Glycoengineered extracellular vesicles released from antibacterial hydrogel facilitate diabetic wound healing by promoting angiogenesis Authors: Kewei Wang, Qin He, Mengmeng Yang, Qincheng Qiao, Jun Chen, Jia Song, Nan Zang, Huiqing Hu, Longqing Xia, Yingyue Xiang, Fei Yan, Xinguo Hou, Li Chen Published: 2024 ABSTRACT AbstractDiabetic wounds have become a global healthcare burden owing to impaired angiogenesis and persistent infections. Extracellular vesicles (EVs) can improve diabetic wounds, though their targeting ability is limited. In this study, we investigated the performance of a novel hydrogel dressing comprised of gelatin methacryloyl, glycoengineered EVs, and polylysine in treating infected diabetic wounds. High‐throughput single‐cell RNA sequencing (scRNA‐seq) and immunofluorescence staining revealed that E‐selectin (SELE) levels were higher in diabetic wounds than in non‐diabetic wounds. Mesenchymal stromal cells (MSCs) were transfected with a lentivirus containing fucosyltransferase VII (FUT7) and a CD63‐P19‐Nluc vector to enhance the expression of sialyl Lewis X (sLeX), the ligand of E‐selectin, on the surface of EVs (s‐EVs) derived from transfected MSCs (s‐MSCs). s‐EVs can target human umbilical vein endothelial cells (HUVECs) under lipopolysaccharide stimulation and promote the function of stimulated HUVECs in vitro. To promote and sustain the release of s‐EVs, we fabricated a gelatin methacryloyl (Gel)/poly‐L‐lysine methacryloyl (PL)‐5 hydrogel with good antibacterial ability, biocompatibility and mechanical properties. In a mouse experiment, s‐EV@Gel/PL‐5 exhibited excellent angiogenesis and anti‐inflammatory abilities and further promoted the healing of infected diabetic wounds. Our findings demonstrated the potential of the s‐EV@Gel/PL‐5 hydrogel in the clinical treatment of diabetic infectious wounds. INTRODUCTION With rapid societal and economic development, the incidence of diabetes is gradually increasing, and refractory diabetic wounds such as diabetic foot ulcers have attracted considerable attention as serious and costly chronic complications (Armstrong et al., 2023; McDermott et al., 2022). Angiogenesis is important for sustaining the blood supply required for skin repair. However, angiogenesis is inhibited in chronic diabetic wounds, leading to reduced capillary density and, in turn, reduced nutrient and oxygen supply, ultimately delaying wound healing (Yuan et al., 2023). Infection can further prolong diabetic wound healing and exacerbate amputation and mortality rates (Fang et al., 2022; Yu et al., 2023). Accordingly, more attention should be paid to the research and development of biocompatible materials that can promote angiogenesis and alleviate inflammation in the clinical management of infected diabetic wounds. Increasing evidence indicates that mesenchymal stromal cells (MSCs) and extracellular vesicles (EVs) derived from paracrine MSCs can improve diabetes and its complications (Song et al., 2023; Wang et al., 2018; Xiang et al., 2020). Human umbilical cord MSCs (HucMSCs) have low immunogenicity, making them ideal for clinical research and applications (Xie et al., 2020). In contrast to MSCs, EV‐based therapy is a cell‐free approach that avoids the risks of immune rejection and tumour formation after stromal cell transplantation, given the inability of EVs to self‐replicate. In addition, EVs are relatively stable and convenient to store and transport (Yuan et al., 2017). Various recent studies investigated the therapeutic effects of native MSC‐derived EVs and reported low specificity for target cells (Jafari et al., 2021; Lotfy et al., 2023). Therefore, several studies have aimed to enhance the therapeutic efficacy of EVs by glycoengineering EVs‐producing cells or EVs themselves (Phan et al., 2018; Zhang et al., 2023). In this study, we re‐analysed high‐throughput single‐cell RNA sequencing (scRNA‐seq) data from a previous study that investigated gene expression profiles at the single‐cell level in chronic foot ulcers of diabetic and non‐diabetic patients (Justynski et al., 2023). After a series of data processing steps, we performed a differential gene analysis of endothelial cells in the dermis and found that some endothelial injury‐related genes were differentially expressed between the diabetic and non‐diabetic groups. Based on the scRNA‐seq results and an extensive literature review, we identified E‐selectin as a key factor in EV engineering. E‐selectin, an adhesion molecule expressed by SELE, is rapidly and specifically expressed on the surface of endothelial cells in response to inflammatory stimulation (Chen et al., 1997). Sialyl Lewis‐X (sLeX) is expressed by leukocytes and binds to E‐selectin‐expressing endothelial cells under inflammatory stimulation (Pachón‐Peña et al., 2017). Therefore, we aimed to enhance the expression of sLeX on the surface of EVs by genetic engineering and thereby increase the ability of EVs to target the inflammatory vascular endothelium. Fucosyltransferase VII (FUT7) catalyses the transfer of L‐fucose to N‐acetylglucosamine at the distal unit of the alpha 2,3 sialylated lactosamine and participates in sLeX biosynthesis (Lowe, 2002; Shi et al., 2021). Previous studies have shown that exogenous FUT7 expression only slightly increased the cell‐surface expression of sLeX, whereas co‐expression of FUT7 and the sLeX protein vector PSGL‐1 significantly increased its expression (Zheng et al., 2022). Therefore, co‐expression of FUT7 and sLeX protein vectors is essential for increasing the cell‐surface expression of sLeX. The surface protein CD63 serves as an EV scaffold protein, and its large extracellular loop can be used for genetic engineering and the surface expression of functional components (Corso et al., 2019; Zheng et al., 2022). The large extracellular loop of CD63 was fused with P19 (19‐mer peptide), a short glycosylation domain derived from PSGL‐1. Consequently, we constructed a lentivirus encoding FUT7 and CD63‐P19, transduced the lentiviral vectors into MSCs, selected successfully transduced cells (s‐MSCs), and collected the supernatant to obtain the EVs (s‐EVs) secreted by the transfected cells. Biomedical hydrogels have good biocompatibility, mechanical properties and biodegradability. Hydrogels can preserve EVs and sustain their release, and represent effective modes for delivering EVs to disease sites (Zou et al., 2021). Gelatin methacryloyl (GelMA) is a photo‐crosslinked hydrogel formed by the methacrylation of gelatin. GelMA hydrogels share basic properties with the extracellular matrix (Yue et al., 2015). GelMA is widely used in the fields of tissue engineering and regenerative medicine (Wang et al., 2022). Poly‐L‐lysine (ε‐PL), polymerised from 25–35 lysine residues, is an antimicrobial polypeptide produced by Streptomyces. ε‐PL has a broad spectrum of antimicrobial activities against bacteria and fungi, such as Staphylococcus aureus (S. aureus) and Escherichia coli (E. coli) (Li et al., 2014). ε‐PL contains a large number of positively charged amino groups, which bind easily to the bacterial cell wall through electrostatic adhesion, resulting in the release of intracellular fluids and bacterial death (Yu et al., 2019). ε‐PL and its derivatives have good biocompatibility along with antibacterial and biodegradable properties and are widely used in the field of biological tissue engineering, such as for antibacterial wound dressings. Polylysine methacryloyl (PLMA) is also a photo‐crosslinked hydrogel modified by methylacryloylation of ε‐PL with glycidyl methacrylate to introduce double bonds in the ε‐PL molecule. Therefore, PLMA is expected to fulfil an antibacterial role in wound treatment. In this study, HucMSCs were used as EV sources. The glycan profile of the s‐EVs was engineered to target inflammatory vascular endothelium. Lipopolysaccharide (LPS)‐stimulated human umbilical vein endothelial cells (HUVECs) had a higher surface expression of E‐selectin, which mediates the uptake of glycoengineered EVs. We subsequently mixed s‐EV with the GelMA/PLMA hydrogel to develop a new dressing (s‐EV@Gel/PL) for infectious diabetic wounds to alleviate inflammation and promote angiogenesis (Scheme 1). Schematic illustration of multifunctional hydrogel fabrication for diabetic wound healing. (a) Design of plasmid and generation of modified EV. (b) Activation of endothelial cells and their uptake of modified EVs. (c) Fabrication of GelMA and PLMA. (d) Preparation of s‐EV@Gel/PL. (e) Application of hydrogel on infected diabetic wound. EVs, extracellular vesicles; GelMA, gelatin methacryloyl; PLMA, polylysine methacryloyl. MATERIALS AND METHODS Ethics approval statement All animal experiments were approved by the Institutional Animal Care and Use Committee of the Qilu Hospital of Shandong University (Approval No. DWLL2023‐133). The study involving human samples was approved by the Ethics Committee of the Qilu Hospital of Shandong University (Approval No. KYLL‐2022(ZM)‐898). All participants provided written informed consent for the collection of tissue samples, and all methods were performed in accordance with the Declaration of Helsinki. Cell culture Fresh umbilical cords were obtained from healthy full‐term newborns after obtaining written informed consent from the mothers. MSCs were isolated as previously described (Willis et al., 2020). Briefly, we removed the outer membrane and blood vessels of the umbilical cord and cut the Wharton's jelly into small pieces, which were then placed in cell culture flasks containing fresh α‐MEM (Gibco, Billings, MT, USA) supplemented with 10% EV‐free fetal bovine serum (FBS, Gibco). The medium was changed every 2 days. MSCs from passages 3–6 were used in this study. HUVECs at the third passage were purchased from Anwei‐sci Cell Center (Shanghai, China) and cultured in an endothelial cell medium (ECM; ScienCell, Carlsbad, CA, USA). The next passage of HUVECs was used in this study. All cells were cultured in a humid 37°C incubator at 5% CO2. Isolation and identification of EVs EVs isolated from MSCs (EVs) and s‐MSCs (s‐EVs) were isolated as previously described. Upon reaching 80%–90% confluence, the cells were rinsed with PBS and cultured in serum‐free medium. After 24 h of incubation, the medium was collected and EVs were extracted. The culture supernatant was centrifuged at 300 × g for 10 min, the supernatant was collected and centrifuged at 2000 × g for 20 min, further centrifuged at 10 000 × g for 30 min, and filtered through 0.22‐µm pore filters (MilliporeSigma, Burlington, MA, USA). The supernatant was subjected to ultracentrifugation (Optima XPN‐100; Beckman Coulter) at 100,000 × g for 70 min. The acquired pellets were rinsed with PBS before undergoing another round of ultracentrifugation to obtain purified EVs. Nanosight tracking analysis (NTA, ZetaVIEW Twin, Particle Metrix, Inning am Ammersee, Germany) was used to measure the size distribution and zeta potential of EVs. Transmission electron microscopy (HT‐7700, Hitachi, Tokyo, Japan) was used to observe EV morphology. Western blotting was used to detect the surface expression of the EV‐specific markers CD9 (Cat. No. CY5337; Abways, Shanghai, China), Alix (Cat. No. 12422‐1‐AP; Proteintech, Shanghai, China) and TSG101 (Cat. No. CY5985, Abways). Expression of sLeX (Cat. No. 551344; BD Pharmingen, San Diego, CA, USA) of EVs and s‐EVs was assessed by western blotting. Preparation and transfection of lentiviral vectors The well‐recognised EV scaffold protein CD63 has a large extracellular loop that can be fused with a potential sLeX carrier, P19 (Table S2), and a luciferase reporter, Nluc, at its C‐terminus, to construct the CD63‐P19‐Nluc carrier for the sLeX protein. The whole gene FUT7 and CD63‐P19‐Nluc template were synthesised and cloned into the multiple cloning site (MCS) of a GV348 (Ubi‐MCS‐SV40‐puromycin) lentiviral vector (GeneChem, Shanghai, China). The concentrated and purified lentiviruses were stored at –80°C until use. Lentiviral infection was induced in MSCs at the third passage. MSCs (3–5 × 104 cells/mL) were seeded in a T25 cell culture flask and cultured in a humid incubator at 37°C and 5% CO2 until a confluence of ∼20% was reached. MSCs were transfected with the lentivirus overnight. After reaching ∼80% confluence and growing well, the transfected MSCs were selected using puromycin (3 µg/mL; Cat. No. IA0530, Solarbio, Beijing, China) for 48 h. The surviving MSCs were further cultured with a medium containing puromycin (1 µg/mL). The medium was changed and collected every 2 days. The transfected cells were subjected to flow cytometry (DxFLEX, Beckman Coulter, USA) to identify surface markers, adipogenic and osteogenic induction, and stained to determine the differentiation capacity of the s‐MSCs. The effect of viral transfection on MSC viability was assessed using a CCK‐8 assay. The expression of sLeX (Cat. No. 563526, BD Pharmingen, San Diego, USA) on the surface of s‐MSCs was estimated by flow cytometry. Preparation of hydrogels Lithium phenyl‐2,4,6‐trimethylbenzoyl phosphinate (LAP) photo‐initiator standard solution was dissolved in PBS (0.25%, w/v) and stored in the dark. GelMA and PLMA solutions were passed through 0.22 µm pore filters to ensure sterility and mixed in different proportions with LAP solution to obtain 5% (w/v) GelMA (Gel), 5% (w/v) GelMA + 2% (w/v) PLMA (Gel/PL‐2), and 5% (w/v) GelMA + 5% (w/v) PLMA (Gel/PL‐5). The hydrogels were formed by crosslinking the precursor solution under ultraviolet radiation for 15 s. Animal models and treatments We injected 8‐week‐old C57BL/KsJ mice intraperitoneally with 150 mg/kg streptozotocin (Cat. No. S0130, Sigma‐Aldrich, St. Louis, MO, USA). Random blood glucose levels were measured twice, and mice with levels of 16.7 mmol/L were considered to have diabetes mellitus and included in the experimental group. Mice were anaesthetised with isoflurane (2%), their dorsal hair was shaved, and the underlying skin was sterilised with 75% ethanol. A full‐thickness wound (8 mm) was created on the dorsal skin using a skin biopsy perforator. A mixed solution of S. aureus and E. coli (1 × 108 CPU/mL) was inoculated into the wounds for 2 days. Four mice were subsequently euthanised under anaesthesia, and skin tissues were collected to determine E‐selectin expression in the infected diabetic wounds. EVs and s‐EVs were stained with PKH67 (Cat. No. MINI67, Sigma‐Aldrich) according to the manufacturer′s instructions, and then ultracentrifugation was performed to obtain the PKH67‐labelled EVs and s‐EVs. Twelve mice were randomly categorised into four groups to receive treatment of hydrogels loaded with PKH67‐labelled s‐EVs, hydrogels loaded with PKH67‐labelled EVs, hydrogels loaded with PKH67 dye, and local injection of PKH67‐labelled s‐EVs, respectively. The twelve mice were euthanised under anaesthesia on the third day of treatment, and skin tissues were collected, frozen, sectioned and stained. The remaining mice were randomly divided into five groups (n = 6 per group) receiving (1) PBS, (2) EV (100 µg), (3) s‐EV (100 µg), (4) Gel/PL‐5, or (5) s‐EV@Gel/PL‐5. The day on which mice received the dressing treatment was recorded as day 0. The treatments in groups (1), (2), and (3) were injected around the wound (100 µL total, 25 µL per site). The hydrogel treatments in groups (4) (Gel/PL‐5) and (5) (s‐EV@Gel/PL‐5, containing 100 µg s‐EV) were dropped onto the wounds (100 µL total) and cross‐linked under ultraviolet radiation for 15 s. The dressing was changed every 3 days. Wounds were photographed at a fixed time and distance (calibrated using a ruler) to calculate the wound area using ImageJ software. The Remaining area was calculated as follows: remaining area (%) = St / S0 × 100%, where St represents the wound area at different times and S0 represents the initial wound area. The mice were euthanised under anaesthesia on day 12. The collected skin wound tissues were preserved for subsequent use. Histology and immunostaining Healing wound samples were collected on Day 12 of the hydrogel treatment, fixed in 4% paraformaldehyde, embedded in paraffin, and cut into 5‐µm‐thick sections. Samples were de‐paraffinised and stained with H&E (Cat. No. BA4907, Baso, Wuhan, China) and Masson's trichrome stain (Cat. No. G1006, Servicebio, China). To determine the angiogenic and anti‐inflammatory effects of the hydrogels, the wound samples were subjected to immunohistochemical staining with anti‐CD31 antibody (Cat. No. ab182981, Abcam, Cambridge, UK), anti‐α‐SMA (Cat. No. ab7817, Abcam), anti‐IL‐1β (Cat. No. ab283818, Abcam), and anti‐IL‐6 (Cat. No. ab290735, Abcam) according to the manufacturer's instructions. Epidermal thickness and mean integrated optical density values were quantified using ImageJ Pro software. For immunofluorescence staining of E‐selectin expression in wounds, wound tissues obtained from humans and mice were fixed in 4% paraformaldehyde, embedded in paraffin, cut into 5‐µm‐thick sections, de‐paraffinised, antigen repaired, blocked with 10% goat serum for 60 min at room temperature, and cultured with anti‐CD31 (Cat. No. sc‐376764, Santa Cruz Biotechnology, Dallas, TX, USA) and anti‐E‐selectin (Cat. No. ab18981, Abcam) at 4°C overnight. The next day, sections were incubated with Alexa Fluor‐labelled secondary antibody for 60 min and stained with DAPI for 5 min at room temperature. Images were collected using a fluorescence microscope. Toxicity test of hydrogels in vivo Peripheral blood was collected on day 12 of the hydrogel treatment. EDTA anticoagulant was added to part of the blood to obtain whole blood and detect blood indices. The other part of the blood was centrifuged to obtain serum for the detection of different protein markers. The heart, liver, kidney, and lungs were also obtained, fixed in 4% paraformaldehyde, embedded in paraffin, and cut into sections for H&E staining for a histological analysis of hydrogel toxicity. scRNA‐seq of wound samples We acquired and reanalysed data based on a previously reported expression matrix (Justynski et al., 2023). Data were obtained from 12 wound samples provided by nine participants, including five with diabetic foot ulcers and four with non‐diabetic foot wounds, of whom one provided two samples and another provided three samples (Table S1). Data were analysed using ScanPy (Wolf et al., 2018). We first filtered low‐quality cells and genes and performed doublet filtering using Scrublet (Wolock et al., 2019). After the data were normalised and log‐transformed, we identified the cell cycle and regressed it to eliminate its effects. After removing batch effects with the BBKNN method (Polański et al., 2020), PCA and UMAP (McInnes & Healy, 2020) dimensionality reductions were conducted, and Leiden cluster (Traag et al., 2019) analysis was used to identify different cell populations. Cell types were determined by annotation based on a previous human skin atlas (Reynolds et al., 2021) using the CellTypist software package (Domínguez Conde et al., 2022). Independent t‐tests were used to assess the intergroup differences in gene expression. RNA‐sequencing (RNA‐seq) analysis Total RNA was extracted from HUVECs treated with control, LPS, LPS + EV, and LPS + s‐EV using Trizol reagent (Cat. No. 15596018CN, Invitrogen, USA) according to the manufacturer's instructions. The library construction was conducted through the steps of RNA quality control (QC), mRNA isolation and fragmentation, cDNA synthesis, end repair, adaptor ligation addition, PCR amplification, and library QC. The constructed library was sequenced on the Illumina NovaSeq X Plus platform at Tsingke Biotech Co., Ltd. (Beijing, China). The sequencing data was quality‐controlled with fastp to obtain the clean reads which were stored in FASTQ format. The clean reads were then mapped to the reference genome Homo sapiens (GRCh38.110) using HISAT2. The expression levels of genes were calculated by StringTie. Comparison of transcripts with reference annotations was detected by Gffcompare software to obtain the final assembly annotation results. The ballgown package was used to provide file input for FPKM quantification. Differential expression analysis was performed using DESeq2 for downstream pathway analysis. Human phospho‐kinase array The Human Phospho‐Kinase Array Kit (Cat. No. ARY003C, R&D systems, Minnesota, USA) was used to analyse the phosphorylation profiles of kinases and their protein substrates of HUVECs treated with control, LPS, LPS + EV, and LPS + s‐EV. HUVECs were stimulated with different treatments, lysed in the provided lysis buffer, and possessed according to the manufacturer's instructions. Statistical analysis Data are presented as the mean ± standard error of the mean (SEM) and analysed using GraphPad Prism 8.0 software (GraphPad Software, La Jolla, CA, USA). Student's t‐test was used to analyse differences between groups, while one‐way analysis of variance (ANOVA) was used to analyse the differences between three or more groups. All experiments were repeated at least three times. Statistical significance was set at p < 0.05. Other materials and methods used in this study are elucidated within the supporting information. RESULTS scRNA‐seq data suggest gene‐driven endothelial injury After filtering and processing the data, the cell cycle was identified and regressed (Figure S1). A final matrix comprising 44,426 cells expressing 24,771 genes was constructed. Clustering after dimensionality reduction resulted in 19 cell populations (Figure 1a), annotated as 13 different cell types (Figure 1b). There are two types of endothelial cells in the dermis of healthy adults, vascular endothelial cells (VEC) and lymphatic endothelial cells (LEC), which express PECAM1 and LYVE1, respectively (Figure 1c). We analysed gene expression in these two types of endothelial cells and found that endothelial damage‐associated genes such as ‘SELE’, ‘RPS4Y1’, ‘NFKBIA’, ‘C2CD4B’, and ‘JUNB’ were highly expressed in the diabetic group compared to those in the non‐diabetic group (p < 0.001, Table S3, Figure 1d,e). High endothelial damage‐associated gene expression in diabetic wounds. (a) Grouping on UMAP dimensionality reduction graphs using Leiden clustering. (b) Different cell types on the UMAP dimensionality reduction map after annotation based on the skin cell atlas. (c) Marker gene expression in two endothelial cell types on UMAP dimensionality reduction plots. (d) Heatmap of top 20 differentially expressed genes in endothelial cells of samples from diabetic and non‐diabetic patients. (e) Expression of endothelial injury‐related genes in endothelial cells from samples obtained from two patient groups. (f) Representative immunofluorescence staining images of E‐selectin expression in wounds of non‐diabetic and diabetic patients. Scale bar: 20 µm. (g) Representative immunofluorescence staining images of E‐selectin expression in wounds of non‐diabetic and diabetic mice. Scale bar: 20 µm. UMAP, uniform manifold approximate projection. The E‐selectin adhesion molecule, encoded by SELE, is present in inflammation‐stimulated endothelial cells and is thought to promote the accumulation of blood leukocytes at sites of inflammation by mediating cell adhesion to the vascular endothelium. To design E‐selectin‐targeting glycoengineered treatments for diabetic wounds, we validated the expression of E‐selectin—an expression product of the vascular endothelium in response to inflammatory activation. In human samples, E‐selectin was co‐localised with CD31, a marker of vascular endothelium, in diabetic wounds, whereas E‐selectin was not expressed in non‐diabetic wounds (Figure 1f). Immunofluorescence staining of samples from the mouse model showed that E‐selectin was highly expressed in the wounds of diabetic mice compared to that in non‐diabetic mice (Figure 1g). Preparation and identification of MSCs, s‐MSCs, EVs and s‐EVs The viability of passage‐5 MSCs was not affected by lentiviral infection, as determined by the CCK‐8 assay (Figure 2a). Flow cytometry revealed that sLeX was highly expressed in s‐MSCs (Figure 2b). Notably, the expression of FUT7 and CD63‐P19‐Nluc significantly enhanced the concentration of sLeX on s‐MSCs. To demonstrate the adipogenic differentiation capacity of MSCs and s‐MSCs, the cells were cultured in an adipogenic medium for 3 weeks and stained with Oil Red O (Figure 2c). To demonstrate the osteogenic differentiation capacity of MSCs and s‐MSCs, the cells were cultured in osteogenic medium for 4 weeks and then stained with Alizarin Red (Figure 2d). Flow cytometry of surface markers showed that MSCs and s‐MSCs were positive for CD105 and CD73 expression and negative for CD34 and HLA‐DR expression (Figure 2e). These findings confirmed that MSCs were isolated successfully and that cells transduced with lentiviral vectors retained the properties of stromal cells. Identification of MSCs and s‐MSCs. (a) Cell viability assay of MSCs and s‐MSCs. (b) Detection of sLeX expression on s‐MSC surface by flow cytometry. (c)–(d) Oil Red O and Alizarin Red staining of MSCs and s‐MSCs following adipogenic and osteogenic induction. Scale bar: 20 µm. (e) Identification of MSCs and s‐MSCs by flow cytometry. Data are shown as the mean ± SEM. ns, no significant difference; s‐MSCs, MSCs overexpressing sLeX. EVs and s‐EVs were characterised based on NTA, western blotting, and transmission electron microscopy (TEM). The NTA showed that the median diameters of EVs and s‐EVs were 132.4 and 141.4 nm, respectively—both within the normal range (Figure 3a). The zeta potentials of EVs and s‐EVs were –29.55 and –33.44 mV, respectively (Figure 3b). Western blotting confirmed the expression of markers CD9, Alix, and TSG101 of EVs, whereas the cellular endoplasmic reticulum marker calnexin was not expressed in EVs and s‐EVs (Figure 3c). TEM revealed the cup‐shaped morphology and bilayer membrane structure of EVs and s‐EVs (Figure 3d). Western blotting revealed that the sLeX levels were higher in s‐EVs than those in EVs (Figure 3e,f). EVs were extracted successfully and EVs generated with glycoengineering retained the characteristics of EVs. Additionally, the surface expression of sLeX was significantly increased in the paracrine EVs from s‐MSCs. Identification of EV and s‐EV. (a) Diameter and (b) zeta potentials of EVs and s‐EVs measured by NTA. (c) Western blot analysis of cellular endoplasmic reticulum marker Calnexin and EV markers, CD9, Alix and TSG101. (d) Morphology of EVs observed by TEM. Scale bar: 200 nm. (e) Western blot analysis of sLeX expression on EV and s‐EV. (f) Statistical analysis of sLeX relative to TSG101 expression. NTA, nanosight tracking analysis; s‐EV, EVs secreted by MSCs overexpressing sLeX; TEM, transmission electron microscopy. ** p < 0.01. s‐EVs exhibit strong binding towards activated endothelial cells Once bound to the ligand sLeX, E‐selectin expressed on cells mediates the rapid internalisation of sLeX‐expressing EVs through endocytosis (Sun et al., 2013; Zheng et al., 2022). To verify the cellular uptake of s‐EVs, HUVECs were first stimulated with LPS to determine E‐selectin expression. Previous studies indicated that the expression of E‐selectin on the surface of HUVECs peaked after a 4‐h exposure to 1 µg/mL LPS (Dayang et al., 2019; Gong et al., 2017). Flow cytometry indicated that E‐selectin expression on HUVECs was significantly higher after exposure to LPS for 4 h than that in the control group (Figure 4a). Lentiviral vector encoding CD63‐Nluc was prepared and transduced to MSCs (n‐MSCs). The secreted EVs (n‐EVs) from n‐MSCs carried the Nluc tag with no targeting function, which can be used as control groups when using Nluc Luciferase to detect the uptake of Nluc‐labelled EVs by cells (Figure S2) (Zheng et al., 2022). There was no significant difference in the uptake of non‐targeted EVs between un‐activated and activated HUVECs, whereas activated HUVECs showed significantly higher uptake of s‐EVs (>3.9‐fold) than non‐activated HUVECs (Figure 4b). Fluorescence microscopy indicated that EVs and s‐EVs were taken up by HUVECs with and without LPS stimulation, while LPS‐stimulated HUVECs could take up more s‐EVs (Figure 4c). LPS‐stimulated HUVECs rapidly take up s‐EVs. (a) Expression of E‐selectin on HUVECs after exposure to LPS for 4 h. (b) Quantification of cellular Nluc expression after 4 h of EVs treatment. (c) Representative images of normal and activated HUVECs taking up DiO‐labelled‐EVs and ‐s‐EVs. Scale bar: 20 µm. (d) Representative image of transwell migration assay of HUVECs under different treatments. Scale bar: 100 µm. (e) Quantitative analysis of cell migration in Transwell migration assay. (f) Representative image of cell scratch assay of HUVECs under different treatments at 10 and 20 h. Scale bar: 100 µm. (g) Quantitative analysis of migration area in cell scratch assay. * p < 0.05, ** p < 0.01, *** p < 0.001. HUVECs, human umbilical vein endothelial cells; LPS, lipopolysaccharide; ns, no significant difference; s‐EV, EVs secreted by MSCs overexpressing sLeX. The migration capacities of four groups of HUVECs (control, LPS, LPS + EV, and LPS + s‐EV) were determined using transwell assays (Figure 4d). HUVECs in the LPS + s‐EV group exhibited the highest migration rates (Figure 4e), consistent with the scratch assays (Figure 4f). The treated HUVECs were pre‐cultured with the negative control and LPS for 4 h, after which the cells were scratched, rinsed with PBS, and supplied with a new culture medium. The scratch assay revealed that more HUVECs migrated in the LPS + s‐EV group than those in the other three groups at 10 and 20 h (Figure 4 g). Characterization of Gel, Gel/PL‐2 and Gel/PL‐5 Previous studies indicated that a 5% concentration of GelMA is suitable for wound healing (Zhu et al., 2022); thus we selected 5% (w/v) GelMA as the applied concentration. In addition, 2% and 5% PLMA were selected to determine the mechanical and antibacterial properties of the gel. An inversion assay was performed to evaluate the sol‐to‐gel transformation of the Gel/PL‐5 hydrogel system. The solution exhibited flow characteristics after inversion of the tube but showed a rapid sol‐to‐gel transition under ultraviolet radiation, with no fluidity noted after inversion (Figure 5a). Gel/PL‐5 had a lower swelling ratio than Gel and Gel/PL‐2, indicating that the swelling capacity of the hydrogel decreased gradually with increasing crosslinking density (Figure 5b). The degradation ratio (measured every 2 days) showed that Gel was degraded more rapidly than the other hydrogels. The degradation of Gel/PL‐5 was significantly slower than that of the low concentration PL (Figure 5c). The hydrogels were immersed in PBS to obtain the s‐EV release curves. Gel released s‐EVs rapidly and explosively in the first 3 days, whereas Gel/PL‐5 released s‐EVs continuously and steadily (Figure 5d). The storage moduli (G′) values of the three groups increased over time, exceeding the loss moduli (G′′) values, and finally reached an equilibrium, indicating that the precursor solutions formed solid hydrogels with stable internal structures (Figure 5e). With an increase in the oscillation frequency, the G′ of the three hydrogels was always higher than the corresponding G′′ value, indicating a stable internal structure (Figure 5f). The scanning electron microscope images revealed porous structures inside the three hydrogels (Figure 5g). With an increase in the PL concentration, the internal crosslinking of the hydrogel was enhanced and the hydrogel pore size gradually decreased (Figure 5h). Confocal microscopy imaging revealed the successful loading and homogenous three‐dimensional distribution of s‐EVs in Gel/PL‐5 (Figure 5i). Characteristics of hydrogels. (a) Gelation of precursor solution to form hydrogels under ultraviolet radiation. (b) Swelling properties of hydrogels. (c) Degradation properties of hydrogels. (d) s‐EV release curves. (e) Rheological behaviour of hydrogels. (f) Rheological analysis of hydrogels under frequency scanning. (g) SEM images of all hydrogels. (h) Mean pore size and pore diameter distribution of all hydrogels. (i) 3D reconstruction of PKH67‐labelled s‐EV in Gel/PL‐5. Scale bar: 100 µm. SEM, scanning electron microscopy; s‐EV, EVs secreted by MSCs overexpressing sLeX. Biocompatibility of hydrogels Excellent biocompatibility is important for the clinical use of wound dressings. We cultured HUVECs in conditioned media containing different concentrations of hydrogels for 48 h and found that none of the hydrogels affected cell viability (Figure 6a). A haemolysis test showed that, compared with the positive control group treated with Triton‐X 100, the haemolysis rate in blood co‐cultured with hydrogel extract was <5%, indicating good blood compatibility (Figure 6b,c). Cell cytotoxicity was further determined by live/dead cell staining after culturing in hydrogel‐conditioned media for 48 h. Compared with the positive control group treated with palmitate acid (PA, 0.5 mmol/L), fewer dead HUVECs (with red fluorescence) were observed in the hydrogel‐treated groups (Figure 6d, Figure S3), indicating that Gel, Gel/PL‐2, and Gel/PL‐5 exhibit excellent biosafety. Biocompatibility and antibacterial activity of hydrogels. (a) Cell viability of HUVECs after incubation with hydrogel extract for 48 h. (b) Haemolysis ratio and (c) haematological performance of hydrogels. Triton X‐100 served as positive control, and PBS as negative control. (d) Live/dead cell staining of HUVECs after incubation with hydrogel extract for 48 h. Scale bar: 100 µm (e) Photographs and (f) survival ratio of S. aureus and E. coli on agar plates after incubation with different hydrogels. (g) Live/dead staining of S. aureus and E. coli from extract after incubation with different hydrogels. Scale bar: 20 µm. * p < 0.05, ** p < 0.01, *** p < 0.001. HUVECs, human umbilical vein endothelial cells; ns, no significant difference. Antibacterial activity of hydrogels in vitro Diabetic wounds are often accompanied by multiple bacterial infections, the most common of which are caused by S. aureus and E. coli (Prasathkumar & Sadhasivam, 2021). The antibacterial activity of Gel/PL is mainly attributed to PL, which contains numerous positively charged amino groups that interfere with the membrane potential of bacteria. Gel alone had no antibacterial properties but showed an increase in antibacterial properties with increasing PL concentration, as evidenced by the gradual decrease in the number of bacterial colonies (Figure 6e). The survival rates of S. aureus and E. coli in the Gel/PL‐5 group were 1.109% ± 0.7262% and 4.381% ± 0.6922% (Figure 6f). The live/dead staining assay showed that the bacteria in the control and Gel groups were mainly live (green), whereas those in the Gel/PL‐2 and Gel/PL‐5 groups were mainly dead (red, Figure 6g). Therefore, the Gel/PL‐5 group exhibited good antibacterial properties, which could inhibit diabetic foot infections and promote diabetic wound healing. s‐EV@Gel/PL‐5 accelerates diabetic wound healing in vivo Prior to application of s‐EV@Gel/PL‐5 as a wound dressing, we performed experiments to ensure the sustained release and the vascular targeting function of s‐EV in vivo. We stained EV and s‐EV with PKH67 and obtained PKH67‐labelled EVs and s‐EVs. The release of s‐EVs from the hydrogel into wound tissue could still be detected on day 3 after the hydrogel dressing was administered (Figure S4). On day 3 of hydrogel treatment, fluorescent signals of s‐EV could be observed locally in the wound area (Figure S5a). However, weak fluorescence signals of s‐EV were observed at the local injection site only in the s‐EV injection group (Figure S5b). In the hydrogel loaded with normal EVs, the fluorescence of EVs was distributed in different structures of the wounds without tissue specificity (Figure S5c). To rule out the effects of the PKH67 dye itself, the hydrogel loaded with PKH67 dye was applied to the wound tissue, and a small amount of fluorescence signal was found scattered in the skin tissue without particular pattern (Figure S5d). Moreover, immunofluorescence staining images of mouse wound tissue treated with hydrogel loaded with PKH67‐labeled s‐EVs revealed that s‐EVs with green fluorescence were co‐localised with vessels expressing CD31 in the vascular region, indicating the vascular targeting function of s‐EV (Figure S6a). However, EVs with green fluorescence were not limited to the vascular endothelium but was also distributed in other tissues outside the vascular tissue (Figure S6b), indicating that normal EVs cannot specifically reach the vascular endothelium. We determined the effect of s‐EV@Gel/PL‐5 in full‐thickness diabetic wounds infected with S. aureus and E. coli. Figure 7a shows the experimental timeline of wound infection and treatment. Figure 7b shows representative images and traces of wounds for each group on days 0, 3, 6, 9 and 12 of the treatment. The infected open wound appeared red and swollen. On day 12, the s‐EV@Gel/PL‐5 group had the smallest remaining wound area (1.353%) compared to the PBS, EV, s‐EV and Gel/PL‐5 groups (24.73%, 12.06%, 7.05% and 16.45%, respectively; Figure 7c). Infected wound healing in diabetic mice. (a) Schematic illustration of diabetic wound treatment. (b) Representative photographs of diabetic wounds and wound traces in the PBS, EV, s‐EV, Gel/PL‐5 and s‐EV@Gel/PL‐5 group on days 0, 3, 6, 9 and 12 after dressing treatment. (c) Quantification of remaining wound area in different groups on days 3, 6, 9 and 12. (d) Representative H&E and Masson's trichrome staining images of wound tissue in different groups on day 12. Scale bar: 500 µm. (e) Quantification of epithelial thickness in different groups. Data are shown as the mean ± SEM. * p < 0.05, ** p < 0.01, *** p < 0.001. HUVECs, human umbilical vein endothelial cells; s‐EV, EV secreted by MSCs overexpressing sLeX. Skin re‐epithelialisation and appendage generation were observed by H&E staining at 12 days after treatment (Figure 7d). The s‐EV@Gel/PL‐5 group had the thickest skin epithelium, followed by the s‐EV group, which was better than that of the PBS, EV and Gel/PL‐5 groups (Figure 7e). Meanwhile, s‐EV@Gel/PL‐5‐treated wounds exhibited denser collagen fibres, thicker granulation tissue and higher nascent collagen deposition. We assessed the changes in angiogenesis on day 12 after treatment by performing immunohistochemical staining for vascular markers (CD31 and α‐SMA; Figure 8a). The s‐EV@Gel/PL‐5‐treated group exhibited more new blood vessels, indicating a better angiogenic effect than that in the other groups (Figure 8b,c). In addition, we evaluated the expression of inflammatory cytokines (IL‐1β and IL‐6) and found that the s‐EV@Gel/PL‐5‐treated group showed a lower expression of IL‐1β and IL‐6 than other groups (Figure 8d,e). Angiogenesis and inflammation of diabetic wound. (a) Representative images of CD31, α‐SMA, IL‐1β and IL‐6 staining of wound tissue in different groups on day 12. Scale bar: 500 µm. Quantitative analysis of mean IOD of (b) CD31, (c) α‐SMA, (d) IL‐1β and (e) IL‐6. * p < 0.05, ** p < 0.01, *** p < 0.001. IOD, integrated optical density. Systematic detection of in vivo toxicity The hydrogel had no pathological effect on haematological composition in vivo (Figures S7 and S8), with no damage or pathological changes observed in the vital organs (heart, liver, lungs and kidneys; Figure S9). s‐EVs promote angiogenesis through the p53 signal pathway To elucidate the mechanism of the s‐EV effect on HUVECs, we performed RNA‐seq analysis of HUVECs receiving control, LPS, LPS + EV and LPS + s‐EV treatment (Figure S10). We identified 214 differentially expressed genes (DEGs), with 71 genes were upregulated in the LPS + s‐EV group compared with the LPS group, and 143 genes were downregulated (Figure 9a,b). The Kyoto Encyclopedia of Genes and Genomes (KEGG) was conducted to elucidate the pathway enrichment of s‐EV on inflammatory HUVECs and demonstrated the significant enrichment of genes in the p53 signal pathway (Figure 9c). Heatmap analysis revealed that p53 signal pathway‐related genes were differentially expressed in LPS group and the LPS + s‐EV group (Figure 9d). Since cytokine signalling pathways are typically regulated by phosphorylation of signal‐dependent transcription factors (Ciummo et al., 2023), we performed the human phospho‐kinase array on HUVECs to explore the phosphorylation status of transcription factors. LPS stimulation could increase the phosphorylation level of p53 at the S392 site, while the phosphorylation level decreased significantly after receiving s‐EV treatment (Figure 9e,f, Figure S11). Some proteins, including GSK3α/β, WNK and Yes, showed significantly different phosphorylation levels in HUVECs receiving LPS treatment and LPS + s‐EV treatment; the degree of phosphorylation of these genes has also been reported to be associated with endothelial inflammation and injury, and vascular leakage (Erfinanda et al., 2022; Garrud et al., 2024; Xu et al., 2023). Both RNA‐seq analysis and human phospho‐kinase array experiments pointed to changes in the p53 signal pathway. Therefore, we considered that s‐EV could improve the function of inflammatory endothelial cells mainly through the p53 signal pathway and the reduction of the phosphorylation p53. s‐EVs promote angiogenesis through the p53 signal pathway. (a) Heatmap of differentially expressed genes of the RNA‐sequencing data of HUVECs receiving LPS treatment and LPS + s‐EV treatment. (b) Volcano plot of HUVECs receiving LPS treatment and LPS + s‐EV treatment. (c) KEGG analysis of the enrichment pathways. (d) Heatmap of gene expression related to p53 signal pathway. (e) Human phospho‐kinase array performed on HUVECs receiving LPS treatment and LPS + s‐EV treatment. (f) Statistical analysis of relative mean pixel density of the human phospho‐kinase array. The arrays were quantified using ImageJ software. Values were normalized to the LPS group. HUVECs, human umbilical vein endothelial cells; KEGG, Kyoto Encyclopedia of Genes and Genomes; LPS, Lipopolysaccharide; s‐EV, EV secreted by MSCs overexpressing sLeX. DISCUSSION An effective treatment is urgently needed to treat diabetic wounds and reduce the associated amputation and death rates. Diabetic wounds exhibit reduced angiogenesis, leading to insufficient capillary and vascular densities, and decreased transport of oxygen and nutrients, ultimately delaying wound healing. In addition, diabetic wounds are highly susceptible to infection, especially during the inflammatory phase. Inflammation‐suppressing therapy plays a crucial role in final wound closure and scar formation (Tu et al., 2022). Therefore, wound dressings with anti‐inflammatory and pro‐angiogenic effects represent a promising clinical avenue. scRNA‐seq technology can accurately analyse the cellular composition of tissues and cell suspensions. Combined with high‐throughput sequencing, scRNA‐seq can reveal gene structure and expression in individual cells among large populations and reflect the heterogeneity between cells (Nie et al., 2023). scRNA‐seq is an important modern tool for describing the molecular characteristics of lesions (Venteicher et al., 2017). We searched the scRNA‐seq of the GEO database GSE245703 (original PMID: 38127424). In the original manuscript, the analysis of endothelial cells mainly focused on apoptosis and efferocytosis, and the DEGs of endothelial cells shown were confined to apoptosis and efferocytosis‐related signalling pathways. Additionally, due to different data processing methods, such as using BBKNN method for batch correction and Leiden cluster analysis to identify different cell populations, the results of our study were slightly different from those of the source manuscript. In this study, endothelial cells were selected from the 13 cell types for subsequent analysis. Differential gene analysis of endothelial cells proved that the expression of endothelial damage‐associated genes, such as ‘SELE’, ‘RPS4Y1’, ‘NFKBIA’, ‘C2CD4B’ and ‘JUNB’, were upregulated in diabetic wounds compared to that in non‐diabetic wounds, which reflects the vascular damage in the diabetic wounds. Based on the literature review, we selected ‘SELE’ as the key factor for glycoengineering of EVs. E‐selectin (expressed by the SELE gene) is necessary for endothelial progenitor cell adhesion, endothelial cell homing in vitro, and neovascularization in vivo (Liu et al., 2010). We revealed that E‐selectin expression was significantly higher in infected human and mouse wounds than in normal wounds, which is consistent with the findings of a previous study and reflects the potential of E‐selectin as a target for diabetic wound healing (Van Der Laan et al., 2001). We demonstrated that E‐selectin was highly expressed in stimulated HUVECs and that it mediated the uptake of EVs expressing sLeX by endothelial cells. The cell and mouse experiments confirmed the expression of E‐selectin in the wound vascular endothelium, reflecting the importance of EV glycosylation and the feasibility of the proposed glycoengineered hydrogel. MSCs are multipotent cells with self‐renewal and multidirectional differentiation capacities. With the emergence and development of regenerative medicine, MSCs are considered a promising strategy for treating diabetes and its complications, including diabetic wounds. However, there are increasing reports that MSCs promote diabetic wound healing via paracrine EVs (Shi et al., 2017; Yang et al., 2020; Yu et al., 2020). Compared to MSCs, EVs are unable to self‐replicate, thus reducing the potential for tumour formation after transplantation. EVs also exhibit low immune reactivity, stable performance, easy storage, and small particle sizes for easy filtration and sterilisation. However, treatment using native EVs is mostly nonspecific and the number of lesion‐homing EVs is limited. Therefore, it is necessary to engineer EVs to increase their ability to target lesion sites and increase their therapeutic efficiency. Based on previous findings, we modified the dense glycocalyx coating of the outermost layer of EVs using rational glycoengineering to increase EV targeting. MSCs were transduced with lentivirus vectors encoding FUT7 and CD63‐P19, selected with 3 µg/mL puromycin, and further selected with a maintenance concentration of 1 µg/mL puromycin to obtain cells stably expressing the transduced vectors. Our results revealed that the surface expression of sLeX was significantly increased in the paracrine EVs from s‐MSCs. sLeX‐expressing EVs specifically bind to HUVECs under inflammatory stimulation, thereby promoting their function in vitro and neovascularization in vivo. GelMA is formed by replacing the amino and hydroxyl groups of gelatin with methacrylate group. The degree of methacryloyl substitution determines the mechanical properties and pore sizes of GelMA (Chen et al., 2012). GelMA with 60% methacrylic anhydride substitution is commonly used for wound dressings (Li et al., 2023; Yuan et al., 2023). GelMA has a porous structure suitable for the loading and slow release of EVs (Li et al., 2023). The extracellular matrix‐like properties of GelMA provide good biocompatibility and degradability. GelMA has recently become a popular biomedical material in applications as a drug carrier or scaffold for the regeneration and repair of various tissues and organs, such as hair (Huang et al., 2023), nerves (Cheng et al., 2022), vascular (Wang et al., 2018), muscle (Hwangbo et al., 2021), cartilage (Hu et al., 2020) and bone (Sun et al., 2024). PLMA is the product of ε‐PL methylacryloylation by glycidyl methacrylate and the introduction of double bonds on ε‐PL molecules. ε‐PL, a cationic homopolyamide produced by actinomycetes, is widely used as an antibacterial material in infected wound healing (Wang et al., 2021). ε‐PL and its derivatives avoid the drug resistance and metal ion deposition associated with traditional antibacterial drugs or metal ions (e.g., Ag+) (Hyldgaard et al., 2014; Wang et al., 2021). In this study, we infected diabetic wounds with S. aureus and E. coli to determine the antibacterial effects of the hydrogels. Through the colony count and live/dead assays of the bacteria, we revealed that Gel/PL‐5, containing 5% PL, exhibited a better bacteriostatic ability than that of Gel/PL‐2. GelMA and PLMA are photopolymerizable hydrogels that can be crosslinked under UV irradiation for a few seconds. LAP is the most commonly used photo‐initiator because of its outstanding biocompatibility and minimal immunogenicity, making it possible for GelMA and PLMA to cross‐link under blue light at a wavelength of 405 nm (Lee et al., 2014). To confirm the sustained release of s‐EV@Gel/PL‐5, we conducted the experiment that diabetic wounds received the treatment of hydrogels loaded with PKH67‐labelled s‐EVs and local injection of PKH67‐labelled s‐EVs respectively. Our study revealed that the inflammatory vascular endothelium of diabetic wounds expressed E‐selectin, and the s‐EV‐loading could be released by the hydrogel directly to enter the vascular tissue of the wound, rather than entering the systemic blood circulation before eventually reaching the wound site. On day 3 of s‐EV injection, the fluorescence signal of s‐EV was observed only at the local injection site in the local s‐EV injection group, and the intensity of the fluorescence signal was significantly weaker than that in the hydrogel‐loaded s‐EV treatment group, indicating that injected s‐EV were quickly consumed and cleared up. In this case, it took longer to heal the wound and might need more injections. On day 3 of hydrogel treatment of mice wounds, fluorescent signals of s‐EV could be observed locally in the wound area, indicating that the s‐EV released by the hydrogel could still play a therapeutic role in wound healing. In addition, immunofluorescence staining in the hydrogel loaded with PKH67‐labeled s‐EVs group revealed that s‐EVs could specifically reach the vascular endothelium. However, in the hydrogel‐loaded EV group, the fluorescent signals were not concentrated around certain tissue, demonstrating the non‐targeting ability of normal EV. We used the hydrogel containing the PKH67 dye as a negative control group to rule out the possibility of abnormal fluorescence signals caused by dye. The controlled release and vascular targeting experiment in vivo provided feasibility for further study on the effect of s‐EV@Gel/PL‐5 on wounds. In vivo, we found that s‐EV@Gel/PL‐5 had the best healing effect on infected wounds compared to the other treatments. The wound‐healing ratio, regenerated epidermal thickness, and collagen deposition rate demonstrated that EV performance was significantly improved by glycosylation and Gel/PL‐5 encapsulation. The expression of CD31 and α‐SMA was higher in the wound tissue of the s‐EV group than in the EV group, suggesting that the surface expression of sLeX in EVs enhanced angiogenesis in the infected diabetic wounds. The s‐EV@Gel/PL‐5 group showed better anti‐inflammatory and angiogenic effects than the s‐EV group, indicating that EV‐loaded hydrogels exert superior therapeutic effects over EVs alone. The s‐EV@Gel/PL‐5 group exhibited the lowest pro‐inflammatory cytokine levels and highest microvessel density. The favourable therapeutic effect of the s‐EV@Gel/PL‐5 group demonstrated that s‐EVs in hydrogels could still maintain functional biological activity and that the combined effect of s‐EV@Gel/PL‐5 was much greater than that of the single components. Biosafety is critical for the clinical acceptance of hydrogel dressings. We tested the biocompatibility and toxicity of the hydrogels both in vitro and in vivo and found that GelMA and Gel/PL exhibited good biocompatibility, as evidenced by the CCK8 and live/dead staining assays after 48 h of co‐culturing HUVECs and hydrogel extracts. After co‐culturing mouse blood and hydrogel extracts for 1 h, we found that haemolysis rates were significantly lower than those in the positive control group, reflecting the good blood compatibility of the hydrogels in vitro. We observed no significant histological or haematological changes in the peripheral blood, heart, liver, kidney and lung tissues of hydrogel‐treated mice, suggesting that the hydrogels both improved diabetic wound healing and had no toxic effect in vivo. In terms of use, subcutaneous injection of EVs may cause subcutaneous bleeding, while hydrogels loaded with EVs can be applied to wound surfaces noninvasively without the risk of subcutaneous bleeding and further infection (Wang et al., 2022). When used by diabetic wound patients, considering the convenience of patient use, patients can apply the hydrogel and EVs mixture to the wound surface themselves, without requiring them to go to a hospital or clinic for subcutaneous injection. Based on the above reasons, the hydrogel‐loaded EV method for wound treatment is a good choice for both basic research and clinical applications. Tumour suppressor p53 has been reported to be associated with endothelial dysfunction, and its phosphorylation could lead to impairment of the endothelial barrier system (Ciummo et al., 2023). Phosphorylation of the six key serine residues (Ser6, Ser15, Ser20, Ser37, Ser46 and Ser392) of p53 is associated with instability of p53. The residue Ser392, a highly preserved residue, is a major phosphorylation site of p53 (Cox & Meek, 2010). The phosphorylation level of p53 at the Ser392 site is low under normal conditions, and the phosphorylation level significantly increases under stress and stimulation. Phosphorylation of p53 at S392 was associated with endothelial apoptosis and angiogenesis inhibition. LPS stimulation could increase the phosphorylation of p53 at Ser392, and the phosphorylation could be reversed after the improvement of endothelial function, which is consistent with our results (Betapudi et al., 2016). Therefore, the mechanism of s‐EV is related to the p53 signalling pathway, and the specific mechanism will be further explored in future studies. CONCLUSION This study demonstrated that the enhanced expression of sLeX on the surface of EVs could promote wound angiogenesis. The photopolymerizable hydrogel Gel/PL‐5 exhibited excellent antibacterial activity, biocompatibility and mechanical properties. Gel/PL‐5 carrying s‐EVs showed sustained release of s‐EVs and accelerated healing of infected diabetic wounds. The newly developed s‐EV@Gel/PL‐5 hydrogel represents a promising new dressing for treating infected diabetic wounds. AUTHOR CONTRIBUTIONS Kewei Wang: Conceptualization; investigation; methodology; visualization; writing—original draft; writing—review and editing. Qin He: Conceptualization; funding acquisition; software; writing—original draft; writing—review and editing. Mengmeng Yang: Investigation; methodology. Qincheng Qiao: Software. Jun Chen: Investigation; methodology. Jia Song: Investigation; methodology. Nan Zang: Investigation; methodology. Huiqing Hu: Investigation; methodology. Longqing Xia: Project administration. Yingyue Xiang: Supervision. Fei Yan: Supervision. Xinguo Hou: Supervision; Data curation; visualization; writing—review and editing. Li Chen: Conceptualization; funding acquisition; methodology; supervision; writing—review and editing. CONFLICT OF INTEREST STATEMENT The authors declare no competing interests. Supporting information Supporting Information