Adipose-derived stromal cells for osteoarticular repair: trophic function versus stem cell activity

The identification of multipotent adipose-derived stromal cells (ASC) has raised hope that tissue regeneration approaches established with bone-marrow-derived stromal cells (BMSC) can be reproduced with a cell-type that is far more accessible in large quantities. Recent detailed comparisons, however, revealed subtle functional differences between ASC and BMSC, stressing the concept of a common mesenchymal progenitor existing in a perivascular niche across all tissues. Focussing on bone and cartilage repair, this review summarises recent in vitro and in vivo studies aiming towards tissue regeneration with ASC. Advantages of good accessibility, high yield and superior growth properties are counterbalanced by an inferiority of ASC to form ectopic bone and stimulate long-bone healing along with their less pronounced osteogenic and angiogenic gene expression signature. Hence, particular emphasis is placed on establishing whether stem cell activity of ASC is so far proven and relevant for successful osteochondral regeneration, or whether trophic activity may largely determine therapeutic outcome.


Introduction
Established strategies for cartilage and bone repair, such as autologous chondrocyte transplantation (ACT) (Ref. 1) and bone grafting (Ref. 2), have reached broad clinical application and yield satisfactory results due to continuous improvement. These therapies, however, require the excision of healthy tissue from a nonlesioned site, necessarily incorporating the disadvantages of additional medical procedures, donor site morbidity and further rehabilitative burden on the patient (Ref. 3). Repair strategies that are based on autologous bone-marrow-derived stromal cells (BMSC) do not circumvent these problems, but harvesting bone marrow from the iliac crest is generally judged as less invasive (Ref. 4). The discovery that multipotent stromal cells can be isolated from lipoaspirates (Ref. 5) and that the number of adherent cells in an equal volume of adipose tissue exceeds the content of bone marrow aspirate by about 300-fold (Refs 6, 7, 8) challenged the assumption that bone marrow would be the most appropriate source for cellbased therapies of skeletal injuries and diseases.
In order to verify whether adipose-derived stromal cells (ASC) represent an easily accessible cell type that may substitute for BMSC completely in cell-based approaches for osteochondral regeneration, they were characterised in terms of in vitro performance (Refs 9, 10), in vivo localisation (Refs 11,12) and their ability to differentiate into various mesenchymal cell types (Refs 13,14,15,16). This review summarises current knowledge of ASC and BMSC plasticity and in vivo function, describing similarities and differences between both cell types that have been determined upon expansion. Furthermore, an overview is provided on osteoarticular regenerative approaches that have thus far been conducted using ASC. In summary, data on ASC-based osteoarticular repair strategies indicate that ASC do not possess intrinsic osteochondral potential, such as BMSC, but require reprogramming for in vivo development towards the osteochondral lineage. These observations stress the concept of equivalent mesenchymal progenitors in bone marrow and adipose tissue (Ref. 8). In view of a long list of successful experimental intervention studies in distinct models, trophic functions of ASC may be more relevant than stem cell potential in mediating osteoarticular repair.

Stemness of BMSC and ASC Criteria for stem cell definition
Thus far absent from the literature is a comprehensive, general convention that defines intrinsic properties for stem cells of any given tissue (Ref. 17). From a functional point of view, a well-accepted interpretation would be that a single stem cell possesses the capacity to build up a physiological, multicellular tissue that is capable of autonomous regeneration in vivo. Specific cellular functions such as asymmetric cell division, prolonged self-renewal and differentiation capacities are needed to fulfil this requirement. Most importantly, in vitro detection of these properties in a particular cell type alone, however, does not necessarily prove stemness. It is self-explanatory that a stem cell only deserves this designation if the observed fundamental capacities represent intrinsic features of the native cell in vivo, rather than being achieved by artificial treatments or molecular reprogramming. These stringent criteria for stem cell definition (Ref. 18) are met by haematopoietic stem cells (HSC), which reconstitute bone marrow when clonally derived HSC are transplanted into lethally irradiated mice (Ref. 19 and subjected to an adipogenic pre-induction protocol prior to subcutaneous implantation. As expected after pre-induction, in vivo adipose tissue formation was reported in these studies. Evidence that transplanted clonal ASC can generate adipose tissue in vivo without such a pre-induction is still missing, but such a demonstration would not only be encouraging for their use in adipose tissue engineering but would also further clarify if ASC may indeed represent tissue-specific stem cells distinct from BMSC. In vitro characteristics of expanded ASC and BMSC Similar morphological features but different growth behaviour A thorough review of the literature on in vitro performance of culture-expanded ASC and BMSC reveals strong similarities between stromal cells of both sources, factually overweighing the differences. For instance, no morphological differences were reported to date, and the same spindle-shaped phenotype was frequently described (Refs 39, 40, 41). Upon isolation, adherent human and mouse ASC seem to exhibit a higher proliferation rate (Refs 41, 42, 43, 44, 45), but equal growth behaviour compared to BMSC has also been reported (Refs 40,46 By means of a quality control check that is commonly performed at the beginning of studies, ASC were frequently analysed for CD34-negativity, since the absence of this marker is a prerequisite to meet the minimal criteria for multipotent mesenchymal stromal cells (Refs 25, 47). However, several reports that dealt with a comparison of ASC and BMSC described that adherent ASC included a substantial CD34-positive fraction, whereas BMSC that were analysed in parallel were completely CD34-negative (Refs 48, 50, 52, 60,61). In these studies, the selection of adherent cells from the adipose tissue-derived SVF at first led to a twofold enrichment of CD34-positive cells (Ref. 61), followed by a gradual decrease in subsequent passages (Ref. 24). Nevertheless, a considerable number of CD34-positive cells was still detected in passage 4 (Refs 48, 52), and Yoshimura et al. even described that after 20 weeks of cultivation, almost 20% of the ASC population was still CD34-positive. These inconsistent data on CD34 expression in ASC cultures may simply reflect that, depending on tissue source and isolation protocol, CD34positive endothelial cells were occasionally included in primary isolates and gradually disappeared with culture time, due to unfavoured growth conditions. It remains to be determined, however, if this hypothesis or the choice of an antibody of the appropriate subclass (Ref. 47) accounts for the conflicting results. In any case, the general absence of CD34 in BMSC cultures represents another noticeable difference compared to ASC preparations.
Compared to the CD markers that were discussed above, considerably less experimental data indicate differential expression of CD10

Reduced performance of ASC in osteochondral in vitro differentiation assays
In line with indications of an intrinsic osteogenic potential of BMSC, exposure to common osteogenic differentiation media induced more mineralisation (Refs 40, 50, 68, 69), higher alkaline phosphatase activity (Refs 40, 44, 68) and stronger gene expression of osteogenic markers, such as runx2, osteocalcin, osterix, alkaline phosphatase and collagen-1 (Refs 40, 44, 52), compared to ASC. In turn, and corresponding to their physiological origin, ASC seem to exhibit a higher affinity to adipogenic differentiation, since inclusion of lipid droplets (Refs 44, 50, 53) and expression of the adipogenic marker gene peroxisome proliferator-activated receptor (PPARγ) (Refs 44, 53) were more intense than in BMSC upon induction. However, similar adipogenic in vitro differentiation capacities of adipose and bone marrow-derived cells were also reported (Refs 52, 69, 70), but no study described a higher adipogenic potential for BMSC. In line with better in vitro osteogenesis, BMSC also showed better performance in common chondrogenesis assays. In vitro differentiation of BMSC in 3D-pellet culture under treatment with TGF-β resulted in more intense collagen-II staining (

Comparison of trophic activity
One main path to tissue reconstruction by cellbased therapeutic strategies involves stem cell activity to establish and build new tissue by proliferating and differentiating cells, which are progeny of the implanted cells. A second way to regeneration is the stimulation of endogenous healing capacity by trophic activity of implanted cells, which attract host progenitor cells and organise repair by local and invading cells. Implanted cells may even disappear after this task has been successfully fulfilled. In this second scenario, even transient stem cell activity or differentiation capacity within target tissues may be dispensable as long as trophic activity is high.
Similar to the established trophic role of BMSC (Refs 72, 73), cultured ASC were shown to secrete a wide range of proteins (Ref. 74

In vivo comparison of ASC and BMSC
The extent of the described in vitro differences between ASC and BMSC gives the impression that cells of both sources may fundamentally differ from each other. This point of view must be carefully considered, since in vitro variances may stem from dissimilar donor tissue processing, cell isolation protocols, cell yield and culture methods. In the context of osteochondral regeneration, the proof of in vivo exchangeability of ASC and BMSC is far more important, and aspects of in vivo stem cell activity like trophic activity should be considered, as long as precise healing mechanisms are unclear for the diverse application settings.

Untreated ASC do not form ectopic bone
Ectopic bone formation is a standard activity of human BMSC on calcium phosphate ceramics such as β-tricalcium phosphate (β-TCP) and hydroxyapatite (HA)/TCP in immunodeficient mice (Refs 20,92,93,94,95,96,97) with no osteogenic pre-induction protocols required, in line with skeletal stem cell activity of BMSC. Ectopic transplantation of ASC reliably led to de novo generation of bone when cells were subjected to osteogenic pre-induction before implantation (Refs 45, 98, 99, 100, 101). Overexpression of BMP-2/RUNX2 (Ref. 102) or BMP-7 (Ref. 103) in ASC allowed the omission of the pre-differentiation step.
Whether ASC possess the same intrinsic ability to form ectopic bone without any of these pretreatments in standard assays, and to what extent they build up new bone themselves, largely remained elusive until the issue was recently addressed by Brocher  All in all, beyond their reduced performance in osteochondral in vitro differentiation assays, ASC showed no intrinsic osteochondral in vivo differentiation potential and, thus, seem to possess no skeletal stem cell properties as seen with BMSC, providing a strong argument for fundamental functional differences regarding their use for in vivo osteochondral repair. Since nonclonal cells are widely used for tissue regeneration, the benefit of enhanced availability of ASC, therefore, appears currently to be balanced by an enhanced need for inductive conditions via timely and intensive in vitro culture efforts, if their physical contribution to the new skeletal tissue is desired.
ASC and BMSC require pre-differentiation for ectopic cartilage formation The most convincing demonstration of spontaneous chondrogenic in vivo potential of ASC and BMSC derives from observations of

Missing evidence for physical ASC contribution to the repair of damaged cartilage
The most direct and least invasive approach to use ASC for the treatment of cartilage defects is by intraarticular injection of cells. Studies that started with an induction of osteoarthritis (OA) by anterior cruciate ligament transection (ACLT) or collagenase treatment, followed by intra-articular injection of autologous ASC, have been conducted in mouse (Ref. 114) and rabbit (Refs 115,116) (Table 3). Different histological evaluations and OA scoring scales were used to measure OA progression, but in all cases, positive effects of ASC compared to the injection of cellfree solvent were reported. Labelled ASC were detectable in the synovial membrane and medial meniscus 20 days after injection (Ref. 115) and at the synovial lining and cruciate ligaments up to 5 days after injection (Ref. 114). Human ASC injected into unimpaired mouse knee joints showed long-term persistence in joint tissue in 60% of all mice up to 186 days after injection, but a substantial fraction of ASC seemed to have migrated to the bone marrow, adipose tissue and muscle. Thus, while a certain degree of persistence of injected cells can therefore be assumed, evidence for in vivo differentiation of donor ASC or long-term integration into articular cartilage tissue is missing, and contributions by trophic activity cannot be judged.
Besides artificial OA induction, the capacity of ASC to repair surgical cartilage incisions has been investigated (Table 3). In a scheme similar

Site-dependant bone repair capacity of ASC
The majority of ASC-based tissue engineering approaches are directed at orthotopic in vivo formation of bone (Table 4). Across all of these studies, a well agreed upon point is that repair of defective bone by ASC can be achieved when transplantation is preceded by extensive predifferentiation protocols ( Refs 101,123,124,125,126,127,128) or genetic manipulation with genes encoding for bone inducers ( Refs 129,130,131,132,133). Overexpression of BMP-2 represents the most common strategy for the latter technique, using transgenic ASC for local growth factor delivery, rather than expecting spontaneous differentiation into osteoblasts. BMP-2 overexpression in ASC has also been used to substitute strategies that include immobilisation   Controversial data exist regarding the performance of expanded but otherwise untreated ASC in the context of long bone repair, although only a limited number of studies is available (Table  4). When nontransgenic or mock-transduced ASC were loaded on carrier matrices and transplanted into long-bone defects, no bone formation was observed (Refs 129,131,132). In a sheep longbone model, ASC were unable to induce defect bridging, while BMSC facilitated defect regeneration in the same setting (Ref. 143). A closer look at the ASC control groups of the above studies further confirms the impression that pre-differentiation or genetic manipulation of ASC is a prerequisite for stimulation of bone formation. This applies even for orthotopic sites in long bones, where the microenvironment is rich in osteoinductive proteins which are released from the defect endings. To our knowledge, the only exception when nontransduced ASC led to substantial bone formation in the context of long-bone repair is a study by Han and Li, in which ASC were used as a control to Runx2-overexpressing cells. Possibly, the surgical connection of the implant to the vascular network was the key to the positive results of this study (Ref. 130).
Dissimilar to long-bone repair, healing of critical size defects in the cranium appears to be less challenging with untreated ASC, since bone formation without any in vitro predifferentiation was reported in at least four studies ( Refs 144,145,146,147). Furthermore, untreated ASC that were transplanted as controls for newly established repair strategies also generated considerable amounts of bone ( Refs 125,128,142), although complete absence of defect repair by ASC control groups has also been described (Refs 123, 126). Thus, orthotopic bone formation by uninduced ASC appears to be site-dependent and favoured by characteristics of the cranial microenvironment that are not present in long bones. Origin from the ectodermal germ layer, development via the intramembranous pathway and enhanced blood supply differentiates bone in the cranium from long bones. Thus, it is tempting to speculate that one major advantage in the cranium may be the denser vascular network of skull bones, which is especially interesting in the context that, beyond an absent osteochondral commitment, a lower angiogenic signature was noted for ASC (Refs 42,59,66,148) and orthotopic bone formation by ACS can be triggered by co-transplantation of endothelial cells (Ref. 127).
If the trophic activity of ASC is the most crucial for stimulation of bone repair, a lower requirement for attraction and stimulation of endothelial progenitors could explain the better performance of ASC in the cranium. In line with this, a persistence of donor ASC could not be detected for more than 2 to 4 weeks after transplantation, even in settings where complete cranial defect repair was observed ( Refs 131,142). In sharp contrast, a single study by Cowan et al. reported that transplantation of uninduced ASC led to stable engraftment of the cells in a cranial defect and over 95% of nuclei in the newly formed bone were donor-derived after 12 weeks (Ref. 144). As it is not apparent which specific experimental parameters have enabled this exceptional engraftment, analogous success is waiting for repetition. Overall, particular success of ASC in cranial but not long-bone defects suggests that, in view of their low osteochondral and angiogenic signature, ASC affect bone regeneration most probably via their trophic activity than by in situ differentiation to osteoblasts with long-term persistence. Additional precise studies must unravel the contribution of host and donor cells to tissue repair as well as the influence of scaffolds, precultivation, species and defect site in order to reach consensus on the main mechanisms driving ASC-dependent promotion of osteoarticular repair despite lower osteogenic and angiogenic signatures and an apparent lack of skeletal stem cell properties.

Conclusion
More than 50 in vivo studies have been performed to date in order to verify the potential of ASC to be used for osteoarticular regeneration. In each of the quite heterogeneous experimental setups, specific protocols were established that either enabled chondrogenic or osteogenic differentiation of the cells or that resulted in positive effects on defect healing. Regarding the greater accessibility of ASC compared to BMSC, these data are entirely encouraging for the future use of ACS in skeletal regenerative medicine. However, it is now clear that ASC do not exhibit the same degree of osteoarticular predetermination as BMSC and more manipulation is required to drive ASC into the chondrogenic or osteogenic lineage (Fig. 1). The observations that the spontaneous formation of an ectopic bone organ by BMSC cannot be reproduced with ASC and that orthotopic bone formation is only stimulated at favoured sites confirm this issue and thereby exclude a skeletal stem cell identity for ASC. Altogether, a review of the literature suggests that mainly trophic functions determine the therapeutic outcome after ASC application. Future research is needed on a direct comparison of BMSC and ASC in osteoarticular therapy to decide where and how successful BMSC protocols have to be modified to achieve promising results with ASC.

Conflicts of Interest
None. Features associated with this article Figure  Figure 1. In vivo cartilage and bone formation potential of ASC Tables  Table 1. Summary of ectopic bone formation studies with ASC. Table 2. Summary of ectopic cartilage formation studies with ASC. Table 3. Summary of orthotopic cartilage formation studies with ASC. Table 4. Summary of orthotopic bone formation studies with ASC.