Mechanisms of Action of Multipotent Mesenchymal Stromal Cells in Tendon Disease

Multipotent mesenchymal stromal cells (MSCs) are a promising therapeutic tool to treat tendon disease. Aiming to establish successful treatment approaches and to fully exploit the regenerative potential of the MSC, it is crucial to understand their mechanisms of action. However, these can be multifaceted and strongly context-sen-sitive and are still not well-understood in the context of tendon disease. This review aims to shed light on the different possible mechanisms, including engraftment, tenogenic differentiation, extracellular matrix synthesis and remodeling, immunomodulation, pro-angiogenetic effects, trophic support, and protection of resident tendon cells. Evidence from experimental and clinical (veterinary) case studies was compiled and interpreted in conjunction with the respective in vitro and animal models used.


Introduction
Tendinopathy is a common cause of recurring pain and long-term impairment in leisure and professional athletes, increased age being an additional risk factor. The prevalence of clinically manifest conditions in risk groups is high: in a cohort of football players, 21% suffered from Achilles tendon problems [1]. Moreover, even in clinically healthy volunteers, ultrasonographic evidence of Achilles tendon alterations was found in 16% [2]. This indicates that clinical manifestation is only the tip of the iceberg, the basis of which is a long-term interplay of inflammatory and degenerative changes.
Tendons have to withstand high mechanical loads and serve as an energy storage with elastic properties. The required biomechanical properties are provided by the extracellular matrix (ECM) [3], which is largely composed of hierarchically structured, cross-linked, and crimped collagen type I fibrils. The tenocytes, while representing only 5% of the tissue volume, maintain the ECM structure by constant remodeling. This normally enables biochemical and biomechanical adaptations to exercise [4]. Recurrent overuse impairs this physiological adaptation.
The onset of tendinopathy is currently understood as the result of a failed healing response to repeated tissue trauma. Microruptures, oxidative, mechanical, and heat stress activate resident cells and trigger a cascade of inflammation and degeneration, culminating in ECM deterioration. Key molecules involved include vascular endothelial growth factor (VEGF), interleukin (IL)-1, tumor necrosis factor (TNF)-α, prostaglandin (PG)E2, glutamate, and substance P [5,6]. These mediators foster the ingrowth of blood vessels and nerves and the activation of nociceptive pathways. They are also implicated in the upregulation and activation of matrix metalloproteinases (MMP) and downregulation of their endogenous inhibitors (tissue inhibitors of matrix metalloproteinases; TIMP) [7]. This entails ECM degradation which successively alters and weakens the ECM structure [6]. When the accumulated damage and sensitization reach a threshold, clinical manifestation of tendinopathy comprises classical signs of inflammation including pain. Furthermore, provoked by new overload events, massive tissue trauma can occur. The resolution of inflammation is crucial to limit tissue damage, yet this mechanism often fails. Promoting fibrosis, a lack of pro-resolving signals, and persistence of macrophages entails the continuing activation of fibroblasts [8,9]. Furthermore, macrophages could further contribute to ECM degradation via MMP secretion. Once at a diseased state, the intrinsic regenerative capacity of tendons is poor. Although endogenous mesenchymal stem-like cells with high tenogenic potential reside within tendons [10][11][12], these are susceptible to damage and suffer age-related changes [13,14]. In pathological states, they could even contribute to fatty degeneration, fibrosis, and heterotopic ossifications [15,16].
Treatment of tendinopathy still represents an unsolved challenge. Mainly, the use of strict rehabilitation exercise regimens is sufficiently evidence based [17,18]. Anti-inflammatory drugs are frequently used, but they do not only counteract the active inflammation but also its resolution [19]. Biologicals such as platelet rich plasma have also received much attention, but clinical evidence is not convincing [17,20,21]. Research also focuses on the potential of endogenous tendon progenitor cells [22], which may be a promising strategy but will not be addressed in this review.
Multipotent mesenchymal stromal cells (MSCs) represent a therapeutic tool which might meet the clinical need of an adaptive treatment that simultaneously addresses different aspects of the disease. MSCs reside in virtually any tissue, in close proximity to the vasculature [23,24]. MSCs derived from bone marrow and adipose tissue (BMSC and ASC, respectively) have been most extensively characterized [25,26]. The fibroblast-like cells have been defined by a set of inclusion and exclusion antigens, their plastic-adherence, and trilineage differentiation potential in vitro [26]. While their differentiation potential into mesenchymal cell types, including tenocytes [27], has led to their extensive use in tissue engineering, it has become evident that their therapeutic potential by far exceeds cell replacement [24,28]. While proof of MSC engraftment is often lacking, MSC-based cell therapy has shown beneficial effects in diverse scenarios in animal models, mostly mediated by immunomodulatory and trophic mechanisms [29][30][31][32][33]. Particularly, the immunomodulatory potential is extensively being researched and already exploited clinically, e.g., for treatment of graft-versus-host disease [34][35][36].
The use of MSC for tendon repair was first suggested in 1998 [37] and, interestingly, has been published as a case report on an equine patient as early as 2003 [38]. Since then, several experimental animal studies-the recent ones being reviewed here-and case series in equine patients [39][40][41] have raised hope that local implantation of MSC into acute tendon defects improves healing. However, translational progress into human orthopedics is underwhelming, and although equine patients are being treated and few first-in-man clinical trials have been performed or initiated [42][43][44], convincing evidence from randomized, controlled clinical studies has neither been obtained in equine nor in human patients so far [45]. This Mechanisms of Action of Multipotent Mesenchymal Stromal Cells in Tendon Disease DOI: http://dx.doi.org/10.5772/intechopen.83745 may in part be due to our still limited understanding of the MSC mechanisms of action in tendon healing, which delays the development of targeted treatment approaches.
The aim of this review was to collect the evidence for the different possible MSC mechanisms of action in the treatment of tendon disease. In vitro and in vivo studies published within the last 5 years were screened and their results were compiled, focusing on MSC-based cell therapy using BMSC or ASC.

Tendon regeneration and defect models 2.1 In vitro and ex vivo models
In vitro and ex vivo models relevant to MSC mechanisms of action in tendon regeneration comprise two major groups, with some overlap (Figure 1) Typically assessed parameters following tenogenic differentiation include the expression of tenogenic transcription factors (scleraxis and, in the more recent studies, mohawk), the transmembrane glycoprotein tenomodulin, as well as the expression and deposition of extracellular matrix components (e.g., collagen I, collagen III, decorin, and tenascin-C) and biomechanical parameters in case of tissue engineered constructs. Upregulation of matrix components such as collagen I or tenascin-C and improved construct strength do not only suggest tenogenic differentiation but also indicate ECM-modulating activities of the MSC. However, it should be acknowledged that no truly specific tendon marker has yet been identified, and that only expression patterns of combined marker sets, e.g., collagen I, scleraxis, and tenascin-C, discriminate healthy tendon from diseased tendon or other musculoskeletal tissues [95].
The figure gives an overview of the in vitro models included in this review, illustrating the overlap between tenogenic differentiation models and coculture models, and summarizes the most commonly assessed outcome parameters. Note that in this context, the term "coculture" is used to summarize the models investigating the interplay between tenocytes and MSC, thus it does not exclusively refer to cocultures of different cell types but also includes cell culture models using conditioned media or tendon explants.

In vivo models
In vivo studies on MSC-based tendon therapies need to be discriminated with respect to the animal model used (small vs. large, type of disease or defect model) and the treatment approach (strategy for MSC delivery, possible adjuvant treatments, timing of treatment, MSC source, and cell numbers applied).
Interestingly, there appears to be a fair balance between small and large animal studies. This suggests preclinical progress, but it is also due to the interest in the equine species within the veterinary community. The tendon defects were created surgically in the majority of studies, with full thickness transections or segmental defects (mostly in the Achilles tendon) used in small animals or dogs and surgically created core lesions in the superficial digital flexor tendons in the equine model. Although there is reason to believe that enzymatical induction of tendon lesions better mimics the ECM degeneration and inflammation in tendon disease, only few among the recent studies used collagenase-based tendinopathy models [106,108,110,129,137,139]. Still, neither surgical nor enzyme-based approaches fully reflect the complex tendon pathophysiology. In this light, providing particularly valuable information, some studies in the equine species were performed using horses suffering from naturally occurring tendinopathy [134,138,141] (Figure 2).  The diagram displays the numbers of studies performed in different animal species which were included in this review and indicates the types of tendon defect models used in the respective species.
Approaches for MSC implantation include local delivery of MSC suspensions, mostly via (ultrasound-guided) injection [106-112, 119, 120, 127-133, 136-141], coating of suture materials with MSC [113], MSC delivery in fibrin-based vehicles [54,114,124] or cell sheets [54,123,125,126], and the use of diverse constructs of MSC and scaffold materials [115-118, 121, 122]. Interestingly, while the delineation between MSC-scaffold constructs for MSC delivery and for tendon replacement is sketchy, it is remarkable that construct-based approaches are almost exclusively used in small animals. This indicates that translational progress using these approaches is poor, possibly due to their incapability to meet the biomechanical demands in large animals or humans.
Further aspects of the treatment approach are likely to influence MSC mechanisms of action and complicate the coherent interpretation of findings from different studies. Adjuvant treatments, e.g., simultaneous growth factor delivery, or pre-treatment of the MSC, such as pre-differentiation or inflammatory licensing before cell delivery, may support certain mechanisms synergistically but may negatively interfere with other mechanisms. For example, bone morphogenetic protein (BMP)-12 promotes MSC tenogenic differentiation but reduces their immunomodulatory potential [93]. Next, the timing of the treatment is of great importance as different mechanisms of action of MSC are likely to be relevant during different stages of tendon healing. Furthermore, the dosage, i.e., the numbers of MSC applied, may not only play a role with respect to treatment efficacy but also with respect to supporting specific mechanisms of action [120]. For example, interactions between MSC and immune cells depend on the ratio of MSC to leukocytes present [142].
Last not least, the MSC source is likely to influence their mechanisms of action, which is an issue with equal relevance for in vitro findings. On the one hand, this applies to the choice of donor in terms of age and health status [143] and in terms of autologous, allogeneic or, in case of many small animal models, even xenogeneic use of MSC. On the other hand, the tissue origin of MSC as well as the donor species impact on the cell characteristics [57, 140,144] and thus potentially on their mechanisms of action. Therefore, mainly studies focusing on the well-characterized BMSC and ASC were included and their tissue origin discriminated where appropriate. Furthermore, it was attempted to compile only studies which enabled the discrimination of MSC effects from those of possible additional treatments. In this line, in vivo studies using genetically engineered MSC for other purposes than cell tracking were not included in this review.

Engraftment and tenogenic differentiation
The assumption that MSC engraftment and their tenogenic differentiation after implantation into a tendon lesion lead to the replacement of damaged tenocytes dates back to the earlier days of MSC research and mirrors the general conception of MSC at that time [27,38]. In the following years, the fact that MSC persistence at the site of tissue damage could not be achieved in models for a wide variety of diseases led to the assumption that differentiation and cell replacement might not even contribute to the regenerative effects observed after MSC transplantation [28]. This hypothesis was fostered by the compelling finding that paracrine factors released by the MSC can lead to similar beneficial effects as the MSC themselves, leading to the concept of cell-free MSC-based therapies [145]. Still, the situation Tendons 6 might be slightly different in tendon pathologies, and at the moment, it cannot be excluded that tenogenic differentiation of engrafted cells could contribute to regeneration, perhaps as a basis for further trophic and ECM-modulatory mechanisms.

In vitro evidence
An extensive body of recent literature describes the tenogenic differentiation of MSC in response to a wide range of stimuli, although unfortunately, no generally accepted in vitro model or standard tenogenic differentiation assay exists. Current concepts of tenogenic differentiation are reviewed in detail elsewhere [146,147]. The most commonly used stimuli to induce tenogenesis in MSC include growth factors, scaffolds with specific topography, and cyclic mechanical loading, with most studies combining two or more of these approaches, based on earlier studies in the field of tissue engineering [37, [148][149][150].
Growth factors used for induction of tenogenic differentiation mainly include transforming growth factor-β family members ( on physiological stimuli, such as mechanical loading, biomimetic scaffolds, or cross-talk with tenocytes, are most insightful. In contrast, the use of growth factors (typically at concentrations exceeding those found in vivo) or genetic modifications is suitable for mechanistic studies and may be helpful for tenogenic pre-differentiation prior to MSC implantation but does not reflect the in vivo situation. To understand if physiological stimuli could promote the same distinct tenogenic phenotype as artificial TGF-β concentrations, it would be helpful to gain further insight into the downstream signaling networks and their possible interfaces. So far, however, tenogenic signaling has mainly been investigated following growth factor stimulation [67, 82, 89, 90]. Only few studies have attempted to elucidate the signaling pathways activated in MSC in response to mechanical load or scaffold topographical cues, focusing on the role of rho/ ROCK [154,155].
Yet, although physiological stimuli have repeatedly been shown to induce tenogenic differentiation in MSC, it should not be anticipated that this mechanism is analogously activated when MSCs are implanted into a tendon lesion. Selfevidently, the tendon lesion does not provide a physiological but rather a pathophysiological environment, which may have an entirely different impact on the MSCs. Unfortunately, this issue is still underrepresented in the current literature. Recently, we investigated ASC tenogenic properties in response to physiological tenogenic and simultaneous inflammatory stimulation [84]. This study demonstrated that ASC tenogenic properties are compromised not only in the presence of the pro-inflammatory cytokines IL-1β and TNF-α but also in the presence of leukocytes. Similarly, IL-1β and IL-6 inhibited tenogenic differentiation in tendonderived stem cells [156,157]. Furthermore, again in tendon-derived stem cells, stiff matrices impeded tenogenic differentiation [158]. Together, these findings suggest that MSC tenogenic differentiation may be impaired in a pathophysiological in vivo environment, which can comprise inflammatory stimuli as well as stiff (fibrotic) ECM, depending on the stage of disease.

In vivo evidence
Although extensively investigated in vitro, there is no distinctive evidence of tenogenic differentiation following MSC implantation in vivo. One conceivable explanation is that MSC differentiation is in fact impaired in the pathophysiological lesion environment. Nevertheless, in contrast to studies in other disease models, MSCs have been repeatedly localized in treated tendon lesions, providing a basis for long-term regenerative effects, possibly including differentiation and cell replacement. Furthermore, there is some evidence of homing of MSCs to tendon lesions, although not unambiguous. The mechanism of homing may be of minor importance with respect to cell delivery at the macroscale, as the cells are almost exclusively delivered locally in MSC-based tendon therapies. Yet, the capability of homing is still indicative of MSCs that are capable of identifying regions of tissue damage at the microscale, where they would actively integrate.
None of the small animal studies included in this review specifically addressed MSC homing to tendon lesions. However, when bursal tissue was implanted in rotator cuff tendon lesions in a rat model, the green fluorescent protein-labeled mesenchymal stem cells from this tissue infiltrated the healing tendons [159], demonstrating the presence of homing signals. Accordingly, ASC infiltration into the tendons was also evident when cell sheets were used as delivery vehicle in a canine model [126]. However, when injected into the tendon sheath, BMSC homed to synovial structures but were not attracted Tendons to the tendon lesions in an ovine model of intrasynovial tendon healing [127]. In the equine large animal model, homing of MSC to tendon lesions has been addressed in more detail. Scintigraphic short-term in vivo tracking of technetium-labeled BMSC showed that the cells homed to the tendon lesion after administration by regional limb perfusion, although local administration by direct intralesional injection was more effective, and no homing was observed after intravenous administration. These findings were consistent between artificial tendon lesions [135] and natural tendinopathy [134]. Interestingly, intraarterial limb perfusion showed greater accumulation of BMSC in the lesion on day 10 after surgical lesion induction than on day 3 [135]. This finding illustrates that the stage of tendon disease is of importance to MSC homing mechanisms. However, scintigraphic tracking also revealed that even after local injection, only a relatively small proportion of the injected BMSC remains at the injury site (24% after 24 h) [134]. In accordance with this, we and others demonstrated that ASCs are distributed via the bloodstream within the first few days after their injection into equine tendon lesions, possibly as they are washed away before they can home and attach [136,139]. We additionally observed that the ASCs were subsequently also found in nontreated tendon lesions, indicating their capability of homing [139].
Engraftment of MSC within treated tendon lesions was demonstrated in several studies, albeit results are not conclusive as to the numbers of surviving cells in relation to the cell numbers administered. In rat Achilles tendon defects, BMSC or ASC could be identified histologically at 2, 4, and 8 weeks after cell implantation (injection) [107,109,112], as well as 3 weeks after implantation of a BMSC-seeded collagen scaffold [116]. Complementing these small animal studies, MSCs have been traced in large animal studies, including longitudinal in vivo cell tracking. In sheep, green or red fluorescent protein-labeled BMSCs were detected histologically at 1, 2, 3, 4, and 6 weeks following their implantation [128,129]. In the equine model, we and others could trace superparamagnetic iron oxide-labeled ASC by magnetic resonance imaging during follow-up periods of up to 24 weeks after implantation into artificial tendon lesions [132,139] and umbilical cord tissuederived MSCs during a follow-up period of 8 weeks in naturally occurring tendinopathy [138]. In the experimental tendon lesions, histological results confirmed the presence of the simultaneously fluorochrome-labeled ASC until week 24 [132,139]. This provides evidence for a remarkable long-term persistence of part of the locally injected MSC, yet it has neither been proved nor disproved whether these cells commit to a tenogenic fate.

Extracellular matrix modulation
The restoration of the ECM architecture and functionality is a major goal in regenerative tendon therapies. Based on the early hypothesis of MSC engraftment and tenogenic differentiation, it was assumed that the differentiated cells would subsequently synthesize new tendon ECM. Indeed, MSCs are capable to synthesize a considerable amount of extracellular matrix even in an undifferentiated state [160]. Furthermore, the composition of the ECM synthesized by differentiated MSC reflects the respective tissue lineage, which is well-established for their chondrogenic or osteogenic differentiation. Corresponding in vitro data exist for the differentiation into the tenogenic lineage, although not always consistent between studies. There is also in vivo evidence that MSC transplantation improves tendon ECM structure. However, this is not necessarily due to ECM synthesis by the MSC themselves but might also be a consequence of protective and stimulatory effects on tenocytes, which in turn might be capable to synthesize the new ECM. Moreover, importantly, there is not simply a lack of ECM in tendinopathy but rather a dysfunctional ECM composition and structure, due to the imbalance of remodeling activities. Particularly, in later stages of the disease, chondroid degeneration and fibrosis impair ECM functionality, thus effective ECM regeneration would also comprise its remodeling and the restoration of physiological remodeling activity within the tendon.

In vitro evidence
As most tenogenic differentiation studies investigated the expression and/ or deposition of tendon-specific extracellular matrix molecules as a marker for successful differentiation, there is quite extensive evidence that the ECM synthesis by MSC is altered during tenogenic differentiation. However, there is some discrepancy between different studies as to whether the ECM molecule expression pattern of tenogenic MSC truly corresponds to that of healthy tendon tissue.
Collagen . While these molecules are important components of native tendon ECM, contributing to collagen organization and fibrillogenesis, their increased presence is also indicative of tendon degeneration or fibrosis [161][162][163]. Therefore, in order to achieve a beneficial ECM replacement by MSC, their ECM synthesis would have to be highly balanced. It is not yet sufficiently proven that this can be achieved by inducing tenogenic differentiation.
With respect to the hypothesis of active ECM remodeling by MSC, comparatively few data exist so far. Treatment with BMP-12 induced an enhanced secretion of MMP-1 and -8 by ASC [93]. Similarly, ASC culture in collagen scaffolds increased MMP-1, -2, -8, -9, and -13 gene expression and MMP activity compared to two-dimensional culture [46]. For tendon-derived stem cells, it was also found that cyclic mechanical loading did not only upregulate ECM-related genes but also the integrins α1, -α2, and -α11, as well as MMP-9, -13, and -14 [164]. Thus, tenogenic stimuli may increase expression and activation of MMP by MSC. Furthermore, it was found that BMSC inhibits MMP activity in the cell culture medium through secretion of TIMP-1 and TIMP-2, even in an inflammatory environment [165], but that BMSC as well as ASC accumulate active MMP at their cell surface [166]. Although these latter two studies did not focus on tendon therapies, they suggest that MSCs could contribute to matrix remodeling in a highly targeted manner.

Tendons
Some studies also provide first insight into the interplay of MSC and tenocytes/ tendon ECM in matrix remodeling and will therefore be addressed in more detail. In direct co-cultures of ASC and tenocytes, a different temporal regulation of MMP and ECM components was observed compared to tenocytes alone [105]. This included the upregulation of collagen I and tenascin-C gene expression at day 7 and downregulation of tenascin-C and collagen III at later time points (14 and 21 days, respectively) and a higher collagen I to collagen III ratio on protein level at day 7. MMP-1, -2 and -3, as well as TIMP-1 gene expression, increased over time in tenocytes alone but showed a different temporal regulation pattern in the cocultures with a significantly increased MMP-3 expression at day 7 [105]. A different study from the same group investigated the indirect co-culture of ASC and tendon explants [104]. Here, total protease activity was increased in the co-cultures at day 3, as were the collagenases (putatively MMP-1 and -14) but not the stromelysins MMP-3 and -10. Furthermore, collagen III and tenascin-C deposition by ASC were reduced at day 7. Histology also suggested that ASCs had protective effects on the explant structure, but this was not consistent between donors [104]. However, seemingly in contrast to these findings, MMP-8, -9, and -13 expression by ASC in collagen scaffolds was lower upon stimulation with tendon ECM extract [46], and microvesicles from amniotic membrane mesenchymal cells induced a downregulation of MMP-1, -9, and -13 in tenocytes [101]. Thus, while it can be assumed that MSC actively contribute to and/or modulate tendon ECM remodeling, the exact temporal regulation and context-sensitivity of this mechanism need to be addressed in future studies.

In vivo evidence
Several in vivo studies have investigated the effect of MSC treatment on tendon ECM composition and structure, as well as on tendon biomechanical parameters. In most of these studies, including an equine large animal study with a follow-up of 45 weeks, the ECM composition was improved by BMSC and ASC treatment, with higher expression of collagen I on gene and/or protein level [106,114,120,122,140]. Collagen III expression was found to be decreased after ASC implantation [110,125,126] but increased after BMSC implantation [106,122]. Tenascin-C and decorin were found to be increased following BMSC and ASC treatment [112,114,140], and glycosaminoglycans were decreased after BMSC treatment [141]. Based on these data, MSCs appear to increase collagen I deposition in healing tendons. Furthermore, as an increase of human-specific collagen I and tenascin-C was demonstrated in a rat model after human ASC implantation, there is also some evidence that MSCs actively contribute to the synthesis of new ECM [114]. The contribution of collagen III, tenascin-C, and decorin synthesis/modulation to tendon healing is to be considered controversially, as illustrated above, and certainly depends on its balance with regard to other ECM components. Yet, beyond mere collagen I synthesis, BMSC and ASC have also repeatedly been shown to improve the structural organization of healing tendons, again including the study with a 45-week follow-up, as well as an experimental trial in horses with naturally occurring tendinopathy [108,115,121,140,141]. In conjunction with the synthesis and protection of desired ECM components such as collagen I, this could be due to active ECM remodeling and the contribution of synthesized small ECM molecules to collagen fibrillogenesis. Still, it should be acknowledged that some studies in the equine model could demonstrate only few compositional or structural improvements 5 months after ASC treatment [133,137]. Moreover, despite generally improved ECM structure and collagen I synthesis, collagen II deposits and areas staining positive for alizarin red were found in BMSC-treated tendons [106], suggesting that erroneous MSC differentiation toward the chondrogenic and osteogenic lineage had occurred. Nevertheless, functional testing of BMSC-and ASC-treated tendons indicated an improvement of functional parameters in the majority of studies [107, 108, 112-115, 117, 119, 121, 122], which represents a beneficial effect that can be attributed to ECM regeneration [3].
So far, very few in vivo studies have investigated the effect of MSC on the presence and activation of matrix-remodeling enzymes and their endogenous inhibitors. In the equine model, MMP-13 activity was decreased 6 months after BMSC treatment [141], and MMP-3 gene expression was upregulated in the healing tendons 45 weeks after BMSC treatment [140]. Together, these results might suggest that collagen degradation could be inhibited while degradation of small ECM components is promoted. However, there is much overlap regarding MMP substrates [167], and other studies found no significant differences in MMP and TIMP expression due to ASC treatment [112]. Further studies have to substantiate this hypothesis.
When MSCs were combined with tenogenic growth factors, conflicting results were reported. Treatment with ASC and GDF-5 decreased MMP-2 and TIMP-2 expression and resulted in inferior biomechanical properties compared to ASC treatment alone [112]. Treatment with ASC and BMP-12 promoted ECM degradation, which was interpreted as a side effect of the fibrin-based delivery vehicle [124], but improved tendon ECM regeneration when delivered as cell sheets without fibrin [123]. Interestingly, the latter study showed that this may have been mediated by modulating the ECM remodeling activity of macrophages [123]. A further study from the same group demonstrated beneficial effects of combined ASC and CTGF treatment, although not evaluating effects of ASC alone [125]. A different study showed that predifferentiated BMSC sheets, induced by stepwise stimulation with TGF-β1 and CTGF, resulted in superior tendon regeneration, including improved biomechanical properties than BMSC alone [54]. However, in this study, again, fibrin was used for delivery of noninduced cells, which may have contributed to the differences observed. Thus, although some data suggest that the additional use of growth factors potentiates the beneficial effects of MSC on ECM regeneration, more evidence supporting this hypothesis is required. It should also be acknowledged that growth factor supplementation might impair other regenerative mechanisms of MSC at the same time [93].

Immunomodulation
There is a substantial body of evidence that demonstrates the immunomodulatory potential of MSC. While not all underlying mechanisms have been elucidated in detail yet, it is well-understood that MSCs suppress T cell proliferation and promote the modulatory M2 macrophage phenotype [168]. Furthermore, small ECM molecules synthesized by the MSC, such as tenascin-C and decorin, could contribute to immunomodulation [163,169]. Therefore, it is likely that immunomodulation plays an important role in MSC-based tendon therapies. Against that background, it appears surprising that relatively few studies have addressed the interplay between MSC and the immune system in the context of tendon disease. This may be due to the long-existing perception that inflammation is absent during most stages of tendon disease, which, however, has been changing [5,170]. While so far existing findings are summarized in the following, immunomodulation in the context of tendon disease will remain a promising field of future research.

In vitro evidence
In vitro evidence for MSC immunomodulation in tendon disease is scarce. The most comprehensive study investigated whether ASCs influence the effects of differently polarized macrophages on tenocytes in a tri-culture system [98]. In co-cultures of M1 macrophages and tenocytes, release of inflammatory mediators, such as PGE2 and IL-1β, was increased compared to M1 macrophage cultures alone or compared to co-cultures with M0 or M2 macrophages, suggesting inflammatory tenocyte activation. When ASCs were directly co-cultured with the macrophages for 5 days, with the tenocytes added for the last 24 h, tenocyte activation was decreased, with significantly lower release of TNF-α and IL-1β in tri-cultures with M1 macrophages. At the same time, the presence of ASC had increased CD206 expression in M0 and M1 macrophage populations, indicating a switch toward the anti-inflammatory M2 macrophage phenotype and providing insight into the suppressive mechanism. However, ASCs did not effectively counteract inflammatory activation of tenocytes by IL-1β, even when ASCs had been primed with IFN-γ [98].
Interestingly, it has also been shown that tenogenic differentiation of BMSC induced by GDF-5 involves arachidonic acid production and signaling pathways [67], suggesting a link between differentiation and inflammatory processes. In this line, addition of BMP-12 increased IL-6 secretion by ASC and attenuated the suppressive effect of ASC in a mixed lymphocyte reaction [93]. Microvesicles from amniotic membrane mesenchymal cells downregulated TNF-α expression in tenocytes but in contrast to conditioned medium, they had no effect on peripheral blood mononuclear cell proliferation [101,171]. These studies provide preliminary insight into the modulation of inflammatory tenocyte activation by MSC, while they also suggest that their immunomodulatory potential may be higher when not tenogenically differentiated. Yet, MSC immunomodulation is highly contextspecific and influenced by a variety of factors including three-dimensional culture environments as well as inflammatory priming/licensing [172,173]. Therefore, it remains crucial to perform further studies specifically mimicking aspects of tendon pathophysiology.

In vivo evidence
The most insightful studies were performed by the same group, shedding light on ASC-mediated immunomodulation in tendon healing in the canine model [123][124][125][126]. Corresponding to the group's in vitro findings, ASC alone, delivered via cell sheets, stimulated the anti-inflammatory M2 macrophage phenotype in healing tendons and reduced total mononuclear cell infiltration. The M2 macrophage markers CD163, MRC1, and CD204 were increased on mRNA and/or protein level, as well as IL-4, prostaglandin reductase-1, and VEGF [123,126]. Combined administration of ASC and BMP-12 promoted these effects, particularly with respect to IL-4 expression [123]. Furthermore, combined treatment with ASC and CTGF decreased IL-1β, IL-6, and IFN-γ and increased IL-4 expression [125]. These latter findings challenge the hypothesis that tenogenic differentiation decreases the MSC immunomodulatory potential. However, when the inflammatory reaction at the tendon repair site was promoted by a fibrin-based delivery vehicle, ASC and BMP-12 further fostered these unwanted effects [124]. This might indicate that strong inflammation alters the MSC immunomodulatory properties toward a proinflammatory phenotype. In contrast, priming with TNF-α increased the anti-inflammatory effects of BMSC: While nonprimed as well as primed BMSC increased IL-10 and reduced IL-1α, primed BMSC also reduced IL-12 and the numbers of M1 macrophages and increased IL-4 and the numbers of M2 macrophages in rat Achilles tendon defects [118]. Further evidence of anti-inflammatory effects of BMSC in tendon healing was demonstrated in a rat model, in which TNFα, IFN-γ, and IL-1β were reduced, along with an increase of IL-2 and growth factors, including VEGF [111]. Apparently in contrast to most of these findings, however, we observed that clinical signs of inflammation were increased by ASC treatment in the equine model, although this effect was transient [137]. This again illustrates that MSCs can also adopt a pro-inflammatory phenotype and raises questions as to how and whether this should be controlled. When addressing this issue, it should be acknowledged that a certain extent of inflammation is required to drive resolution. In this respect, macrophages and their M2 polarization driven by MSC may play a particularly important role.

Trophic support and pro-angiogenetic effects
In addition to the direct effects of MSC on ECM composition and immune cells, trophic support and protection of resident cells are likely to contribute to beneficial effects of MSC in tendon healing. Tenocytes and tendon stem cells rescued by the MSC may be enabled to promote ECM regeneration and counteract inflammation. Furthermore, a MSC-mediated increase in vascularity may be beneficial at least in some stages of tendon healing, as it would improve energy and oxygen supply, as well as disposal of metabolites, thus reduce oxidative and metabolic stress. The presence of vascular endothelial cells, as well as the combination of tenogenic growth factors with VEGF, has also been shown to promote tenogenic differentiation [60, 74]. However, increased vascularity is also associated with tendinopathy pathogenesis and may foster neurogenic inflammation [6], thus this issue is discussed controversially.
Trophic effects on tenocytes were demonstrated in vitro, when ASC and BMSC, as well as BMSC-conditioned medium, promoted the proliferation of tenocytes [94, 102,103]. Furthermore, ASC as well as BMSC-conditioned medium promoted tenocyte migration [102,103], and ASC promoted healing in a microwound model [92]. In vivo, results are inconsistent as to whether BMSC and ASC decrease [137,141] or increase [117] cellularity within healing tendons. However, the rate of apoptosis was lower following BMSC treatment [107], suggesting protective effects of the MSC. Moreover, ASC combined with CTGF locally increased the numbers of CD146-positive tendon stem cells, suggesting an activation and possible rescue of this endogenous cell population [125].
Pro-angiogenetic effects were observed in small, as well as large animal studies, which demonstrated that BMSC and ASC implantation increased vascularity [106,129,131], likely mediated by an increase in VEGF (see below). Yet, the opposite effect was observed in horses suffering from naturally occurring tendinopathy following implantation of BMSC [141].
With respect to possible growth factor signaling, in vitro, higher TGF-β bioactivity was found in the BMSC secretome compared to tenocytes [100]. Upon tenogenic differentiation of ASC using BMP-12, VEGF secretion was significantly increased, although no effect on TGF-β was observed [93]. First in vivo evidence regarding the contribution of growth factors in tendon healing following BMSC or ASC implantation was obtained in rat models, in which VEGF, TGF-β, and hepatocyte growth factor expression were increased in the MSC treatment groups [106,111,112]. Yet, these studies did not comprehensively reveal whether these factors were released by the MSC or other cells within the tendon lesion.
The brevity of this subsection illustrates that the insight into trophic and protective mechanisms, as well as growth factor release by MSC, in the context of tendon therapies is still limited. Further research is crucial to improve our ability to exploit these effects and, last not least, to prevent potential negative effects associated with some growth factors, such as hypervascularization in response to VEGF or fibrosis in response to TGF-β.

Discussion
This review aimed to compile the evidence supporting specific mechanisms of action that may contribute to tendon regeneration in MSC-based cellular therapies. The analysis of the recent literature demonstrated an imbalance between the numbers of studies investigating tenogenic differentiation in vitro and ECM regeneration in vivo and the numbers of studies elucidating other potential mechanisms. This is conceivable as most studies investigating MSC in the context of tendon disease did not specifically aim at clarifying the mechanisms of action. Particularly, the in vivo studies mostly addressed MSC efficacy, at which ECM characteristics are reasonable outcome parameters. Still, despite the overlap with tissue engineering, the overrepresentation of tenogenic differentiation studies may reflect a delay in the field of tendon research. Tendon pathophysiology itself is still not well-understood, making it challenging to transfer the rapidly changing perception of MSC into experimental settings relevant to tendon disease in a timely manner. Yet, it can be anticipated that the general understanding of MSC mechanisms will be successively incorporated into tendon research in the following years.
Taking into account the existing data, the best-evidenced beneficial effect of MSC in tendon regeneration is the improved ECM regeneration. MSCs may also protect and rescue resident tendon cells, but only few data support this hypothesis so far. Both, ECM regeneration and tendon cell protection, are likely to be mediated by a range of mechanisms acting in concert. These may be active over long periods of time, as the engraftment of MSC within tendon lesions was repeatedly demonstrated.
The possible mechanisms mediating ECM regeneration include ECM synthesis and targeted remodeling by the engrafted MSC, inhibition of MMP over-activation, modulation of immune cells with suppression of macrophagemediated matrix degradation, and modulation of growth factor signaling. Last but not least, the rescue of resident tendon cells could prevent ongoing ECM degeneration, and their trophic support and stimulation by MSC-derived growth factors could re-initiate ECM synthesis and a healthy state of ECM remodeling driven by the tenocytes. A varying extent of evidence supports these different mechanisms, with the collectively most convincing data available for ECM synthesis, immunomodulation, and VEGF-mediated angiogenesis. Figure 3 illustrates the possible interplay between the different mechanisms and their potential synergies.
The figure summarizes the currently known mechanisms of MSC that may contribute to tendon regeneration. Mechanisms for which there is conclusive evidence from in vivo studies are designated in bold typeface.
However, there may also be antagonisms between different mechanisms, although the evidence is not yet entirely conclusive. Perhaps, tenogenic differentiation and immunomodulation may not occur at the same time. Tenogenic differentiation was shown to interfere with the immunomodulatory potential of MSC [93], and inflammatory environment compromised tenogenic MSC properties [84]. Yet, some in vivo studies revealed anti-inflammatory effects of combined MSC and tenogenic growth factor administration [123,125], although it remained unclear if the MSCs had undergone tenogenic differentiation. It is possible that the context, i.e., the stage of tendon disease, may favor one mechanism over the other. For example, immune cells such as macrophages are not predominating during subclinical stages [6], and the macrophage polarization pattern is distinct in acute vs. chronic disease [8], which will certainly impact on the activation of MSC immunomodulatory mechanisms.
A range of limitations impedes a coherent interpretation of the existing data. These include the different treatment approaches chosen and models used, which make it difficult to elucidate specific reasons for contradictory findings. Inter-donor variability is a further issue that may obscure clarity of findings in studies using human or large animal MSCs [100,143]. Furthermore, although tenogenic differentiation has extensively been studied, there is neither a consensus on differentiation protocols nor have specific markers for tenogenic differentiation been used consistently. Next, the limited understanding of tendon (patho)physiology makes it difficult to judge whether certain effects observed are beneficial or rather detrimental, e.g., with respect to MMP or TGF-β activity. Last but not least, the illustrated imbalance between evidence levels for particular mechanisms makes it difficult to draw a comprehensive picture at the moment.

Conclusion
This review demonstrates progress but also substantial weaknesses which still exist in our understanding of MSC-based cellular tendon therapy and the MSC mechanisms of action in tendon healing. Therefore, considering the low level of clinical evidence, at the moment, MSC-based treatment of tendinopathy appears only justified in the framework of clinical studies. Otherwise, although clinical translation appears temptingly close, it may be wiser to slow down the pace and focus on research into MSC mechanisms in relevant disease models to eventually be able to coax the MSCs toward targeted tendon regeneration.