and centrifuged at 16,000 × g at 4 °C for 30 min. The resulting pellets were ApoVs, which were resuspended in PBS and stored for further use.

PS⁺ liposomes were prepared using a modified thin-film hydration method as previously described³². Lipid composition was optimised as 1,2-Distearoyl-sn-glycero-3-phosphocholine (DSPC) (Solarbio, ID3300)/cholesterol (Solarbio, C8280)/1,2-distearoyl-sn-glycero-3-phospho-L-serine (DSPS) (Avanti Polar Lipids, 840029) (40:40:20 molar ratio). Briefly, lipids were dissolved in chloroform and subjected to rotary evaporation to form a homogeneous lipid film. Hydration was achieved by rehydration with PBS buffer, followed by sonication. The resulting mixture was then extruded three times through a 200 nm filter membrane to obtain uniform liposomes.

Nanoparticle tracking analysis (NTA)

NTA was performed on a Nanosight NS500 system (Malvern Panalytical) with a 488 nm laser. Following daily calibration, samples diluted in PBS were analysed at 25 °C. All samples were recorded under standardised settings. Each sample underwent 3 cycles (1 s/cycle) across 11 positions.

Transmission electron microscopy

EV samples were examined using a Tecnai G2 spirit 120 kV transmission electron microscope (Thermo FEI) following negative staining with uranyl acetate-oxalate solution for 5 min.

NPVs preparation and characterisation

According to previously reported methods, platelets were isolated from the blood of volunteers¹⁸. The collection, processing and research use of all blood specimens involved in this study were reviewed and formally approved by the Ethics Committee of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, and informed consent was obtained from all participants. Briefly, platelet-rich plasma (PRP) derived from healthy donors was first centrifuged at 100 × g for 10 min, and the supernatant was collected. This supernatant was then centrifuged at 900 × g for 30 min, and the pellet was resuspended in PBS containing 500 nM PGE1, resulting in a suspension of platelets (PLT). To prepare NPV, the PLT suspension (1 × 10⁶ cells/ml) was extruded sequentially through polycarbonate filters with pore sizes of 0.4 μm and 0.2 μm. Each extrusion step was performed using a mini-extruder, and the process was repeated 6 times for each step. After extrusion, the solution was centrifuged at 25,800 × g for 1 h at 4 °C, and the resulting pellets were the NPVs. The diameter of NPV was determined using Nanosight NS500 (Malvern Panalytical, UK). NPVs with or without centrifugation were stored at 4 °C for different durations, then stained with Annexin V-APC (1:200) and analysed for phosphatidylserine (PS) positivity via flow cytometry.

siRNA

All siRNAs were purchased from Genepharma (Shanghai, China). The sequence of siStx1a was sense 5′-GAGCUCAUGUCGGACAUUATT-3′, antisense 5′-UAAUGUCCGACAUGAGCUCTT-3′. The sequence of siStx4a was sense 5′-GGGAUAACAUCUCAGACGATT-3′, antisense 5′-UCGUCUGAGAUGUUAUCCCTT-3′. The sequence of siSCR was sense 5′- UUCUCCGAACGUGUCACGUTT-3′, antisense 5′- ACGUGACACGUUCGGAGAATT-3′. siRNA transfection was performed using Lipofectamine RNAiMax (Invitrogen, 13778030). qPCR was performed 48 h post-transfection, and Western blot analysis was conducted 72 h post-transfection to validate the knockdown efficiency of siRNA.

Assaying membrane repair kinetics following B/B induced Injury

The experimental procedure for assessing cellular membrane repair kinetics using calcein was similar to previously reported literature⁴⁰. hMLKL^1-140 −2Fv-OE NIH3T3 were first co-incubated with the indicated vesicles (1 × 10¹⁰ particles/ml) for 24 h. Subsequently, the vesicles were removed, and 1 μM calcein-AM (Dojindo, C326) was added for staining at 37 °C for 10 min. After staining, the cells were washed twice with cell imaging buffer (CIB: HBSS with 10 mM HEPES, pH 7.4). In subsequent experiments, the calcein loaded cells were maintained in CIB and placed in TCS SP8 X incubator (Leica, Germany), with conditions set at 37 °C and 5% CO₂. Subsequently, 150 nM B/B (MCE, HY-13992) was added to induce hMLKL^1-140 −2Fv oligomerization, leading to pore formation in the plasma membrane. By this method, when PM is damaged, intracellular impermeable calcein leaks out until the damage is repaired. Thus, we can reflect the cellular membrane repair capacity by tracking the temporal changes in calcein fluorescence. Live-cell fluorescence imaging was performed using the TCS SP8 confocal microscope (Leica, Germany), with images captured every 15 s. Finally, the ratio of the average fluorescence intensity at each time point to the initial fluorescence intensity (∆F/F) was calculated to reflect the cellular membrane repair kinetics. To observe the chronological order of calcein leakage and nucleic acid dye entry following lethal membrane damage, cells were co-treated with 150 nM B/B and 100 nM Sytox orange, with imaging performed at 15-s intervals.

To verify whether NPVs enhance cellular membrane repair capacity through SNARE-mediated membrane fusion, hMLKL^1-140 −2Fv-OE NIH3T3 were first transfected with the indicated siRNA. After 56 h, 75 μg/ml NPV (measured by BCA) was added to the cells, followed by incubation for 16 h. Subsequently, Calcein-AM staining was performed, and 100 nM B/B was added to assess the membrane repair kinetics via live-cell fluorescence imaging.

Immunogold staining for electron microscopy

To observe the distribution of NPVs in agonal cells, RIPK3-NIH3T3 cells were co-incubated with 75 μg/ml NPVs (measured by BCA) and TSZ for 4 h. Agonal cells were then sorted using flow cytometric sorting (CytoFLEX SRT, Beckman Coulter). For the normal group, cells were co-incubated with 75 μg/ml NPVs for 4 h. After incubation, the cells were washed three times with PBS to remove residual NPVs. Subsequently, the cells were fixed with fixative (HaoKe) (4% paraformaldehyde and 0.1% Glutaric dialdehyde in 0.1 M phosphate buffer) for further transmission electron microscopy (TEM) analysis. The immunogold labelling was performed as previously described³⁸. After fixation, the cells were first incubated with blocking solution (0.1 M PBS pH 7.4 + 0.1% BSA) for 30 min. Anti-CD41 antibody (1:50, abcam) was then added and incubated on a shaker at 4 °C overnight. Following this, 1.4 nm Nanogold®-lgG (1:50, Nanoprobes, 2003-1 ML) was added and incubated on a shaker at 4 °C overnight. The cells were then subjected to silver enhancement for 2 min, followed by gradual dehydration and embedding for subsequent analysis. Ultrathin sections (70 nm) were cut and placed onto 150 mesh nickel grids with a Formvar membrane after polymerised. The images were acquired by in an Tecnai G2 Spirit 120 kV transmission electron microscope (Thermo FEI, The U.S.).

Total internal reflection fluorescence (TIRF) imaging

To monitor whether NPV exhibits exocytosis, we performed live-cell TIRF imaging based on previously reported methods⁴⁰. hMLKL^1-140 -2Fv-OE NIH 3T3 were first co-incubated with 75 μg/ml DiI-labelled NPV (measured by BCA) for 16 h. Subsequently, the cells were washed three times with CIB, and all subsequent experiments were conducted with cells maintained in CIB. After placing the cells in the live-cell imaging station, 100 nM B/B was added to induce membrane injury. Live-cell TIRF imaging was then performed using the Olympus IX83 microscope (Olympus, Japan), with images captured at a rate of 2 frames per second. DiI labelled NPVs were briefly assessed for their fluorescence characteristics in the TIRF field. The ratio of the fluorescence intensity (peak or total) at each time point to the initial fluorescence intensity (∆F/F) was calculated to reflect the change in fluorescence intensity. We use the Akima spline method to fit the curve.

Detection of NPV escape from the endolysosomal system

After adding TSZ to the mAG-Gal3/RIPK3-OE NIH 3T3 for 2 h, the medium was replaced with CIB containing TSZ, and live-cell imaging was performed using the TCS SP8 confocal microscope, with images captured every 2 min. To monitor potential changes in lysosomal pH following necroptosis induction, RIPK3-OE-NIH 3T3 cells were treated with TSZ for 3 h to induce necroptosis, followed by incubation with 1 μM Lysosensor (Invitrogen, L7535) for 5 min at 37 °C. After replacing the medium with CIB, the cells were imaged under a confocal microscope (Nikon A1).

To detect the co-localisation of NPV with lysosomes, 75 μg/ml DiI-labelled NPV (measured by BCA) was co-incubated with Lamp1-zsGreen/RIPK3-OE NIH3T3, either treated or untreated with TSZ for 3 h. Cells were then fixed with 4% paraformaldehyde, and fluorescence imaging was performed using the TCS SP8 confocal microscope. The Manders overlap coefficient between NPV and Lamp1 was calculated using ImageJ software v9.5.1.

Agonal cell resuscitation assay

When inducing necroptosis in the NPV group, NPVs (75 μg/ml) were added simultaneously with 30 ng/ml TNF-α, 10 μM Lcl-161 and 100 μM z-VAD-fmk. In the control group, only TSZ was added. After 4.5 h, RIPK3-OE NIH3T3 were digested, stained with Annexin V-APC (1:100) and 50 nM Sytox Green, and then sorted for ANX V⁺ Sytox Green^- cells (agonal cell) by CytoFLEX SRT. Agonal cells from the NPV group were further incubated with NPVs (75 μg/ml), while the control group was incubated with complete medium alone. After 16 h, the cells were digested again, and cell death was analysed by flow cytometry. To verify that NPVs resuscitate agonal cells via SNARE-mediated membrane fusion, RIPK3-OE NIH3T3 were transfected with the indicated siRNA for 56 hours, followed by treatment with TSZ to induce cell death. Subsequent procedures were performed as above. For MEF, the cell death was induced by TSZ for 2.5 h, after which cells were sorted, treated and analysed as above. MEFs that had undergone resuscitation were cultured for an additional 7 days, followed by senescence-associated β-galactosidase staining using a commercial kit (Beyotime, C0602) according to the manufacturer’s instructions.

Clonogenic growth

Agonal cells were seeded into 12-well plates with or without NPVs, at a density of 50000 cells per well. After 16 h, NPVs were removed and replaced with complete medium without NPV. After 7 days of culture, cells were stained with crystal violet (Solarbio, G1062). The plates were scanned using a document scanner and images are shown. After crystal violet staining, the dye absorbed by cells was dissolved with 95% ethanol, and the absorbance at 590 nm (the peak absorption wavelength of crystal violet) was measured using a microplate reader. MEF cells undergoing one or three rounds of resuscitation were plated at a density of 10,000 cells/well in 12-well plates. After 10 days of culture, crystal violet staining was performed.

Subcutaneous tumour formation assays

3 × 10⁶ MEF undergoing one or three rounds of resuscitation were prepared in 100 μl PBS supplemented with 100 μl Matrigel (Corning, 354234) and injected subcutaneously into 4-week-old female BALB/c nude mice. The volume of the tumour was measured on day 16.

IncuCyte analysis

Cells from the respective sources were first seeded into a 96-well plate overnight at a density of 10,000 cells per well. The following day, TSZ and 50 nM Sytox Green were added simultaneously. The cells were then transferred to the IncuCyte live-cell imaging system (Sartorius, Germany), and images were captured every 15 min. The SytoxGreen labelled cells were quantified by the IncuCyte FLR.

Supernatant generation for in vitro assay

MEF were seeded into 6-well plates at a density of 1 million cells per well and cultured in DMEM containing 10% FBS on the first day. On the second day, the serum-containing DMEM was removed, and the cells were washed three times with PBS before being replaced with 1 ml of serum-free DMEM. After 24 h, the culture medium was collected, referred to as MC. For MD supernatant, on the second day, TSZ was added for 2 h, followed by replacement with of 1 ml of serum-free DMEM. After 24 h, the medium was collected. For MP supernatant, NPVs (75 μg/ml) were added on the first day and incubated for 16 h. The NPV-containing medium was then removed, and the cells were washed three times with PBS before being replaced with serum-free DMEM. After 24 h, the medium was collected. For the MR supernatant, 1.7 million sorted agonal cells derived from MEF were seeded into 6-well plates and treated with NPVs (75 μg/ml) simultaneously. After 16 h of incubation, the NPVs-containing medium was removed, and the cells were washed three times with PBS (remove terminally dead cells). Serum-free DMEM was then added, and the medium was collected after 24 hours.

After collecting the four types of supernatants, they were first centrifuged at 300 × g for 5 min at 4 °C. The supernatants were then collected and centrifuged again at 3000 × g for 5 min at 4 °C. The final supernatants were used for subsequent in vitro experiments. When used for in vitro wound healing assays, the supernatants were supplemented with 10% FBS. For co-culture with macrophages, the supernatants were supplemented with 10% FBS and 20 ng/mL M-CSF. If separation of different components in the supernatants was required, the supernatants were added to 3 kDa centrifugal filters and centrifuged at 2000 × g for 30 min at 4 °C. We evaluated the separation efficiency of ultrafiltration using a commercial silver staining kit (Beyotime, P0017S), following the manufacturer’s protocol. The concentration of PGE2 in the supernatant was determined using a PGE2 ELISA kit (Elabscience, E-EL-0034) according to the manufacturer’s instructions.

In vitro wound healing assay

In a 12-well plate, 2-well culture inserts were placed. Each side of the insert was seeded with 8000 primary mouse TF. Once the confluence of cells on both sides reached 80%–90%, the inserts were removed, and the medium was replaced with the indicated supernatant. The closure of the wound created by the removal of the insert was observed and photographed using CKX53 microscopy (Olympus, Japan).

qPCR

Total RNA was extracted from the cells using an RNA extraction kit (Qiagen). The RNA was then converted into cDNA using reverse transcription enzymes (Promega, Madison). RT-qPCR assays (catalogue no. A6002, Promega) were conducted following the manufacturer’s protocol. The primer sequences for the indicated genes are provided in the Supplementary Table  2. The qPCR experiments were performed on a Rotor-Gene Q thermal cycler (Qiagen). β-Actin was selected as the reference gene for normalisation.

Western blot analysis

Protein samples were separated via SDS-polyacrylamide gel electrophoresis (SDS-PAGE), and then electrotransferred onto nitrocellulose membranes. The membranes were blocked with 10% skim milk dissolved in Tris-buffered saline containing 0.05% Tween-20 (TBST) for 1 h at room temperature. The following primary antibodies were incubated: anti-CD42b (Proteintech, 67857-1-Ig, 1:5000), anti-CD41 (abcam, ab134131, 1:1000), anti-STX4 (ABclonal, A5996, 1:1000), anti-STX1A (abcam, ab272736, 1:1000), anti-pMLKL (Ser345) (CST, 37333, 1:1000), anti-MLKL (CST, 37705, 1:1000), anti-PARP (CST, 9542, 1:1000), anti-Cleaved Caspase-3 (CST, 9661, 1:1000), anti-β tubulin (proteintech, 66240-1-Ig, 1:20000), anti -α-SMA (proteintech, 80008-1-RR, 1:20000), anti-β actin (proteintech, 66009-1-Ig, 1:20000), anti-active β catenin (CST, 8814, 1:1000), anti-β catenin (proteintech, 66379-1-Ig, 1:5000), anti-Erk1/2 (CST, 4695, 1:1000), anti-pErk1/2 (Thr202/Tyr204) (CST, 4370, 1:2000), anti-^hypeIF5A (antibodysystem, RGK08101, 1:1000), anti-eIF5A (abcam, ab32443, 1:5000), anti-CD81(abcam, ab79559, 1:1000), anti-TSG101 (abcam, ab125011, 1:1000), anti-CANX (proteintech, 66903-1-Ig, 1:5000) and anti-Alix (proteintech, 67715-1-Ig, 1:4000) at 4 °C overnight. After washing steps, the membranes were incubated with the corresponding horseradish peroxidase (HRP)-conjugated secondary antibody (1:5000 dilution, Diagbio) for 1 h at room temperature. Protein bands were visualised using enhanced chemiluminescence (ECL) reagents (FDbio) and captured with a chemiluminescence imaging system. The gray intensity of the target bands was quantified using ImageJ (Fiji) software. The uncropped images of Western blots are provided in the Source data.

Untargeted metabolomics

Samples were transferred to EP tubes, and 4-fold extraction solution (methanol: acetonitrile = 1:1, v/v) containing isotopically labelled internal standards was added. The samples were then incubated at –40 °C for 1 h. Subsequently, the samples were centrifuged at 4 °C at 13,800 × g for 15 min. The supernatants were collected and transferred to vials for instrumental analysis. Additionally, equal volumes of supernatants from all samples were pooled to prepare quality control (QC) samples for analysis. For polar metabolites, Vanquish ultra-high-performance liquid chromatography (UHPLC) system (Thermo Fisher Scientific, the U.S.) was employed. Separation of target compounds was achieved using a Waters ACQUITY UPLC BEH Amide column (2.1 × 100 mm, 1.7 μm). For non-polar metabolites, this study utilised Vanquish UHPLC system equipped with a Phenomenex Kinetex C18 column (2.1 × 100 mm, 2.6 μm) for chromatographic separation of target compounds.

The Orbitrap Exploris 120 mass spectrometer was operated under the control of Xcalibur software (version 4.4, Thermo Fisher Scientific) to acquire both MS and MS/MS data. Metabolomics was performed with the assistance of Biotree (Shanghai, China). The raw data were converted into mzXML format by ProteoWizard software, and metabolites were identified based on the BiotreeDB (V3.0) database⁸⁵. The data were subsequently imported into SIMCA 18.0.1 software for multivariate analysis. Metabolites with VIP > 1 and p < 0.05 (student t test) were considered as differential metabolites.

Animals

Four-week-old female BALB/c nude mice, 6-week-old female C57BL/6J and 8-week-old male ICR mice used in this study were purchased from Gempharmatech. Animals were housed in groups of 4–6 mice per individually ventilated cage in a 12 h light-dark cycle (06:30-18:30 light; 18:30-06:30 dark), with constant room temperature (21 ± 1 °C) and relative humidity (40–60%). Animals had access to food and water ad libitum. All animal studies were performed according to the ethical regulations and protocols approved by the Institutional Animal Care and Use Committee of Zhejiang University.

Diabetic wound model

The establishment of Streptozotocin (STZ)-induced diabetic mice and a 6-mm full-thickness skin defect wound model was performed as previously reported¹⁸. Briefly, 8-weeks-old male ICR mice were intraperitoneally injected with 100 mg/kg STZ (sigma, S0130) for two consecutive days. Two weeks after the first STZ injection, if the blood glucose level of the mice exceeded 16.7 mM under normal conditions, the diabetic model was considered successfully established. Subsequently, a 6-mm circular full-thickness skin defect was created in the central dorsal skin of ICR mice using a skin punch. For the NPV group, 150 μL NPV (75 μg/mL) was administered using a combined approach of topical application and intradermal injection around the wound on days 0, 2, 4 and 6 after wound modelling. From day 0 to day 12 post-wounding, wound and ruler images were captured every 2 days, and the wound area was analysed using ImageJ. Nec-1s (10 mg/kg, MCE, HY-14622A) was administered via tail vein injection on days 0, 2, 4 and 6 after wound modelling⁸⁶. Additionally, 1 × 1 cm² square skin tissues centred on the wound were collected on days 4, 6, 8 and 12 post-wounding for subsequent histological analysis, immunofluorescence studies, RNA-seq and LC–MS/MS. For LC–MS/MS, the tissue samples were pulverised and resuspended in PBS (1:9, with 1% protease inhibitor) and fully dissolved via ultrasonication for 5 minutes. The sample was then centrifuged at 12,000 × g for 10 min and the suspension was transferred to a new collection tube.

To knockdown STX1A and STX4 in skin tissue, we performed grid-pattern intradermal injections in a 2.5 × 2.5 cm² area on the dorsal skin of mice, with a total of 5 injection points, each receiving 100 μL of AAVDJ-shStx1a and AAVDJ-shStx4 at a titre of 5 × 10¹¹  vg/mL. AAVDJ-shStx1a-mCherry and AAVDJ-shStx4-mCherry were constructed by Genechem (Shanghai, China). The sequence of shStx1a was sense 5′-GAGCUCAUGUCGGACAUUATT-3′, antisense 5′-UAAUGUCCGACAUGAGCUCTT-3′. The sequence of shStx4a was sense 5′-GGGAUAACAUCUCAGACGATT-3′, antisense 5′-UCGUCUGAGAUGUUAUCCCTT-3′. The inserted sequence of AAVDJ-ctrl was 5′-TTCTCCGAACGTGTCACGT-3′. The shRNA was cloned into the GV865 vector and driven by the U6 promoter for expression. Three weeks after AAVDJ injection, skin tissue was harvested, and immunofluorescence staining for mCherry was performed to assess the expression efficiency of AAVDJ.

Random-pattern skin flap model

The skin flap model was established following a method similar to previously reported studies¹⁵. Under aseptic conditions, full-thickness skin on the back of 8-week-old male ICR mice was incised. The short edge near the tail was preserved, and a skin flap (1.5 × 3.5 cm²) was elevated. All vessels connecting the flap to the underlying skin were completely severed. The flap was then reinserted into the donor bed and sutured in place using 4–0 non-absorbable sutures. Postoperatively, mice received intramuscular injections of penicillin (30,000 U/kg) daily to prevent infection. For the NPV group, 200 μL of NPV (75 μg/mL) was administered locally via intradermal injection at the flap site on days 0, 2 and 4 post-surgery (three injection points per application). On day 7 post-surgery, blood perfusion of the flap was assessed using a laser speckle imaging system (RFLSI III, RWD Life Science, China). Additionally, the flap was divided into three regions: area I, representing the proximal viable tissue; area II, the transitional zone between viable tissue and visibly necrotic tissue; area III, the distal apparently necrotic tissue. For subsequent histological analysis and immunofluorescence staining, 1 × 0.5 cm² samples of area II were collected.

Spinal cord clip compression injury model

The spinal cord clip compression injury model was established as previously described⁸⁷. Six-week-old female C57BL/6J mice are anaesthetised and placed in a prone position. A dorsal midline incision is made to expose the T9-T11 vertebrae. Following a T10 laminectomy, the dura mater is kept intact. A 30-g aneurysm clip is applied vertically to the exposed spinal cord for 1 min. The clip is then removed, and the wound is closed in layers. Tail vein injection of 75 μg/ml NPVs (150 μl, in saline) was administered to mice in the NPVs group every two days, starting on day 3 post-injury. Both the SCI and sham groups received an equivalent volume of saline on the same schedule. Post-operative care includes manual bladder expression twice daily until reflex bladder recovery. Functional recovery is assessed using the Basso Mouse Scale. The hind limb motor function was observed and evaluated by two unbiased experts. The BMS scoring ranges from 0 (complete paralysis) to 9 (normal hind limb motor function). The final BMS score was calculated based on the average score given by the two experts. On days 14 and 28 post-injury, a 1-cm spinal cord segment centred on the lesion site was collected for subsequent H&E staining and immunofluorescence analysis.

In vivo biological safety experiments

One hundred fifty microliter NPV (75 μg/mL) was administered using a combined approach of topical application and intradermal injection around the wound on days 0, 2, 4 and 6. The control group received an equivalent volume of saline administered in the same manner. On day 12, blood samples and major organs were collected from the mice for subsequent experiments.

Histological analysis and immunostaining

Following fixation with 4% paraformaldehyde for 24 h, the tissues were processed for paraffin embedding. Serial sections were prepared and subjected to H&E staining. In the case of skin tissues, the scar width was measured. Additionally, skin sections were stained with Masson trichrome (Servicebio) to evaluate collagen deposition. Immunohistochemistry was performed to detect α-SMA (1:500, proteintech, 80008-1-RR) and Ki67 (1:1600, CST, 9449), enabling the assessment of micro vessel formation and proliferative activity of cells. Inflammatory activity was analysed via immunofluorescence staining for iNOS (1:400, CST, 13120) and Arg1 (1:400, ProteinTech, 66129-1-Ig). Necroptosis level was analysed via immunofluorescence staining for pMLKL(1:1600, CST, 37333).

Statistical analysis

Descriptive statistics are presented as the mean ± standard deviation (mean ± s.d.), mean ± standard error of the mean (mean ± s.e.m.) or mean ± 95% confidence interval. Statistical significance between two independent groups was determined using a two-tailed, unpaired Student’s t test. For multiple group comparisons, one-way analysis of variance (ANOVA) with Tukey’s two-sided post hoc test or two-way ANOVA with Šidák’s (or Tukey’s) two-sided multiple-comparison test was performed, as appropriate. For integer data, comparisons between two groups were performed using the Mann-Whitney U test, and comparisons among multiple groups were conducted using the Kruskal-Wallis test. The sample sizes (biological replicates), specific statistical tests and main effects of our statistical analyses for each experiment are detailed in each figure caption. A p-value less than 0.05 was considered statistically significant. All statistical analyses were conducted using GraphPad Prism software.

Ethics

Every experiment involving animals, or clinical samples have been carried out following a protocol approved by an ethical commission. Each participant gave informed written consent.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Z.H., P.C., and X.L. designed the project. Z.H., Q.W., Y.C., J.Z., Y.W., Y.Y., K.Y., and J.X. performed the experiments. M.L., H.Y., L.SG., A.Y., and C.G. provided technical help on this project. Z.H. and Y.C. made the visualisation. Z.H., Q.W., P.C., S.F., and X.L. wrote the manuscript. P.C., S.F., and X.L. supervised the project. All the authors analysed and interpreted the data, and reviewed and edited the manuscript before submission.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the article and its supplementary files. Any additional requests for information can be directed to and will be fulfilled by the corresponding authors. The five scRNA-seq databases of necroptosis-related disease are publicly available from GEO database (Alzheimer’s disease, GSE174367; Acute live failure, GSE223561; Acute kidney injury, GSE183279; Fatal COVID-19, GSE171524; Chronic wound, GSE176417 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE176417]). The single-nucleus RNA sequencing(snRNA-seq) data of acute myocardial infarction were downloaded from Cellxgene [https://cellxgene.cziscience.com/collections/8191c283-0816-424b-9b61-c3e1d6258a77]. The transcriptomics data generated in this study are available at NCBI BioProject under accession numbers PRJNA1426858, PRJNA1426900, and PRJNA1426918. The metabolomics data generated in this study have been deposited in the OMIX database with dataset identifier OMIX015266. The uncropped images of Western blots are provided in the Source data. Source data are provided with this paper.

Competing interests

Z.H., P.C., X.L., S.F., and C.G. have applied for patent (application number CN202610259166.8) related to this study. The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71653-z.