Efficacy and safety of small extracellular vesicle interventions in wound healing and skin regeneration: A systematic review and meta-analysis of animal studies

Small extracellular vesicles (sEVs) have been proposed as a possible solution to the current lack of therapeutic interventions for endogenous skin regeneration. We conducted a systematic review of the available evidence to assess sEV therapeutic efficacy and safety in wound healing and skin regeneration in animal models. 68 studies were identified in Web of Science, Scopus, and PubMed that satisfied a set of prespecified inclusion criteria. We critically analyzed the quality of studies that satisfied our inclusion criteria, with an emphasis on methodology, reporting, and adherence to relevant guidelines (including MISEV2018 and ISCT criteria). Overall, our systematic review and meta-analysis indicated that sEV interventions promoted skin regeneration in diabetic and non-diabetic animal models and influenced various facets of the healing process regardless of cell source, production protocol and disease model. The EV source, isolation methods, dosing regimen, and wound size varied among the studies. Modification of sEVs was achieved mainly by manipulating source cells via preconditioning, nanoparticle loading, genetic manipulation, and biomaterial incorporation to enhance sEV therapeutic potential. Evaluation of potential adverse effects received only minimal attention, although none of the studies reported harmful events. Risk of bias as assessed by the SYRCLE's ROB tool was uncertain for most studies due to insufficient reporting, and adherence to guidelines was limited. In summary, sEV therapy has enormous potential for wound healing and skin regeneration. However, reproducibility and comprehensive evaluation of evidence are challenged by a general lack of transparency in reporting and adherence to guidelines. Methodological rigor, standardization, and risk analysis at all stages of research are needed to promote translation to clinical practice.


Introduction
Poor skin healing continues to have a substantial impact on the quality of life of millions of individuals around the globe. Skin is the body's first line of defense. In response to injury, skin activates a series of intricately orchestrated events controlled by numerous signals [1], with the goal of restoring the multi-layered structure and the continuum of the skin and reinstating its protective, thermogenic, endocrine and sensory functions [2]. Generally, wounds heal through four distinct but overlapping phases. These phases are: 1) hemostasis (platelet aggregation and fibrin clot formation); 2) inflammation (recruitment of inflammatory cells); 3) tissue regeneration (restoration of skin structure via cell proliferation, extracellular Ivyspring International Publisher matrix deposition, new blood vessel, and appendage formation, resulting in granulation and re-epithelialization); and 4) remodeling (long-term maturation of the newly formed tissue to closely resemble the native equivalent) [2][3][4]. Disruption of any of these phases-due to systemic or local causes-may result in a prolonged healing process or suboptimal recovery, marked by a failure to restore the architecture and function of the healing tissue [5]. Due to population aging and comorbidities, the prevalence of chronic non-healing wounds has risen dramatically, affecting millions of individuals each year. This imposes an increasing burden on health systems and economies [5]. Acute wounds are also widespread, accounting for millions of medical treatment facility visits and hospital admissions annually. Deep wounds can result in permanent disability and scarring, while burn injuries require lengthy hospitalization, incur high costs, and have high morbidity and fatality rates [6]. Unfortunately, currently available remedies for skin wound healing are incapable of meeting the urgent clinical needs [3,7]. Even though standard therapies such as routine debridement, infection management, and dressings may demonstrate some benefits, they fall short of addressing the pathophysiology of dysfunctional healing. Hence, researchers have placed great emphasis on developing biologically active formulations to rescue inadequate repair [3]. Of these, single bioactive factors that target specific wound indications-such as cytokines [8] and growth factors [9]-have garnered research interest, with a few gaining regulatory approval [3]. However, therapeutic modalities that target multiple inherent deficits in non-healing lesions might be more effective in addressing their complex pathophysiology that may include vascular, neurologic, inflammatory, and metabolic impairments [10].
Extracellular vesicles (EVs), which transfer cocktails of functional cargo (such as proteins, lipids, miRNAs, other RNAs, and DNA) horizontally between cells [11,12] may be multipotent stimulants of endogenous tissue repair [13]. EVs are a class of natural anuclear cell-released particles delimited by a phospholipid bilayer. As colloid members of the cell secretome [14], EVs are produced by almost all types of cells, in varying sizes and with different subcellular origins [15]. Each EV displays surface molecules that may target recipient cells. EVs are believed to communicate signals by fusing with target cells or simply binding to cell receptors [16], ultimately causing recipient cells to undergo phenotypic changes [12]. EVs can interact with target cells residing in the microenvironment or be carried to distant cells via biological fluids, and their internal and external cargo contribute to intercellular communication [17]. Recent studies have recognized the role of EVs in the pathogenesis of diseases [18] and in various natural physiological processes [19]. Indeed, the potent effects that were once credited to stem cells, for instance, are now thought to be partially mediated by EVs [20], making EVs a promising alternative to potentially risky cell therapies [21]. Moreover, EVs from certain sources may benefit from relative immunological tolerance in cross-species and interindividual transfer [22]. In the absence of functional definitions, EVs were classically categorized according to combinations of size, biogenesis, and biophysical separation process. For example, as microvesicles (100-1000nm, budding from the plasma membrane, also called ectosomes); apoptotic bodies (1-5μm, released from fragmented apoptotic cells) and exosomes (30-150nm, endosomal multivesicular body-derived nanovesicles) [23,24]. However, due to the increasingly recognized overlap in size between these categories [25] and the absence of universal differentiating markers, the term EVs is preferred [14,26]. This systematic review will focus on the therapeutic applications of a nanosized subclass termed small EVs (sEVs, ~30-200 nm), which includes but is not limited to endosome-origin exosomes. sEVs have been demonstrated to enhance tissue regeneration [27] and to modulate the immune system [28]. They have also been used for drug delivery [29,30], as vaccines [31] as biomarkers [32], and as therapeutic targets in "vesicle-mediated pathogenesis" [33]. EVs mediate signaling in all phases of physiological cutaneous wound healing (reviewed extensively in [34]). Platelet- [35] and monocyte-derived EVs [36] regulate clot formation and thus hemostasis. Neutrophil-derived EVs modulate inflammation [37]. Macrophage- [38] and endothelial progenitor cell-derived EVs [38] drive angiogenesis, and myofibroblast-derived EVs remodel the extracellular matrix (ECM) [39]. In recent years, the number of studies examining the therapeutic potential of sEVs in wound healing and skin regeneration has expanded dramatically.
The rapid progression of sEV therapeutic modalities toward clinical applications prompted us to critically appraise the available preclinical evidence for the benefits and adverse effects of sEVs in skin healing and regeneration. In our approach, we emphasized methodological rigor and reporting quality in accordance with field guidelines, including the Minimal Information for Studies of Extracellular Vesicles 2018 (MISEV2018) [14] and the criteria for MSC identification of the International Society for Cell and Gene Therapy (ISCT) [40]. We used a systematic review methodology for inclusive, bias-free coverage of existing studies, which could not be achieved by a conventional narrative review approach [41]. We further performed a meta-analysis for a quantitative pooled estimate of sEV efficacy across a vast body of literature, while assessing the heterogeneity of study outcomes and the likelihood of publication bias. Our work thus informs the scientific community of the available evidence from preclinical animal research and provides insights into the likelihood of clinical translation.

Search results
This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. On November 11th, 2019, a search on Web of Science, Scopus, and PubMed retrieved a total of 664 articles. All articles were pooled into Endnote X9.3.3 software, and 315 duplicates were removed. Titles and abstracts were screened to include articles investigating the therapeutic application of sEVs in skin repair, rejuvenation, and wound healing in mammalian models. We excluded 273 studies that were in vitro studies, reviews, reports, commentaries, conference proceedings, or articles written in languages other than English. The remaining 76 articles were read in full to determine satisfaction of the eligibility criteria. As a result, 48 studies were excluded, of which two studies were not in English (Chinese), 20 studies did not characterize sEVs by size and/or at least one sEV protein marker, and 26 studies exclusively reporting in vitro findings. Additionally, on March 1st, 2021, we updated our search to include another 40 studies, bringing the total number of manuscripts eligible for this systematic review to 68. The flow chart in Figure 1 summarizes the study selection approach.

General characteristics of the included studies
The 68 studies identified as eligible for inclusion were published between 2015 and March 1 st , 2021. Approximately 56% (n = 38) were published in 2020 or later, reflecting a surge in interest in sEVs to promote wound healing and skin regeneration. The studies originated from nine different countries, with China accounting for 84% (n = 57). Figure 2 depicts year of publication (2a) and region according to the corresponding author's affiliation (2b).

Animal species
Animal models have been used to reveal the intricate physiological and biochemical processes involved in wound healing and skin regeneration, as well as to assess the efficacy and safety of proposed therapeutic interventions. Rodents were used in 66 studies (97%): mice (36 studies) and rats (30 studies). One study tested a non-human primate model (macaque) [42], while another used the New Zealand rabbit model [43] ( Figure 3A).

Loading sEVs with nanoparticles
Two studies loaded superparamagnetic iron oxide nanoparticles (Fe 3 O 4 -NPs) into sEVs by incubating the nanoparticles with the parent cells before sEV isolation [46,69]. Following intravenous administration of nanoparticle-loaded sEVs, Li et al. magnetized the nanoparticles using an external magnetic guide positioned beneath the wound site to improve targeting and distribution capabilities [46]. In another study, Wu et al. applied static magnetic fields (SMF) to parent cells to enhance the therapeutic properties of the secreted nanoparticle-loaded sEVs [69]. In that experiment, nanoparticle-loaded sEVs were introduced locally to the wound via subcutaneous injection.
sEV preparation sEV collection conditions 62 studies (91.2%) separated sEVs from conditioned medium. Since serum contains sEVs, 22 studies (32.4%) collected sEVs from serum-free medium. Others prepared culture medium with sEV-depleted FBS (n = 22, 32.4%) or platelet lysate (n = 1, 1.5%). However, only six studies revealed the details of FBS-EV depletion protocols, and without reporting before-and-after particle counts. Chemically defined serum replacements were used in nine studies (13.2%), while autologous serum was the supplement of choice in a single study [82]. 11 studies (16.2%) did not report how they dealt with the issue of contaminating sera. 15 studies (22.1%) did not disclose the duration of cell culture conditioning before harvest. In the remaining studies, sEVs were collected after 24 hours (10.3%, n = 7, 11%) or 48 hours (n = 33, 48.5%) of conditioning.  sEV separation techniques There is no gold standard separation technique for sEVs, and sEV separation methods varied considerably across the studies ( Figure 5A). Ultracentrifugation (n = 43, 63.2%) was the most widely used technique, but with various centrifugation protocols. Ultrafiltration by membranes of pore size 0.22 µm (n = 43, 63.2%) or 100 kDa (n = 27, 39.7%) was often done as an adjunct to other separation steps. Commercial precipitation-based isolation kits, density gradient ultracentrifugation, and size exclusion chromatography (SEC) were used in 16 (23.5%), six (8.8%), and two (2.9%) studies, respectively. Additional washing steps were reported in 20 studies (29.4%). No study used tangential flow filtration (TFF), asymmetrical flow field flow fractionation, or microfluidics. 32 studies (47.1%) combined two or more separation techniques to achieve higher purity ( Figure 5B).

Dose
The administered sEV dose differed widely. sEV amount was approximated as protein amount in most studies, ranging from 2 µg to 5 mg (n = 45, 66.2%). In 7 studies (10.3%), dose was reported as number of particles, ranging from 2×10 10 to 2×10 12 particles (n = 7, 10.3%). However, only one study explicitly took into account the size of the animal, reporting sEV dose as protein per animal weight (5 mg/kg) [85], and amount was not reported at all in 15 studies (22.1%). Dose-response was assessed in one trial with three doses of 25, 50, and 100 µg/ml of PBS [91] and in five studies (7.4%) with low and high doses [44,52,82,92,93]. In these studies, wound healing was reported to be positively associated with dose.

Dosing frequency and intervention duration
The majority of studies (n = 51, 75%) involved a single dose. Of the multi-dose studies (n = 17, 25%) ( Table 2), two compared repeated-dose vs single-dose administration, concluding that repeated administration of low doses outperformed a single high dose [44,82]. The intervention period was diverse ranging mostly from eight to 28 days (Table 2).

Immuno-biocompatibility
Human sEVs were administered to immunocompetent animals in 57 studies (83.8%) ( Figure 3C). Allogeneic sEVs (from the same species) were used in 11 studies (16.2%), whereas autologous sEVs (from the same subject) were investigated in only one study [42]. Direct comparisons of the efficacy and immune response to autologous, allogeneic, or xenogeneic sEVs were inadequately considered in the included studies in this review, with only a single study comparing the therapeutic efficacy of sEVs from xenogeneic and allogeneic sources [51], and another comparing autologous sEVs with allogeneic sEVs [42]. No difference in efficacy was found between xenogeneic and allogeneic sources [51]. Autologous sEVs were reported to be more effective and viable in treated tissues than allogeneic sEVs, even though the latter had a sufficient therapeutic effect when compared with placebo [42]. More studies are needed to verify these findings and investigate the mechanism behind.

Labelling and tracking of sEVs in vivo
Only six studies (8.8%) reported tracking of transplanted sEVs [42,44,82,86,87,94]. In these studies, sEVs were pre-labeled with lipophilic fluorescent dyes, namely PKH26 (n = 3), PKH67 (n = 1), DiR (n = 1), and lipid conjugated Cy7dipalmitoylphosphatidylethanolamine (DPPE). All but one of these studies [82] administered a single dose of sEVs. A wide span of time points (1 hour to 21 days) was investigated. sEVs were tracked in vivo in four studies [44,86,87,94], ex vivo (post-mortem) in one study [42], and in vitro and ex vivo in another study [82]. None of these studies examined the biodistribution of sEVs to organs other than the skin tissue around the wound area. See Table 4 for detailed findings about sEV bioavailability.

Quality assessment
We sought to evaluate the quality of both methodology and reporting. Methodological biases may skew the outcomes of studies, resulting in misleading estimates of therapeutic efficacy and flawed inferences. Poor reporting impedes experiment evaluation and reproducibility. We thus evaluated several methodological aspects and compliance with established guidelines.

Quality of reporting
Quality of reporting was generally low. Most of the reviewed studies did not report pre-processing details such as donor number, age, and gender. For EV production, the number, seeding density, and passage number of EV-secreting cells were poorly reported, along with cell viability at harvest. EV depletion protocols were reported for only six of 22 studies that depleted exogenous EV collection medium. For EV separations, centrifugation details such rotor type, adjusted K factor, and the volume centrifuged were deficient. For in vivo experiments, no study indicated how the sample size for animal models was calculated. In 14 studies (20.6%), the sample size was not even disclosed. 15 studies did not reveal administered dose (22.1%). Several studies reported results for experiments that did not have a corresponding methodology section and thus could not be reproduced. In terms of outcome reporting, only seven studies (10.3%) reported the actual numerical data. All comparisons were depicted exclusively as graphical data presentations. In terms of statistical analysis, most studies did not report the absolute p-value and confidence interval of the measured outcomes.

Risk of bias assessment
We used the SYRCLE's ROB tool [95] to assess the risk of bias in animal experiments ( Figure 6). Overall, there was an unclear risk of bias for most of the elements investigated. Randomization of animals was reported in 44 studies (64.7%) but without disclosing the randomization method. 24 studies (35.3%) did not report randomization. While 42 studies (61.8%) reported comparable baseline characteristics between control and experimental groups, judgment was not possible in 26 studies (38.2%) due to insufficient reporting of certain animal characteristics, particularly age, which is a determinant factor in wound healing.
None of the studies clarified if allocation was concealed, or if animals were randomly housed. Blinding while performing the experiments was reported for only two studies [96,97]. Six studies conducted random outcome assessment (mostly angiogenesis experiments) [43,65,69,84,88,98]. We identified a high risk of attrition bias in six studies (8.8%) [46,55,75,76,84,99], low risk in 17 (25%), and an uncertain risk in the remaining 45 (66.2%). Blinding while assessing the outcomes was reported for seven studies (10.3%). Low risk was captured for all studies in relation to the selective reporting item, based on what was reported in the methods, although none of these studies reported publishing an a priori protocol to verify this judgment. The summary of the risk of bias assessment is shown in Figure 6.

Adherence to ISCT criteria for MSC characterization
To ensure comparability of studies of mesenchymal stromal cells (MSCs), the International Society for Cell and Gene Therapy (ISCT) has proposed minimal criteria to characterize and define these cells [40]. Specifically, cells should: 1) show ability to adhere to plastic; 2) be positive for surface markers CD105, CD73, and CD90, and negative for CD45, CD34, CD14 or CD11b, CD79a or CD19, and HLA-DR; and 3) show in vitro multi-lineage differentiation capacity into osteogenic, adipogenic,  One study investigated only two ISCT criteria, namely MSC adherence and surface antigen expression, and three focused exclusively on one ISCT criterion, namely surface antigen expression.
Adherence to MISEV2018 for sEV characterization, purity, and nomenclature sEV characterization To verify the identity of the isolated preparations, MISEV2018 indicates that EVs should be characterized by 1) concentration (such as protein and particle count); 2) at least two positive EV protein markers (including at least one transmembrane and one cytosolic marker), plus at least one source-appropriate negative, non-EV protein marker; and 3) two complementary single-vesicle analysis techniques to assess morphology and biophysical properties such as counts and size distribution [14]. Of the 68 studies analyzed here, only 14 (20.6%) satisfied the above criteria [49, 52-55, 69, 82, 83, 88, 90, 96, 100-102].
sEV preparation purity estimation MISEV2018 also suggests reporting protein:particle, lipid:particle, or lipid:protein ratios as surrogates of EV purity. Only two studies (2.9%) quantitatively estimated sEV purity according to MISEV2018, reporting particle:protein ratio [44,83]. Additionally, only 17 studies (25%) checked for the presence of negative/depleted (non-EV) markers that indicate the presence of non-EV contaminants.

sEV nomenclature
Since cells release EVs of varying sizes via different biogenesis pathways, and in the absence of specific, universal markers to distinguish EV subtypes, MISEV2018 recommends the term "extracellular vesicles" [14]. MISEV2018 also encourages that EVs be further described by physical properties (such as size), biochemical makeup, source cell, or culture condition. Historical and variously defined terms such as exosome and microvesicle are discouraged unless biogenesis can be proven. Here, we included studies that investigated the therapeutic potential of "small" extracellular vesicles (30-200nm) in wound healing and skin regeneration. 56 studies (82.4%) used the term "exosome" to describe the preparation without presenting clear justification for use of this term. This included 78% (39/50) of the studies published in 2019 onwards, following the MISEV2018 release. 12 studies (17.6%) used the term "extracellular vesicles" [43,44,49,51,53,86,[100][101][102][103][104][105], 11 of which were published in 2019 or later. Of these, four followed MISEV2018 nomenclature by size identification and specified that they were small extracellular vesicles [44,49,51,100], while the remaining studies described them by the cell of origin and culture condition.
Reporting to EV-TRACK MISEV2018 highly encourages submitting methodological details to EV-TRACK, a crowdsourcing tool developed to enable reproducibility and understanding methodology and experimental outcomes [14,106]. An "EV-METRIC" is assigned to each submitted study based on the proportion of required methodology details that are submitted. Here, only one study (1.5%) reported submitting details to the EV-TRACK knowledgebase [82].
-Both allogenic and autologous iPCderived exosomes promoted wound healing with no immune rejection.
1-Sig difference in wound healing capacity compared to non-exo control (day 6 and 17). With one application of 12.5 g was less beneficial than 2 applications of 7.6 g.
2) Reduction of the granulation tissue. At the 17-day, the hypertrophic epithelium detected in all groups was decreased to a close to normal size layer.
3) Normal collagen localization and deposition. 1-Both rat-sEV-AT and p-sEV-AT equivalently sig* promoted wound closure. Biggest difference with PBS-PVA control was at day 10 (60% vs 20% closure respectively). 2-Both rat-sEV-AT and p-sEV-AT equivalently sig* promoted re-epithelialization and enhanced the thickness and order of granulation tissue and supported hair follicle growth compared to control.

Re-epithelialization
Of the 38 studies (56%) that evaluated re-epithelialization, 37 indicated improvements as a result of sEV intervention. These studies also noted enhanced granulation, culminating in well-formed tissue that resembled the native tissue in thickness and cellularity. Several studies demonstrated that sEV-treated wound beds had significantly enhanced cellularity [45,84], which gradually reduced towards the end of the study [96]. Higher levels of proliferation and migration protein markers such as cytokeratin 14 [44,52], cytokeratin 19 [45,46,108], cytokeratin 10 [99], PCNA [45,46,100,108], Ki67 [61,71,72,74,84], and CXCR4 [100], as well as lower p21 expression [100] were detected in vivo. Additionally, twelve studies reported neogenesis of appendages including hair follicles and sebaceous glands. In contrast, a single study observed no clear improvement in re-epithelialization after treatment with sEVs [83]. However, the same study showed considerable improvement in mature collagen deposition in the sEV-treated group.

Collagen deposition
The degree of collagen deposition was investigated in 45 studies (66.2%). While 25 studies quantified either collagen deposition or expression in sEV treated tissues, the remaining 20 studies described it qualitatively. 34 investigations evaluated total collagen deposition, organization, and maturation using Masson's trichrome staining. Nine studies determined the presence of messenger RNA (mRNA) encoding collagens type I and III in sEV-treated tissues by RT-qPCR. Other techniques were Western blotting (n = 7), immunohistochemistry (n = 6), picrosirius red staining (n = 3), and Herovici staining (n = 1). All 45 studies reported enhanced collagen deposition, maturation, and organization following sEV administration in diabetic and non-diabetic rats. Immature collagen fibers were observed near the wound bed shortly post-treatment and improved significantly in alignment and maturity as healing progressed compared with the control, with an increased mature-to-young fiber ratio as early as five days after wounding [83].
Only 20 studies examined collagen deposition at multiple points, and most frequently at days 7, 14, 21, and 28, while a single study investigated collagen deposition for up to five weeks [46]. Some studies reported a trend towards increased collagen deposition towards the endpoint [52,70,76,92,98], but others showed a declining trend [63,94]. Similarly, mRNA expression of collagen type 1 showed a spike followed by a gradual decline towards the endpoint [74,96,109], while collagen type III expression varied among studies [48,74,94,96,109].
Ex vivo: positive signal was detected at day 6 after the sacrifice of animals.

Angiogenesis
45 studies (66%) examined vascularization of newly formed tissues, reporting a significant improvement in both diabetic and non-diabetic wounds that was associated with sEV treatment. These studies detected a rise in new vessel density, quantified by positive markers like CD31 (n = 35), α-SMA (n = 18), CD34 (n = 4), and Meca32 (n = 2). 12 investigations identified mature blood vessels as those positive for both CD31 and α-SMA, with substantial enhancement approaching the study endpoint. Several studies also showed upregulated vascular endothelial growth factor (VEGF) [67,72,78,79,96,104]. Marker detection techniques included immunohistochemistry (n = 20), immunofluorescence (n = 25), and Western blotting (n = 3). Techniques to assess neovascularization were microfilm perfusion and micro-CT scanning (n = 9) and small animal doppler detection (n = 1). Hettich et al. observed an enhanced vascularization localized to the wound margins where sEVs were injected, implying a possible enhanced local impact near the injection site [83].

Adverse events
No harmful events were reported by any study. 12 studies (17.6%) presented information related to adverse event evaluation, four of them having pre-specified in advance that the potential for harm would be investigated [57,59,61,96]. Yu et al noticed no erythema, edema, or irritation in skin tissues receiving sEV local injections, and no increase in renal injury markers (creatine and BUN) or liver function indicators (ALT and AST) [57]. Su et al. evaluated CD4+ and CD8+ T cell counts in the spleen and lymph nodes proximal to the treated sites in mice and found no difference between treatment with unmodified EVs and a negative control, although they observed a slight decrease of CD8+ T cells in the lymph of animals treated with EVs overexpressing PD-L1 [61]. One study monitored overall well-being and behavior of mice [96], whereas another checked for indicators of degeneration and necrosis [59]. However, neither study reported the outcome. Three studies assessed apoptosis in treated skin tissues using TUNEL assay [67,68] or by evaluating Bcl-2 and iBax levels [78], all of which showed a reduction in apoptosis. Four studies made generic statements that no negative effects or discomfort [60,62,80,104] or no impact on body weight [58,82] were detected. Lastly, one study reported that allogeneic iPSC-derived sEVs did not endow recipient cells with pluripotency or elicit immune rejection following repeated doses over 14 days of treatment [42].
The effect of modifications on the "characteristics" and "therapeutic outcomes" of sEVs
Also, Sung et al. detected an increase in mitochondrial cytochrome C expression in sEVs from H2O2-and hypoxia-primed cells, but not in LPS, naïve, and thrombin groups [53]. In that study, a comparison of the different preconditioning agents revealed a superior effect of thrombin, highlighting the importance of optimizing preconditioning regimens.
* The arm was not specified.

Nanoparticle loading
Superparamagnetic iron oxide nanoparticles (Fe3O4-NPs) were reported to be efficiently loaded into parent cells and sEVs [46,69]. Accumulation of membrane-encapsulated nanoparticles was demonstrated in the cytoplasm [46,69] and nucleus [69] of sEV-producing cells. The use of nanoparticles and static magnetic fields significantly enhanced the beneficial impact of sEVs when administered locally [69]. Analyzing sEV content revealed an abundance of miRNA content, predominantly miR-21-5p, compared with their naïve-sEV counterparts [69], as well as increased sEV release but with no changes to size distribution [69]. However, loading NPs into sEVs alone did not appear to improve their performance when introduced intravenously, as noted by Li et al. Improved wound closure rate, re-epithelialization, neovascularization, and collagen deposition was observed only after employing magnetic guidance, which apparently enhanced targeting and localization [46].

Incorporation of sEVs into biomaterial scaffolds
20 studies examined the efficiency of sEV-functionalized biomaterial scaffolds on wound healing. sEV-loaded scaffolds demonstrated superior therapeutic potential compared with either sEVs or scaffolds alone. A considerable improvement in wound closure (n = 20), new blood vessel formation (n = 14), re-epithelialization (n = 16), and collagen deposition (n = 17) was reported. 16 studies compared gel-only with gel-sEV preparations, while only three studies included sEV-only preparations, demonstrating superior effectiveness of gel-sEV preparations. Interestingly, Henriques-Antunes et al. indicated that light-triggered hydrogel-sEV preparations significantly outperformed preparations of hydrogel-sEV alone [44]. In addition, hydrogels enriched with sEVs were as potent as hydrogels loaded with bFGF in promoting wound healing [61].
A head-to-head comparison with source and conditioned media (secretome) Few studies attempted to compare the performance of enriched sEV preparations in promoting skin healing and regeneration with that of their source and conditioned medium, i.e., the total secretome. Interestingly, seven studies compared sEV efficacy with that of their source (menstrual blood-MSCs, platelet-rich plasma (PRP), iPSCs, saliva, hADMSCs, and hucMSCs) [42,45,80,88,96,99,107]. Four studies detected an overall equivalent effect of sEVs and their source cells on wound closure, re-epithelialization, and collagen deposition [42,45,99,107]. However, Dalirfardouei et al., observed that sEVs were superior at inducing wound closure rate, neovascularization, and M1 to M2 polarization [96]. Similarly, PRP and saliva-derived sEVs appeared to outperform PRP and saliva respectively in promoting wound closure, angiogenesis, and neoepithelialization [80,88]. Just two studies examined the efficacy of sEVs versus conditioned medium. While unfractionated conditioned medium was found to have a similar beneficial effect on wound closure as sEVs, with enhancement of angiogenesis and epidermis thickness [99], EV-depleted conditioned medium had an impact inferior to sEVs but comparable to that of a control [99,109].

Meta-Analysis
26 studies were eligible for meta-analysis of wound closure outcome, involving 174 animals from 12 diabetes and 14 non-diabetes model studies. An overall significant enhancement of wound closure rate was scored for wounds treated with sEVs (SMD = 4.25, 95% CI: 3.39 to 5.11, p < 0.00001) in comparison with control (Figure 7). The heterogeneity index was relatively high (I 2 = 72), reflecting the variability in sEV source, preparation, and dosage regimen. Similarly, subgroup meta-analyses in diabetic and non-diabetic groups demonstrated that sEV therapy was significantly more effective than control in accelerating wound closure in both models (SMD = 4.72, 95% CI: 3.25 to 6.18, p < 0.00001; SMD = 3.94, 95% CI: 2.87 to 5.00, p < 0.00001 in diabetes and non-diabetes models, respectively). Heterogeneity was likewise high in subgroups regardless of the disease model (I 2 = 74%, and 71% in diabetes and non-diabetes models, respectively).
Among all included studies, 22 assessed scar formation, of which 13 studies used scar width (in µm) as the scar assessment metric for examining the influence of sEV interventions. Of the 13 studies that used this metric, nine studies reported sample size and were included in the meta-analysis (a total of 60 animals; two studies used diabetes models, and seven used non-diabetes models). Overall, sEV therapy resulted in a substantial decrease in scar width compared with controls (SMD = -5.85, 95% CI: -7.98 to -3.73, p < 0.00001). However, when subgroup analysis was performed, the difference in scar width between the control and experimental groups in diabetes studies was insignificant (SMD = -12.78, 95% CI: -33.75 to 8.19, p = 0.23). Indeed, the results of only two studies (11 animals) are insufficient to draw conclusions from a meta-analysis, particularly given the high heterogeneity index (I 2 = 92%). In comparison, sEVs significantly inhibited scar development in the non-diabetic group (SMD = -5.69, 95% CI: -7.79 to -3.58, p < 0.00001). Overall heterogeneity of the effect was high (I 2 = 77%) ( Figure  8).
Nine studies assessing blood vessel density (number of blood vessels/mm 2 ) to evaluate the effect of sEV transplantation on angiogenesis were eligible for meta-analysis (44 animals), and subgroup analysis (5 diabetes model studies of 30 animals; 4 non-diabetes model studies of 14 animals) was thus performed. A meta-analysis revealed an overall significant impact of sEVs in supporting blood vessel development (SMD = 5.03, 95% CI: 3.17 to 6.88, p < 0.00001). The heterogeneity index was moderate (I 2 = 60%). In subgroup analysis, both diabetic and non-diabetic subgroups demonstrated a significant positive effect of sEV treatment compared with control (SMD = 5.42, 95% CI: 2.97 to 7.88, p < 0.0001; and SMD = 4.94, 95% CI: 1.26 to 8.63, p = 0.008; in diabetes and non-diabetes models, respectively) ( Figure 9). Heterogeneity indices in the two groups were moderate (I 2 = 66% and 64%, in the diabetic and non-diabetic groups, respectively).   We also performed meta-analyses of the wound closure rate outcome for studies that 1) characterized their EV preparation as required by MISEV2018 (section 2.7.4.a) and 2) disclosed the number of animals used in the experiments. A total of ten studies were considered eligible by these criteria (4 studies of 21 diabetic animals and 6 studies of 50 non-diabetic animals). We performed a sensitivity analysis that resulted in excluding one study [82], which produced considerable heterogeneity in the meta-analysis. Thus, only nine studies were included in the meta-analysis (4 studies of 21 diabetic animals and 5 studies of 40 non-diabetic animals) ( Figure 10). Consistent with our earlier findings, sEV intervention had a substantially favorable influence on wound closure across all studies (SMD = 3.50, 95% CI: 2.61 to 4.38, p < 0.00001) and in subgroup analyses (SMD = 3.13, 95%CI: 1.49 to 4.78, p < 0.0002; SMD = 3.80, 95% CI: 2.85 to 4.76, p < 0.00001) for diabetes and non-diabetes animal models, respectively ( Figure 10A). The sEV interventions, on the other hand, had a more homogeneous effect, as demonstrated by the lower I 2 statistics in each subgroup (I 2 = 48, and 22%, respectively) as well as in the overall meta-analysis (I 2 = 41%), in contrast to the higher heterogeneity observed in the overall analysis of wound closure that also included studies that did not comply with MISEV2018 (Figure 7). The funnel plots for the four meta-analyses showed no evidence of publication bias ( Figure 11).

Discussion
We have systematically reviewed the available evidence on therapeutic efficacy and safety of sEVs in wound healing and skin regeneration using animal models. We summarize recent research in this area and critically appraise the quality of the included studies, with an emphasis on methodology, reporting quality, and compliance with related guidelines. By doing so, we inform the scientific community of the main findings and the quality of evidence provided by the current literature. We detected a recent exponential surge in publications exploring sEV therapeutic potential in wound healing and skin regeneration, highlighting growing interest and enthusiasm towards sEV research in this area. 68 studies met our inclusion criteria, exploring a diverse spectrum of sEVs: not only from MSCs, but also from other cell types, as well as from complex tissues and biofluids. Although separated EV preparations contain mixed populations of sEVs, and not only exosomes (of endosomal origin), the bulk of the analyzed studies continued to describe their preparations as "exosomes," to which they related the observed functionality. In absence of techniques capable of identifying EVs from distinct intracellular origin [110], it is likely that these particles were instead a broader population of sEVs [14,26]. Throughout this discussion we will highlight unresolved issues and address several crucial questions, beginning with this central, fundamental question:

Was sEV intervention therapeutically effective?
Overall, our systematic review and metaanalysis concluded that sEV intervention had significant efficacy in promoting skin regeneration in diabetic and non-diabetic animal models. This finding agrees with earlier systematic reviews that examined the therapeutic efficacy of MSC-derived sEVs in wound healing in general [111] and diabetic wound healing in particular [112]. Based on our analysis, sEV intervention targets multiple features of the intricate healing process, resulting in enhanced regeneration and suppressed fibrosis.
As prior research established that deficient vascularization is a key contributor to the chronicity of diabetic lesions [113], it was remarkable to observe the proangiogenic effect of sEVs on diabetic wounds, which was on par with the effect on non-diabetic wounds. This was supported by our meta-analysis findings that quantitatively revealed that there was no substantial difference between diabetic and non-diabetic models. Diabetic lesions usually exhibit diminished levels of VEGF, which contributes to compromised angiogenesis [114]. sEV interventions boosted blood vessel regeneration and maturation with increased expression of VEGF.  [82], causing considerable heterogeneity in the meta-analysis ( Figure 10A, without the study, and Figure 10B with the study, I 2 = 41% vs 79%). Only a few studies addressed the immunomodulatory effects of sEVs. Nonetheless, those studies provided evidence for sEV modulation of the inflammatory milieu, favoring the transcriptional transition of pro-inflammatory M1 to anti-inflammatory M2 macrophages and reduction of immune cell infiltration. This is crucial for the inflammatory phase to resolve and the subsequent proliferative phase to commence, as persistent inflammation is a typical feature of non-healing wounds [115]. Accordingly, sEVs reduced the production of pro-inflammatory mediators such as TNFα, IL-1, and Toll-like receptor 4 (TLR4), whilst elevating levels of anti-inflammatory counterparts such as IL-10. It is worth noting that this immunomodulatory activity was not limited to MSC-derived sEVs. More studies are needed to explore the potential and mechanisms of sEV interventions on inflammation in this context. Furthermore, sEVs promoted re-epithelialization by enhancing skin cell proliferation and extracellular matrix secretion. sEVs fostered deposition of collagen, especially collagen type I, the major structural protein of the skin, and supported collagen maturation and organization. Although collagen production is critical for efficient wound closure, excessive production may lead to tissue fibrosis and scarring [116]. In ideal healing scenarios, collagen production increases during the proliferation stage, then decreases and matures during the remodeling stage. Nonetheless, our evaluation of the reports on collagen expression levels and ratios of collagen type I/III at later phases of healing revealed discrepancies among studies, demanding further investigation. Even so, by examining scar formation macroscopically and histologically, a number of studies noted a considerable reduction in scar indices following sEV application, supporting the anti-scarring role of sEVs. Myofibroblasts, which are key contributors in collagen deposition and wound contraction [117], were reported to be suppressed, and there was no evidence of fibrosis or hypertrophic scar development. Most studies had relatively short follow-up periods (14-or 21-days post-wounding), though, limiting their ability to thoroughly assess collagen and scar development and maturation in healing tissues. It is estimated that scars in rodent models mature within 70 days of injury [118]. Hence, it is necessary that relevant endpoints be defined and validated for future studies evaluating sEV usefulness in minimizing scarring and promoting tissue maturation. We observed that the protein marker α-smooth muscle actin (α-SMA) was used both to examine myofibroblast abundance and to mark blood vessel formation, revealing suppressed levels in the former and elevated levels in the latter. This creates some uncertainty over the actual expression of this marker in response to sEVs. Further examination of the utility of α-SMA as a differential marker may be needed.
Apparently, sEV modulation of the different healing mechanisms improved wound closure and tissue regeneration, with no differences in closure times for diabetic and non-diabetic wounds. sEVs likely improved the wound microenvironment, eventually encouraging tissue repair [118]. However, some studies found wound closure acceleration throughout the follow-up period, and others only during the early or late stages of healing. This could be due to heterogeneity of sEV sources, preparation methods, and delivery strategies, necessitating additional head-to-head comparisons. Given the challenges of restoration of appendages in adult skin [119,120], it was remarkable that sEVs supported the regrowth of hair follicles and sebaceous and sweat glands, indicating high-quality skin repair [119].
Comparisons with other treatment modalities were very limited. Interestingly, one study reported superior reparative activity of sEVs compared with the FDA-approved PDGF-BB therapy [44]. Also, comparisons of sEV preparations with their cell sources demonstrated comparable, if not superior, efficacy in stimulating skin repair and regeneration. Such comparisons are necessary to establish the relative value of sEVs over cells. Theoretically, cells may be more beneficial than EVs: although EVs have a "set message" that may not further respond to the microenvironment, cells might perceive signals from the milieu and respond by releasing various factors, including EVs. Thus, observations of sEV efficacy versus cells are crucial to establishing comparative value. Efficacy could be due in part to EVs being less immunogenic and carrying a unique payload that may be delivered in trace levels, yet exert a profound impact [121].
To summarize, our in-depth analysis of the various aspects of wound healing and skin regeneration indicates the usefulness of sEVs as regenerative agents to promote skin repair. In the context of diabetic wounds, the evidence reviewed demonstrated that sEV interventions may overcome the barriers to lesion repair. Nevertheless, longer durations of examination may be required to effectively establish the influence on tissue maturation.
While some studies evaluated native sEV efficacy, others modified EV content and functionality, bringing us to the following question:

Could modifying sEVs influence their therapeutic efficacy?
Modifying cells to improve their therapeutic qualities has been extensively investigated in recent years, with encouraging results. Several studies examined if manipulation of EV-producing cells by a variety of strategies translated into improved therapeutic functionalities of released sEVs in wound healing and skin regeneration. Endogenous loading of sEVs was achieved by physical loading with nanoparticles or genetic modification to overexpress certain proteins or nucleic acids. However, efficiency of loading and quantitation of cargo were seldom assessed. Furthermore, priming cells with assorted physiological and pharmacological cues showed a differential influence on the release profile, payloads, and hence downstream functionality of released sEVs [53], emphasizing the need for careful protocol and agent selection. Culturing MSCs with sEVs from neonatal versus adult sources resulted in more potent regenerative potential of released EVs [58], showing the importance of stimulant selection in sEV production. Notably, sEVs derived from modified cells had improved healing properties compared with their non-modified counterparts at all levels (re-epithelialization, proliferation, angiogenesis, collagen deposition, and maturation). It was not clear, however, if these modifications affected the stability, half-life, targeting (tropism), or internalization of these sEVs in injured tissues. On the other hand, encasing sEVs with biomaterials boosted their efficacy, possibly due to an enhanced release profile or other properties.
Indeed, the capability to manipulate EVs has sparked interest in creating "designer EVs" as a means to achieve desirable properties compared with native EV forms, including using them as drug delivery vehicles [29]. Besides manipulating parent cells (i.e., endogenous loading) as described by most studies here, manipulating sEVs post isolation is also possible. Such exogenous loading has been attempted passively [122] and actively (e.g., by sonication [123]), with varying success rates and loading efficiencies. In an exhaustive review aimed at developing a preliminary set of guidelines to standardize reporting of exogenous EV loading, Rankin-Turner et al. were surprised at the paucity of research into fundamental parameters such as incubation time and temperature, along with the variability and inconsistency in strategies [124]. They stressed the need for optimizing loading procedures, which we also fully endorse.
In the midst of numerous studies aimed at improving sEV function, none reviewed here explored altering the sEV surface to enhance targeting, retention, or delivery, or to prevent non-target uptake to prolong EV half-life. For example, conjugating EVs with polyethylene glycol (PEG) might enhance stability in the circulation and reduce build-up in the liver [125], most probably by masking scavenger receptors. Likewise, EV CD47 extends half-life by signaling macrophages, "do not eat me" [126]. EV surface modification has been done by chemical modification (CLICK chemistry and enzymatic conjugation) [127] and membrane cloaking [128]. These strategies may be worth investigating in the future to increase sEV stability and targetability in skin tissues.

Translational challenges to sEV therapy development
Despite encouraging findings, we have identified translational challenges that must be addressed before moving forward with human clinical trials.

Source
What is the ideal source of sEVs? In our review, sEVs were derived from tissues, biofluids, and cultured cells (Figure 4), but research comparing the efficacy of different preparations is lacking. Only a single study compared the efficacy of human umbilical cord-derived-MSCs with human lung fibroblasts (HFL1), concluding that MSC-EV preparations were more potent in enhancing skin regeneration [45] and suggesting source-dependent variation in activity. This emphasizes the need for evidence-based source selection. No studies examined the effect of 3D culture or bioreactors as scalable culture approaches for generating sEV skin therapeutics. Additionally, production of biological therapeutics requires verification of source identity. We examined compliance with ISCT guidelines for minimal identification of MSCs [40], finding that the criteria were met by just over half of the MSC-EV studies. To exclude heterogeneous and non-MSC cell populations, these criteria are important [129,130], and ongoing efforts to develop standardized EV potency assays are crucial [130].

Depletion of exogenous EVs
In vitro, serum-derived EVs from culture medium additives may contaminate EV isolates. These EVs might not negatively affect the potency of therapeutic EVs, but their removal may be necessary to ensure reproducibility. Also, contamination with xenogeneic EVs (e.g., from FBS) could complicate translation. Almost half of the studies included in this systematic review collected sEVs from serum-free media. While this is a good strategy to minimize contaminating EVs, switching to a serum-free condition might starve and stress the cells [131], modifying EV release profile and biological properties [132]. Depleting FBS of EVs before supplementation, primarily using ultracentrifugation, can also alter the activity of cultivated cells [133], although gradual transitioning can help [129]. However, depletion is rarely complete. Driedonks et al. revealed that various depletion strategies had a varying influence on the concentration and types of fetal calf serum RNA contaminants in medium and that optimizing purification techniques can result in lower contaminant levels [134]. While a minimum of 18 hours of ultracentrifugation was recommended to remove bovine EVs [135], depletion of only 70 minutes to three hours was reported in studies we examined. Additionally, contaminants can be present even in serum-free and so-called "chemically defined" media [136]. Because some of the reviewed studies did not disclose their depletion protocols, we cannot fully assess the degree of purity of EV preparations. Indeed, MISEV2018 asserts the importance of specifying the precise source, procedures, and reference for depleted components, as well as the significance of verifying the "EV-free" status (or otherwise) of all supplements [14].

Separation and purification
Separating a specific EV subpopulation effectively while eliminating non-EV impurities may be one of the biggest technical hurdles in developing EV therapeutics [137,138]. Since "the process is the product" [139], different preparation procedures, even starting with the same source, might yield a diverse mixture of co-isolates and sEV subsets [140] with varying attributes such as size and biogenic origin. Thus, different separation techniques could lead to functional and physical differences [140]. While it is still not clear how the different sEV subgroups with varying roles (e.g. disposal vs. signaling) and origin (endosome vs. plasma membrane) interact to influence the final healing outcomes [141], contaminants, on the other hand, were found to affect the therapeutic effect of EV preparations [142]. Interestingly, it cannot be excluded that so-called contaminants might even contribute to desirable effects.
In part for this reason, MISEV2018 does not recommend a specific separation strategy, and combining multiple strategies may improve on single technique approaches. If the goal is to attribute all therapeutic effects to EVs only, high purity preparations should be obtained [14]. Multiple pre-processing steps are needed to isolate EVs from complex biofluid or tissue sources, and thus specific protocols must be tailored to the particular source to remove contaminants before EV separation [143]. The studies we reviewed used diverse separation procedures. Heterogeneity in the separation methods reflects a lack of standardization, hampering comparability and possibly delaying clinical translation [141]. Classical methods like the various protocols of ultracentrifugation still predominate, even though this technique may result in aggregation [144] or functional [140] of EVs and has limited scalability [145]. Precipitation-based concentration kits were also popular among the analyzed studies despite high protein contamination [146] and debatable usefulness [129,146]. Liangsupree et al. found a trend towards size exclusion chromatography (SEC) in recent years [147], but we did not observe this trend for the field that we reviewed. We also did not see uses of size-based separation techniques such as tangential flow filtration (TFF), SEC, or asymmetrical flow field-flow fractionation, or affinitybased technologies. Nonetheless, there was an increased reliance on method combinations and washes to improve purity.
Separation methods for sEV therapeutics should be selected for their scalability, automation potential, ability to optimize the purity and recovery of target sEVs [147], and translatability. Given the parallels between EVs and viruses, established virus purification techniques could be useful for separation and purification of EVs on a large-scale [145]. Concerningly, none of the research reviewed here reported producing GMP-compliant sEVs, mirroring a recent systematic review of EV therapies for lung conditions [148]. Unfortunately, high cost is not a valid excuse for lack of GMP compliance, as EVs will otherwise not be usable in the clinic.

Characterization
Given the inherent difficulties of standardizing sEV source and separation techniques, quantitative identity measures are required for improved post-separation sEV identification, quality control, and comparability. Characterization covers three aspects: identity, integrity, and purity [14]. MISEV2018 guidelines recommend that, at a minimum, the concentration of EV preparations (such as protein and particle count) and EV markers (proteins expected to be enriched or depleted in EVs) should be determined. Of these markers: (1) at least two EV-protein specific markers (referred to as positive or enriched markers) must be detected to confirm EV isolation: one that is an integral membrane or membrane-anchored protein, signifying the presence of the lipid bilayer, and another that is an established cytosolic protein, implying the presence of enclosed cytoplasmic content and not just membrane fragments; and (2) at least a negative or depleted protein marker including source-related protein contaminants (such as albumin in EV plasma or uromodulin in EV urine samples), subcellular compartments, or cell death artifacts. Moreover, MISEV2018 suggests two complementary single vesicle analysis approaches to visualize and evaluate EV biophysical features. Here, preparations were mostly evaluated for protein concentration, particle count, morphology, and EV protein markers, employing multiple techniques; but just a few satisfied all MISEV2018 characterization criteria. To determine size distribution, nanoparticle tracking analysis (NTA) was the most-used method. Since our inclusion criteria stipulated that studies of sEVs with a size distribution beyond 30-200 nm were to be excluded, the studies that were included fell within this range. As different sizing platforms may have varying detection limits resulting in diverse outputs, employing multiple orthogonal technologies may be useful [149].
Although NTA was used to measure particle concentration, most studies reported protein concentration, but not particle concentration, usually by colorimetric assays like BCA. However, data on total protein yield were mostly lacking. Total protein assays do not distinguish EV and non-EV proteins, i.e., sEV preparation purity [150]. As a result, the MISEV2018 guidelines advocate assessing protein-toparticle, lipid-to-particle, or lipid-to-protein ratios to indicate sample purity, as well as negative or depleted markers. Since ratios and negative markers were rarely documented in the reviewed research, the purity of the tested preparations cannot be assessed. It is important to note that co-isolated impurities should not be confused with the extravesicular cargo or loosely-associated factors [17], which may contribute to EV bioactivity [151][152][153], although the distinction between the two is still being defined [17].
Interestingly, our meta-analysis revealed a lower heterogeneity index for the wound closure rate outcome for the studies that adhered to MISEV2018. However, the reviewed articles emphasized "positive" EV markers, particularly tetraspanins (CD63, CD9, and CD81), over cytosolic markers (TSG101, Alix, and HSP70). While tetraspanins may be functional [154], their involvement in sEV-mediated skin wound healing is still largely unknown. Western blot was the preferred approach in these studies due to its utility for bulk analysis and protein marker identification, but it is less helpful to analyze EV subpopulations or to understand marker allocation within positive subpopulations [150]. Single-particle approaches like flow cytometry were less popular here. Morphology was assessed mainly by TEM. Collectively, these findings are consistent with previous reports, in that these analysis techniques remain the most favorable by the EV community [155]; however, expanded approaches are needed to understand the contributions of EV subtypes and non-EV components.
Encouragingly, opportunities for advancement abound. Despite the unique signature of lipids in EVs, they were rarely characterized in the covered studies. The current state of knowledge on the structure and function of EV lipids is still rudimentary [156]. EV lipids may play a part in mediating EV bioactivity [157], including during senescence induction [158]. Here, only a single study identified sEV lipids as key players in the mechanism of sEV-mediated stimulation of wound regeneration [83], highlighting the need for more investigation. Similarly, standardized reference materials are needed to facilitate benchmarking. Nanoparticles and EVs are being engineered or recombinantly synthesized to allow calibration, validation, and quality control [150]. As EV research advances, characterization technologies are also being applied in novel ways to EVs: for example, ion-mobility spectrometry (IMS), Raman spectroscopy (RS), and nano-flow cytometry (Nano-FCM) [159].

Dosage regimen
Optimization of dosing (method of delivery, amount per administration, frequency, interval between doses, and duration of the intervention [160]) is crucial to achieve full therapeutic potential while mitigating off-target effects. For wound healing and skin regeneration, the optimal sEV dosage regimen has yet to be established.

Methods of administration
In most studies we reviewed, sEVs were directly infused to the wound site, primarily via injection, and to a lesser extent by direct application after incorporation into biomaterial scaffolds. Local delivery of sEVs has several advantages, circumventing phagocyte and circulatory clearance (primarily in the liver and spleen [126]) while increasing bioavailability in target tissues and reducing the required therapeutic dose. sEVs were found to exhibit tropism for specific organs (lung, liver, kidney, spleen) when delivered to the systemic circulation [161], but the single study that compared sEV intravenous (IV) with subcutaneous (SC) administration produced an unexpected finding: IV-administered sEVs promoted skin regeneration more effectively than SC-administered equivalents, homing to the wound bed starting from day one of treatment [94]. While prior studies suggested preferential affinity of sEVs for injured tissues [162], the reason for the observed variation in therapeutic impact between IV and SC requires additional investigation.
Non-healing lesions requiring protracted therapy may have limited clinical translatability because of patient compliance. Incorporating sEVs into biomaterial scaffolds might reduce the frequency of application and enable delivery of sEVs in a targeted and concentrated manner. Here, hydrogels (both synthetic and natural) were the most preferred biomaterials. In these biomaterials, EVs are shielded against rapid clearance and destruction in the hostile wound environment, prolonging their duration of release and bioavailability [44]. Hydrogel dressings are typically used to rehydrate the injured tissues, insulate against infections, and provide a temporary framework for host cells to penetrate and adhere to before being replaced by native ECM [163], or for their antibacterial qualities [164]. Combining these qualities with the proangiogenic, pro-regenerative, antiinflammatory, and anti-scarring features of sEVs was beneficial. Several studies in this review utilized "smart" biomaterials: temperature, light, and pHresponsive, enabling spatially and temporally orchestrated release of sEVs. Light-triggerable hydrogels loaded with sEVs accelerated wound closure better than non-triggered hydrogels or even several doses of sEVs, underscoring the critical role of the controlled release system in boosting healing kinetics [44]. Advances in 3D bioprinting will enable scaffolds to be tailored to each patient's unique demands. It is critical, however, to investigate the possible interaction of biomaterials with sEVs, and their effect on the functionality and physical attributes of the vesicles.

Dose and frequency
To date, there are no guidelines for EV dose selection. Here, experimental doses ranged from 2 to 5000 µg of protein. In most studies, dose selection was unclear and did not account for wound size, condition, type of injury, skin type, animal weight, or route of administration. Since protein levels are influenced by purity, dosage estimation may also be unduly influenced by impurities, which were mostly unreported. Dose-response studies were also lacking, and dose was based only on in vitro experiments in 2D in some studies, although the dose needed in vivo to present the therapeutic benefits may differ substantially from that projected in vitro [16]. Two studies concluded that sequential administration of small doses was superior to a single administration of a larger dose [44,82]. Furthermore, sustained release from a light-responsive biomaterial scaffold outperformed high-dose and sequential administration regimens [CITE]. Both head-to-head comparisons and quantifiable measures of potency and quantity of biotherapeutic molecules in EVs are needed to establish dosage [165].

Labeling and tracking of sEVs
Despite the importance of in vivo stability, only six studies tracked transplanted sEVs (Table 4). In these studies, sEVs were found to be localized to the wound bed and were detected in the cytoplasm of skin tissue cells. However, the wide range of follow-up time points and duration hampered comparability and conclusions about how long a single sEV dose remained detectable in tissue. While one study observed that sEV signals lasted up to 21 days (IV delivery) [94], another noted that they disappeared after only five days [87]. Sjöqvist et al. reported a loss of signal four days after injection due to interference by hair growth and scab formation; however, the signal could be detected later after extraction of tissue (ex vivo) [82]. Another study noted that the drop in sEV detection signal was less pronounced for light-triggered hydrogels than for free sEVs, pointing, as previously highlighted, to the potential of sustained-release systems in prolonging sEV bioavailability.
Accelerated clearance of systemically introduced sEVs after repeated administration has been reported in the literature, probably due to development of immune responses [166]. Further investigation is needed to determine if this happens in skin tissues. Additionally, the depth to which sEVs penetrated damaged tissues was not investigated. When Zhang et al. assessed the potential of sEVs to permeate an ex vivo model of intact human skin, they noticed that sEVs could not penetrate beyond the stratum corneum of the epidermis, proposing it as a possible area of activity [167]. However, sEVs would plausibly penetrate deeper in skin injury versus intact skin because of barrier disruption. This should be tested in skin injury.
Beyond general dose and tracking questions, we believe that upcoming studies should address fundamental questions including: 1) What proportion of sEVs are internalized by cells? What are the target cells? Is internalization required to exert the therapeutic effect? How about surface ligand-receptor interactions? Does uptake of sEVs necessarily mean functional delivery of payloads (membranemembrane fusion)? Are sEVs similarly internalized by different types of cells in the skin, or do they show different tropisms? What is the influence of cell source or various preparation techniques on sEV uptake and tissue distribution? Importantly, sEVs were predominantly labeled with fluorescent lipophilic dyes in the analyzed studies. These dyes sequester into the lipid areas of the phospholipid membranes and are highly photostable [168]. As this is a widespread approach for in vivo tracking of labeled cells with negligible alteration of functional and mechanical properties of cells [168], this approach may not be preferable for tracking sEVs due to the lack of specificity.
Unfortunately, lipid dyes also label cellular membranes [168], and some have been found to have low EV staining efficiency [169], bind to non-vesicular impurities such lipoproteins and soluble proteins [170], or aggregate into nano-sized micelles that can be confused with sEVs [130,169]. They could also leak from EVs and persist in tissues, greatly outlasting the labeled EVs in vivo due to long half-life, causing misinterpretation of EV fate in vivo [171]. This lack of specific labels makes tracking EVs a challenging task. New labeling techniques with improved performance are emerging, such as radiolabeling [172] and magnetic labeling [173], yet all with pros and cons, mandating use of complementary technologies for a better understanding of EV distribution in tissues [130].

Relevance of animal models and external validity aspects
The studies here investigated rodent models for sEV efficacy (97% of studies), including diabetic and non-diabetic mice and rats. Type 1 diabetes was predominantly modeled, while type 2 diabetes, the most prevalent of the two, was underrepresented. Similarly, full-thickness excisional wounds were assessed more frequently than other lesions, such as diabetic foot ulcers, pressure ulcers, incisional or partial thickness wounds, and burns.
Although rodent models have been crucial in the initial stages of new drug screening and testing, inherent structural and physiological differences between rodent and human skin have spurred heated arguments on the translational relevance of these models [174]. Indeed, 90-95% of drugs that show promise in pre-clinical trials fail to translate to the clinic [175]. Despite similarities in portraying the four overlapping phases of wound healing, variances in life expectancy, the number of skin layers and thickness, and healing mechanism (e.g., contractionbased) are among the many factors that might undermine clinical applicability [174]. Furthermore, the extent to which existing diabetes models accurately represent chronic, non-healing wounds remains an open subject [3]. For example, in the reviewed studies, diabetic wounds in control groups were capable of independent healing. Also, relevant non-healing wound scenarios such as infected wounds or aged-animal wounds were understudied. Collectively, these concerns challenge the "external validity" of the studies, that is, the degree to which the outcomes can be generalized to human disease [176].
Since each strategy has limitations, we recommend a combination of approaches. Apart from rodent models, which have already demonstrated significant success in promoting cutaneous regeneration, larger animal models such as swine and non-human primates might be important to study. Together, evidence from diverse species might better predict human response [177]. Supporting data from emerging approaches that aim to achieve the 3R principles (replacement, reduction, refinement) for animal experiments are also valuable. This includes evidence from technologies such as microfluidic tissues-on-a chip that involve multiple system interactions and enable real-time mechanistic readouts [178]. Other approaches include in silico experiments, artificial intelligence, and machine learning [179]. Additionally, more representation of lesions such as burns and diabetic ulcers is needed. Consistent evaluation of the testing system's predictability and applicability is critical for determining the effectiveness of each approach.

Internal validity aspects
Successful translation requires not only external, but also "internal validity" [176]: the rigor with which experiments are designed, conducted, and reported [176]. Inadequate reporting impedes assessing experiment validity and is a common issue in preclinical research [180]. In some domains, the scientific community has developed guidelines and checklists of a minimum set of items to be documented in comprehensive and transparent reporting, such as the MISEV2018 recommendations [14], EV-TRACK, and the MSC-EV criteria mentioned previously. Compliance with these guidelines has been correlated with enhanced reporting, experimental design, and conduct [181,182]. Here, we observed that incomplete reporting affected many aspects of research design and methodology. For instance, there was an overall poor reporting of pre-processing and EV source, the number and age of animals, how sample size was determined, and of numerical data for means or variation measures for the outcomes. In some cases, experiments were reported that were not covered in the method section. Only a single study submitted procedures to EV-TRACK [82]. This incomplete reporting limited our ability to examine key elements that are identified by the SYRCLE's ROB tool as plausible sources of bias in animal experiments [95]. As a result, we concluded an unclear risk of bias for most of the elements ( Figure  6), presenting reasonable grounds for concern about the true level of bias in the included studies. Failing to handle internal validity threats is likely a key contributor to overestimated effect sizes in preclinical research [183]. Adoption and documentation of better reporting measures would help to mitigate or identify bias and increase confidence in the evidence and quality of the experiments [184]. The non-mandatory nature of the reporting guidelines may explain their low uptake in published studies. While authors should strive to provide detailed experiment reports, journals and funding agencies are the gatekeepers and should endorse compliance with reporting guidelines [180].
Is sEV intervention a safe option in skin regeneration therapy?
The possibility of adverse reactions to sEV therapy received scant attention, although the few studies that examined this possibility detected no serious adverse effects. Interestingly, no reports indicated immune rejection even of allogeneic or xenogeneic preparations, which constituted the majority of the tested preparations. Of note, Lu et al. observed a mild immune response to allogeneic iPSCs but not to their sEVs, although allogeneic sEV viability in treated tissues was inferior to that of their autologous counterparts [42]. These findings collectively are in line with prior research supporting sEV safety in treating various conditions [20]. Although hundreds of animal studies and clinical trials have demonstrated the biosafety profile of sEVs, safety assessment is integral to any novel therapeutic product development. As emphasized earlier, each EV preparation is a unique product due to the heterogeneous components involved in its manufacture [139]. In the absence of guidelines by regulatory bodies to assess the associated risks, verification of each product's safety is necessary [29]. This entails evaluating the potential risks associated with each stage, from source selection to delivery to patients [185].
Given that sEVs are messengers for their parent cells, caution is needed in source selection. While MSCs have an established general safety profile [20,129], sEV therapies are derived from various sources, not just from MSCs. Some of these sources have unclear safety profiles. Earlier studies in the literature showed examples of immune responses to immune cell-derived-EVs, for instance, when delivered to patients [186]. Additionally, immortalized cell lines and cancer cells may pass along oncogenic components through EVs [141,187]. EV-associated extravesicular cargo [17] such as protein coronae may also elicit immune responses, which should be considered [188].
As discussed earlier, adverse effects associated with manipulation of source cells or EVs should also be assessed. For instance, extending EV half-life might have the potential to induce fibrosis. Additionally, the risks associated with recurring sEV administration, overdosing, and off-target interactions merit further examination. An improved understanding of sEV biodistribution and bioavailability in treated tissues will surely aid in deciphering these interactions [141]. It is also imperative that regulatory approval is obtained before the initiation of clinical trials and application in clinics to confirm safety and efficacy [29]. Unlicensed "exosome"-containing products that might result in devastating consequences for patients led the FDA to issue a "Public Safety Notification on Exosome Products" [189], stressing the need for regulatory approval to ensure safety and efficacy before initiating clinical trials and application in clinics [29].

Concluding remarks and future prospects
Providing effective therapeutic options to promote skin regeneration is a key goal that has yet to be met, placing a continual burden on public health. This health burden is projected to continue to grow as diabetic and elderly populations expand. sEV therapy is proposed as a promising biological therapeutic approach that is able to potentiate skin regeneration. In this review, we have systematically examined studies evaluating sEV potential to induce wound healing and skin regeneration. We critically analyzed findings, reporting, and methodology and discussed strengths, pitfalls, and challenges. Collectively, our critical review of the available preclinical evidence supports the therapeutic efficacy and safety of sEVs in cutaneous wound healing and skin regeneration across all outcomes (wound closure, angiogenesis, anti-inflammation, re-epithelialization, scarring reduction, collagen production), regardless of cell source, production protocol, or disease model. Modifying sEV cargo appears to potentiate therapeutic functionality, resulting in better healing compared with unmodified sEVs. Moreover, biomaterial scaffolds could offer a promising tool for targeted sEV delivery.
However, these exciting findings should be interpreted with caution for several reasons. First, most studies have unclear risk of bias, and insufficient reporting hampered our ability to make an informed judgment. Second, the absence of purity assessment makes it difficult to attribute the observed efficacy to sEVs versus co-isolates. Third, the animal models used may not be faithful representatives of human disease, although this is a general limitation of animal testing and not unique to sEVs. Lastly, heterogeneity in sEV production, sources, characterization, and dosage regimens challenge direct comparisons.
We believe that adherence to established guidelines such as MISEV2018 will facilitate the conduct of rigorous research and enhance its validity. The high failure rate in translating preclinical discoveries to the clinic represents a waste of time, effort, and resources. Therefore, applying rigor and analyzing risks from the early stages of development will largely contribute to greater translatability and success stories. As a result, we urge researchers to follow best reporting practices, including a detailed description of their rigorous work in order to enable reproducibility, comparability, and evaluation of the studies' evidence. Submission to EV-TRACK is a great approach to enhance transparency and may also serve as a guide for authors regarding which items should be reported. To account for the shortcomings of confining research to rodent models, we propose examining a variety of models, including non-rodent and non-animal models. We further recommend that the follow-up period be extended to adequately assess the effect of sEVs on the healed skin maturation and scarring, as well as the immunological response. Future research should cover other skin lesions as well, such as burns, ulcers, incisional ischemic lesions, and more representative models of diabetic type 2 and non-healing wounds.
Moreover, there are several unresolved issues that warrant future research. These include establishing the optimal sEV source with the minimal possible risks and greatest therapeutic efficacy as well as identifying the components responsible for the ascribed functions. In this regard, the evaluation and reporting of the preparations' purity should not be overlooked and should be regularly identified. Furthermore, more research is needed to establish the basis for dosage regimen selection to achieve the best outcome; and to evaluate the in vivo pharmacokinetics and pharmacodynamics of sEVs in skin tissues. Since the evaluation of possible unwanted effects has received less attention, it is critical for future research to conduct a full review of the safety aspects associated with sEV therapy of skin lesions. Comprehensive characterization is key in evaluating the role of EV cargo safety. Indeed, the variability of sEV production processes, sources, and characterization presents a challenge, however, the development of quality controls and suitable reference materials might provide the foundation for standardization, allowing for lab-to-lab comparability. Nonetheless, the current expansion of novel and intriguing approaches is encouraging. Also, personalized sEV skin therapy products might be the future direction as each skin type, condition, and defect may have different therapeutic demands.
The future for sEV skin therapies seems bright if we learn the right lessons. Along the way, decades of experience with enveloped viruses and synthetic nanoparticles, which have many similarities with sEVs, have much to teach the EV field, particularly regarding the former's large-scale production and purification and the latter's delivery. Safety and efficacy evaluation of biologics (such as cell therapy) may also serve as reference models for EV therapy. However, clinical applications should not be hastened prior to clinical trials and regulatory permissions, as this will put patients at risk and may jeopardize the field. Although the road to the clinic is still paved with obstacles, sEV therapy holds tremendous potential as a biological cell-free therapeutic modality capable of promoting wound healing and skin regeneration. Collaborative efforts are needed to realize this potential and successfully translate it into practice.

Literature search strategy
The protocol of this study was developed a priori, peer-reviewed, registered, and published in the International Prospective Register of Systematic Reviews (PROPSPERO; protocol ID: CRD42020159 994). The current review covers only the in vivo part of the intended study. The search was initiated by forming a query (i.e., keywords) sourced from a number of related published studies and the Medical Subject Headings (MeSH) Database. The formed query was further checked by experts in the field (JX.L., M.A.A., JB.F.). Three bibliographic databases containing peer-reviewed journals were searched, namely, Web of Science (Science Citation Index Expanded), Scopus, and PubMed. Where applicable, filters were applied to include only original research publications, that were written in English and had the search terms in the title and abstract. No restriction on the date was applied. All returned articles were pooled from the three databases to EndNote, and duplicates were removed. Initially, the search covered studies that were published until 11 th November 2019. Later, the search was updated to include related studies published from 11 th November 2019 to 1 st March 2021, obtained through an activated search-alert that was set earlier. Study selection was carried out at two stages by two reviewers independently (M.E.A., J.X.L.). Disagreements were resolved by discussion and by consulting the third and fourth reviewers (M.A.A. and J.B.F.). The first search stage involved screening titles and abstracts of retrieved studies, while the second stage was based on reviewing the full text of articles according to the predefined exclusion and inclusion criteria. The search query was adjusted as needed to function with each respective database. Keywords used were as follows: Web of Science: (TS= (exosome* OR "extracellular vesicle*" OR nanovesicle*) AND TS=("skin regeneration" OR "skin rejuvenation" OR "wound healing" OR "skin repair")) AND LANGUAGE: (English) AND DOCUMENT TYPES: (Article).

Eligibility criteria
For the qualitative synthesis of studies, at the first stage we included only those articles that were 1) peer-reviewed original research articles; 2) written in English; 3) evaluated exosome/sEV therapeutic roles in wound healing and skin regeneration; and 4) conducted in mammalian animal models. Human trials were not covered in this review. We excluded studies that 1) were not original research, such as reviews, letters, commentaries, and conference proceedings, as well as those that were non-peer-reviewed (including preprints); 2) were written in languages other than English; 3) were unrelated to sEV applications in wound healing; and 4) were on non-mammalian animal models.
At the second stage of search, only studies that had 1) controlled interventional design; 2) examined at least one sEV positive marker (of the markers recommended in MISEV2018 [14]); 3) experimentally confirmed the size of isolated sEVs (30-200nm); 4) investigated wound healing and skin regeneration either macroscopically or microscopically, qualitatively or quantitatively; and 5) of which full text could be accessed either online or after request from the authors. There were no restrictions on sEV source, source cell manipulation, scaffold or vehicle use, or control type. We excluded studies that: 1) did not establish sEV identity by determining the size; 2) did not establish sEV identity using at least one EV positive protein marker; 3) did not assess wound healing or skin regeneration macroscopically or microscopically, qualitatively or quantitatively; 4) were not controlled; 5) did not include the pre-specified primary outcomes or reported insufficient data on the outcomes; or 6) their full text could not be retrieved despite contacting the authors.

Data extraction and synthesis
Data were extracted by two independent groups of reviewers (M.E.A and CY.N. as well as U.V. and R.S.) and arranged in a pre-designed data extraction form prepared by the reviewers. Disagreements were resolved by discussion and by consulting the third and fourth reviewers (JX.L. and M.A.A.). Data were extracted from texts, tables, figures, supplementary materials and references cited for methods. These included intervention design, dosage (amount), dosing frequency, route of administration, doseresponse assessment, biocompatibility, vehicles, and comparators; characteristics of the animal models used, including species/strain, gender, age, body weight, disease model, wound model, size, and location; as well as the number of animals per group. Data on the EV source cells were also collected including cell type, modification type, if any, cell passage at extraction. Also, data on exogenous EV depletion regimens, methods of sEV separation, and characterization. In addition, any data on labeling and tracking of sEVs after administration were also collected. The primary outcomes extracted from studies were quantitative and qualitative data related to wound healing process evaluation, including wound closure rate; re-epithelialization; collagen deposition and maturation; angiogenesis; and scar assessment. The secondary outcome data extracted were qualitative data related to any reported adverse effects and inflammation. Additionally, general study characteristics such as authors, year, and country of corresponding author/s were obtained.

Quality and risk of bias assessment
The Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE's) risk of bias tool was used to assess each study's risk of bias [95]. This tool comprises a ten item-checklist: (1) random sequence generation, (2) baseline characteristics, and (3) allocation concealment, to evaluate selection bias; (4) random housing and (5) researcher blinding, to evaluate performance bias; (6) random outcome assessment, and (7) blinding of outcome assessment, to evaluate detection bias; (8) incomplete outcome data, to evaluate attrition bias; (9) selective reporting, to evaluate reporting bias; and (10) other source(s) of bias, if any. We modified the tool to include another item, declaration of the randomization method, but we excluded point (10) and thus did not check for other sources of bias.
We further assessed sample size calculation, quality of reporting, adherence to MISEV2018 characterization criteria, nomenclature, purity assessment, and EV-TRACK submission [106]. We also evaluated the adherence to ISCT minimal criteria to characterize mesenchymal stem cells (MSCs), where applicable. We evaluated the animal and disease models used to examine the external validity of the studies.

Meta-Analysis
The studies identified through our comprehensive search were checked for eligibility for a meta-analysis. We performed meta-analysis for three outcomes: wound closure rate, scar reduction, and angiogenesis, using Review Manager 5.4.1 (Cochrane) [190], comparing naïve sEVs (unmodified) with placebo controls. Meta-analysis was performed only when three studies or more reported the same outcome using the same scale. We retrieved means, standard deviations, or standard error of the mean as stated in the studies and emailed the authors to obtain missing data (including sample size), where necessary. We extracted data from figures in studies where numerical data were not available using an online application (WebPlotDigitizer, https://apps. automeris.io/wpd/). We excluded from the meta-analysis the studies for which the sample size was not provided even after contacting the authors. Standard mean differences (SMD) and pooled size effects were estimated using random effects model, taking into account the diversity in sEV preparation, source, and wound size. Statistical heterogeneity (a measure of the heterogeneity of intervention effects across studies) was measured using I 2 index. A subgroup analysis of naïve sEVs versus placebo control was performed to assess if the effect of sEVs varied between diabetes and non-diabetes models. Additionally, we conducted another meta-analysis of wound closure rates for studies that reported characterization using MISEV2018 criteria, with subgroup analysis of diabetes and non-diabetes models. Statistical significance was considered when p < 0.05. To note, outcomes of studies that were not included in the meta-analysis were described qualitatively.