Mesenchymal stem cells are a heterogeneous subset of pluripotent stromal stem cells that are easily isolated from various tissues, including adipose tissue, peripheral blood, bone marrow, synovial fluid, muscle, placenta, umbilical cord, and dental pulp (Uccelli et al., 2008). The minimal criteria for defining MSCs are: the ability to self-renew and differentiate into classical mesodermal lineage cells such as osteoblasts, adipocytes, and chondrocytes in vitro and in vivo; a CD105+, CD73+, CD90+, and CD45–, CD34–, CD11b–, CD79a–, CD19–, and HLA class II- expression profile; a fibroblast-like morphology; and, adherence to tissue culture plastic in vitro (Horwitz et al., 2005; Dominici et al., 2006). Among the various MSC sources, bone marrow (BMMSCs) and adipose (ADMSCs) are the two most commonly used in preclinical and clinical tissue regeneration applications. While, umbilical cord- derived MSCs (UCMSCs) have also been widely employed in research and clinical trials, their use in many applications is limited since they are not practical for autologous administration in adults (Kern et al., 2006). Although BMMSCs were the first MSC type to be characterized and are the most widely used (Caplan, 1991), ADMSCs are an attractive alternative as they are higher in frequency, more easily obtained and cause less donor site morbidity (Reumann et al., 2018). Furthermore, ADMSCs display a higher proliferation rate than BMMSCs in vitro and show a greater ability to maintain their stem cell characteristics, including self-renewal, proliferation, and differentiation potential, after repeated passaging (Zhu et al., 2008).

While both BMMSCs and ADMSCs have been successfully employed in preclinical tissue repair and disease models to promote angiogenesis (Jin and Lee, 2018; Zhang et al., 2019; Ryu et al., 2020), induce bone regeneration (Jin and Lee, 2018) and modulate the immune system (Tao et al., 2016; Zhao et al., 2016; Waldner et al., 2018), there appear to be several differences between cell types. In vitro studies have shown that BMMSCs exhibit significantly higher chondrogenic differentiation capacity (Noël et al., 2008; Mohamed-Ahmed et al., 2018), while ADMSCs show significantly higher adipogenic capacity in vitro (Mohamed-Ahmed et al., 2018). ADMSCs also display a higher endothelial differentiation capacity in vitro than BMMSCs (Fan et al., 2016), and superior angiogenic capacity in several preclinical ischemic injury models (Ikegame et al., 2011; El-Badawy et al., 2016). However, it remains unclear which MSC source exhibits greater osteogenic capacity or immunomodulatory potential. While some in vitro studies showed higher osteogenic differentiation in BMMSCS than ADMSCs (Park et al., 2012), others showed the opposite (Kang et al., 2012). More significantly, no significant differences in bone regeneration ability were observed in vivo between the two MSC types in rat cranial defect models (Wen et al., 2013) or canine radius defect models (Kang et al., 2012). Similarly, both MSC types showed comparable immunomodulatory potential in an immunocompetent myocardial infarction (MI) model (Paul et al., 2013), while BMMSCs displayed greater immunomodulatory potential in an endotoxic shock model (Elman et al., 2014), and ADMSCs demonstrated more effective immunosuppression of peripheral blood mononuclear cells and T-cells in vitro (Waldner et al., 2018).

While the therapeutic effects of transplanted MSCs were originally thought to be due to direct cell replacement (Friedenstein et al., 1968), research soon showed that intravenously administrated MSCs were largely caught in capillaries and/or cleared (Fischer et al., 2009), and that remaining MSCs contributed to short-term therapeutic effects (Caplan and Dennis, 2006). It is now widely theorized that the therapeutic effects of MSCs are mainly due to paracrine secretion of various growth factors, glycosaminoglycans, cytokines and EVs which modulate angiogenesis (Pankajakshan and Agrawal, 2014), apoptosis (Pan et al., 2012), proliferation (Di Nicola et al., 2002), differentiation (Chiossone et al., 2016), and the immune response (Dyer et al., 2014) to create a reparative microenvironment (Phinney and Pittenger, 2017). Secreted by the majority of cell types, EVs are phospholipid vesicles of different sizes, including micro-vesicles (MVs) (200 nm–1 μm) and exosomes (50–200 nm), that transport proteins, lipids, and nucleic acids (Hunter et al., 2008). Exosomes are generated in multivesicular bodies by the endosomal compartment and express endosomal markers (CD9, CD61, CD83, ALIX, TSG101) (Cosenza et al., 2017) and surface molecules that allow them to be targeted to recipient cells (Mathivanan et al., 2010). Meanwhile, MVs are the outcome of direct outward budding of the cell plasma membrane and thus carry cytoplasmic contents (Heijnen et al., 1999). EVs are recognized and internalized by recipient cells through receptor-ligand interactions (Raposo et al., 1996), endocytosis and/or phagocytosis (Morelli et al., 2004), or they can fuse with the target cell membrane and deliver their contents into the cytosol (Tkach and Théry, 2016). Recent research suggests that the paracrine efficacy of MSC-based therapies can largely be attributed to EVs. For example, conditioned MSC culture media was found to have therapeutic effects similar to direct delivery of MSCs in rodent models (Gnecchi et al., 2005; Aslam et al., 2009). Subsequently, Timmers et al. (2008) demonstrated that it was the EVs within the conditioned media that actually were effective.

Extracellular vesicles can be harvested via a variety of methods from cell culture media or clinical samples such as blood plasma, urine, and saliva. The most frequently employed isolation methods include differential ultracentrifugation and density gradient ultracentrifugation, both of which involve centrifugal forces greater than 100,000 × g and can fractionate EVs from their liquid sample of origin into subsets based on size, density, and mass (Zarovni et al., 2015; Li et al., 2017). EVs harvested from different tissues display varying content profiles, depending on their origin, age, state and environment (Tkach and Théry, 2016). For example, the microRNA (miRNA) profile of MSC-derived EVs from myoplastic syndrome patients is significantly different compared to that of EVs from disease-free patients (Muntión et al., 2016). Among EV contents, the function of bioactive lipids and proteins have been well-studied (Toh et al., 2018; Skotland et al., 2020). However, nucleic acid cargo, including mRNA, miRNAs, and other non-coding RNAs has become an increasingly hot topic in EV research. miRNAs, which are small (19–23 nucleotide) non-coding RNAs (Lau et al., 2001) that regulate gene expression via specific binding to messenger RNAs (mRNAs) (Lai, 2002), make up a large portion of the cargo within EVs (Valadi et al., 2007). miRNA transfer to recipient cells via EVs contributes significantly to paracrine signaling and has been found to be a main mediator of therapeutic effects in many preclinical studies. For example, miR-223 from BMMSCs-derived EVs contributed to cardioprotection in a surgically induced sepsis model (Wang X. et al., 2015).

The use of MSC-derived EVs in place of MSC transplantation in clinical treatments provides a number of potential advantages. EV therapies increase the accessibility of damaged tissues, since cultured MSCs are approximately 20 μm in diameter and thus tend to be caught and cleared by the circulation (Crop et al., 2010), whereas EVs are significantly smaller and have demonstrated transport through the pulmonary circulation and the blood-brain barrier (Batsali et al., 2020) (Bang and Kim, 2019). Unlike MSCS, which may undergo changes during in vitro culture that make them a clearance target of NK cells and macrophages (Eggenhofer et al., 2014), EVs are more likely to avoid immune rejection due to their low expression of membrane histocompatibility complexes (Lai et al., 2019). EVs are also more easily modified than MSCs to encapsulate desired therapeutic cargos, and are more easily stored than cells, since they are more stable when freezing and thawing (Lai et al., 2019). However, before the clinical application of MSC-derived EVs can be achieved, the optimal cell source for a given therapeutic application needs to be determined.

This review will compare the therapeutic effects of EVs isolated from BMMSCs and ADMSCs in various in vivo tissue repair and regeneration models (Figure 1). More specifically, the capacity of BMMSC- and ADMSC-derived EVs to induce angiogenesis, osteogenesis and immunomodulation will be investigated. EV cargos and any signaling pathways involved, where characterized, will also be detailed.

Studies employing BMMSC-derived EVs in preclinical models, including calvarial defects (Liang et al., 2019), myocardial infarctions (MI) (Teng et al., 2015; Xu H. et al., 2020), random pattern dorsal skin flaps (Xie et al., 2019), intracerebral hemorrhages (ICH) (Han et al., 2019b), fracture non-unions (Zhang et al., 2020), focal cerebral ischemia models (Doeppner et al., 2015), traumatic brain injury (TBI) models (Zhang et al., 2017), STZ-induced diabetic rat models (Yu M. et al., 2020), and subcutaneous implantation models (Narayanan et al., 2016), generally demonstrated that EV treatment stimulated localized vasculogenesis and/or angiogenesis (see Table 1). Meanwhile, ADMSC- derived EVs promoted neovascularization and angiogenesis in fat grafting models (Han et al., 2019a), acute ischemic stroke models (Chen et al., 2016), acute kidney ischemia/reperfusion (I/R) models (Lin et al., 2016), and MI models (Xu H. et al., 2020) (see Table 1). With far fewer studies employing ADMSC-derived EVs than BMMSC-derived EVs, it remains unclear whether once source displays greater angiogenic potential than the other. In one study that directly compared the effects of EVs derived from both cell sources, human ADMSC-derived EVs displayed significantly increased therapeutic potential compared to BMMSC-derived EVs in a rat MI model, as indicated by improved cardiac function, reduced cardiomyocyte apoptosis and infarction area and increased microvessel density (Xu H. et al., 2020). By contrast, when comparing two separate brain ischemia model studies, BMMSC-derived EVs appeared to display increased angiogenic potential compared to ADMSCs-derived EVs, with the former exhibiting an approximately 4-fold increase in the number of endothelial cells compared to controls, while the latter showed a 1.5-fold change (Doeppner et al., 2015; Chen et al., 2016).

Few of these preclinical studies have investigated the mechanisms and signaling pathways underlying the observed increase in angiogenesis induced by MSC-derived EV therapies. In a calvarial defect model, enhanced angiogenesis due to BMMSC-derived EV treatment was coupled with endogenous MSC migration (Takeuchi et al., 2019). While, in a rat full-thickness skin wound model, human BMMSC-derived EVs accelerated angiogenesis and the cutaneous wound healing process via inhibition of the TGF-β/Smad signaling pathway, as verified by RT-qPCR and western blotting analysis (Yang et al., 2018). Similarly, in a STZ-induced diabetic rat model, BMMSC-derived EVs accelerated wound closure and increased blood vessel area and number by activating the AKT/eNOS pathway (Yu M. et al., 2020). Increased angiogenesis induced by BMMSC-derived EVs is often linked to increased VEGF signaling, as was observed in a random pattern dorsal skin flap model (Xie et al., 2019) and a subcutaneous implantation model (Narayanan et al., 2016). Similarly, HIF-1α-VEGF signaling was found to be associated with enhanced angiogenesis at the fracture site in a non-union model after treatment with BMMSC-derived EVs (Zhang et al., 2020). Meanwhile, in an acute kidney I/R model (Lin et al., 2016) and an acute ischemia stroke model (Chen et al., 2016), ADMSC-derived EVs increased expression of the angiogenesis markers CD31, vWF, VEGF, CXCR4, SDF-1α, and angiopoietin. In another study that directly compared the angiogenic capacity of BMMSC-derived and ADMSC-derived EVs, similar numbers of new blood vessels were observed in a rat MI model (Wang K. et al., 2017). However, ADMSC-EVs yielded higher numbers of CD31+ cells, while EVs derived from endometrial MSCs displayed higher angiogenic capacity than EVs from either BMMSCs or ADMSCs (Wang K. et al., 2017).

In vitro experiments further showed that EVs isolated from both BMMSCs and ADMSCs possessed great potential for inducing angiogenesis and enabled more detailed study of the pathways underlying these effects. ADMSC-derived EVs enhanced angiogenic tube formation in human brain microvessel endothelial cells via increased expression of miR-181b-5p, which, in turn, directly targeted expression of the ion channel protein TRPM7 (Yang et al., 2018). Similarly, BMMSC-derived EVs induced angiogenic tube formation in HUVECs (Teng et al., 2015; Xie et al., 2017; Kang et al., 2020; Yu M. et al., 2020; Zhang et al., 2020) and enhanced expression of the angiogenesis-related genes VEGF, ANG1, and ANG2 in hBMMSCs (Takeuchi et al., 2019) and the mRNA expression of PDGF, EGF, and ANG1 in HUVECs (Yu M. et al., 2020). In a rare study that directly compared EVs derived from the three most commonly employed clinical MSC sources, i.e., BMMSCs, ADMSCs and UCMSCs, ADMSC-derived EVs yielded the highest in vitro protein expression levels of VEGF, bFGF, and HGF in rat neonatal cardiomyocytes and also showed the strongest inhibitory effect on apoptosis (Xu H. et al., 2020).

In order to improve angiogenic therapeutic capacity, many studies isolated EVs from MSCs cultured in hypoxic conditions or in the presence of dimethyloxalylglycine (DMOG), which enhances activation of HIF-1a. In a nude mouse model of fat grafting, for example, hypoxic ADMSC-derived EVs (hyp-ADSC-EVs) dramatically promoted neovascularization and increased the protein expression of VEGF/VEGF-R compared to EVs derived in normoxic conditions (Han et al., 2019a). Similarly, in other studies hyp-ADSC-EVs were found to express significantly higher levels of VEGF and VEGF-R2/R3 and promote increased HUVEC proliferation, migration and tube-formation in vitro (Xue et al., 2018; Han et al., 2019a), compared to EVs from cells cultured in normoxia. These hyp-ADSC-EVs also dramatically changed HUVEC expression levels of the angiogenic genes Angpt1, Flk1 and Vash1, and increased activation of the PKA signaling pathway (Xue et al., 2018). Interestingly, EVs from hypoxia-preconditioned ADMSCs exhibited increased diameters (by 59 nm) compared to those derived in normoxia (Han et al., 2019a). Meanwhile, EVs derived from DMOG-stimulated hBMMSCs increased HUVEC angiogenesis in vitro and decreased expression levels of PTEN (Liang et al., 2019), a tumor suppressor gene found to promote neovascularization by inducing HUVEC migration (Zhang et al., 2016). This decreased PTEN expression was further accompanied by increased expression levels of its corresponding downstream AKT/mTOR signaling pathway members, p-AKT, mTOR, and p-mTOR (Liang et al., 2019). Increased angiogenesis in a calvarial defect model treated with EVs isolated from DMOG-stimulated hBMMSCs was also observed (Liang et al., 2019).

Although both BMMSC- and ADMSC-derived EVs have been shown to promote osteogenesis in vivo, most preclinical studies use BMMSC-derived EVs to induce bone formation and fracture healing. BMMSC-derived EVs were shown to promote bone regeneration in rat calvarial bone defect models (Qin et al., 2016; Takeuchi et al., 2019), murine femoral fracture models (Furuta et al., 2016; Xu T. et al., 2020), rat models of distraction osteogenesis (Jia et al., 2020), and subcutaneous bone formation models in nude mice (Narayanan et al., 2016; Xie et al., 2017) (see Table 2). By contrast, the bone regeneration capacity of ADMSCs-derived EVs has only been explored in two studies to date in calvarial defect models (Li et al., 2018); one of which involved EVs derived from hADMSCs engineered to overexpress miR-375 (Chen et al., 2019) (see Table 2). With so few studies of ADMSC-derived EVs, it is difficult to compare their osteogenic capacity to BMMSC-derived EVs. However, in separate rat calvarial defect studies, treatment with BMMSC-derived EVs led to a greater increase (fourfold) in bone volume [i.e., (BV)/(TV)] compared to controls (Qin et al., 2016), than that induced by ADMSC-derived EVs (approximately 1.33-fold) (Chen et al., 2019).

The mechanisms and pathways underlying the osteogenic effects of MSC-derived EVs have not been widely reported. In a non-union model, BMMSC-derived EVs enhanced osteogenesis via the activation of the BMP-2/Smad1/RUNX2 signaling pathway (Zhang et al., 2020). In a femoral fracture model, BMMSC-derived EVs from young rats yielded increased bone formation and expression of Runx2, ALP and Col I compared to those from older rats; with EV osteogenic capacity linked inversely to levels of miR-128-3p expression, which was found to negatively modulate Smad5 signaling (Xu T. et al., 2020). Significantly, additional studies also found that miRNAs played an important role in promoting EV-mediated bone regeneration, with miR-196a regulating differentiation of osteoblasts (Qin et al., 2016), while miR-375 stimulated osteogenic differentiation of MSCs by inhibiting IGFBP3 (Chen et al., 2019).

In vitro studies of MSC-derived EVs further illustrated their role in bone regeneration through promotion of MSC proliferation, migration, and osteogenic differentiation (Narayanan et al., 2016; Takeuchi et al., 2019). Many studies have observed that treatment with BMMSCs-EVs increased osteogenic differentiation and upregulation of related genes, including RUNX2, Osterix, BMP9, TGF-β1, BMP2, TGF-β, OCN, ALP, Col I, and PDGF in BMMSCs (Narayanan et al., 2016; Xu T. et al., 2020) and osteoblastic MC3T3 cells (Zhang et al., 2020). Similarly, ADMSC-derived EVs were also found to promote osteogenic differentiation of human BMMSCs in vitro by significantly increasing the expression of RUNX2, ALP, and COL1A1 (Li et al., 2018; Chen et al., 2019). Interestingly, BMMSC-derived EVs isolated from young (2-week old) SD rats promoted proliferation and enhanced the osteogenic capacity of older BMSCs, and significantly upregulated the expression of ALP, Runx2, and OCN (Jia et al., 2020). In human osteoblasts, BMMSC-derived EVs enhanced differentiation, likely due to differential expression of miR-196a (Qin et al., 2016). In another study, miR-26 was found to be crucial to the in vitro capacity of BMMSC-derived EVs to induce osteogenic differentiation via silencing experiments (Luo et al., 2019).