Phospholipid-grafted PLLA electrospun micro/nanofibers immobilized with small extracellular vesicles from rat adipose mesenchymal stem cells promote wound healing in diabetic rats

Abstract Small extracellular vesicles (sEVs) derived from mesenchymal stem cells (MSCs) can deliver a variety of bioactive factors to create a favorable local microenvironment, thereby holding huge potential in chronic wound repair. However, free sEVs administrated intravenously or locally are usually cleared rapidly, resulting in an insufficient duration of the efficacy. Thus, strategies that enable optimized retention and release profiles of sEVs at wound sites are desirable. Herein, we fabricated novel functional phosphoethanolamine phospholipid-grafted poly-l-lactic acid micro/nanofibers (DSPE-PLLA) to carry and retain sEVs from rat adipose MSCs, enabling the slow local release of sEVs. Our results showed that sEVs@DSPE-PLLA promoted the proliferation, migration and gene expression (Col I, Col III, TGF-β, α-SMA, HIF-1α) of fibroblasts. It also promoted keratinocyte proliferation. In addition, sEVs@DSPE-PLLA helped polarize macrophages toward the M2 phenotype by increasing the expression of anti-inflammatory genes (Arginase 1, CD 206, IL-10) and inhibiting the expression of pro-inflammatory genes (IL-1β, TNF-α). Further in vivo study in diabetic rat models showed that sEVs@DSPE-PLLA improved the wound-healing process by alleviating the inflammatory responses, stimulating cell proliferation, collagen deposition and angiogenesis. These results highlight the potential of using DSPE-grafted scaffolds for extracellular vesicle immobilization and suggest sEVs@DSPE-PLLA micro/nanofibers as promising functional wound dressings for diabetic wounds.


Introduction
Wound healing is a dynamic and interactive process, which requires various cells, chemokines and cytokines [1,2] to help the process of inflammation, new tissue formation and eventual wound closure [3]. With the high prevalence of diabetes mellitus all around the world, complications such as chronic wounds have been a major threat to patients and bring a heavy economic burden [4,5]. In diabetic chronic wounds, severe imbalance in the local physiological environment, which leads to multiple pathological changes of the wound tissue and disturbs the normal repair process, is one of the key reasons for the poor healing outcome. For instance, higher levels of multiple inflammatory cytokines, including tumor necrosis factor a (TNF-a), granulocyte colony-stimulating factor and monocyte chemoattractant protein-1, are detected at diabetic wound sites [6][7][8]. Meanwhile, the levels of many growth factors and cytokines, including epidermal growth factor, fibroblast growth factor, macrophage inflammatory protein-1 a (MIP-1a) and MIP-1b, are reduced. As a result, a variety of cell functions are impaired. For example, abnormal endothelial cell metabolism leads to vascular insufficiency [9]. The migration of keratinocytes is blocked and the epidermal structure cannot be restored promptly [10][11][12]. Fibroblasts at the wound edge become senescence, exhibiting limited migratory ability and weakened response to growth factor signals [13]. Besides, due to the elevated inflammatory cytokines, inflammatory cells continue to flood into the wound and cause undesirable damage to the tissue. Therefore, delivery of a single bioactive factor to the diabetic wound sites can hardly ameliorate the microenvironment and improve the healing outcomes [6].
Mesenchymal stem cells (MSCs) have been shown to promote skin tissue repair [14,15], but their application has been limited by the relatively low survival rate and chromosomal variation [16]. Recently, accumulated evidence demonstrates that MSCs exert their therapeutical effects mainly through the paracrine mechanism, that is, secreting various bioactive factors to create a favorable local microenvironment [17,18]. One important kind of MSCs paracrine product is the small extracellular vesicles (sEVs), which are small membrane particles with a diameter of <200 nm [19][20][21]. sEVs contain a complex cargo rich in proteins, lipids, RNAs (e.g. microRNAs, lncRNAs and mRNAs) and DNAs and thus can modify the microenvironment and regulate cells in the neighborhood. Compared to MSC treatment, MSC-sEVs have several potential advantages, such as lower immunogenicity and higher stability, and no potential to transdifferentiate into a different cell type [22]. Many studies have confirmed that MSC-sEVs benefit many physiological processes in wound healing, such as helping immunoregulation [23,24], inducing angiogenesis [25] and accelerating re-epithelialization [26], making them attractive candidates for skin tissue repair. However, in most cases, free sEVs were given intravenously, which leads to rapid clearance in vivo. Besides, the systemic administration of sEVs leads to an accumulation of sEVs in the spleen and liver [8,27,28]. As a result, very few sEVs reach the wound sites, which severely limits their clinical use. Local delivery can increase the concentration of sEVs at wound sites. Nevertheless, the sEVs are usually cleared very quickly due to the high metabolic activity of wounds. Thus, strategies that enable optimized retention and release profiles of sEVs at wound sites are desirable.
Herein, we propose to immobilize sEVs onto electrospun micro/ nanofibers, which have been widely used as wound dressings because of their similarity with skin extracellular matrix (ECM) in morphological and dimensional features [29,30]. sEVs isolated from adipose-derived MSCs (ASCs) were used in this study because ASCs are easy to obtain and have a rich source in the body [31]. Two strategies were selected for the immobilization of sEVs on electrospun micro/nanofibers. The first one is to deposit a polydopamine (PDA) layer on the surface of electrospun fibers (sEVs@PDA-PLLA). This mussel-inspired immobilization strategy has been proved to be effective in loading bioactive factors on almost all materials and exhibits good biocompatibility [32]. The second one is to graft PEG-phospholipids (DSPE-PEG 5000 -NH 2 ) on the surface of electrospun fibers (sEVs@DSPE-PLLA). Many studies have shown that lipid chains could insert into the cell membrane through the hydrophobic interaction with the phospholipids of the plasma membrane [33][34][35]. Since sEVs had a lipid bilayer similar to cells, we hypothesize that the conjugated phospholipids on the surfaces of the fibers are able to insert into the membrane of sEVs, thereby immobilizing the sEVs (Fig. 1). The biological functions of functional electrospun fibrous wound dressings were evaluated at the molecular, cellular and animal levels to investigate their regulation on diabetic chronic wounds in rats.
Isolation and characterization of rat adiposederived mesenchymal stem cell-sEVs After culturing rat adipose-derived mesenchymal stem cells (rASCs) with sEV-free medium for 48 h, the culture supernatant was collected. The sEVs derived from rASCs (rASC-sEVs) were isolated by differential ultracentrifugation as described previously [36]. In brief, the collected supernatant was centrifuged sequentially at 300 Â g for 10 min, 2,000 Â g for 10 min and 10 000 Â g for 30 min to remove cells, dead cells, apoptotic bodies and cell debris. Then, the supernatant was centrifuged at 100 000 Â g for 70 min to harvest rASC-sEVs. The pellets were further purified by re-suspension in phosphate-buffered saline (PBS, Gibco) and centrifugation at 100 000 Â g for 70 min to remove contaminating protein. The obtained rASC-sEVs were suspended in PBS and stored at À80 C for later use. The total protein content of rASC-sEVs was quantified by BCA Protein Assay Kit (Beyotime, Biotechnology) according to the manufacturer's instructions. The rASC-sEVs were sterilized by the Centrifugal Filter Units 0.22-lm GV Durapore 50 V R (Merck, USA). The morphology of rASC-sEVs was examined by transmission electron microscopy (TEM, ZetaView Particle, PMX, Germany), and the particle size of rASC-sEVs was detected by nanoparticle tracking analysis (NTA, Tecnai G2 Spirit, FEI, USA). Western blotting was used to detect the specific surface markers of rASC-sEVs, including Tsg101

Fabrication of electrospun fibrous wound dressings
PLLA micro/nanofibrous wound dressings (PLLA) were fabricated by electrospinning. In short, PLLA (Mn ¼ 100 000, China) was dissolved in hexafluoroisopropanol (Aladdin, China) at 10% (w/v) concentration and stirred for at least 6 h. Under the voltage of 12 kV, the polymer solution was extruded with a syringe pump through a 21-G needle. The micro/nanofibers were received with tinfoil and had a thickness of approximately 300 lm. Then, the fibers were dried for 48 h under vacuum to remove the residual organic solvent. Polydopamine modified PLLA fibrous wound dressings (PDA-PLLA) were obtained by soaking in Tris-HCl solution (pH ¼ 8.5, 1 mol/l) with dopamine hydrochloride (2 mg/ml) at room temperature for 4 h. To get PEG-phospholipid-grafted PLLA fibrous wound dressings (DSPE-PLLA), PLLA were soaked in NaOH (0.1 M) for 20 min to expose the carboxyl and then immersed in 4-Morpholineethanesulfonic acid (MES) solution (pH ¼ 5.6, Thermo Fisher Scientific, USA) containing EDC (0.05 mg/ml, Aladdin) and NHS (0.025 mg/ml, Aladdin) for 30 min to activate the exposed carboxyl. At last, PLLA were immersed overnight in MES solution (DSPE-PEG 5000 -NH2) (10 mg/ml, Beijing Hua Wei Rui Ke Chemical) to graft DSPE-PEG5000-NH2 on the surface of PLLA fibers. All of the above procedures were conducted at room temperature. All the wound dressings were sterilized by immersing in 75% ethanol for 1 h and exposed to UV light for 2 h before use. The rASC-sEVs were immobilized onto the fibers by incubating the fibers with sterilized rASC-sEVs at 4 C for 6 h. Then, excess rASC-sEVs were washed off with PBS. PLLA, PDA-PLLA and DSPE-PLLA wound dressings loaded with rASC-sEVs were thereafter referred to as sEVs@PLLA, sEVs@PDA-PLLA and sEVs@DSPE-PLLA, respectively.

Characterization of electrospun fibrous wound dressings
The structure and morphology of PLLA, PDA-PLLA and DSPE-PLLA were observed by emission scanning electron microscope (SEM). The average fiber diameter was determined using ImageJ according to the representative SEM images. The formation of amide(-CONH) in DSPE-PLLA was proved by Fourier transform infrared spectroscopy (FTIR; Vector 33, Bruker, Germany). For rASC-sEVs loading assay, the total amount of rASC-sEVs in PBS was determined by BCA Protein Assay Kit and then loaded onto PLLA, PDA-PLLA and DSPE-PLLA. Then, the amount of free rASC-sEVs in PBS was determined. The loading capacity was calculated by total sEVs minus free rASC-sEVs in PBS. To determine the release profile of rASC-sEVs, the rASC-sEVs were first labeled with PKH26 (Sigma) according to the manufacturer's protocol. In short, PKH26 was co-incubated with rASC-sEVs for 4 min and then stopped by 5% BSA. The PKH26-labeled rASC-sEVs were obtained by centrifuging at 100 000 Â g for 70 min. The pellets were further purified by re-suspension in PBS and centrifuging at 100 000 Â g for 70 min to remove the redundant PKH26. The total amount of rASC-sEVs was represented as the fluorescence value of labeled rASC-sEVs detected by the Thermo Scientific Microplate Reader (Thermo, USA). After loading the labeled rASC-sEVs onto PLLA, PDA-PLLA and DSPE-PLLA, respectively, the wound dressings were placed in a 48-well plate containing 200 ll PBS and incubated in 37 C. All 200 ll PBS was sucked out and its fluorescence value was detected as the released amount of rASC-sEVs. The release rate was calculated according to the following equation: release rate (%) ¼ F t /F 0 Â 100% (F 0 : initial fluorescence value, F t : release fluorescence value at time t).

rASC-sEV uptake assay
Fibroblasts, keratinocytes and macrophages were seeded on various fibrous wound dressings, respectively. After 24 h, the cells were fixed with 4% paraformaldehyde (Leagene, China) for 30 min, followed by incubating with the Cell Navigator F-actin Labeling Kit (Abcam, USA) for 1 h and DAPI (Beyotime Biotechnology, China) for 5 min and observed under the confocal laser scanning microscope (CLSM, Leica TCS SP8, Germany).

Cell migration
CytoSelect TM 24-Well Wound Healing Assay Kit from Cell Biolabs Inc. (San Diego, CA, USA) was used to study the effects of different wound dressings on the migration of fibroblasts in vitro. Briefly, sEVs@PDA-PLLA and sEVs@DSPE-PLLA were placed in a 24-well plate, and the cell migration inserts were put upon the wound dressings. A total of 12 Â 10 4 fibroblasts stained with Cell Tracker Green CMFDA (Shanghai Yisheng, China, 1:1000) were seeded on the wound dressings. The inserts were removed after 24 h. The migration of fibroblasts was observed with Axio Observer 7 (Carl Zeiss, Germany) at 0, 24 and 48 h. The migration rate was calculated according to the following equation: migration rate (%) ¼ S t /S 0 Â 100% (S 0 : initial gap area, S t : cell covered area amid the gap at time t).

Cell proliferation
Cell proliferation was determined using Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Japan) according to the protocol of the manufacturer on Days 1, 3 and 5. In brief, 4 Â 10 4 keratinocytes and 2 Â 10 4 fibroblasts were cultured on various wound dressings in a 48-well plate, respectively. After washing three times with PBS, the cells were incubated in the culture medium containing 10% CCK-8 reagent at 37 C for 2 h. The absorbance was measured at 450 nm.

Real-time quantitative polymerase chain reaction
The relative gene levels in fibroblasts and macrophages were determined using quantitative real-time quantitative polymerase chain reaction (RT-qPCR). Total RNA was extracted and reversetranscribed into cDNA using a Prime Script RT reagent kit with gDNA Eraser (TaKaRa Biotechnology, Japan). The RT-qPCR was performed with the Quant Studio 6 Flex system (Life Technologies, USA) using Maxima SYBR Green/ROX qPCR Master Mix (Thermo Scientific, USA). All genes were normalized to GAPDH (n ¼ 4). Gene expression was quantified using the DDCt method, and the fold change was calculated using the formula 2 ÀDDCt . The primer sequences used in this study are listed in Tables 1 and 2.

Induction of type I diabetic rat model
All animal procedures were approved by the Animal Care and Use Committee of South China University of Technology. Healthy Sprague-Dawley rats weighing 250-300 g were purchased from the experimental animal center of South China University of Technology, Guangzhou. Diabetes was induced by intravenously injecting 1% streptozotocin (STZ, 6.5 mg/100 g) dissolved in 0.1 M sodium citrate buffer. After 7 days, when the rats showed symptoms of polydipsia, polyphagia, polyuria and weight loss and the blood glucose value of the rats was higher than 16.7 M, it indicated that the model of type I diabetes was successfully induced. All rats were kept in a specific pathogen-free environment and were fed a standard diet.
The treatment of full-thickness skin wounds in diabetic rats with sEVs@DSPE-PLLA The rats were anesthetized with isoflurane via a respiratory anesthesia machine and operated under sterile conditions. After shaving and sterilization, two circular, full-thickness skin defect wounds (U ¼ 18 mm) were created on the back of each rat. The wounds were randomly divided into four groups, which either received treatment with PLLA, DSPE-PLLA and sEVs@DSPE-PLLA wound dressings or were untreated (blank). Finally, 3M transparent plaster (3M Health Care, Germany) was used to cover the wound surface. All rats were fed in individual cages after surgery.

The measurement of wound-healing rate
The repairing status of the wounds was observed and recorded using a digital camera on Days 0, 3, 7 and 14 after surgery. The wound-healing rate was calculated according to the following formula: wound-healing rate (%) ¼ (A 0 À A t )/A 0 Â 100%, where A 0 represented the initial wound area on Day 0 and A t represented the wound area on Day t. The wound area was calculated by analyzing the wound photos with ImageJ software.

Histological and immunohistochemical analysis
The tissue samples were collected and embedded in OCT (Q800, TA, USA). Five-micrometer-thick cross-sections were cut from the OCT-embedded samples, followed by fixation with cold acetone. Hematoxylin and eosin (HE) staining and Masson staining were performed for histological evaluations. The immunohistochemical staining of Ki67 (GB111141, Servicebio, 1:500), ARG1 (bs-23837r, Bioss, 1:500), iNOS (GB1119, Servicebio, 1:500) and CD31 (GB12063, Servicebio, 1:500) was performed to determine the proliferative cells, macrophage polarization and the formation of new vessels, respectively. The expression levels of iNOS and ARG1were further quantified by calculating the mean optical density via ImagePro Plus 6.0 software. Specifically, the integrated optical density (IOD) of the positive area (Area) of each image was measured and accumulated to obtain the IOD SUM. The mean optical density ¼ (IOD SUM)/Area, which reflected the immunoreaction intensity in this image.

Statistical analysis
Three independent experiments and at least triplicate samples per group were performed to assure repeatability and statistical significance. Statistical analysis was done with GraphPad Prism 6 software (GraphPad). Data are presented as the mean 6 standard deviation. One-way analysis of variance followed by Dunnett's multiple comparisons test was used to determine significant differences between experimental groups. A P-values of <0.05 were considered statistically significant.

Characterization of rASC-sEVs
The morphology of the isolated particles from the supernatant of rat adipose-derived mesenchymal stem cells (rASCs) culture showed a typical cup or sphere shape by TEM (Fig. 2a). NTA revealed that the diameter of the nanoparticles was 50-200 nm with a peak at 110 nm (Fig. 2b), indicating the presence of sEVs. The expression of sEV-specific markers CD9, CD63 and Tsg101, together with the negative expression of non-sEV marker Calnexin, in the isolated nanoparticles, confirmed their identity as sEVs (Fig. 2c). All of the above results demonstrated that sEVs from rASCs (rASC-sEVs) were successfully isolated.  TCCAGGATGAGGACATGAGCAC  GAACGTCACACACCAGCAGGTTA  TNF-a  ACTCCAGGCGGTGCCTATGT  GTGAGGGTCTGGGCCATAGAA  ARG1  TGACCGCCGTCGTGTTACTTTA  TTCTCGCCCACTAGGCAGTTC  IL-10  GCCAGAGCCACATGCTCCTA  GATAAGGCTTGGCAACCCAAGTAA  CD206  AGAGCTGGCGAGCATCAAGAG  TTCCATAGGTCAGTCCCAACCAA  GAPDH GGCACAGTCAAGGCTGAGAATG ATGGTGGTGAAGACGCCAGTA

Characterization of various electrospun fibrous wound dressings
The diameters of PLLA, PDA-PLLA and DSPE-PLLA were 867.7 6 60, 832.7 6 60 and 734.8 6 80 nm, respectively. There was no significant difference in diameter among them, which was consistent with previous reports [37]. The formation of amide was a sign that the amino group of DSPE-PEG 5000 -NH 2 successfully reacted with the carboxyl group on the surface of PLLA. Specifically, the peak of amide I at about 1650-1660/cm is due to stretching vibrations of C-O groups, and the peak of amide II in the range of 1540-1555/cm originates from stretching C\\N and bending N\\H [38]. The FTIR results showed the peak signals of amide I and amide II were generated at 1650-1660/cm and 1540-1555/cm (Fig. 3b), indicating that DSPE-PEG 5000 -NH 2 was successfully grafted onto the surface of PLLA. As shown in Fig. 3a, there was no obvious attachment of sEVs on the fibers in the sEVs@PLLA group due to the low affinity of PLLA fibers for sEVs. Both sEVs@PDA-PLLA and sEVs@DSPE-PLLA groups had clusters of sEVs attached to the fibers, indicating a strong sEVs adhesion of the modified fibers. The loading capacities of rASC-sEVs on PLLA, PDA-PLLA and DSPE-PLLA were 0.54 6 0.08, 0.76 6 0.06 and 0.62 6 0.05 lg/mm 2 , respectively (Fig. 3c). Apparently, PDA-PLLA loaded more rASC-sEVs than PLLA, which might be due to the strong adhesion ability of PDA. There was no significant difference between the quantity of rASC-sEVs loaded on DSPE-PLLA and PLLA. Figure 3d demonstrates that rASC-sEVs exhibited different kinetic release profiles on PLLA, PDA-PLLA and DSPE-PLLA. A bust release of 35% sEVs was detected during the initial 24 h in the PLLA group. Meanwhile, only 6% sEVs were released from the DSPE-PLLA group. The sEVs loaded on PDA-PLLA were barely released. After 12 days, <40% of sEVs were left on PLLA, while approximately 80% sEVs were retained on DSPE-PLLA. The release rate of sEVs on PDA-PLLA was the slowest and retained almost 90% after 12 days.

Effects of various electrospun fibrous wound dressings on fibroblasts
The uptake of sEVs by fibroblasts was determined by culturing fibroblasts on sEVs@PDA-PLLA and sEVs@DSPE-PLLA for 24 h, respectively. The fluorescence microscopy analysis showed that the majority of rASC-sEVs on both PDA-PLLA and DSPE-PLLA were internalized by fibroblasts, evidenced by the cytoplasmic location of rASC-sEVs (Fig. 4a).
Fibroblasts cultured on sEVs@DSPE-PLLA exhibited a higher proliferation rate compared to those on PLLA and DSPE-PLLA (Fig. 4d). Surprisingly, significant inhibition of fibroblast proliferation was observed in the PDA-PLLA and sEVs@PDA-PLLA groups.
Concerning the expression of genes related to wound healing, the levels of Col I, Col III, VEGF, HIF-1a and a-SMA were greatly upregulated in fibroblasts on sEVs@DSPE-PLLA compared to those on PLLA or DSPE-PLLA on Day 1 (Fig. 5a). In addition, the expression of TGF-b1 in fibroblasts on sEVs@DSPE-PLLA tended to be higher. This higher trend continued for the rest of 7 days

Effects of various fibrous wound dressings on keratinocytes
Like fibroblasts, keratinocytes were also capable of uptaking sEVs efficiently. As shown in Fig. 6a, almost all rASC-sEVs were internalized by keratinocytes, distributed in the cytoplasm of keratinocytes. Compared to PLLA and DSPE-PLLA, sEVs@DSPE-PLLA stimulated keratinocyte proliferation on Days 3 and 5 (Fig. 6b). However, significant inhibition of keratinocyte proliferation was observed on PDA-PLLA and sEVs@PDA-PLLA compared to PLLA, and this trend continued for the rest of 5 days.

Wound-healing rate in diabetic rats
Because of the inhibitory effects of sEVs@PDA-PLLA on the proliferation of both fibroblasts and keratinocytes, we only explored the effects of sEVs@DSPE-PLLA on wound healing in diabetic rats. Representative images of the full-thickness wounds treated with different wound dressings on Days 0, 3, 7 and 14 were taken (Fig. 8a), which clearly achieved a faster, better healing outcome in wounds treated with sEVs@DSPE-PLLA. Specifically, 3 days after surgery, a significantly shrunk wound was observed in the wound treated with sEVs@DSPE-PLLA. After 7 days, >60% wound closure was achieved in the sEVs@DSPE-PLLA group, followed by the PLLA (51.1%), blank (50.4%) and DSPE-PLLA (40.4%) groups. The wound size in all groups was significantly reduced after 14 days, especially in the sEVs@DSPE-PLLA group, with a 94% wound closure achieved (Fig. 8b).

Histological analysis of the wounds in diabetic rats
The results of HE staining are shown in Fig. 9a. On Day 3, the wound beds in all groups were filled with fibrin clots with a large number of inflammatory cells in all groups except the blank. This might be due to the fact that the electrospun micro/nanofibers provided a temporary matrix to assist cells to migrate. On Day 7, typical epidermal tongues (green line) could be seen in each group, indicating keratinocytes at wound edges were migrating to the center. On Day 14, the newly regenerated epidermis almost fully covered the wound site with a neat structure in the sEVs@DSPE-PLLA group. However, the re-epithelization was severely delayed in other groups, especially in the blank group.
According to the results of Masson's trichrome staining (Fig. 9b), collagen fibers were detected in all groups on Day 7, with most collagen being deposited in the sEVs@DSPE-PLLA group. On Day 14, the deposited collagen fibers in each group increased, especially in the sEVs@DSPE-PLLA group. Moreover, collagen fibers in the sEVs@DSPE-PLLA group were arranged more orderly. Our results showed that sEVs@DSPE-PLLA accelerated the reepithelization and collagen deposition during the wound-healing process.

Immunohistochemical analysis of the wounds in diabetic rats
Ki67 immunohistochemical staining was performed on Days 3 and 7 to detect the proliferating cells in the wound beds. As shown in Fig. 10a, Ki67-positive cells were observed in the wound beds of all groups on Day 3, with the sEVs@DSPE-PLLA group being the most. Consistently, the quantitative analysis showed that the sEVs@DSPE-PLLA group had 408 6 83 cells, followed by the PLLA group and the DSPE-PLLA group, which had 194 6 55 cells and 188 6 29 cells, respectively. The number of proliferating cells in the blank group is the least, and only about 80 6 20 cells were detected (Fig. 10b). On Day 3, some Ki67-positive cells are not Figure 5. Gene expression of fibroblasts on various electrospun wound dressings on Day 1 (a) and Day 7 (b) (*statistically significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 vs PLLA, # P < 0.05, ## P < 0.01, ### P < 0.001 vs DSPE-PLLA, & P < 0.1, && P < 0.01 vs PDA-PLLA).
fibroblastic-like cells, and they may be some immune cells, such as neutrophils and macrophages. But on Day 7, while some Ki67positive immune cells were still detected in the blank, PLLA and DSPE-PLLA groups, the majority of Ki67-positive cells in the sEVs@DSPE-PLLA were fibroblasts. On Day 7, the number of Ki67positive cells in the sEVs@DSPE-PLLA group notably increased, much higher than that in all other groups.
The immunohistochemical staining of ARG1 and iNOS was performed on Days 3 and 7 to determine the macrophage phenotypes (Fig. 11). The expression levels of ARG1 were higher on Day 7 than on Day 3 in all groups. Apparently, higher ARG1 expression was detected in the sEVs@DSPE-PLLA group compared to the other three groups on both Days 3 and 7 (Fig. 11a and c). Conversely, the lowest expression of iNOS was detected in sEVs@DSPE-PLLA on Days 3 and 7 (Fig. 11b and c). On Day 7, the expression of iNOS in sEVs@DSPE-PLLA was further reduced to a nearly undetectable level. However, the expression of iNOS in the blank, PLLA and DSPE groups was still high, indicating that the wound beds were still in an inflammatory environment.
In addition, the expression of CD31 on Days 7 and 14 was examined to illustrate the angiogenesis at the wound sites (Fig. 12). The number of newly formed microvessels in the sEVs@DSPE-PLLA group was the largest (19 6 3.53) among all groups, followed by DSPE-PLLA (16 6 5.65) and PLLA (13 6 2.12). The microvessels in the blank group were the least (7 6 1.41). On Day 14, significantly more microvessels formed in all groups. Similar to the results on Day 7, the largest number of microvessels was observed in the sEVs@DSPE-PLLA group (41.5 6 4.94).

Discussion
To exert optimal therapeutical efficacy of sEVs, a local and slowrelease pattern that mimics the paracrine function of parent cells should be achieved. Several strategies have been developed to improve the local retention and release profile of sEVs. For example, chemical conjugation realizes the long-term binding of sEVs to the substrates. However, it may impair the functions of sEVs due to the alteration in the modified surface proteins. Therefore, mild tethering methodologies that can anchor sEVs on the wound dressings without compromising their integrity and functionality are in high demand [39]. Li et al. [40] constructed a lamininderived peptide PPFLMLLKGSTR-modified HA hydrogel (pGel), which can capture exosomes through the integrins on the membrane surface of exosomes. Li et al. [40] designed a novel biospecific peptide (TKKTLPTAHLHNRS), which specifically bind to type I collagen scaffolds with its N-terminal heptapeptide   (TKKTLRT) and the transferrin receptor of exosomes with the other segment (AHLHNRS). It is noted that these strategies used nature receptors of exosomes as anchor points. Such occupation of the nature receptors may impair the functions of extracellular vesicles, considering that many nature receptors mediate intracellular communication across a wide range of physiological and pathophysiological settings. PDA, which is highly adhesive for biological reagents, is supposed to be able to anchor extracellular vesicles as well. For example, Li et al. [41] coated PLGA scaffolds with PDA to enable a slow local release of exosomes. Similarly, we also found that PDA modification effectively improved the sEVs' loading capacity of PLLA and greatly slowed the release rate of sEVs from PLLA. Besides, sEVs@PDA-PLLA stimulated the VEGF expression in fibroblasts (Fig. 5) and helped the M2 polarization of macrophages (Fig. 7b). However, PDA modification exhibited remarked inhibitory effects on cell proliferation and migration (Figs 4 and 6). Previous studies have shown that PDA-modified surfaces inhibited the proliferation and migration of vascular smooth muscle cells probably through the negative regulation of catechol on cell growth and motility [42], but further investigation is required to elucidate how PDA affects cell behaviors. It is also worth mentioning that many conflict results regarding the effects of PDA on cell proliferation exist, indicating the interaction between PDA and cell is cell type dependent. Another strategy we developed in this study was to modify the electrospun PLLA fibers with phospholipids, which can easily insert into the lipid bilayer of sEV membrane as a result of the hydrophobic interaction between lipids, thereby achieving the retention and sustained release of sEVs on the PLLA fibers. As expected, DSPE-PLLA showed improved sEVs' loading and retention capacities compared to PLLA (Fig. 3c and d). Unlike chemical conjugation, the anchoring of phospholipids into the sEV membrane is noncovalent and reversible; thus, the anchored phospholipids dissociate from the membrane constantly, which explains the faster release rate of sEVs from DSPE-PLLA than that of from PDA-PLLA. One critical factor that determines the dissociation rate is the degree of the interaction force between the phospholipids and the membrane, which depends on the properties of the phospholipids. For example, phospholipids with double oleyl chains, such as DSPE used in this study, exhibit a stronger hydrophobic interaction than those with single oleyl chain [43]. Thus, it is possible to adjust the sEVs' release profile by using different phospholipids. Besides, since no surface proteins are required for the insertion of DSPE into the sEV membrane, the biological functions of sEVs are unlikely impaired. Indeed, sEVs@DSPE-PLLA effectively regulates multiple key players involved in skin repair, including fibroblasts, keratinocytes and macrophages, and enhances the wound-healing process in diabetic rats.
Wound healing is a multifaceted process governed by sequential yet overlapping phases, including hemostasis, inflammation, proliferation and remodeling [28]. The inflammatory phase begins almost immediately after injury. Prompt resolution of inflammation is considered beneficial to wound repair. In contrast, a wound trapped in a persistent inflammatory state usually leads to chronic wounds. One of the most important immune cells during the inflammatory phase is macrophages, which can roughly be divided into two groups: the M1 phenotype that stimulates the secretion of pro-inflammatory cytokines and the M2 phenotype that promotes cell division and collagen synthesis in the wound bed [44]. In the microenvironment of chronic wound healing, most macrophages maintain the M1 phenotype, resulting in the excessive secretion of inflammatory factors [45]. According to our results, sEVs@DSPE-PLLA alleviated the inflammatory response by promoting the expression of anti-inflammatory genes (ARG1, CD 206, IL-10) and inhibiting the expression of pro-inflammatory genes (IL-1b, TNF-a) in macrophages in vitro (Fig. 6). Consistently, sEVs@DSPE-PLLA drove a switch in the macrophage phenotype from M1 to M2 in the wounds of diabetic rats, as evidenced by the notably increased expression of ARG1 and the decreased expression of iNOS 7 days after surgery (Fig. 10). This phenotypic transformation was mainly regulated by sEVs as DSPE-PLLA did not show similar effects. Studies have shown that sEVs derived from ASCs not only induce the transition of macrophages from a proinflammatory M1 phenotype to an anti-inflammatory M2 phenotype but also affect the cellular behavior of T cells [24,46]. This effect may be attributed to the microRNAs enclosed in the sEVs and their roles in intracellular communication. For example, it has been reported that sEVs secreted by MSCs contain miR-223, which suppresses the classic pro-inflammatory pathways by targeting pknox1 [47]. Besides, our previous study detected various immunomodulatory microRNAs, such as miR-132-3p, miR-125-5p, miR-133a-3p and let-7b-5p in rat rASCs-sEVs [48], which may contribute to the regulatory effects of sEVs on macrophage polarization.
Skin cells at the wound edges play active roles during the proliferation and remodeling phases after the inflammation. For example, fibroblasts migrating into the wound beds and proliferating are critical because they produce a new collagenrich ECM and secrete a variety of growth factors [49,50]. Under the stimulation of TGF-b1, fibroblasts differentiate into contractile myofibroblasts, which promote wound closure [51,52]. Except for fibroblasts, keratinocytes, the most abundant epidermal cells, start to migrate within a few hours after injury to fill the wound defect, which is particularly important because it helps restore the epidermis structure and the barrier function [53]. Our results showed that sEVs@DSPE-PLLA promoted the proliferation of both fibroblasts and keratinocytes in vitro (Figs 4 and  6). Consistently, more proliferating fibroblasts and keratinocytes existed in the wound beds of diabetic rats treated with sEVs@DSPE-PLLA 7 days after surgery (Fig. 10). Besides the trophic factors in the sEVs, sEVs may stimulate cell proliferation via their surface proteins. A recent report demonstrates that Wnt3a on the exterior surface of CD63þ exosomes contributes to the promoted fibroblast and endothelial cell proliferation [54]. Our results also showed that sEVs@DSPE-PLLA promoted the remodeling of collagen and wound contraction in vivo (Fig. 9). This probably resulted from the increased secretion of collagen and TGF-b1 in fibroblast by sEVs@DSPE-PLLA, together with the enhanced myofibroblastic differentiation of fibroblast, as evidenced by the upregulated gene expression of Col I, Col III, TGF-b and a-SMA (Fig. 5). According to our results (Fig. 9), the newly regenerated epidermis almost fully covered the wound site in the sEVs@DSPE-PLLA group 14 days after the surgery. In contrast, reepithelialization was not complete in other groups. Similarly, Figure 11. Immunohistochemical staining of ARG1 (a) and iNOS (b) in macrophages in the wounds of diabetic rats on Days 3 and 7. (c) Quantification of ARG1 and iNOS in macrophages in the wounds of diabetic rats on Days 3 and 7.
sEVs derived from MSCs have been reported to promote the synthesis and remodeling of ECM and accelerate the reepithelialization in wound healing [55,56].

Conclusions
In this study, we successfully fabricated phospholipid-grafted PLLA electrospun micro/nanofibers immobilized with sEVs (sEVs@DSPE-PLLA) for wound repair. By slowly releasing sEVs, sEVs@DSPE-PLLA effectively regulated the functions of key cells involved in wound healing, including fibroblasts, keratinocytes and macrophages. Further in vivo study revealed that sEVs@DSPE-PLLA enhanced the wound-healing process by resolving inflammation, promoting cell growth and collagen deposition and enhancing angiogenesis at the wound sites in diabetic rats. Our study suggested the potential application of sEVs@DSPE-PLLA as a novel wound dressing to comprehensively improve the unbalanced microenvironment and promoted the repair process of chronic wounds.