Cartilage is a unique connective tissue that plays essential roles in vertebrate development and adulthood. Cartilage anlagen develop in the embryo before bone, and thus provide the first skeleton of the embryo. They fulfill essential roles in patterning of the head, trunk, and limbs, and in protecting vital organs. Furthermore, the development of cartilage growth plates constitutes the principal drive for body growth, and also provides structural templates and induction signals for the formation of most bones through a process called endochondral ossification (Olsen et al.,2000). Finally, cartilage structures that persist through life in airways, joints, and ears are essential to our breathing, articulation, locomotion, and hearing. It is therefore not surprising that cartilage malformation diseases, which account for a large proportion of birth defects in humans, can in some cases have severe consequences for a person's expectancy and quality of life, and can lead to embryonic or perinatal lethality or life-long handicaps. The spectrum of their manifestations is wide, ranging from minor local deformities (chondrodysostoses) to generalized malformations (chondrodysplasias). Some occur as unique defects, while others occur as a part of complex syndromes involving skeletal and nonskeletal pathways. Together they illustrate the importance, uniqueness, and complexity of the chondrogenesis process, and strongly suggest the involvement of many genes and regulatory mechanisms. To understand the genetic causes of cartilage malformation diseases, and improve prevention and therapeutic strategies, we must reach a thorough understanding of this process and identify all genes involved, as well as their functions, interactions, and modes of regulation.
Cartilage is avascular and non-innervated, and the primary function of its only residing cells, the chondrocytes, is to build, maintain, and remodel the abundant extracellular matrix of the tissue. Many types of cartilage malformation diseases are caused by mutations in genes for specific components of this matrix (Mundlos and Olsen,1997a; Cohn,2001). Many laboratories have therefore been studying the transcriptional mechanisms governing their cartilage-specific expression, with the long-term goal of identifying the molecular mechanisms that drive chondrocyte differentiation and underlying cartilage diseases. The framework of the cartilage matrix is a collagen fiber network that is comprised primarily of type II collagen (encoded by the Col2a1 gene) and secondarily of type IX collagen (Col9a1, Col9a2, and Col9a3) and type XI collagen (Col11a1, Col11a2). Collagen type X (Col10a1) is also produced in abundance but exclusively by prehypertrophic and hypertrophic chondrocytes. This collagen composition of cartilage contrasts with that of bone and most other connective tissues, which are built on a fibrillar network of collagen types I, III, and V. The cartilage collagen network entraps a highly hydrated gel of proteoglycans and glycoproteins. Aggrecan (Agc1) is a large, very abundant proteoglycan that is almost unique to cartilage. It forms enormous aggregates by binding to linear chains of the glycosaminoglycan hyaluronan with the help of the link protein (Crtl1). Glycoproteins, such as cartilage oligomeric protein (Comp) and matrilin 1 (Crtm), and small proteoglycans, such as fibromodulin (Fmod) and perlecan (Hspg2), vary in abundance according to the types of cartilage and regions of the cartilage elements involved. In addition to being responsible for the biomechanical properties of the tissue, the cartilage matrix also significantly modulates chondrocyte differentiation and activity (So et al.,2001; Kirn-Safran et al.,2004).
Chondrocytes fulfill their functions in cartilage by undergoing a complex differentiation process (Fig. 1A). At the onset of skeletogenesis (at about mid-gestation in the mouse embryo), mesenchymal precursor cells commit to chondrogenesis and differentiate into prechondrocytes and then early chondroblasts. They form cartilage anlagen that prefigure future skeletal elements. In the center of the diaphyses of future long bones, the cells rapidly progress toward prehypertrophy, hypertrophy, and terminal maturation, and ultimately undergo apoptosis. Bone-forming cells then invade their lacunae and form primary ossification centers. On either side of these centers, early chondroblasts develop cartilage growth plates. They start by assuming a flattened shape and organizing into longitudinal columns. They proliferate at a high rate until, one layer at a time, they exit the cell cycle and start to increase in size, undergoing prehypertrophy followed by full hypertrophy. They ultimately terminally mature and undergo apoptosis to allow primary ossification centers to expand. The columnar chondroblast and hypertrophic chondrocyte zones contribute together to the rapid elongation of long bones. Cells located in the middle of the epiphyses of future long bones undergo a similar maturation to lead to the formation of secondary ossification centers, but do so several days or weeks after birth in the mouse, after a long pause at the early chondroblast stage. Articular joints start to form at either end of developing long bones in the fetus when mesenchymal cells located in presumptive joint areas differentiate up to a prechondrocytic stage and then undergo apoptosis to create joint cavities. Early chondroblasts that line the joint cavities develop into articular chondroblasts, and postnatally into articular chondrocytes. Each step of the chondrocyte differentiation pathway is characterized by specific histological features, cellular activities, and gene expression profiles (Fig. 1B). From a transcriptional point of view, the chondrocyte differentiation pathway thus corresponds to a succession of major genetic program switches. It was postulated years ago that these switches were likely controlled by specific sets of transcriptional activators, repressors, and associated factors. Studies in many laboratories over the world during the last two decades have started to confirm this hypothesis by uncovering a number of factors with essential roles in determining cell fate and differentiation. Some of these factors were uncovered through the identification of genes that are mutated in severe cartilage and bone malformation diseases (reviewed in Olsen and Mundlos,1997b). The goal of this review is to revise, one at a time, each of the major steps of the chondrocyte differentiation pathway, and to analyze our current knowledge and gaps in knowledge regarding the control of this process at the transcriptional level.
COMMITMENT OF MESENCHYMAL CELLS TO CHONDROGENESIS AND PRECHONDROCYTE DIFFERENTIATION
Chondrocytes derive in the early embryo from mesenchymal cells that migrate into presumptive skeletogenic sites from the cranial neural crest, paraxial mesoderm, and lateral plate mesoderm. The cells become tightly packed at these sites and form cell mass condensations that prefigure the future skeletal elements (reviewed in Hall and Miyake,2000; DeLise et al.,2000). They express extracellular matrix and cell adhesion molecules such as N-cadherin, N-CAM (Ncam1), tenascin C (Tnc), versican, and thrombospondin-4 (Fig. 1B). In the center of these condensations, prechondrocytes emerge that turn off expression of mesenchymal and condensation markers, and start to express Col2a1 and other early cartilage markers.
The transcription factor Sox9 has essential, non-redundant roles in specifying the commitment and differentiation of mesenchymal cells toward the chondrogenic lineage in all developing skeletal elements. Sox9 features an Sry-related HMG box domain, through which it binds and also bends DNA. It also features a potent transactivation domain and a dimerization domain that stabilizes binding to multiple sites on DNA. The first clue that Sox9 plays a role in chondrogenesis came with the identification of heterozygous mutations in and around SOX9 in human patients with camptomelic dysplasia (CD), a severe form of chondrodysplasia that is often associated with XY sex reversal and malformations in several internal organs (Wagner et al.,1994; Foster et al.,1994). Sox9 is turned on in chondrogenic and osteogenic mesenchymal cells prior to condensation, remains highly expressed in prechondrocytes and chondroblasts, and is turned off when the cells undergo prehypertrophy (Wright et al.,1995; Ng et al.,1997; Zhao et al.,1997). In a study of mouse chimeric embryos generated by injecting Sox9+/+ blastocysts with Sox9–/– embryonic stem cells, Sox9–/– chondroprogenitor cells were mixed with Sox9+/+ cells in presumptive sites of skeletogenesis (Bi et al.,1999). However, the cartilage anlagen that subsequently developed featured Sox9+/+ cells exclusively. A few Sox9–/– cells remained in the outskirts of condensations but did not express Col2a1, Agc1, or other typical chondrocyte markers. Complementing this study, the inactivation of Sox9 in mouse embryos using a Prx1Cre transgene, which is expressed in early limb bud mesenchymal cells prior to chondrogenesis, resulted in a virtual absence of appendicular cartilage elements (Akiyama et al.,2002). The mesenchymal cells located in the presumptive sites of chondrogenesis underwent extensive apoptosis instead of forming precartilaginous condensations. Expression of N-cadherin and N-CAM was normal, suggesting that Sox9 may primarily promote cell survival or may control other condensation genes, which remain to be identified. An additional experiment in which Sox9 was inactivated specifically in neural crest cells by means of a Wnt1-Cre transgene, demonstrated that Sox9 is not required for mesenchymal cell differentiation into osteoblasts in developing intramembranous bones, but is required to specify mesenchymal cell differentiation into chondroblasts rather than osteoblasts in the developing cartilage of future endochondral bones (Mori-Akiyama et al.,2003). Together, these experiments demonstrate that Sox9 is required to form prechondrocytic condensations.
Sox9 has master roles in the onset of cartilage development, and as yet no other transcription factors have been identified that might control early chondrogenic cell fate and differentiation upstream or in the same steps as Sox9 in all developing cartilage elements. However, many transcription factors have been identified that control mesenchymal cell migration, proliferation, survival, and condensation in one or a subset of cartilage elements. These factors include members of the Pax, Hox, forkhead-helix, homeodomain, and other protein families, and a number of them are mutated in various types of cartilage and bone malformation diseases (Mundlos and Olsen,1997b). Each factor is generally expressed in a complete or partial spatial and temporal overlap with other family members with which it shares redundant roles. Together these factors control the diversity of shapes and sizes of the skeletal elements in each organism and across species. Classically, they are therefore referred to as patterning factors. At least some of them may also play direct roles in determining chondrogenic cell fate and differentiation. Describing all of these factors is beyond the scope of this review. We will limit our descriptions to those of Pax1 and Pax9, Nkx3.2 and Nkx3.1, and Barx2. Pax1 and Pax9 are transcriptional activators with a paired box DNA-binding domain, and Nkx3.1 and Nkx3.2 are transcriptional repressors related to the Drosophila bagpipe factor. The four genes have largely overlapping expression patterns during skeleton development. They are activated in the paraxial mesoderm-derived sclerotomes, and expression of Pax1 and Pax9 is maintained in chondrogenic mesenchymal cells until they reach the prechondrocyte stage, whereas expression of Nkx3.1 and Nkx3.2 is maintained throughout chondroblast differentiation (Tribioli et al.,1997; Deutsch et al.,1988; Peters et al.,1998 and1999). Expression of Pax1 and Pax9 is induced by the signaling molecule Sonic hedgehog (Shh), and expression of Nkx3.2 (possibly Nkx3.1 as well) is induced by Pax1 and Pax9 (Rodrigo et al.,2003). While Pax9–/– mice have no apparent axial skeleton defect, Pax1–/– mice feature severely malformed vertebral bodies and intervertebral discs, and Pax1–/–Pax9–/– mice virtually lack a vertebral column (Wallin et al.,1994; Peters et al.,1998 and1999). In these double mutants, sclerotomal cells activated Sox9 and Col2a1 and correctly migrated toward the notochord, but subsequently failed to proliferate, condense properly, and maintain expression of Sox9 and Col2a1. They then underwent apoptosis. Similarly, Nkx3.1–/– mice showed no skeletal defect, Nkx3.2–/– mice showed severe axial skeletal defects, and Nkx3.1–/–Nkx3.2–/– mice showed more severe defects similar to those of Pax1–/–Pax9–/– mice (Lettice et al.,1999; Tribioli and Lufkin,1999; Herbrand et al.,2002). These four transcription factors are thus needed to maintain the chondrogenic fate of sclerotome-derived skeletal precursors and allow them to form prechondrogenic condensations. Interestingly, experiments using chicken somitic explants in culture have shown that Nkx3.2 and Sox9 are able to induce each other's expression in the presence of Shh, and that this positive autoregulatory loop can be maintained by BMP signaling and results in robust chondrogenesis (Zeng et al.,2002). The homeobox transcription factor Barx2 is highly expressed in precartilaginous mesenchymal condensations of the limb bud, and remains expressed in developing joints and articular cartilage (Meech et al.,2005). It occupies the cartilage-specific enhancer of the Col2a1 gene in vivo, possibly in cooperation with Sox9, and experiments in which its expression was manipulated in a culture model of limb bud precartilaginous cell condensation have shown its ability to regulate expression of the condensation marker N-Cam and the early cartilage matrix marker Col2a1. Barx2 may thus cooperate with Sox9 in the early development of cartilage. In vitro experiments have shown that its expression is likely under the control of BMP signaling. Its actual role in vivo, however, remains to be demonstrated.
EARLY CHONDROBLAST DIFFERENTIATION
Overt differentiation of prechondrocytes into fully committed and active chondrogenic cells is manifested by a strong increase in cell proliferation and deposition of cartilage matrix. Col2a1 expression is upregulated and a different splice variant of the gene is produced (type IIB instead of type IIA (Sandell,1994)). The cells also start to express Agc1, Cartl1, the genes for the collagen types IX and XI, and other cartilage extracellular matrix components at high levels. We previously proposed to refer to these very active cells as chondroblasts instead of chondrocytes, in analogy with the terminology used in other cell lineages (osteoblast/cyte, adipoblast/cyte, myoblast/cyte, and fibroblast/cyte). Accordingly, blasts are distinguished from cytes by the fact that they rapidly proliferate and build new tissue, whereas cytes are essentially growth-arrested and limit their activity to maintaining the functional integrity of mature tissues (Smits et al.,2001).
Sox9 remains highly expressed in chondroblasts and is required for their overt differentiation. Indeed, when it was inactivated in the mouse embryo using a Col2a1 transgene, precartilaginous condensations formed normally, but prechondrocytes were unable to undergo chondroblast differentiation (Akiyama et al.,2002). They failed to maintain expression of Col2a1 and to activate other cartilage-specific extracellular matrix genes. Two other members of the Sox family, L-Sox5 and Sox6, are critical effectors of chondroblast differentiation (Lefebvre,2002). L-Sox5 is a longer product of the Sox5 gene than the Sox5 protein initially identified in adult testis, and it is highly identical to Sox6. L-Sox5 and Sox6 share with Sox9 only partial identity in the HMG box DNA-binding domain and none outside this domain. They have no transactivation or transrepression domains, and may thus act mainly to facilitate organization of transcriptional complexes. Based on their structure and roles in vivo, it is likely that their molecular roles are virtually identical, but different from those of Sox9. They are activated in prechondrocytes and highly expressed in chondroblasts in all developing cartilage elements of the mouse embryo (Lefebvre et al.,1998). Sox5–/– and Sox6–/– mice were born with limited skeletal abnormalities, while Sox5–/–Sox6–/– mutants died in utero with rudimentary and poorly developed cartilage anlagen (Smits et al.,2001). Chondroprogenitor cells developed normal precartilaginous condensations in Sox5–/–Sox6–/– embryos, revealing that Sox5 and Sox6, in contrast to Sox9, are not needed for lineage commitment and prechondrocyte differentiation. Prechondrocytes then failed to undergo proper chondroblast differentiation. They maintained expression of Col2a1 at a low level and activated other cartilage extracellular matrix genes, including Agc1, Crtl1, Comp, and Crtm, only after a long delay and only at low to undetectable levels, and thus accumulated very little cartilage extracellular matrix. They also failed to actively proliferate. This severe impairment of chondroblast differentiation occurred despite normal expression of Sox9, indicating that Sox9 requires Sox5 and Sox6 to drive overt chondrogenesis. In another study, Sox5 and Sox6 were not expressed when Sox9 was inactivated using a Prx1Cre or Col2Cre transgene (Akiyama et al.,2002), indicating that Sox9 is required to activate these two genes. However, a direct role for Sox9 in activating these genes has not yet been demonstrated. A recent study showed that prechondrocytes in Sox5–/–Sox6–/– embryos activate specific markers of tendon fibroblasts instead of typical chondroblast markers (Brent et al.,2005), revealing that in addition to effecting chondroblast overt differentiation, Sox5 and Sox6 also secure the fate commitment of mesenchymal cells to the chondrocyte lineage.
Consistent with the results of studies in vivo, in vitro experiments have suggested that Sox9 and L-Sox5/Sox6 cooperate with each other to directly activate Col2a1 (Bell et al.,1997; Lefebvre et al.,1997,1998). The three proteins bind in vitro to an enhancer region in the first intron of Col2a1 that features several Sox binding sites and is sufficient to drive cartilage-specific expression of reporter genes in transgenic mouse embryos. Sox9 is sufficient in transient expression assays to activate constructs that feature multiple copies of this minimum Col2a1 enhancer. However, L-Sox5 and Sox6 are needed in addition to Sox9 to activate, at a significant level, constructs containing a single copy of large Col2a1 promoter and intron-1 segments, and thus resembling the Col2a1 endogenous gene more closely than the shorter constructs. It remains to be shown whether Sox9 and L-Sox5/Sox6 contact this enhancer in vivo and how they cooperate with each other. Other cartilage matrix and regulatory genes have also been shown to feature functional Sox binding sites in cartilage-specific cis-acting elements, including CD-RAP (Xie et al.,1999), Col11a2 (Bridgewater et al.,1998; Liu et al.,2000), Agc1 (Sekiya et al.,2000), and Cartm (Rentsendorj et al.,2005). Most of these studies demonstrated the ability of Sox9 to bind and activate these regulatory elements. Since L-Sox5 and Sox6 are needed, and Sox9 is not sufficient for significant expression of these genes in vivo, it is anticipated that future studies will show that L-Sox5 and Sox6 also have the ability to bind and contribute to the activation of these regulatory elements.
While gene inactivation experiments in the mouse have demonstrated essential roles for Sox9 and L-Sox5/Sox6 in chondroblast differentiation, they have not addressed the important question as to whether they are sufficient for chondroblast differentiation. To address this question, an experiment was conducted in which various cell types in vitro, and subcutaneous tissue and periosteum in vivo were forced to express these Sox genes (Ikeda et al.,2004). The data showed that Sox9, L-Sox5, and Sox6 were needed together, and were sufficient to confer on nonchondrogenic cells the ability to activate Col2a1, Agc1, and a number of other cartilage markers, and to become surrounded by a proteoglycan-rich extracellular matrix resembling cartilage tissue. Moreover, the three Sox genes together were also able to suppress expression of markers for hypertrophic chondrocytes and osteoblasts. This experiment thus confirmed that L-Sox5, Sox6, and Sox9 constitute a master chondrogenic trio. It was also noted that the addition of transforming growth factor-β (TGF-β) and BMP-2 further increased the chondrogenic potential of the Sox trio. This suggests that additional, as yet unidentified transcription factors are likely to work together with the Sox trio in promoting chondroblast overt differentiation. In addition, the fact that the three Sox genes are coexpressed in cell types other than chondroblasts, including areas of the developing brain and heart, leaves open the question of how the Sox trio fulfills its specific chondrogenic functions.
DEVELOPMENT OF COLUMNAR CHONDROBLAST ZONES
The early chondroblasts in the metaphyses of developing long bones start to form columnar zones once the cells in the diaphyses have undergone prehypertrophy and proceed toward terminal maturation. Initially small and round, they become flattened and organized into parallel, longitudinal columns. They proliferate at the highest rate at the top of the columns (away from the primary ossification center) and progressively decrease their proliferation rate as they move down the columns (Smits et al.,2004). They undergo irreversible growth arrest as they convert into prehypertrophic chondrocytes, one layer at a time, at the bottom of the columnar zone. Two models have been proposed to account for the maintenance of columnar zones in established growth plates of embryonic long bones. According to the first model, columnar chondroblasts derive from epiphyseal chondroblasts located in the periarticular region and differentiate from round to flattened cells when they enter the columnar zone (Kobayashi et al.,2005). According to the second model, columnar chondroblasts find their source immediately at the top of the columnar zone, and thus in the metaphysis rather than the periarticular region (Smits et al.,2004). The first model was proposed to interpret the results of experiments in which parathyroid hormone-related peptide (Pthrp) and Indian hedgehog (Ihh) signaling pathways were modified in various ways, such that changes were observed in the relative height of the epiphyses and columnar zones. The changes were interpreted as premature or delayed maturation of periarticular chondroblasts into columnar chondroblasts. Another interpretation is that modulations in Ihh and Pthrp signaling result in ectopic (rather than premature or delayed) differentiation of periarticular chondroblasts into columnar chondroblasts or vice versa. One argument supporting the second model is that the source of columnar chondroblasts is likely the same before and after the secondary ossification centers form (in the epiphyses). After the formation of these centers, the source of columnar chondroblasts clearly resides just above or at the top of the columnar zone (Abad et al.,2002). Another argument is that chondroblasts proliferate at the highest rate at the top of the columnar zone and may thus ensure self-renewal (Smits et al.,2004). As of today, however, fate-mapping or other experimental evidence needed to definitively validate either model is still lacking. At the gene expression level, columnar chondroblasts distinguish themselves from early chondroblasts by quantitative, rather than qualitative, changes. No specific markers have been identified for these cells, but it is well recognized that they progressively upregulate extracellular matrix genes, including Agc1 and Comp, and regulatory genes, such as Nkx3.2 and the gene for the fibroblast growth factor receptor-3 (Fgfr3), as they move down the columns. It is also clear that Ihh, Pthrp, Fgf, Bmp, and Wnt signaling molecules and transcriptional mediators interact together to control the formation and maintenance of the columnar zone, and that the basic-region leucine-zipper Atf2 transcription factor contributes to stimulate columnar chondroblast proliferation (Reimold et al.,1996; Kronenberg,2003).
L-Sox5 and Sox6 are absolutely required for the development and maintenance of columnar chondroblasts, as proved by the total absence of chondroblast columns in Sox5–/–Sox6–/– fetuses despite the development of prehypertrophic chondrocytes (Smits et al.,2001), and by a reduction in the length of columnar zones in Sox5/Sox6 compound mutants (three alleles are inactivated) and total loss of chondroblasts in some growth plates of these mutants by birth (Smits et al.,2004). The columnar zone length reduction in compound mutants has been explained by a lower rate of proliferation of chondroblasts at the top and within the columnar zone, and by premature chondrocyte prehypertrophy. Although Ihh stimulates columnar chondroblast proliferation (St-Jacques et al.,1999), its expression was not affected in Sox5/Sox6 compound mutants, and its signaling was even upregulated, as determined by increased expression of its transcriptional target and mediator Ptc1 (Smits et al.,2004). Sox5/Sox6 may thus promote columnar chondroblast proliferation independently or downstream of Ihh. Interestingly, while Sox5 and Sox6 are needed for optimal proliferation of the columnar chondroblasts, Sox9 may inhibit cell proliferation. The columnar zone indeed appeared slightly shorter than normal in mice that overexpressed Sox9 in chondroblasts, and a decrease in chondroblast proliferation and expression of cyclin D1, which is required for optimal proliferation of these cells, was measured in these mice (Sox9/Col2Cre knock-in (Akiyama et al.,2004)).
The Runt domain transcriptional activator Runx2 (also referred to as core-binding factor α 1 (Cbfa1)) and its close relative Runx3 promote columnar chondroblast proliferation and organization in columns. Although Runx2 is expressed in chondrogenic mesenchymal cells, it is no longer expressed in chondroblasts. It is reactivated, and Runx3 is activated, when chondroblasts are about to become prehypertrophic, and Runx2, but not Runx3, remains expressed through chondrocyte hypertrophy and terminal differentiation (Kim et al.,1999; Yoshida et al.,2004). Runx2 is also expressed in the osteoblast lineage, while Runx3 is not. Essential roles for Runx2 in skeletogenesis were uncovered when it was shown that haploinsufficiency of RUNX2 causes cleidocranial dysplasia (CCD) in humans, that Runx2–/– mice lack bones, and that Runx2 is required for osteoblast differentiation (Ducy et al.,1997; Komori et al.,1997; Mundlos et al.,1997; Otto et al.,1997). Runx2–/– mice also exhibited a disturbance of chondrocyte maturation, as will be described later in this review (Inada et al.,1999; Kim et al.,1999). While Runx3–/– mice had minor skeletal defects, Runx2–/– mice showed a significant reduction in proliferation of chondroblasts in some skeletal elements, and Runx2–/–Runx3–/– mice showed a similar reduction in chondroblast proliferation and a complete lack of columnar columns in all skeletal elements. These important roles of Runx2/Runx3 in columnar chondroblasts were shown to be indirect, and largely (if not entirely) mediated by their ability to activate Ihh expression in prehypertrophic chondrocytes (see below).
PREHYPERTROPHIC AND HYPERTROPHIC CHONDROCYTE DIFFERENTIATION
The differentiation of chondroblasts into prehypertrophic and hypertrophic chondrocytes represents a major phenotypic switch. In addition to exiting the cell cycle, the small and round or flattened cells progressively increase their cytoplasmic volume up to 10 times. At the prehypertrophic stage, they contain higher levels of RNA for Col2a1, Agc1, and most other early cartilage matrix genes than chondroblasts, and sequentially activate the genes for the parathyroid hormone and parathyroid hormone-related peptide receptor (Pthr1), Ihh, and Col10a1. At the hypertrophic stage, they cease to express early cartilage matrix genes and also terminate expression of Pthr1 and Ihh. They upregulate expression of Col10a1 and activate the gene for the vascular endothelial growth factor (Vegf). Chondrocyte prehypertrophy occurs very early for the cells located in the center of the cartilage anlagen of endochondral skeletal elements. These cells essentially differentiate from a prechondrocytic or very early chondroblastic stage directly to prehypertrophy. Several studies convincingly demonstrated the virtual dispensability of the chondroblastic stage in the progression of these cells toward prehypertrophy by showing that prehypertrophic chondrocytes developed in these regions in Sox5–/–Sox6–/– fetuses (Smits et al.,2001,2004), as well as in Sox9/Col2Cre conditional null mutants (Akiyama et al.,2002), despite a lack of chondroblast overt differentiation. In the rest of the cartilage anlagen, the Pthrp/Ihh feedback loop, as well as Fgf, Bmp, and other signaling pathways, play critical roles in controlling the rate at which chondroblasts undergo prehypertrophy (reviewed in Kronenberg,2003).
Sox5/Sox6 delay chondrocyte prehypertrophy, but are needed for hypertrophy. In Sox5/Sox6 compound mutants, chondroblasts underwent bona fide differentiation and then prematurely underwent prehypertrophy (Smits et al.,2004). They prematurely upregulated Fgfr3, which promotes chondrocyte maturation, and overexpressed Runx2, which is needed for prehypertrophy. In Sox5–/–Sox6–/– fetuses, prehypertrophic chondrocytes failed to become morphologically hypertrophic and to express Col10a1, but did reach terminal differentiation (Smits et al.,2001), and in Sox5/Sox6 compound mutants, hypertrophic zones, delineated by the Col10a1 expression domain, were shorter than normal (Smits et al.,2004). The mechanism whereby Sox5 and Sox6 promote hypertrophy is unknown, but it is likely indirect since they are no longer expressed beyond prehypertrophy. It has been suggested that Sox9 delays chondrocyte hypertrophy, based on the observations that Sox9+/– mouse fetuses featured prematurely mineralized cartilages and expanded hypertrophic zones in cartilage growth plates (Bi et al.,1999), and that mice that overexpressed Sox9 under the control of Col2a1 regulatory elements (Sox9/Col2a1 knock-in) delayed cartilage mineralization (Akiyama et al.,2004). It is not clear from these data, however, whether Sox9 delays prehypertrophy and/or hypertrophy. The notion that Sox9 may delay chondrocyte prehypertrophy is supported by the observations that Sox9 is phosphorylated by the protein kinase A (PKA) downstream of Pthrp signaling, which delays prehypertrophic differentiation, and that PKA-mediated phosphorylation of Sox9 may increase Sox9 activity (Huang et al.,2001). The abrogation of Sox5, Sox6, and Sox9 expression that occurs in prehypertrophic chondrocytes should be sufficient to explain the loss of expression of early cartilage matrix genes that is seen in hypertrophic cells. Interestingly, however, several transcriptional repressors have been suggested to contribute to downregulation of these genes. This was suggested to be the case for the Snail family zinc finger proteins Snail and Slug, when their expression in prehypertrophic and hypertrophic chondrocytes in vivo and ability to repress expression of Col2a1 reporter genes in vitro was demonstrated (Seki et al.,2003). This was also suggested to be the case for the Krüppel-associated box zinc finger protein NT2, when its expression in hypertrophic chondrocytes and ability to repress a Col11a2 reporter gene in vitro was demonstrated (Tanaka et al.,2002). However, the exact roles played by these factors in chondrogenesis in vivo remain to be determined.
Runx2 and Runx3 have essential roles in inducing chondrocyte prehypertrophy and hypertrophy. As indicated above, they are activated in chondroblasts that reach prehypertrophy, and Runx2 remains expressed through hypertrophy and terminal differentiation. Runx2–/– mice exhibited a disturbance of chondrocyte maturation in some skeletal elements (Inada et al.,1999; Kim et al.,1999). Chondrocytes were blocked before prehypertrophy in the humerus and femur, as illustrated by the absence of Pthr1, Ihh, and Col10a1 expression, whereas prehypertrophic and hypertrophic differentiation was only delayed in the tibia, radius, and phalanges. Further evidence that Runx2 is able to induce prehypertrophy and hypertrophy was obtained in studies that showed that forced expression of Runx2 in chondroblasts of transgenic mice resulted in ectopic maturation of these cells, whereas expression of a dominant negative form of Runx2 prevented chondroblasts from undergoing prehypertrophy (Takeda et al.,2001; Ueta et al.,2001). The fact that chondrocyte maturation occurred in some skeletal elements of Runx2–/– mice prompted a search for additional transcription factor(s) involved in inducing chondrocyte maturation. This search led to the discovery that Runx3 has roles in chondrocyte maturation. Runx3–/– mice and Runx2+/–Runx3–/– mice showed a delay in endochondral ossification, and Runx2–/–Runx3–/– mice showed a complete absence of prehypertrophic and hypertrophic chondrocytes (Yoshida et al.,2004), thus revealing that Runx2 and Runx3 interact closely in inducing chondrocyte maturation. Runx2 is likely a direct transcriptional activator of chondrocyte maturation markers. It binds in vivo to multiple recognition sites in the Col10a1 promoter and activates Col10a1 reporter constructs through these elements in vitro (Zheng et al.,2003). Runx2 also stimulates Ihh expression and binds directly to and activates the Ihh promoter in vitro (Yoshida et al.,2004). These data, together with the observation that Pthrp inhibits Runx2 expression in vitro (Iwamoto et al.,2003; Li et al.,2004), strongly suggest that Runx2 is part of the Ihh/Pthrp feedback loop that controls chondrocyte maturation in the growth plate. Interestingly, Runx3 was able to bind, but not to activate, the Ihh promoter in vitro, which suggests that it likely contributes directly to Ihh activation but acts in cooperation with transcriptional activators (Yoshida et al.,2004).
The binding of Runx proteins to DNA is stabilized by heterodimerization with the coactivator core-binding factor-β (Cbfb). Cbfb is expressed in mature chondrocytes, and it is therefore not surprising that Cbfb–/– mice displayed disruption of chondrocyte maturation as seen in Runx2–/– mice (Kundu et al.,2002; Yoshida et al.,2002). Two Distal-less-related homeobox transcriptional activators, Dlx5 and Dlx6, also promote chondrocyte maturation and may cooperate with Runx2 and Runx3 in this function. They are expressed in chondrocyte mesenchymal precursors, turned off in chondroblasts, reexpressed in prehypertrophic chondrocytes, and may remain expressed in hypertrophic and terminal chondrocytes (Ferrari and Kosher,2002; Robledo et al.,2002). They are also expressed in the osteoblast lineage. Dlx5–/– mice displayed no major defects in chondrocyte maturation, while Dlx5–/–Dlx6–/– mouse embryos featured a severe delay in skeleton mineralization, apparently due to delayed maturation of chondrocytes from the prehypertrophic stage (Robledo et al.,2002). Complementing that study, ectopic expression of Dlx5 in chicken limbs was shown to result in short skeletal elements with precocious maturation of chondroblasts into prehypertrophic cells, and with extended zones of hypertrophic and terminal chondrocytes (Ferrari and Kosher,2002). Dlx5 and Dlx6 may cooperate with Runx2 in directly activating mature chondrocyte markers, as suggested by in vitro experiments that showed the ability of Dlx5 to bind to a recognition site in the Col10a1 promoter and to activate a Col10a1 promoter construct (Magee et al.,2005), as well as to interact physically with Runx2 (Roca et al.,2005). A negative modulator of Runx2 function is the histone deacetylase Hdac4 (Vega et al.,2004). Hdacs modulate cell growth and differentiation by deacetylating histone proteins, thereby favoring chromatin condensation and thus transcriptional repression. Hdac4 expression in cartilage overlaps with Runx2. Hdac4 inactivation in the mouse and ectopic expression in chondroblasts of transgenic mice led to phenotypes resembling those obtained upon ectopic expression of Runx2 in chondroblasts and Runx2 inactivation, respectively, suggesting that Hdac4 may repress Runx2 activity. Confirming this notion, Hdac4 was found to physically interact with Runx2 and to inhibit its DNA binding capacity.
While these data, taken together, point to essential roles for Runx2 and its related and associated factors in driving chondrocyte maturation, they also raise questions. For example, it remains unknown why Runx2 activates Col10a1 and Ihh in chondrocytes but not in osteoblasts, why Ihh expression is restricted to a few layers of prehypertrophic chondrocytes while the Runx2 expression domain extends deeper into the hypertrophic zone, and why Runx2 does not activate osteoblast-specific genes until chondrocytes undergo terminal maturation.
CHONDROCYTE TERMINAL DIFFERENTIATION
Chondrocytes undergo a last dramatic phenotypic change when they progress from hypertrophy to the terminal stage. During this transition, they lose expression of Col10a1 and activate a new set of genes, including those for the matrix metalloproteinase-13 (Mmp13), osteopontin (Spp1), and alkaline phosphatase (Alpl), which are also markers of osteoblasts. Like osteoblasts, they also induce extracellular matrix mineralization. After years of controversy as to whether they end up dying or fully differentiating into endochondral osteoblasts, the current view is that most, if not all, undergo apoptosis (reviewed in Adams and Shapiro,2002).
The common features of terminal chondrocytes and mature osteoblasts strongly suggest that these two cell types may be governed by a common or very similar set of transcription factors. Runx2 is required for expression of most osteoblast markers in vivo, and several studies have shown that it may directly activate such genes as Spp1 and Mmp13 (Sato et al.,1998; Hess et al.,2001). One can therefore postulate that Runx2, which remains expressed in terminal chondrocytes, also controls these genes in these cells. The transcriptional activator c-Maf was shown to be specifically expressed in late hypertrophic and terminal chondrocytes, as well as in osteoblasts (MacLean et al.,2003). It belongs to the Maf family of the basic leucine zipper (bZIP) proteins, and has the ability to form homodimers or heterodimers with bZIP proteins such as Fos and Jun (Kerppola and Curran,1994). Prehypertrophy and early hypertrophy were not affected in c-Maf–/– mice, as determined by the normal onset and level of expression of Pthr1, Ihh, and Col10a1, and the normal enlargement of chondrocytes (MacLean et al.,2003). However, the first terminal chondrocytes, which expressed Spp1 and Mmp13, appeared later in c-Maf–/– fetuses than in control littermates and were removed later, resulting in expanded chondrocyte maturation zones and a delay in endochondral ossification. The expression levels of terminal markers, as well as the extent of matrix mineralization were normal in c-Maf–/– mice. These results indicate that c-Maf facilitates both the initiation of terminal differentiation and the completion of the chondrocyte differentiation program. Since the transcriptional control of chondrocyte terminal differentiation remains largely unknown, further studies will have to address questions regarding the factors that, in conjunction with c-Maf, contribute to the switch from hypertrophy to terminal maturation, and allow the activation of some osteoblast markers while others, such as the collagen type I genes, remain silent.
ARTICULAR CHONDROCYTE DIFFERENTIATION
Once the presumptive joint regions are specified in the embryo, the first step in joint formation is cavitation. This step occurs when mesenchymal cells that are located between or within precartilaginous condensations (depending on the joints) and have developed to the prechondrocyte stage revert to a mesenchymal stage and undergo apoptosis (Archer et al.,2003). At the same time, prechondrocytes that line the new cavities differentiate into early chondroblasts and then into articular chondroblasts that activate typical markers, including the lubricin gene (Prg4) (Rhee et al.,2005). At the end of postnatal development these cells become articular chondrocytes, which maintain articular surfaces throughout life. These cells rarely divide, and they maintain high expression of Agc1 and Prg4 and low expression of Col2a1. They likely express other specific markers that have not yet been identified. Normal articular chondrocytes never undergo hypertrophic differentiation, except in the tidemark, the cartilage zone that abuts the subchondral bone. Tidemark chondrocytes express Col10a1, induce mineralization of the cartilage matrix, and are swollen (Girkontaite et al.,1996). Articular chondrocytes can revert to an immature chondroblastic stage and even undergo prehypertrophic to terminal chondrocyte maturation in osteoarthritis, a common joint degenerative disease, indicating that their differentiation state, although normally permanent, is not terminal. It is generally believed, but remains to be proven, that the qualitative and quantitative differences between the gene expression profiles of growth plate and articular chondroblasts reflect differences in expression and activity of specific transcription factors, and that repression mechanisms may be operating in normal articular chondrocytes to prevent them from undergoing hypertrophic maturation.
Sox9, Sox5, and Sox6 are expressed in articular cartilage throughout life (Davies et al.,2002) (Smits and Lefebvre, unpublished data). Studies using Sox5/Sox6 and Sox9 mutant mice have demonstrated the importance of these genes in early joint development (Smits et al.,2001; Akiyama et al.,2002), but the early lethality of these mice has prevented their use postnatal studies to prove the postulated roles of the Sox genes in articular cartilage formation and postnatal maintenance. Not surprisingly, Runx2 is not expressed in normal articular cartilage, except in the tidemark (Kuboki et al.,2003), but it has been detected in osteoarthritic chondrocytes (Wang et al.,2004). The chicken Erg (chErg) transcriptional activator, which features an Ets-related DNA-binding domain, has been suggested to play a role in directing the fate of chondrocytes toward an articular chondrocyte phenotype or a growth plate phenotype (Iwamoto et al.,2000,2005). While full-length chErg is preferentially expressed in prehypertrophic chondrocytes, a short splice variant, which lacks an internal segment of the protein and is named C-1-1, is preferentially expressed in developing and adult articular cartilage. Virally driven expression of C-1-1 in chicken leg buds yielded cartilages in which chondroblasts failed to undergo hypertrophic differentiation, as shown by the absence of Col10a1 expression, and expressed Tnc, a marker of prechondrocytes and immature articular cartilage. In contrast, virally driven expression of full-length Erg stimulated chondrocyte maturation. These data suggest that C-1-1 and full-length Erg may have opposite roles, preventing chondrocyte maturation in articular cartilage and stimulating chondrocyte maturation in growth plates, respectively. The molecular mechanism that underlies the differences in these biological functions of the two splice variants remains unknown. The roles of Erg3, the mammalian ortholog of chicken Erg, and other Erg proteins in chondrogenesis in vivo, also remain unknown. Our knowledge regarding the transcriptional control of articular chondroblast and chondrocyte differentiation thus remains incomplete. Major questions remain to be addressed regarding the mechanisms whereby articular chondroblasts and chondrocytes “escape” growth plate maturation and express joint-specific markers.
The uncovering of networks of transcription factors with critical roles in determining chondrocyte fate and differentiation is progressively emerging as the reward of extensive research that has been conducted in many laboratories all over the world for about two decades (Fig. 2). The commitment of mesenchymal progenitor cells to chondrogenesis, precartilaginous condensation, and differentiation into prechondrocytes requires the Sry-related HMG box transcriptional activator Sox9. Sox9 ensures cell survival and activates Col2a1 and other cartilage early markers in all chondrogenic cells throughout the embryo. Numerous factors (collectively referred to as skeleton patterning factors, and including members of the Hox, Pax, Nkx, and other families) control the migration, proliferation, and survival of mesenchymal precursor cells in specific subsets of skeletogenic populations. Several of these factors, such as the paired box Pax1 and Pax9 transcriptional activators and the Bagpipe-related homeobox Nkx3.1 and 3.2 transcriptional repressors, act together with Sox9 to maintain the chondrogenic cell fate and promote early differentiation. The overt differentiation of prechondrocytes into early chondroblasts and subsequently into columnar chondroblasts requires the pair of redundant Sox DNA-binding factors L-Sox5 and Sox6. These factors promote cell proliferation and also robustly increase production of cartilage extracellular matrix by boosting expression of the genes for all major components of this matrix. Their expression requires Sox9, and Sox9 may directly cooperate with them in controlling expression of cartilage matrix genes. A major switch in cell fate and differentiation occurs with chondrocyte prehypertrophy and hypertrophy. Characterized by sequential expression of Pthr1, Ihh, and Col10a1, by downregulation of early cartilage matrix genes and by massive cell swelling, this switch is induced by the runt domain transcriptional activators Runx2 and Runx3. These two factors appear to directly activate Ihh and Col10a1, and may cooperate in that function with the Distal-less related homeobox Dlx5 and Dlx6 transcriptional activators. L-Sox5 and Sox6, and probably Sox9 as well, act to delay chondrocyte prehypertrophy, but L-Sox5 and Sox6 are also required for chondrocyte overt hypertrophy, likely through indirect mechanisms. Chondrocyte terminal differentiation represents yet another major switch in fate and differentiation, as the cells convert into an osteoblast-like phenotype and end up undergoing apoptosis. This switch is facilitated by the basic leucine-zipper transcriptional activator c-Maf, and is likely to require Runx2. L-Sox5 and Sox6, again through indirect mechanisms, are also required to delay terminal differentiation. The transcriptional mechanisms responsible for maintaining articular chondrocytes throughout life and for controlling the expression of specific genes, such as Prg4, have not yet been uncovered; however, a role for the Erg family factor C-1-1 in preventing articular chondrocyte maturation toward hypertrophy has been suggested.
In conclusion, much progress has been made in recent years to uncover essential chondrogenic transcription factors and to suggest a number of potentially important factors. We can anticipate that additional breakthroughs will be made over the next couple of years with the identification of new transcriptional players, and with a fuller understanding of the roles of each individual player and its interactions with other transcriptional regulators and with signaling pathways. We can therefore foresee that novel strategies will soon be proposed to prevent cartilage malformation diseases and significantly improve the expectancy and quality of life for many patients suffering from inherited cartilage malformation diseases.
We thank Shunichi Murakami for helpful comments on the manuscript, and apologize to the many colleagues whose contributions we were unable to cite due to space constraints.