Articular cartilage is a highly specialized tissue that reduces joint friction and protects the bone ends from the shear forces associated with high mechanical load. The articular cartilage consists of chondrocytes and few progenitor cells [1] which are organized in various layers, from the fibrotic to the mature and hypertrophic mineralizing layer of chondrocytes in direct contact with the sub-chondral bone. The extracellular matrix (ECM) of chondrocytes is distinct from that of other connective tissues. This ECM is composed of a network of fibrillar collagens that give the tissue its shape, strength and tensile force and, proteoglycans that give resistance to compression [2]. It contains the large aggregating proteoglycan aggrecan which is attached to hyaluronic acid polymers via a link protein and predominantly, the collagens type II (80–90% of total collagens), IX and XI. Once the cartilage is formed in the adult, the turn-over of ECM protein replacement is low with a collagen and proteoglycan half-live of 100 and 3–24 years, respectively [3]. This, low rate of matrix remodelling partly explains why chondrocytes are relatively inactive metabolically although they can respond to various stimuli to maintain normal homeostasis.

After injury due to traumas or osteo-articular diseases, the articular cartilage is frequently damaged resulting in fibrillation and subsequent degradation which can also lay down into the sub-chondral bone. This is the result of the limited capacity of cartilage for repair due to the absence of vasculature that cannot provide the progenitor cells from the blood or the bone marrow to enter the tissue. Accordingly, the resident articular progenitor cells or chondrocytes entrapped within the surrounding matrix do not migrate into the lesions to secrete a reparative matrix. Consequently, the limited repair capacity and the absence of pharmacological agents have prompted researchers and surgeons to develop surgical methods to restore cartilage surfaces using tissue grafts or cell-based therapies (for review, see [4]). However, none of the current cartilage repair approaches allowed the generation of long term hyaline cartilage replacement tissue.

Among the possible explanations for the limited results described with the current methods of cartilage repair is the lack of integration of the chondrocytes within the existing cartilage. This is likely to be due to the insufficient capacity of implanted cells to secrete the cartilaginous matrix and to recapitulate the complex events resulting in the zonal organization of the cartilaginous tissue [5]. The lack of integration may also be due to the incomplete differentiation of chondroprogenitor cells or instability of the chondrocytic phenotype. The implantation of undifferentiated cells has already been applied in humans and although a significantly improved patient outcome was observed after one to five years, the defects were filled with a fibrocartilaginous tissue [6]. Finally, leakage of the cell suspension may be the cause of loss or decreased viability of the implanted cells as currently reported in autologous chondrocyte transplantation [7]. In summary, cell-based therapies have proved their feasibility but showed no superiority over other surgical methods on the long-term highlighting a crucial need for optimizing various combinations of cell types, scaffolds and/or chondrogenic factors (Fig. 1).

Chondrocytes, the resident cells of cartilage, produce the components of the ECM and represent the cells of choice for engineering articular cartilage. Adult chondrocytes have been isolated from various sources like articular cartilage, nasal septum, ribs or ear cartilage [8, 9]. However, ear cartilage is an elastic cartilage, which exhibit different mechanical properties as compared to the hyaline cartilage found in joint and nasal septuml. Isogai et al. have shown that chondrocytes give rise to cartilage tissue having the characteristics of its original tissue [8]. A chondrocyte from ear cartilage will thus give rise to an elastic cartilage. With a view to an application to the repair of articular cartilage, it therefor seems more appropriate to use hyaline cartilage as a source of chondrocytes. A comparison between different sources of hyaline chondrocytes (nasal, costal, and articular) has shown the superiority of costal and nasal chondrocytes on articular chondrocytes in term of quantity of cartilage formed after transplantation in subcutaneous sites [8]. One of the main limits related to the use of chondrocytes, is their instability in monolayer culture resulting in the loss of their phenotype. Indeed, chondrocytes lose the expression of the chondrocytic markers which are collagen II and aggrecan mainly, but also SZP (superficial zone protein) [10]. This loss of the chondrocytic phenotype is accompanied by a phenotypic shift towards a fibroblastic one. This fibroblastic phenotype is characterized by an increased expression of collagen I and the adoption of the spindle-shape characteristic of fibroblasts [11]. This process of dedifferentiation is however reversible. Indeed, if dedifferentiated chondrocytes are placed in a three-dimensional environment, they retrieve their differentiated phenotype [12, 13]. Chondrocytes from osteoarthritic (OA) cartilage have also been considered. However, OA chondrocytes undergo metabolic alterations, which can lead to a low response to inductive environmental factors [14, 15]. Although chondrocytes derived from OA patients appear less appropriate for articular cartilage repair, it has been reported that OA chondrocytes may be able to recover a normal chondrocytic phenotype after in vitro three-dimensional (3D) culture [16]. However, additional studies are required to clearly decipher whether OA chondrocytes could be manipulated in vitro to be suitable for cell therapy of cartilage.

Multipotent mesenchymal stromal cells or mesenchymal stem cells (MSC) are an attractive source of cells for cartilage engineering due to their easy access and high capacity of in vitro expansion. They are mainly isolated from bone marrow or adipose tissue but have been isolated from a number of other tissues including synovium, periosteum, umbilical cord vein or placenta [17]. MSC were first characterized by their clonogenic potential determined by the capacity to form Fibroblast Colony Forming Units (CFU-F). In the bone marrow, the frequency of CFU-F is in the range of 1 cell in 10⁴–10⁵ mononuclear cells [18, 19]. They are now characterized by their capacity to adhere to plastic, their phenotype (CD73⁺, CD90⁺, CD105⁺, CD14⁻ ou CD11b⁻, CD19⁻ ou CD79α⁻, CD34⁻, CD45⁻ and HLA-DR⁻) and their trilineage differentiation potential [20]. MSC also exhibit the potential to differentiate into myocytes, tendinocytes, ligamentocytes [21], cardiomyocytes [22], neuronal cells [23, 24] and other cell types [25]. More recently, these cells have been described as immunoregulatory cells since they were shown to escape the immune recognition and to inhibit the host defence mechanisms (for review, see [26]). The interest of various other progenitor cells, such as multipotent adult progenitor cells (MAPC) or marrow-isolated adult multilineage inducible (MIAMI) cells, for cartilage differentiation has been shown in vitro but has still to be validated in vivo [27, 28].

The TGF-β superfamily of polypeptides includes TGF-β, BMP, activins and inhibins. Because TGF-β and BMP are the best characterized peptides and their role in osteogenesis and chondrogenesis is well documented [3], we will focus on these molecules in the following paragraphs. The TGF-β family includes 5 members (TGF-β1–5) which are produced by many different cell types. However, the concentration of TGF-β is approximately 100-fold greater in bone than in other tissues [29]. Articular cartilage itself contains large amounts of latent TGF-β. Although present in tiny quantities in normal physiological conditions, active TGF-β1, 2 and 3 are generally considered to be potent stimulators of proteoglycans and type II collagen synthesis in primary chondrocytes (for review[3, 29]. In vitro, TGF-β1, 2 and 3 were also shown to induce the chondrogenic differentiation of MSC [30–32]. Recently, rabbit MSC encapsulated with thermo-reversible hydrogel releasing heparin-bind TGF-β3 were shown to differentiate toward chondrocytes [33]. In vivo, a porous gelatin-chondroitin-hyaluronate scaffold in combination with a controlled release of TGF-β1 could induce the chondral differentiation of MSCs to form ectopic cartilage [34]. The same group also reported that implantation of rabbit MSC in a full-thickness defect resulted in better chondrocyte morphology, integration, continuous subchondral bone, and much thicker newly formed cartilage layer when compared to control group [34]. However, several studies have shown that injection of TGF-β or TGF-β-expressing adenoviruses results in side effects in the joints, such as osteophyte formation, swelling and synovial hyperplasia [35–37].

BMPs constitute a large sub-class of polypeptides whose role is essential for chondrogenesis during skeletal development. Indeed, mutations in BMP-5 and BMP-14 genes result in brachypodism in mice and chondrodysplasia in humans. A number of mice deficient for BMP are nonviable but in BMPR-IB, ActR-IA, BMP-7 or BMP-14 deficient mice, severe appendicular skeletal defects have been observed suggesting that they play important synergistic or overlapping roles in cartilage and bone formation in vivo [38]. A number of BMP, including BMP-2, -4, -6, -7, -13, -14 can stimulate the chondrogenic differentiation of MSC [39–42] and enhance the synthesis of collagen type II and aggrecan by chondrocytes in vitro [43]. In vivo, healing of full-thickness cartilage defects in the rabbit was improved when combining microfracture and recombinant BMP-7 [44]. Similarly, the implantation of a type I collagen sponge containing BMP-2-expressing naked plasmid DNA implanted in full-thickness cartilage defects stimulates the transfection of MSC subjacent to the defect and cartilage repair [34]. The use of ex vivo retrovirally transduced muscle-derived stem cells isolated from mouse skeletal muscle to express BMP-4 enhanced chondrogenesis and significantly improved articular cartilage repair in rats [45]. A more recent study indicated that repair of chondral lesions in the knee joints of miniature pigs by periosteal precursor cells is facilitated in deeper hypoxic zones of cartilage repair tissue and is stimulated by BMP-2, and to a lesser extend IGF-1, which enhance HIF-1α activity [46]. However, when implanted in ectopic localizations, BMPs led to bone formation via endochondral ossification which should be avoided for articular cartilage engineering [47]. This observation points to the notion that optimal regulation of BMPs may enhance their efficacy in a regulated tissue engineering strategy. Indeed, Huard’s team showed that Noggin delivery can inhibit heterotopic ossification caused by BMP-4, demineralized bone matrix, and trauma in calvaria defect model [48, 49] suggesting that this strategy may be also useful for inhibing endochondral ossification induced by hypertrophic cartilage. Although these BMP have shown great potential in animal models, no clinical studies have been conducted to validate their potential to enhance cartilage repair in humans.

The Wnt family of secreted ligands contains more than 20 members in vertebrates that are characterized by conserved cysteine residues. These proteins exhibit unique expression patterns and distinct functions in development [50]. The Wnt family members signal through the canonical β-catenin-dependent pathway or β-catenin-independent pathways. Various Wnt members are involved both in early and late skeletal development and play a role in the control of chondrogenesis and hypertrophy. 9 Wnt genes are expressed in the postnatal growth plate, including Wnt-4, -5a, -5b, -10b and -11 at higher levels, and Wnt-2b, -7b, -9a and -10a at very low levels [51]. Wnt-1, Wnt-4, Wnt-7a, Wnt-8 block chondrogenic differentiation but display different effects on hypertrophy. On the contrary, Wnt-5a and Wnt-5b promote chondrogenesis [52, 53]. Wnt-5a together with Wnt-5b were shown to regulate the proliferation of chondrocytes and their maturation into hypertrophic chondrocytes in both the embryonic and postnatal growth plate and Wnt-11 does not affect chondrogenic differentiation. The role of Wnt-3a on chondrogenesis is more controversial. Importantly, we recently identified Wnt-6 as a new factor able to induce the chondrogenic differentiation of primary MSCs while inhibiting both osteogenic and adipogenic differentiation (pers.com.). The key genes identified in both embryonic cartilage development and postnatal endochondral bone formation includes the Wnt-5a/Fz-5 receptor complex and Wnt-7a [54]. It is generally reported that Wnt members responsible for the induction of the osteogenic differentiation activate the β-catenin-dependent pathway. Indeed, in vitro inactivation of β-catenin in MSC causes chondrocyte differentiation under conditions allowing only osteoblasts to form [55]. Consistently, in vitro loss- and gain-of-function analyses reveal that β -catenin activity is necessary and sufficient to repress the differentiation of mesenchymal cells into Runx2- and Sox9-positive skeletal precursors [56]. However, β-catenin was recently shown to be required for both osteogenesis and chondrogenesis in adult mature tissues [57]. Overall, it appears that the Wnt network has dual roles in cartilage, as has been described in other tissues: it is an important regulator of chondrocyte development, but deregulated signaling is detrimental to mature tissues and may lead to disease.

In vertebrates, the FGF family comprises twenty-two structurally related proteins with a molecular mass from 17 to 34 kDA that bind one of four tyrosine kinase FGF receptors (FGFR) [58]. The importance of FGF signalling in skeletal development is highlighted by the number of dysplasias, mainly dwarfing chondrodysplasias and craniosynostosis syndromes, attributed to specific mutation in the genes encoding the FGFR-1, -2 and -3 [59]. During limb development, FGFR-2, Sox9 and collagen type II are among the earliest genes upregulated in condensing mesenchyme. Mice lacking FGFR-3 develop skeletal overgrowth while mice overexpressing an activated form of FGFR-3 develop skeletal dwarfism. These data have shown that signalling through FGFR-3 negatively regulates chondrocyte proliferation through a STAT1 pathway and differentiation through the MAPK pathway. Genetic studies have also identified defects in chondrogenesis in mice lacking FGF18 whereas no apparent defects in chondrogenesis occurs in mice lacking FGF-2, -5, -6, -7, -8 and -17. Additionally, the skeletons of FGF-9^−/− mice are sightly smaller than the wild type littermates [60].

In the adult cells, the chondrogenic effect of FGF and FGFR has been confirmed by very few studies. Indeed, forced expression of FGFR-3 in the murine C3H10T1/2 MSC line was shown sufficient for chondrogenic differentiation [61]. Accordingly, FGF-18 which was shown to be a selective ligand for FGFR-3 in limb bud mesenchymal cells, suppress their proliferation while promoting differentiation to produce cartilage matrix [62]. In cultured chicken chondrocytes, FGF-9, another ligand for FGFR3, rapidly induces the upregulation and secretion of the matrix phosphoprotein osteopontin, known to be associated with chondrocyte and osteoblast differentiation. Unexpectedly, FGF-9-induced osteopontin was accompanied by inhibition of differentiation and increased proliferation of the treated chondrocytes [63]. In adult chondrocytes, FGF-2 is mainly mitogenic although recent studies have shown that it can inhibit the anabolic ativities of other growth factors, such as IGF-1. MSCs expanded in FGF-2-supplemented medium proliferated more rapidly than control MSCs and FGF-2 treatment enhanced subsequent chondrogenic differentiation in a 3-dimensional culture [64, 65]. In addition, in a rabbit model, FGF-2 can stimulate articular cartilage restoration in temporomandibular osteoarthritic defects, although the effective concentration range of FGF-2 would have to be determined [66]. Implantation of a fibrin sealant incorporating FGF-2 successfully induced healing of a mechanically induced defect with hyaline cartilage and concomitant repair of the subchondral bone [67]. One study reports the use of FGF-18 in a setting of rapidly progressive osteoarthritis in rats. The data showed that FGF-18 induced an increase in chondrophyte size and remodeling of the subchondral bone suggesting that it can stimulate repair of damaged cartilage [68]. The important issues coming from the contradictory results reported to date include a better characterization of the signalling pathways activated by FGFs, a better understanding of the interplay between these pathways as well as understanding the contribution of additional factors, such as HS, in regulating FGF activity in cartilage and bone development.

The IGF family is composed of two ligands (IGF-1 and IGF-2), two cell surface receptors (IGF1R and IGF2R), at least six different IGF binding proteins (IGFBP-1 to IGFBP-6), and multiple IGFBP proteases, which regulate IGF activity in several tissues. IGF-1 is the most studied form with respect to cartilage repair. In embryonic development, mice with IGF-1^−/− mutations display severe growth retardation, have developmental defects in various organs while IGF-2^−/− are 60% smaller than their wild-type littermates but grow normally after birth. Accordingly, mice nullizygous for the IGF1R gene demonstrate severe fetal growth retardation [69]. In humans, a reported natural deletion of exons 4 and 5 of the IGF-1 gene results in severe pre- and postnatal growth and developmental deficits, combined with mental retardation [70].

IGF-1 and IGF1R are expressed by chondrocytes, osteoblasts, and osteoclasts. IGF-1 is considered an essential mediator of cartilage homeostasis through its capacity to stimulate proteoglycan synthesis and, promote chondrocyte survival and proliferation [71, 72]. IGF-1 also induces the differentiation of MSC towards the chondrocytic phenotype as shown by the upregulation of the specific markers [33, 73]. In a critical size full-thickness cartilage defect horse model, defects filled with fibrin clots loaded with IGF-1 repaired better than empty defects and contained mainly chondrocytes with predominantly collagen type II rich matrix [74]. In the same model, the combined use of chondrocytes and IGF-1 tend to improve the overall continuity and consistency of the repair tissue [74]. However, an age-related or OA-associated decline in the responsiveness of chondrocytes to IGF-1 appears to be due at least in part to over-expression of IGFBPs. The titration of FGFs by IGFBPs may account for the variable results reported to date with IGF-1 treatment for in vivo cartilage repair studies. Improvement will require optimizing the dose, injection regimen and/or combination with other growth factors.

In mammals, the Hh family comprises 3 members of highly conserved proteins: Sonic hedgehog (Shh), Indian hedgehog (Ihh), and Desert hedgehog (Dhh). In concert with other signaling molecules, Ihh has been found to function as a central regulator of endochondral ossification, coordinating chondrocyte proliferation, differentiation, and ossification of the perichondrium. Expression of Ihh induces the upregulation of a second secreted factor, parathyroid hormone-related protein (PTHrP), which is expressed in distal chondrocytes of the skeletal elements. Mice overexpressing Ihh or a constitutively activated allele of Smo displayed an increased chondrocyte proliferation [75]. Furthermore, as Gli3 is the major mediator of Shh signaling during limb patterning, it is not surprising that mutations in the human Gli3 gene cause a variety of inherited skeletal patterning defects [76]. The requirement of Hh signalling varies in different developmental processes: osteoblast differentiation in the perichondrium and chondrocyte differentiation in the sclerotome are induced by transient exposures to Ihh or Shh, respectively, whereas the onset of columnar and hypertrophic chondrocyte differentiation depends on continuous Ihh signaling to maintain the Ihh-PTHrP interactions. Furthermore, hedgehog signaling has been shown to interact with several other signaling pathways, including those of FGFs, Wnts, and BMPs.

Although hedgedhog signalling has been mainly described in growth plate chondrocytes, the role of Shh during chondrogenesis has been shown. Indeed, in vitro, Shh-treated MSCs showed expression of cartilage markers aggrecan, Sox9, CEP-68, and collagen type II and type X within 3 weeks [77]. In addition, Hh pathway dramatically impaired adipogenesis of MSC, with reduced lipid accumulation, a decrease in adipocyte-specific markers, and acquisition of insulin-resistant phenotype stimulation [78]. In vivo, few studies are available and report the role of Shh on bone regeneration [79].