Regeneration of bone defects often presents significant challenges, particularly in patients with decreased tissue regeneration capacity due to extensive tissue damage, disease, advanced age, and confounding systemic and lifestyle factors (Gruber et al., 2006; Borrelli et al., 2012). Between 5 and 10% of all bone fractures, according to some reports even up to 50%, result in delayed or failed healing (Tzioupis and Giannoudis, 2007; Gómez-Barrena et al., 2015; Ekegren et al., 2018). To overcome these barriers, a number of research studies focused on enhancing bone regeneration by applying mesenchymal stromal cells (MSCs) derived from various connective tissues. Ex vivo processing and culture methods have been developed to obtain sufficient MSC numbers for therapy (Schallmoser et al., 2009; Sensebé et al., 2013; Robey et al., 2015). Furthermore, MSCs have been combined with various scaffolds and signaling factors in order to tissue engineer viable “bone substitutes” recapitulating key features of autologous bone grafts and enhancing bone regeneration (Frohlich et al., 2008; Jakob et al., 2012). In vitro culture of these constructs in order to drive cell differentiation, bone-like matrix deposition, and increased mechanical properties has also been extensively studied (Marolt et al., 2006; Grayson et al., 2011; Bhumiratana et al., 2016; Vetsch et al., 2016; Mitra et al., 2017; Zhao et al., 2018). Recapitulation of mechanisms present during embryonic bone development was proposed as a “developmental (re)engineering” strategy for the preparation of intermediate grafts capable of forming fully functional bone (Jukes et al., 2008; Tonnarelli et al., 2014; Bernhard et al., 2017). Viable, large bone-like grafts in clinically relevant dimensions (several millimeters to centimeters in size) have been achieved using dynamic culture of scaffolds seeded with MSCs in bioreactors (Grayson et al., 2010, 2011; Güven et al., 2011; Sørensen et al., 2012; Bhumiratana et al., 2016). In addition, in some cases these grafts comprised rudimentary vascular networks. Bone marrow and adipose tissue MSCs were used in the majority of preclinical and clinical studies (Marolt et al., 2010; Robey, 2011; Grayson et al., 2015; Nancarrow-Lei et al., 2017) (Table 1). However, various other sources of MSCs have also been investigated, including skeletal muscle, bone, cartilage, tendon, dental pulp, perinatal tissues (e.g., Wharton's Jelly, umbilical vein/cord blood, amnion, placenta), embryonic stem cells and induced pluripotent stem cells. Due to the aging-related decline in tissue regeneration (Kassem and Marie, 2011; Marie, 2014; Baker et al., 2015; Bhattacharjee et al., 2019), involving both intra- as well as extra-cellular mechanisms, perinatal tissues and induced pluripotent stem cells have raised interest as potential sources of “young” MSCs with high regenerative properties (Kern et al., 2006; Baksh et al., 2007; Robey, 2011; De Peppo et al., 2013; Ghasemzadeh et al., 2018; Spitzhorn et al., 2019).

Despite this growing knowledge, the translation of MSC-based bone regeneration strategies from research studies to clinical use has been slow (Jakob et al., 2012; Grayson et al., 2015). Limited mechanistic understanding of MSC therapeutic actions and MSC fate following transplantation (Dupont et al., 2010; Manassero et al., 2016) has made the requirements for therapeutic preparations, such as optimal cell numbers, cell phenotype, maturity, and mechanical properties of tissue-engineered grafts difficult to define (Jakob et al., 2012; Oryan et al., 2017). Technical challenges and high costs related to manufacturing under good manufacturing practice (GMP) guidelines and procedures for regulatory approval of MSC therapies (e.g., under advanced therapy medicinal products (ATMP) classification) pose additional barriers for clinical translation (Sensebé et al., 2013). According to our recent search in one of the clinical trials databases (www.clinicaltrials.gov, July 2019, combinations of search terms “bone regeneration,” “bone fracture,” “alveolar bone,” “maxilla,” “osteonecrosis,” and “stem cells”), a number of clinical trials (mostly phase I and I/II) employing MSCs to enhance bone regeneration have been registered. However, only for a few the research findings have been published (Table 1).

In recent years, a shift from the view of MSCs as being cells that directly contribute to new tissue formation toward seeing MSCs as “medicinal cell factories” that secrete a variety of bioactive molecules with trophic and immunomodulatory activities has steered the research into MSC secretome for bone regeneration (Hofer and Tuan, 2016; Caplan, 2017). MSCs also secrete various types of extracellular vesicles (EVs) which contain proteins, lipids, and nucleic acids with potential pro-regenerative properties (Marote et al., 2016; Börger et al., 2017; Turchinovich et al., 2019; van Balkom et al., 2019). Several recent studies in small animal models suggested the therapeutic potential of unfractionated MSC secretome as well as MSC-derived extracellular vesicles for bone regeneration (Table 2). In the current review, we recapitulate the recent developments in bone regeneration strategies employing MSC transplantation and MSC-based tissue engineering, as well as the use of MSC secretome and vesicular fractions (Figure 1). Finally, we discuss the challenges and future perspective of these various MSC-based bone regeneration strategies.

According to a consensus position statement, the term MSCs is used to describe a population of multipotent mesenchymal stromal cells that: (1) can be isolated based on their ability to adhere and grow on tissue culture plastic surface, (2) exhibit a defined pattern of positive and negative surface markers, and (3) have the ability to undergo differentiation into osteogenic, chondrogenic, and adipogenic lineages under standard in vitro conditions (Dominici et al., 2006). The cells corresponding to this definition were initially isolated from the bone marrow (Friedenstein et al., 1987; Pittenger et al., 1999) and subsequently from many other fetal and adult tissues, including bone, adipose tissue, muscle, blood, dental pulp, placenta, amnion-amniotic fluid, umbilical cord/cord blood etc. (Sudo et al., 2007; Crisan et al., 2008). Importantly, the “stemness” of MSCs, i.e., their ability to self-renew in vivo, has only been shown for some MSC subpopulations from the bone marrow by serial transplantation assays (Sacchetti et al., 2007; Méndez-Ferrer et al., 2010). As a surrogate measure of stem cells present in cell preparations, a colony forming units (CFU) assay was commonly used (Robey, 2011) and this was suggested to correspond to the bone forming capacity in vivo (Braccini et al., 2007). Only recently, the identity of genuine, self-replicating skeletal stem cells has been described (Chan et al., 2018). Nevertheless, the possibility to relatively easily isolate, culture, and differentiate MSCs from various tissues, particularly those remaining as medical waste (e.g., lipoaspirate, perinatal tissues), has fueled the research into MSCs for bone regeneration therapies.

The native process of bone healing proceeds through overlapping stages of inflammation, repair, and remodeling, which involve multiple signaling pathways acting in concert within the bone defect and the surrounding soft tissues (Oryan et al., 2017). In vivo studies examining effects of transplanted MSCs on bone regeneration found limited numbers of transplanted cells surviving and engrafting in the defect sites, and the exact mechanisms of the contribution of exogenous MSCs to new tissue formation are not clear (Geuze et al., 2010; Bhumiratana et al., 2016; Manassero et al., 2016; Oryan et al., 2017). In vitro and in vivo studies suggested that the transplanted MSCs can have multiple paracrine effects on endogenous cell populations, including immune cell modulation, angiogenic activity, MSC and endothelial progenitor recruitment, cell proliferation, stem cell differentiation, anti-apoptotic effects, and wound healing (Ponte et al., 2007; Chen et al., 2008; Ando et al., 2014; Hofer and Tuan, 2016; Oryan et al., 2017). Depending on their source and manipulation (Oskowitz et al., 2011), MSCs produce a variety of signaling factors, including: purines, bone morphogenetic proteins, CCL2, Connexin 43, cyclooxygenase/prostaglandin, CD95/CD95 ligand, galectins, heme oxygenase-1, human leukocyte antigen-G, interleukin-3, interleukin-6, leukemia inhibitory factor, NO, transforming growth factor beta, vascular endothelial growth factor, hepatocyte growth factor, platelet derived growth factor, basic fibroblast growth factor, and others (Ponte et al., 2007; Chen et al., 2008; Ando et al., 2014; Hofer and Tuan, 2016). The growing understanding of MSC immunomodulatory and trophic activities has steered the research toward the potential of therapeutic MSC secretome, prepared by media conditioning on cultured MSCs, to enhance various stages of bone regeneration (Caplan and Dennis, 2006). In a sense, the use of therapeutic secretome mimics more closely the process of native bone healing by involving multiple signaling factors that work in synergy at low concentrations, rather than a single or a few signaling factors in super-physiologic concentrations (potentially leading to serious side effects, as reported for the high, clinically applied doses of recombinant bone morphogenetic protein therapies) (James et al., 2016). In addition, secretome-based cell-free therapeutic approaches present several significant advantages over current cell- and tissue-based therapies. The absence of replicating (allogeneic) cells in secretome fractions significantly improves the patient safety profile, the low metabolic activity allows for improved quality control and quality assurance, and the simplicity of storage provides the basis for cost-efficient shipping of this potentially off-the-shelf therapeutic substance.

A number of studies evaluated the potential of MSC secretome, either as unfractionated conditioned medium or the extracellular vesicles-enriched fraction, for bone regeneration (Table 2). Secretomes from human adult bone marrow MSCs (Osugi et al., 2012; Katagiri et al., 2013; Ando et al., 2014; Ogata et al., 2015), human dental pulp cells (Fujio et al., 2017), and human fetal MSCs (Wang et al., 2016; Xu et al., 2016) have been investigated. Human adult bone marrow MSC-conditioned media contained insulin-like growth factor-1, vascular endothelial growth factor, hepatocyte growth factor, transforming growth factor beta, monocyte chemoattractant proteins-1 and−3, interleukin-3, and interleukin 6, as determined by ELISA assays (Osugi et al., 2012; Katagiri et al., 2013; Ando et al., 2014; Kawai et al., 2015). Extended analysis of the bone marrow MSC-conditioned medium using a cytokine antibody array yielded 43 proteins with levels at least 1.5-times higher compared to a control medium background (Ando et al., 2014). In particular, serum-free conditioning of the bone marrow MSCs, as reported by most studies (Table 2), was shown to result in a highly angiogenic secretome (Oskowitz et al., 2011).

Osugi et al. reported the potential of human bone marrow MSC-conditioned medium to enhance bone healing in a rat calvarial defect model (Osugi et al., 2012). In vitro, this study demonstrated that conditioned medium from human bone marrow MSCs enhanced the migration, proliferation, and expression of osteogenic marker genes, such as osteocalcin and RUNX2, in rat bone marrow MSCs (Osugi et al., 2012). Application of the conditioned medium with agarose gel in calvarial defects resulted in higher bone regeneration compared to the agarose gel mixed with human bone marrow MSCs or vehicle controls (phosphate buffered saline or culture medium) after 4 and 8 weeks, presumably by enhancing rat MSC migration into the defect (Osugi et al., 2012). A related study by Katagiri et al. confirmed the in vitro findings of Osugi et al. and found that the conditioned medium soaked onto collagen sponges significantly increased early bone regeneration in calvarial defects (after 2 and 4 weeks) compared to the control group (Katagiri et al., 2013). A later study of angiogenesis in the newly regenerated bone suggested vascular endothelial growth factor to be the crucial component in the conditioned medium, promoting angiogenesis and migration of endogenous stem cells (Katagiri et al., 2017). Together, these studies suggested that bone marrow MSC-conditioned medium promotes angiogenesis and bone regeneration in a rat calvarial defect model, but also has the potential to mobilize endogenous MSCs.

Ando et al. demonstrated that repeated applications of human bone marrow MSC-conditioned medium accelerated callus formation in a high-speed mouse distraction osteogenesis model through multiple regenerative mechanisms, similar to the transplantation of bone marrow MSCs (Ando et al., 2014). Analysis of conditioned medium identified factors that recruit mouse bone MSCs, endothelial cells, and endothelial progenitors, inhibit inflammation and apoptosis, and promote osteoblast differentiation, angiogenesis, and cell proliferation. In particular, conditioned medium depleted of monocyte chemoattractant proteins-1 and−3 failed to recruit mouse bone marrow MSCs and callus formation (Ando et al., 2014). Further studies with human bone marrow MSC-conditioned medium showed the potential to enhance regeneration of rat periodontal defects and rat bisphosphonate-related osteonecrosis of the jaw (Kawai et al., 2015; Ogata et al., 2015). In periodontal defects, CD31, CD105, and Flk1 positive progenitors occurred more frequently in the conditioned medium group than controls after 2 weeks and regenerated periodontal tissue was found 4 weeks after implantation (Kawai et al., 2015). In the bisphosphonate-related osteonecrosis of the jaw model, conditioned medium resulted in the healing of open alveolar bone sockets in 63% of the rats (with complete soft tissue coverage and socket bones), compared to the exposed necrotic bone with inflamed soft tissue remaining in the control group (Ogata et al., 2015). Histological analyses demonstrated new bone formation and appearance of osteoclasts with conditioned medium treatment, which was significantly higher compared to the non-treatment group, thus indicating that anti-apoptotic and anti-inflammatory effects of conditioned medium regulated the turnover of local bone (Ogata et al., 2015). A related in vitro study indicated that bone marrow MSCs release paracrine factors which directed osteo/odontogenic differentiation of dental pulp cells (Al-Sharabi et al., 2014). Further in vitro studies indicated that bone marrow MSC-conditioned medium promoted osteoclast differentiation and expression of master regulatory transcriptional factors for osteoclastogenesis, as well as showed maintenance of osteoclasts despite the presence of RANKL inhibitors (Ogata et al., 2017). Interestingly, a cytokine mixture composed of recombinant monocyte chemoattractant proteins-1, vascular endothelial growth factor, and insulin-like growth factor-1 in concentrations similar to those found in bone marrow MSC conditioned medium, promoted migration, proliferation, and osteogenic differentiation of rat MSCs in vitro similarly to conditioned medium, and intravenous application improved the healing of medication-related osteonecrosis of the jaw in a rat model (Ogata et al., 2018).

Fujio et al. studied conditioned media from human dental pulp cells cultured in normoxic and hypoxic conditions (Fujio et al., 2017). Significantly higher angiogenic potential of conditioned media from hypoxic compared to normoxic cultures and no enhancement of either conditioned media compared to controls on the mineralization of human fetal osteoblasts were found in vitro. In a mouse distraction osteogenesis model, repeated injections of conditioned medium resulted in increased blood vessel density and higher bone formation compared to the medium control group at 4 weeks (Fujio et al., 2017). In a comparative study of conditioned media from human fetal MSCs, human adult MSCs and rat MSCs, human fetal MSC-conditioned medium showed the highest osteogenic capacity and the lowest immunogenicity in vitro, as well as enhanced bone consolidation after repeated injections in the rat distraction osteogenesis model (Xu et al., 2016). A further in vitro study reported that the secretome of human fetal MSCs ameliorated replicative senescence and enhanced cell proliferation and the osteogenic differentiation potential of human adult MSCs (Wang et al., 2016). Concomitant activation of SIRT1 and FOXO3a expression, upregulation of p21 gene expression, and downregulation of BAX and p53 gene expression were found, and the pre-treatment resulted in restored osteogenic ability of senescent human adult MSCs in a nude mouse ectopic bone formation model (Wang et al., 2016).

Based on the positive outcomes of studies in bone regeneration models, Katagiri et al. reported a first-in-human study and clinical case reports of alveolar bone regeneration using conditioned medium from human bone marrow MSCs (Katagiri et al., 2016). Human bone marrow MSC-conditioned medium was soaked on beta-tricalcium phosphate or on atelocollagen sponge and 8 patients with severe alveolar bone atrophy were treated prior to, or at the same time as dental implant placement. The patients experienced no systemic or local complications and showed early mineralization in the augmented bone according to radiographic evaluations. Calcium phosphate biomaterial structures gradually became indistinct from the surrounding bone 6 months after the surgeries, and biopsies confirmed new bone replacement of the resorbed biomaterial (Katagiri et al., 2016).

Growing understanding of MSC intercellular communication via their secretome offers new options for tissue engineering strategies in bone regeneration (Lamichhane et al., 2015). Extracellular vesicles (EVs) are a heterogeneous group of small, lipid-bilayer enclosed, cell-derived particles, exerting effects on fundamental cellular processes in a pleiotropic manner (El Andaloussi et al., 2013). This mechanism is evolutionary conserved from bacteria to humans and plants (Schuh et al., 2019) and was initially considered a means of eliminating unneeded cellular compounds (Johnstone et al., 1987). However, in 1996, Raposo et al. demonstrated that EVs could modulate an adaptive immune response (Raposo et al., 1996) and research from the last two decades has shown that they facilitate intercellular communication, acting as mobile signaling platforms modulating fundamental biological processes in health and disease (Isola and Chen, 2016). Virtually every cell type secretes EVs and they act by the horizontal transfer of proteins, lipids, mRNAs, miRNAs, and other non-coding RNAs, altering the activity of a neighboring or distant target cell (Van Niel et al., 2018). Thereby, the lipid bilayer protects nucleic acids and proteins from degradation in the extracellular environment, allowing their efficient transport. Taken together, these findings have spurred a tremendous amount of effort to exploit EVs as potential therapeutics for immune response modulation and tissue regeneration.

Currently, the term “extracellular vesicles” refers to all types of secreted vesicles released by different cell types under different conditions (Araldi et al., 2012). Based on their biogenesis, two main categories of EVs can be envisioned for therapeutic use: exosomes and microvesicles. Exosomes are rather homogeneous in size (~40–100 nm) and are generated in the endosome as intraluminal vesicles which subsequently mature to multivesicular endosomes. Cargo is sorted into exosomes by distinct mechanisms resulting in the formation of heterogeneous populations of intraluminal vesicles (Colombo et al., 2014). Sorting either involves ESCRT (endosomal sorting complex required for transport), Alix (or PDCD6IP; programmed cell death 6 interacting protein), and TSG101 (tumor susceptibility gene 101 protein) or takes place via an ESCRT-independent mechanism (reviewed in Van Niel et al., 2018). Finally, exosomes are released into the extracellular space upon fusion of the multivesicular endosomes with the cellular plasma membrane. Some of the commonly used markers to identify exosomes are various tetraspanins (CD63, CD81, and CD9), ALIX, TSG101 (tumor susceptibility gene 101 protein), and flotilin-1. However, a recent study demonstrated that classical tetraspanin-enriched exosomes contain a much more limited repertoire of active molecules than has previously been assumed (Jeppesen et al., 2019), highlighting the necessity of suitable protocols for exosome characterization.

Microvesicles (often also referred to as ectosomes or shedding vesicles) represent a rather heterogeneous population of EVs generated by outward budding and subsequent fission of the plasma membrane. Both cargo sorting and subsequent Ca²⁺-dependent vesicle shedding are regulated by various small GTPases, including members of the ARF (ARF6 and ARF1), Rab20, and Rho (Rac1 and RhoA) families (Tricarico et al., 2017). Microvesicles can range in size from ~100 nm to ~1 μm, are rich in phosphatidylserine and are often characterized by the presence of integrins, selectins, and CD40 ligand. Recently, annexin A1 has been proposed as a specific marker for microvesicles which are shed directly from the plasma membrane (Jeppesen et al., 2019). Once EVs are released into the extracellular environment, uptake into target cells takes place either by a receptor-mediated process, internalization by endocytic uptake, or by simple fusion of the cell and vesicle lipid bilayer. However, the unequivocal classification of EVs remains difficult, as the different types share overlapping characteristics and no unique markers are available. Further, current protocols used to purify EVs result in a heterogeneous vesicle population (Willms et al., 2016) and cannot fully exclude a soluble fraction within the purified product, which is why the term “vesicular secretome fraction” has been coined, encompassing both soluble and vesicular components (Gimona et al., 2017).

Several groups provided evidence that EVs are important regulators of stromal cell maintenance (Ratajczak et al., 2006; Quesenberry et al., 2015) and function (Weilner et al., 2016) and it is now widely accepted that much of the efficacy of stromal cell therapies comes from EVs and/or soluble factors (Caplan and Dennis, 2006; Caplan and Correa, 2011). In particular, EVs secreted by MSCs are considered promising candidates for future cell-free regenerative therapies. The paracrine action of MSCs was shown by Gnecchi et al., demonstrating that application of MSC-conditioned medium ameliorated tissue damage in a rodent model of acute myocardial infarction (Gnecchi et al., 2005). Later studies confirmed these results in both pig and mouse models using fractionated MSC secretome preparations (Timmers et al., 2008; Arslan et al., 2013). Since then, the regenerative and immunomodulatory capacity of MSC-derived EVs have been evaluated in several animal disease models, including e.g., kidney and liver injury, lung disease, cartilage repair, hind limb ischemia, ischemic brain injury, and spinal cord injury (reviewed in Harrell et al., 2019).

As MSCs secrete large amounts of EVs (Yeo et al., 2013), EV-based approaches to boost bone regeneration were soon evaluated. As pointed out earlier, in comparison with direct MSC transplantation, MSC-derived EVs appear prima facie safer, as they are devoid of viable cells. Further, systemically applied EV preparations are less likely to be trapped in the lung or liver and are most likely less immunogenic (Lai et al., 2013). Finally, they can be stored for an extended period of time, offering the possibility of an off-the-shelf product for restoration of bone defects (Webber and Clayton, 2013). Indeed, EVs purified from bone marrow MSCs, umbilical cord MSCs, endothelial progenitor cells, and iPSC-MSCs enhanced healing of bone in rodent models (Furuta et al., 2016; Qi et al., 2016; Jia et al., 2019; Zhang et al., 2019). Furuta et al. isolated EVs from human bone marrow MSC-conditioned medium and injected it into femoral fractures in a CD9 negative mouse strain (which produces low levels of EVs) and wild type mice (Furuta et al., 2016). Preparations of EVs contained low levels of bone-repair related cytokines. However, their application accelerated fracture healing compared to control (vehicle treated) animals (Furuta et al., 2016). Repeated injections of EVs prepared from human ESC-MSCs similarly promoted the regeneration of osteochondral defects in a rat femur model (Zhang et al., 2016b) and enhanced bone regeneration was found in a rat distraction osteogenesis model with injections of EVs prepared from rat endothelial progenitors (Jia et al., 2019).

More recently, MSC-EVs combined with various scaffold materials generated bone in vivo in ectopic sites (Xie et al., 2017) and successfully promoted bone repair in rodent calvarial bone defects (Qi et al., 2016; Qin et al., 2016; Li et al., 2018). Ideally, biocompatible scaffolds should degrade at an appropriate rate and facilitate the controlled release of the extracellular vesicles, as mere loading results in a burst release and might be less efficient in exerting pro-regenerative effects. Qi et al. showed in a calvarial defect model, established in osteoporotic rats, that treatment with tricalcium phosphate scaffolds loaded with EVs prepared from hiPSC-MSCs repaired bone defects through enhanced angiogenesis and osteogenesis (Qi et al., 2016). More recently, Li et al. reported that human adipose MSC-derived EVs immobilized onto poly-lactic-co-glycolic acid scaffolds exhibited a slow release profile in vitro and enhanced bone regeneration in mouse calvarial defects after 6 weeks (Li et al., 2018).

However, the molecular and cellular mechanisms underpinning the osteogenic effect of EVs remain poorly defined and they are certainly cell- and tissue context-dependent (Harrell et al., 2019). Potentially, they can be attributed to a protective effect in necrotic and ischemic environments (Liu et al., 2017), recruitment of endogenous MSCs (Osugi et al., 2012; Furuta et al., 2016), and HIF-1alpha-dependent pro-angiogenic activity (Qi et al., 2016; Zhang et al., 2019). In addition, MSC-EVs seem to directly promote osteogenic differentiation, in part by activating the PI3K/AKT signaling pathway (Zhang et al., 2016a) and via the activity of miRNA-196a released from the EVs (Qin et al., 2016). In general, MSC-derived extracellular vesicles have been shown to exert anti-inflammatory properties by inducing high levels of anti-inflammatory proteins, while concomitantly attenuating pro-inflammatory cytokines in THP-1 monocytes in vitro and inducing regulatory T cells in vivo (Zhang et al., 2014). Taken together, MSC-EVs elicit pleiotropic effects, promoting bone repair. However, it is important to note that the characteristics of the microenvironment influence the function of EVs produced by these cells (Huang and Feng, 2017). Along these lines, Zhu et al. demonstrated that the pro-angiogenic and osteogenic actions of bone marrow MSC-EVs are impaired in type-1 diabetes (Zhu et al., 2019). Therefore, it is imperative to choose appropriate cell sources for EV production for future therapeutic strategies. Taken together, EV-based, cell-free therapies appear to be a promising strategy to repair bone tissue. However, to ensure comparability between preclinical studies and to allow future translation into the clinic, EV purification and characterization protocols urgently need to be harmonized (Witwer et al., 2019).

Production of MSCs for clinical use requires appropriate laboratory procedures adhering to GMP regulations. These procedures have been reported previously by several groups for bone marrow MSCs (Schallmoser et al., 2009; Fekete et al., 2012). Similarly, the manufacturing of MSC-derived therapeutic secretome and EVs for clinical testing requires GMP-compliant strategies, and several groups focused on establishing such protocols (Andriolo et al., 2018; Mendt et al., 2018; Rohde et al., 2019). Existing GMP protocols for MSC expansion can be largely adopted for the first step in secretome/EV production and standard operating procedures can be expanded by the concomitant processing steps of the conditioned medium. Overall, the therapeutic secretome/EV field benefits at this point from the experience gained from MSC manufacturing for clinical applications as well as from the ample experience available from virus-like particle enrichment. In an excellent overview on the current status of EV manufacturing, Whitford and Guterstam (2019) indicated that the technologies and knowledge for GMP-compliant EV production are at hand. However, some adaptation and fine tuning is still required. For instance, it should be noted that large scale production of therapeutic MSCs in order to obtain significant batch sizes of either cells or EVs for preclinical and clinical testing may alter the physiological state of the MSCs. The influence of cell passage number on the therapeutic potential of MSC-derived EVs has been questioned at least for adipose MSCs (Serra et al., 2018). Furthermore, the use of serum-free medium can cause changes in exosome biology and cargo sorting, and the use of coatings (e.g., fibronectin, gelatin) alters the profile of MSC adhesion and mechanotransduction and automatically violates a central requirement for defining MSCs, namely plastic adherence (Brindley et al., 2011; Whitford and Guterstam, 2019).

Bioreactor systems and associated analytical approaches have been studied for scale-up and a more efficient, reproducible, safe and cost-effective production of MSCs and secretome compared to static culture (Carmelo et al., 2015; Mizukami et al., 2019). Carmelo et al. developed a xeno-free microcarrier-based stirred culture system for scalable expansion of bone marrow and adipose MSCs. Their secretome analysis suggested a priming effect of stirred culture conditions toward potentially increased production of specific cytokines by MSCs (Carmelo et al., 2015). Similarly, a study by Teixeira et al. indicated that culture of bone marrow MSCs on microcarriers using computer-controlled suspension bioreactors enhanced the neuroregulatory profile of the secretome compared to static cultures (Teixeira et al., 2016). Secretory profiles of the MSCs can also be modulated by the surface structure of the microenvironment (Leuning et al., 2018), pointing further to the enhanced potential of engineered three-dimensional culture environment for optimizing MSC-derived therapeutic products.

Within the field of bone regeneration, there is a clear need for new cell-based or cell-free therapies for a number of conditions in which bone does not heal, resulting is significant morbidity and burden for the patients. At this time, several MSC-based therapies have reached the stage of clinical trials, but none of the approaches has been accepted in the clinic as the standard of care (Stegemann et al., 2014). Additional research is needed to advance our mechanistic understanding of various cell-, EV-, and tissue engineering-based therapies within specific clinical indications.

For MSC-based cell therapies, some of the currently registered clinical studies are starting to address the questions of dosing, the use of autologous vs. allogeneic cells, and efficacy compared to standard treatments using autologous bone grafts. However, many challenges remain in the standardization, quality control, and potency evaluation of MSC-based cell therapies, as well as in the scale-up, GMP manufacturing, and logistics, in particular when autologous MSCs are used. In addition to inter-patient variability due to age, health, and various risk factors, inherent differences between MSCs from various tissue sources present specific advantages and disadvantages, such as the potential to be harvested in high or low yields, surgical procedures and donor site morbidity, potential of use with minimal intra-surgical manipulation/preparation of the cells, the need for additional cell storage and culture of the cells to increase their numbers, co-application of signals to promote cell phenotype/differentiation, local vs. systemic delivery via injections or in combination with scaffolds (i.e., tissue-engineered therapies). For certain indications, e.g., complex facial bone defects, extended in vitro or in vivo engineering of viable bone substitutes might be required to provide appropriate architecture as well as mechanical properties for defect stabilization and loading, which are key for successful healing. For others, MSC injections or co-application with bone substitute materials might provide the critical components required for successful outcomes.

Due to these complexities with viable cells and tissue transplantation, a number of investigations have also focused on enhanced mobilization and homing of endogenous cells (which has not been extensively investigated in the clinic) (Yao and Lane, 2015; Karlsson et al., 2016; Yao et al., 2016) and on novel, secretome-based therapeutic approaches. Whereas clinical translation of the whole secretome might prove challenging as well, re-constitution of major secreted growth factor components (e.g., by using recombinant proteins) in concentrations similar to those in the MSC secretome might recapitulate the therapeutic effects, as was recently demonstrated (Ogata et al., 2018). Of note, repeated applications of conditioned media were used in several studies to achieve the therapeutic effects, in contrast to single applications in the case of cell- and tissue-engineering-based therapeutic approaches. Surviving transplanted cells/tissues will be exposed to the effects of the local environment in the defect and might change their functions as a result. In contrast, preparation of therapeutic secretome might incorporate in vitro culture modifications to modulate the composition of released components.

Therapies based on EVs face similar problems as cell therapies in translating the promising research data into well-defined clinical trials and subsequent applications to patients. A central issue in the field is the standardization of manufacturing and analytic processes. A prerequisite for this advancement seems to be the awareness that any cell-derived EV product will always comprise of a heterogeneous mixture of extracellular vesicles and in most cases also of co-purifying soluble components. While this heterogeneity may be more of a concern for naive, unmanipulated MSC-EV preparations, loading and engineering of EV products to contain specific signals (Sutaria et al., 2017) might be less affected, with a greater focus on the characterization of the active component (e.g., siRNAs or engineered proteins) and its therapeutic potential.

Intuitively, the discussion on the best cell source for use in bone regeneration will always consider that cells that perform better in in vitro differentiation assays toward osteogenic lineage may be better suited for the manufacturing of EVs that support or enhance bone healing. It must be emphasized though, that these assays are highly artificial and the link between in vitro cell differentiation potential and the regenerative potential of cell-derived EVs must be further evaluated. Current efforts in the EV community are focused on determining the best producer cell source, on potential modes of action, and on standardization of physico-chemical assays and analytical devices for functional EV characterization. At the present time, analytical methods that are specifically tailored to characterize therapeutic EVs are largely missing. Nanoparticle tracking analysis is capable of determining the total number of particles in the heterogeneous (secretome) solution, but neither the particle size nor the particle number can currently predict the functionality of the product. Cell-based potency assays are difficult to establish to the level of GMP compliance and will have to be designed specifically for each clinical indication.

There is also no current consensus on the best method to expand the producer MSCs (two-dimensional vs. three-dimensional culture, batch-feed vs. stirred tank bioreactor, serum supplemented vs. serum free/defined medium) or to enrich/purify EVs or secretome. Therefore, application for clinical trials cannot rely on or reference to already approved investigational new drug/investigational medicinal products. This is well-reflected by the mere four interventional studies with EVs that are listed at clinicaltrials.gov. To some, the current classification of EV-containing therapeutic substances as biologicals by both EMA and FDA seems to be an advantage, since ATMP regulations do not apply. It remains to be seen, however, how the regulatory agencies deal with the application on a case-by-case basis in the absence of clear and binding regulations.

Finally, prospective manufacturers of EV therapeutics have to take into consideration that EV-based products for improved bone healing and regeneration aim, for the most part, at non-life-threatening medical indications. Irrespective of the considerable benefit for the patient and the health care system, this fact impacts the possible pricing of the product and thus the total manufacturing cost. For instance, the current financial requirement for available CAR-T cell therapy (430,000$ US) is an extreme example of the costs of cell-based therapy. Therapies based on EVs toward bone healing should look more at costs in the range of 1,000–2,000$ US to make the treatment both attractive and suitable for refund by the health insurance providers. This, however, will again demand precise calculations of the cost of goods and manufacturing costs, as well as for costs for testing, release, and storage. In this respect, batch size and dosing once again take center stage and must be considered in the development of a GMP-compliant manufacturing process for EV therapeutics and EV-enriched secretome.

In conclusion, further advancements in the science and translation of MSC-based therapies are required, and this growing body of knowledge, in conjunction with the improved identification of clinical indications and patients best suited for these treatments, will be key to creating therapies that will be consistently more successful than current treatments, while also being cost effective and marketable.

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.