A critical evaluation of specific aspects of joint development


  • A.A. Pitsillides,

    Corresponding author
    1. Department of Veterinary Basic Sciences, Royal Veterinary College, London, United Kingdom
    • Reader in Cell Matrix Biology, Department of Veterinary Basic Sciences, Royal Veterinary College, London NW1 0TU, UK
    Search for more papers by this author
  • Doreen E. Ashhurst

    1. Department of Basic Medical Sciences, St. George's, University of London, London, United Kingdom
    Search for more papers by this author


Synovial joint formation has been divided into two phases; the formation of the anlagen of the opposing bones and the interzone and the later formation of the joint cavity. Here we review current theories on the mechanism by which these events are achieved in the joints of the developing limb. Developmental Dynamics 237:2284–2294, 2008. © 2008 Wiley-Liss, Inc.


The development of functional synovial joints is usually divided into two phases. The first is the formation of cartilaginous (cartilage is defined as a type of avascular connective tissue that contains cells, the chondrocytes, in lacunae embedded within a dense, semi-translucent matrix of predominantly type II collagen and the proteoglycan, aggrecan) anlagen and the intervening interzones in which the joints will develop, that is, limb patterning. The second is the formation of the joint cavity, the articular cartilage, synovium and other related structures within the joint. This sub-division has frequently resulted in the examination of one part of the developmental cascade without reference to the other, and may have contributed to confusion in the literature about the mechanisms involved in both phases of this process. Although it has frequently been asserted that interzones are formed by the interruption of a continuous cartilaginous anlagen (or rod) and that cavitation involves localised cell death, our critical evaluation of the evidence, presented below, suggests that the validity of these assertions needs to be reviewed.


It is first necessary to consider how, and when, the positions of the joints are determined along the proximodistal limb axis. This raises one crucial question: are the anlagen of the future bones discrete entities from the outset, or is an uninterrupted cartilaginous matrix divided later? The first possibility is supported by evidence (see below) that cartilage develops within discrete cell condensations. These cartilaginous “anlagen” elongate and approach each other, but remain separated by a distinct, non-cartilaginous region, the interzone of the presumptive joint. The alternative view is that the joints appear at specific times at predetermined sites within a continuous cartilaginous rod and that joint formation requires localized removal of cartilaginous matrix and cellular re-differentiation (Craig et al.,1987; Pacifici et al.,2000; Niedermaier et al.,2005). Many different technical approaches have been used to investigate these possibilities. To identify chondrogenic and pre-chondrogenic tissue, morphological approaches use stains (see Fig. 1) and specific antibodies to identify cartilage matrix constituents, while molecular approaches aim to distinguish cells with appropriate mRNA expression profiles.1

Figure 1.

Photomicrographs of whole mounts of developing chick embryos stained with Victoria blue at various stages (stage 24–32). A: Chick at about Stage 24–25 (day 4.5) showing developing fore and hind-limb zeugopod and stylopod. B: Stage 26–27 (day 5). C: Stage 28 (day 5.5). Both B and C show developing fore-limb zeugopod and stylopod and hind limb zeugopod, stylopod, and early autopod. D: Stage 29 (day 6) showing developing fore- and hind limb zeugopod and autopod. E: (and detail F, fore-limb; G, hind-limb): Stage 30 (day 6.5). H: (and detail I, fore-limb; J, hind limb): Stage 31–32 (day 7). These are provided for orientation and to show the progressive increase in Victoria Blue staining intensity at later stages.

To address whether the anlagen are discrete from their onset, it is pertinent first to consider the changing morphology of developing embryo limbs. Chick limb buds have a central region of undifferentiated prechondrogenic mesenchyme throughout their length by HH stage 24–25 (Figs. 1–3) (Hamburger and Hamilton,1951; Thorogood and Hinchliffe,1975).2 This central mesenchyme is avascular and looser than the peripheral mesenchyme (Singley and Solursh,1981; Wilson,1986; Yin and Pacifici,2001). At stage 26–27, three discrete condensations of cells corresponding to the future bones of the stylopod and zeugopod appear in the prechondrogenic mesenchyme (Figs. 1, 2, and 4). These cells are closely packed with little matrix between them (Singley and Solursh,1981). At stage 28, the cells within these enlarged mesenchymal condensations acquire the appearance typical of chondrocytes separated by cartilaginous matrix that stains intensely with Alcian blue (see Introduction section, Figs. 1, 2, and 4). Thus, at stage 28 it appears that the cartilaginous anlagen of the long bones have formed and are developing as separate, discrete entities.

Figure 2.

The development of the cartilaginous anlagen in chick limb buds. Column A shows the fore-limb zeugopod and stylopod at HH stages 25, 27, and 28. Column C shows the hind-limb autopod at HH stages 29, 30, and 32. Column B shows photomicrographs of the tissues that develop sequentially in the anlagen-undifferentiated mesenchyme (white), prechondrogenic condensations (blue), and cartilage (pink). Column A, stage 25, shows a single Y-shaped element in en face view, in which only undifferentiated mesenchyme is observed; stage 27 shows 3 pre-chondrogenic condensations separated at the putative elbow joint; stage 28+ shows the onset of cartilage formation in the mid-diaphyseal region of each element. Column C, stage 29, shows the developing metatarsal (M) prechondrogenic condensation with a region of undifferentiated mesenchyme extending distally; stage 30 shows a further distal area of prechondrogenic mesenchyme of the proximal phalange (P); stage 32 shows the development of a further phalange and cartilage in the proximal elements.

Figure 3.

Photomicrographs of developing chick embryo limb buds stained with Alcian Blue (pH 2.5) followed by Haematoxylin and Eosin. A: Fore-limb at stage 25 showing the early developing limb bud (×10). B: The developing hind limb bud containing the putative knee joint (arrow) at stage 27 (×5). C: Detail of B, ×10. All magnifications are original.

Figure 4.

Photomicrographs of developing chick embryos stained with Alcian Blue (pH 2.5) followed by Haematoxylin and Eosin. A (and detail B, ×10): Hind limb at stage 28 showing the separate cartilaginous anlagen of the pelvic girdle (p), femur (f), tibia (t), and developing autopod (a) (×5). Arrow indicates position of interzone in B. C (and detail D, ×10): Hip joint at stage 30 showing the pelvic girdle (p) and the head of the femur (hf) (×5). E (and detail F and G, ×10 and ×20, respectively): Hind region of stage-32 embryo showing the developing hip (h) and knee (k) joint regions and presumptive interzones (i) (×2.5). H: Foot at stage 30 showing developing phalanges. The distal phalanx (dp) is separated from its proximal neighbour by presumptive interzone (×10). I: Terminal phalanx of stage-35 foot, separated from its proximal neighbour by interzone (×10).

The anlagen elongate and an intervening putative joint region of prechondrogenic cells, which will later form the interzone, is retained between them. The approximate positions of the elbow and knee joints are thus clearly defined by stage 28; both develop in close to 90° of flexion (Figs. 1 and 2). The anlagen grow and the cartilage matures, but the interzones retain their undifferentiated mesenchyme (Figs. 1 and 2). The hip and shoulder joints develop similarly. The head of the femur, for example, is separated from the pelvis by an interzone from stage 28 (Fig. 4). Between stages 28 to 35, the discrete cartilaginous elements of the autopod develop progressively in a proximodistal sequence (Figs. 1, 2, and 4). Initially, longitudinal undifferentiated mesenchymal digital rays, corresponding to each digit, grow toward the end of the limb bud. Within each ray the metacarpal/tarsal condensation develops proximally, followed later, and distally, by the condensation of the proximal phalanx. These condensations are separated from their outset by a region that will develop into the phalangeal joint interzone. The remaining phalanges are laid down sequentially until the terminal phalanx is formed. This indicates that all the digital bones develop as discrete entities, which in their turn become cartilaginous (Figs. 1, 2, and 4). Indeed, recent elegant studies support the existence of a population of cells in a “phalanx-forming region” at the distal end of the digital rays that initially express a labile fate, which only becomes fixed as they are incorporated into the condensed cartilage of the digital primordia (Suzuki et al.,2008). Anlagen formation in mouse limbs displays a similar sequence of events between days 11–15 in utero (Fig. 5).

Figure 5.

Photomicrographs of developing mouse hind limb stained with Alcian Blue (pH 2.5) followed by Haematoxylin and Eosin. A: Developing pelvis (p), femur (f), and small part of the tibia (t) in 14-day-old fetus (×5). Note that the tibia, of which only a very small part is in the plane of this section, is at right angles to the femur and expanded developing femoral condyles. B: Detail of the hip joint (×20) shows separation of cartilaginous structures at the developing interzone (i). C (and detail D, ×20): Developing fore-limb containing scapula (s), humerus (h), and radius/ulna (r/u) in 13-day-old fetus (×5). E: Hip joint of 15-day fetus (×20) showing the head of the femur (hf) almost completely surrounded by the acetabulum (arrows) highlighting distinct interzone (i). F (×10, and detail G, ×20): Autopod of 16-day fetus showing developing metatarsal-phalangeal joint highlighting distinct interzone (i).

Support for the presence of a continuous cartilaginous structure relies primarily on the distribution of type II collagen. Type II collagen mRNA can indeed be extracted from chick limb buds from stage 20 (Kravis and Upholt,1985). It is found by in situ hybridization throughout the chick forelimb mesenchyme at stage 21, but is confined to the central region at stage 23 (Nah et al.,1988; Swalla et al.,1988). By stage 25, a Y-shaped region of loose mesenchyme, encompassing the proximal three long bones, contains cells expressing type II collagen mRNA and a matrix containing the protein, but at stage 27, the type II collagen–containing matrix appears to be confined to 3 clearly separate zones (Von der Mark et al.,1976; Dessau et al.,1980; Koyama et al.,1995; Lizarraga et al.,2002). This changing distribution of type II collagen suggests that there is continuity across the elbow and knee joints at early developmental stages. From stage 28, cells that express type II collagen mRNA are also found in the anlagen of the autopod, but in contrast, these are found in discrete rather than continuous regions (Koyama et al.,1995; Nalin et al.,1995). It is, however, of special significance that type II collagen and its mRNA are also present at early stages in developing non-skeletal tissues (Kosher and Solursh,1989; Cheah et al.,1991; Wood et al.,1991; Ng et al.,1997), but more importantly, that type II collagen exists in two distinct forms, types IIA and IIB (Ryan and Sandell,1990; Ryan et al.,1990).


It is now known that the collagen in prechondrogenic and many other developing tissues is type IIA, whereas type IIB is the normal cartilage collagen (Ng et al.,1993; Sandell et al.,1994; Oganesian et al.,1997; Zhu et al.,1999). In situ hybridization has demonstrated that transient type IIA collagen mRNA expression precedes type IIB mRNA expression at cartilaginous sites in human and mouse embryos (Ng et al.,1993; Lui et al.,1995; Oganesian et al.,1997; Zhu et al.,1999). Our immunohistochemical observations of chick embryos show that the Y-shaped areas of precartilaginous mesenchyme present at stage 25, and the separating elements at stage 27, contain type IIA collagen protein (Figs. 6 and 7). Nalin et al. (1995) showed clearly that type IIA mRNA expression precedes that of type IIB mRNA in the developing chick autopod. Furthermore, type IIA collagen mRNA was isolated from chick embryos at stage 25, while type IIB mRNA was not detected until stage 28 (Nah and Upholt,1991).

Figure 6.

Localization of type IIA collagen in developing hind-limb of stage-27 chick embryos. A: The position and flexion of the knee joint (arrow) is visible in this sagittal section (Alcian blue and H+E). B: Type IIA collagen is localized in the central mesenchyme of the limb and a region of diminished type IIA collagen indicates the position of the putative joint in an adjacent section. C: An area shown at higher magnification. The localization of type IIA collagen was performed on formaldehyde-fixed, wax-embedded sections. The sections were pre-treated with trypsin and hyaluronidase before exposure to the rabbit anti-type IIA collagen antibody (see Organesian et al., 1987). The second antibody was a goat anti-rabbit IgG conjugated to alkaline phosphatase. Scale bars = 100 μm.

Figure 7.

Schematic diagram showing the anticipated localisation of type IIA and type IIB collagen in the developing fore-limb zeugopod/stylopod. Type IIA collagen shows a widespread distribution (light grey) at stage 25 and is enriched in pre-cartilaginous condensations (hatched areas) at stage 27. We speculate that type IIA is retained only in these regions at stage 29, and lost from developing zeugopod/stylopod at later stages. We also speculate that type IIB is expressed only in the regions of overt cartilage differentiation (dark grey) from stage 28–29 where it is retained at later stages (stage 32).

These apparently contradictory findings relating to the temporospatial distribution of type II collagen mRNA and protein can, nevertheless, be reconciled. Probes for type II mRNA and antibodies against type II collagen will detect both types IIA and IIB mRNAs and proteins. Thus, previous workers, who described cartilaginous continuities across presumptive joints at early stages, have assumed erroneously that the presence of type II collagen mRNA and protein necessarily indicate the presence of cartilage. The introduction of probes specific to types IIA and IIB collagen mRNAs and an antibody specific to type IIA collagen by Sandell and co-workers has enabled regions of type IIA collagen to be distinguished from those of type IIB (Ryan and Sandell,1990; Oganesian et al.,1997; Sandell et al.,1997).3 It follows from the evidence presented above that type IIA collagen is present in the mesenchymal and prechondrogenic regions up to stage 27 in the chick (Figs. 2 and 7) and the transience of this expression is illustrated by its loss from the elbow and knee joint interzones (Fig. 7). Regions of type IIA collagen in developing embryos are, therefore, not necessarily cartilaginous (see above). Type IIB collagen mRNA and its protein are found only from stage 28 when cartilage formation starts. The data presented by Nalin et al. (1995) show that type IIA collagen, as localised by its mRNA, is never present in the interzones of the developing chick autopod. Thus, there is little evidence for the presence of continuous cartilaginous structures in developing limbs (see Craig et al.,1987; Francis-West et al.,1999b; Spitz and Doubule,2001; Doubule,2002). To use type II collagen as an unequivocal marker of cartilage, an antibody specific to type IIB collagen is essential. Collectively, these observations indicate that the long bone anlagen develop as discrete, discontinuous, cartilaginous elements.


Thus, despite the initial equivalence of all the cells that give rise to the cartilaginous elements and presumptive joint regions, the cells must diverge to create two distinct populations before, or during, overt chondrogenesis. This initial equivalence of cells that produce cartilage or other joint tissues makes the identification of early commitment markers vital if novel chondrogenic or joint regulatory factors are to be defined. Sox9, a member of the SRY (sex-determining region of the Y-chromosome) family that contains the HMG (high mobility group) box DNA-binding domain is a key transcription factor that regulates chondrogenesis. Studies of embryo mouse limbs showed a co-localisation of Sox9 and Col2a1 mRNAs in cells of mesenchymal condensations from day 10 onwards; that is, both before and during cartilage deposition (Wright et al.,1995; Zhao et al.,1997). Indeed, it was later confirmed that Sox9 was essential in cartilage formation and chondrocyte differentiation (Bi et al.,1999). Loss of Sox9 from mouse limb mesenchyme prior to condensation in a Cre/loxP recombination system results in the complete absence of both cartilage and bone, whereas Sox9 inactivation at later post-condensation stages leads to diminished chondrocyte proliferation, chondrodysplasia, and defective joint formation (Akiyama et al.,2002). In situ hybridisation during mechanically-induced secondary cartilage formation also suggests that Sox9 contributes to chondrogenesis (Archer et al.,2006). These in vivo data support a strong link between Sox9 and chondrogenesis and its likely role in early differentiation.

In vitro studies showed that combined Sox9, Sox5, and Sox6 expression was sufficient to induce cartilage formation by a broad range of cell types (Ikeda et al.,2004). In addition, Wnt-induced chondrocyte differentiation is Sox9-dependent, as activation of canonical Wnt signalling by constitutively active LEF-1 fails to promote chondrogenesis in Sox9-deficient cells (Yano et al.,2005). These in vitro studies reinforce a link between Sox9 and chondrogenesis. Cell fate-mapping using Sox9-reporter mice revealed, however, that Sox9-expressing cells in limb mesenchyme can also give rise to osteoblasts, tendon, and synovial cells, and that progenitor cells in developing nervous system, intestine, pancreas, and testis also express this gene (Akiyama et al.,2005; Kawakami et al.,2006).

The relationship between Sox9 and type II collagen mRNAs has not been defined in developing embryos. The key issue regarding Sox9 function in cartilage formation may, however, be clarified if we examine the relationship between Sox9 and type IIA, “non-cartilaginous” and type IIB “cartilaginous” forms of collagen type II. For instance, in vitro studies of pluripotent mouse embryo stem cells disclose a close association between Sox9 and collagen type IIA mRNA at early times (5 days of culture), but no such association with type IIB mRNA levels that are found 14 days later after the formation of embryoid bodies (Tanaka et al.,2004). The divergence of Sox9 and type IIB mRNA levels is also seen in BMP2-induced embryonic cell chondrogenesis (zur Nieden et al.,2005). Moreover, a close correlation in vivo between Sox9 and type IIA, but not type IIB, collagen is also seen in chondrosarcomas (Soderstrom et al.,2002). Despite a report of reduced Sox9 mRNA levels associated with elevated Col2A1 mRNA in human osteoarthritic cartilage, recent studies appear to verify the promotion of Col2A1 mRNA and type II protein expression by over-expression of Sox9 (Aigner et al.,2003; Li et al.,2004; Tew et al.,2005). The Sox9:type IIA link is strengthened by their co-distribution during “attempted” cartilage repair, and by the chondrogenic commitment and increased growth capacity exhibited by mouse stem cells in which human-Sox9 is overexpressed (Salminen et al.,2001; Kim et al.,2005). Together, these data suggest that while Sox9 controls differentiation to the chondrocytic fate, it also serves a key role in earlier mesenchymal cell commitment.

It seems unlikely, however, that Sox9 contributes directly to initiate joint formation within interzonal cell populations. Sox9 mRNA was not found in the developing joint region of either Xenopus laevis or chick limbs (Satoh et al.,2005; Barlow et al.,1999). Indeed, the identity of joint commitment markers remains largely a mystery. HMGN1, an abundant member of the HMG class of proteins that loosens chromatin fibres, is a potential candidate for several reasons: (1) loss of HMGN1 accelerates chondrogenic differentiation, (2) HMGN1 expression shows a complimentary distribution to Sox9 and is diminished in all but committed, continually renewing cells, and (3) HMGN1 normally sequesters Sox9 in cells poised in the prechondrogenic state (Furusawa et al.,2006). Thus, while some chondrogenesis initiating factors are now identified, those controlling cell fate in presumptive joint regions that control the initiation, formation, and maintenance of joint cavities remain to be defined.

These observations also highlight distinctions that can be made between cartilage of the expanding anlagen and the articular cartilage, which is derived later from interzonal cells. An awareness of this is important to the concept that articular chondrocytes have a distinct phenotype. Indeed, putative chondrocytes of the interzone differ from those of the epiphyses of the anlagen in that types I, III, and V collagens are synthesized before cavitation, while type II collagen and its mRNA are found only after cavitation when the articular cartilage develops (Von der Mark et al.,1976; Craig et al.,1987; Bland and Ashhurst,1996,2001). Other distinctions are that matrilin-1 is found in the epiphyseal cartilage, but not in the chondrogenous layers, or the articular cartilage of mouse and rabbit joints (Kavanagh and Ashhurst,1999; Murphy et al.,1999; Segat et al.,2000). Similarly, by means of a mouse in which the Cre-recombinase gene was targeted to exon 1 of the matrilin-1 gene, Hyde et al. (2007) demonstrated that matrilin-1-expressing cells are never present in the interzone or the articular cartilage (Hyde et al.,2007). Versican, in contrast, is not found in epiphyseal cartilage of the anlagen, but is found in perichondria, interzones, and articular cartilage (Snow et al.,2005). Further evidence supporting the distinct phenotype of the interzone cells comes from genetic cell fate-maps produced by mating ROSA-LacZ-reporter mice with mice expressing Cre-recombinase at prospective joint sites under the control of Gdf5 regulatory sequences, in which reporter expressing cells form the interzone from E13.5 and were still present in structures that appear to arise from the interzone, namely the articular cartilage, synovium, and other joint tissues, on post-natal day 7 (Koyama et al.,2008). Distinctions between developing articular cartilage and epiphyseal cartilage may be of vital importance now that putative stem cells have been identified in the surface zone of articular cartilage (Dowthwaite et al.,2004). Although it has yet to be finally established through appropriate “tracing” experiments, the notion that the interzonal cells are the precursors of articular chondrocytes and contribute to the formation of articular cartilage would allow a re-evaluation of questions regarding cartilage turnover, regeneration, and repair to be made.


The major unresolved question about the second phase of joint development is whether the formation of the joint cavity, that is cavitation, is dependent upon cell death. How is cavitation achieved? It is frequently suggested that cavitation involves cell death along the joint line (Abu-Hijleh et al.,1997; Spitz and Doubule,2001; Mariani and Martin,2003). Based on the evidence frequently cited, this contention is, however, questionable. Mitrovic (1977) describes degenerating cells in chick metatarsophalangeal joints between 7 and 9 days, but states that the later formation of the joint cavity on day 10 (stage 35 to 37) “was not due to a massive degenerative process, although signs of cell necrosis and macrophagocytosis were occasionally found.” More recently, Nalin et al. (1995) endorsed this view, stating that “there were no TUNEL-positive cells at the time of cavitation at days 12–15 (stages 38–41)” in chick interphalangeal joints, although some were present before day 10 in the chondrogenous layers of the interzones. Mori et al. (1995) and Kimura and Shiota (1996) identified dead cells in the presumptive joint regions of mouse embryo feet between days 12 and 14, but similarly Fernández-Terán et al. (2006) found very restricted cell death in the presumptive interphalangeal joints at days 13.5–14.5 in the fore-limb and days 14.5–15.5 in the hind limb. It should be emphasised that in the mouse, cavitation does not appear to take place in these proximal joints until day 15 and not until about day 18 in the distal joints. Mitrovic (1978) and Ito and Kida (2000) found no morphological or biochemical signs of cell death before, or during, cavitation of rat knee joints. Similarly, while a few apoptotic cells were seen in the chondrogenous layers of rabbit interzones 2 days prior to cavitation, no apoptotic or TUNEL-positive cells were found in these regions during cavitation (Kavanagh et al.,2002a).

Evidence in support of cell death during cavitation comes from the description of “dark,” degenerating cells in the intermediate layer of chick knee joints between stages 30 to 35; none are found at stage 37 when cavitation is complete (Abu-Hijleh et al.,1997). Other studies concluded that these dark, basophilic cells in the rat knee joint are, however, healthy and later form the surface layer of the articular cartilage (Mitrovic,1978; Ito and Kida,2000). Our recent observations also fail to support the contention that these dark cells are degenerating in chick joints. We find that they label intensely for GDF-5, that they exhibit characteristics similar to erythroblasts and granuloblasts, and that their numbers increase in joints of immobilised developing limbs in which cavitation fails (Kavanagh et al.,2006; Anne Osborne, PhD thesis, personal communication). There is, therefore, little evidence that cell death is involved directly in the cavitation process; its timing and distribution preclude its immediate participation (see Pacifici et al.,2005). This does not, however, rule out the possibility that a very restricted amount of localised cell death could act to precipitate and promote later joint-forming activity and that, despite the apparent lack of direct association between cell death and cavitation, these processes are nevertheless linked. This possibility has yet to be examined (Pacifici et al.,2005).

An alternative view based on morphological observations during cavitation is that a cleft appears between the flattened cells of the intermediate layer and, once separated, these cells form the surface layer of the articular cartilage; there is virtually no cell death (Andersen,1961; Mitrovic,1977,1978; Edwards et al.,1994; Pitsillides et al.,1995; Bland and Ashhurst,1996,2001; Dowthwaite et al.,1998; Ward et al.,1999; Ito and Kida,2000; Kavanagh et al.,2002a; Osborne et al.,2002; Lamb et al.,2003). The mechanism by which this precise cleavage is achieved is, as yet, not fully understood. To create planes of separation, changes in composition of the extracellular matrix that are driven by local (re)modelling events must take place. Our studies suggest that enzymatic degradation is unlikely, but they do however confirm that selective increases in local cellular capacity to synthesise, export, and bind the unsulphated glycosaminoglycan, hyaluronan (HA), may contribute to joint cavitation (Edwards et al.,1994,1996; Pitsillides et al.,1995; Dowthwaite et al.,1998). This is in agreement with studies that support a joint formation mechanism that relies on changing local matrix composition (Pitsillides,1999, Kavanagh et al.,2002b). Our recent studies show that selective activation of the MEK-ERK pathway in cells at sites of cavitation may lead to changes in HA synthesis and binding that are consistent with a role for this pathway in joint cavitation (Ward et al.,1999; Lamb et al.,2003; Bastow et al.,2005). Candidate regulatory genes, which operate up-stream of the MEK-ERK pathway to regulate HA-dependent events, are unknown. Recent evidence suggests that FGF-2, -4, and -10 contribute in some way to joint formation (Lovinescu et al.,2003; Kavanagh et al.,2006). Their well-defined potential for promoting activation of the MEK-ERK pathway makes these possible candidates. It is probable that the most appealing candidates will exhibit preferential expression in cells along the developing joint-line. It is important, however, that any analysis aimed at their identification takes account of the fact that some gene products may contribute either to the determination of joint position, or to cavitation, while others may contribute to both stages. It is crucial that careful account is taken of the fact that the developing interzones (and perichondria) exhibit significantly higher cell densities than the neighbouring cartilaginous elements. Interpretation of data that ignores this may erroneously implicate factors in joint formation processes that are not directly involved.


Predominantly through analyses of their distribution, several signalling molecules have been implicated in joint development (de la Fuente and Helms,2005). From analyses of patterning, the positions of the elbow and knee joints of chicks are determined by stage 22/23, the ankle and wrist joints by around stage 26, and those within the autopod from stage 27 (Dudley et al.,2002; Sun et al.,2002). At these stages the anlagen are prechondrogenic, not cartilaginous (see above). The signalling proteins that determine the position of these joints must, therefore, be expressed before the above stages. A primary candidate, capable of contributing to both joint position and cavitation, is Wnt9A (previously designated Wnt14). Wnt9A mRNA is found in chick metatarsophalangeal joints at stage 27 and later, at stage 29, at the sites of the putative phalangeal joints (Hartmann and Tabin,2001). Wnt9A is also expressed in mouse elbow joints at embryonic day 11.5 and at day 13.5 in the metatarsophalangeal joints (Guo et al.,2004). Although evidence from Wnt9A null mice suggests that it does not initiate joint cavity formation, its importance in specifying joint position is supported by experiments that that Wnt9A misexpression leads to alterations of the skeletal pattern in chick limbs (Hartmann and Tabin,2001). At the digit tip, Wnt9A is expressed after FGF-8 expression ceases and it is suggested that this allows tip formation (Sanz-Ezquerro and Tickle,2003).

Wnt9A also appears to exert a later effect in joint formation by promoting the induction of CD44 mRNA expression in the joint region (Hartmann and Tabin,2001). CD44 is the major hyaluronan-binding protein, which has been assigned a role in cavitation (Edwards et al.,1994; Dowthwaite et al.,1998). This suggests that Wnt9A lies up-stream of local changes in HA synthesis and binding. The precise relationship between Wnt9A and HA regulation has not been examined. Establishing the basis of this relationship will determine whether Wnt9A exerts dual roles in joint formation, namely in early joint patterning and later in the coordination of CD44 expression and local increases in HA synthesis during cavity formation. For example, the possibility that MEK-ERK signalling is also required to regulate Wnt9A expression remains to be investigated. Furthermore, conservation of Wnt9A expression at the synovial lining of fully formed joints points to its possible involvement in the maintenance of joint cavities, but again this role has yet to be defined.

Clues to the mechanism of joint formation may arise from elucidating patterns of gene and protein expression in developing limbs. Indeed, roles for Gdf5, FGF-10, Noggin, α5β1 integrins, β-catenin, and the ets transcription factor family member, ERG, and its alternatively spliced variant C-1-1 have been suggested partly from their expression patterns (Storm et al.,1994; Brunet et al.,1998; Merino et al.,1999; Lovinescu et al.,2003; Garciadiego-Cazares et al.,2004; Guo et al.,2004; Pacifici et al.,2006). Multiple joint and skeletal defects are induced by mutations in members of the BMP family, which, in turn, are members of the TGFβ super-family, namely growth and differentiation factors-5 and -6 (Gdf-5 and -6) (Settle et al.,2003). Their overexpression, however, fails to induce joint formation, but results in overproduction of cartilage and loss of joints (Francis-West et al.,1999a). This suggests that although necessary, the expression of Gdf-5 and/or Gdf-6 is not sufficient for joint formation. The results of our examination of GDF-5 protein expression in normal and immobilised limbs support this predominantly regulatory role for GDF-5 in chondrogenesis (Kavanagh et al.,2006). Based on the blockade of joint formation in Noggin mutants and the finding that its overexpression acts only to restrict chondrogenesis, it appears that Noggin plays a similar facilitatory, but not initiatory, role in joint formation (Brunet et al.,1998; Capdevila and Johnson,1998). More recently, it has been shown that in addition to promoting expression of Wnt9A and CD44 mRNAs and decreasing the expression of type II collagen mRNA, blockade of α5β1 integrin function induces ectopic joint formation. Moreover, these studies also established that mis-expression of α5β1 integrin results in joint fusion and the failure of cavity formation (Garciadiego-Cazares et al.,2004). Together, these data suggest that suppression of α5β1 integrin function is both necessary and sufficient for joint formation. Similar conclusions have been drawn from studies that addressed the distribution and role of activated β-catenin. These support an essential role for Wnt9A and for Wnt/β-catenin signalling in the early stages of synovial joint formation (Guo et al.,2004). This linkage between genes involved in early patterning events to the specific processes that control joint cavity formation clearly need to be fully defined. This is particularly apparent for genes, such as Sonic hedgehog, that appear to fulfil a broad range of roles, which extend to a specific influence on joint formation (Niedermaier et al.,2005).

Many other genes, including Wnt-4, that have been implicated in joint formation, display temporal patterns of expression that do not precede joint specification (Storm and Kingsley,1996,1999; Kawakami et al.,1999; Francis-West et al.,1999a,b; Hartmann and Tabin,2000). This suggests that their function is in the control of chondrogenesis. Equally, many genes have been implicated in the retention of non-chondrogenic tissue in the interzone. If the role of these genes in the various stages of joint formation is to be determined, it is vital that in addition to defining patterns of expression, attention is also paid to the timing of their expression and, pivotally, to their function in controlling the events leading to joint cavity formation.


Distinguishing between genes, which impact upon different aspects of these processes, may be facilitated by examining the effects of immobilisation of chicks in ovo; immobilisation is known to inhibit cavitation without affecting joint specification or patterning (Fell and Canti,1934; Lamb et al.,2003; Osborne et al.,2002). This creates a new challenge in the identification of joint-forming genes that can be linked to the cavitation process itself. Recent studies have confirmed this view by showing that skeletal movement promotes the acquisition of the joint line-forming phenotype; that is, cells with elevated CD44 and UDPGD content and activity, and high levels of constitutively active extracellular-regulated kinase (ERK) (Ward et al.,1999; Kavanagh et al.,2002b; Dowthwaite et al.,2003; Bastow et al.,2005). Gain of function/loss of function studies coupled with limb immobilization would also allow factors that are independent of intrinsic movement, but capable of specifying joint position, or of promoting joint cavitation, to be unequivocally established. There is, therefore, ample scope for further experimental studies. Indeed, there is now evidence for mechanical regulation of gene expression at the very early stages of embryonic development (Farge,2003), which suggests that such investigations will be fruitful.


It has been assumed throughout this review that joint specification and cavitation are essentially similar in both avian and mammalian species and in the different synovial joints within the body. We have sought to clarify the possible misconceptions about joint formation current in the literature. Thus, critical evaluation of the many and varied morphological, immunohistochemical, and in situ hybridisation studies of limb development provides strong evidence that cartilaginous models of bones develop as discrete anlagen. The published accounts of joint cavitation afford little evidence of cell death at the time of cavitation, which suggests that alternative mechanisms are involved. Finally, although we acknowledge that we have considered only a few specific aspects of joint development, we have attempted to highlight some of the pertinent issues that could be considered by those researchers aiming to identify the mechanisms involved in the specification of joint position and subsequent cavitation.


We thank Ms. Yvette Bland for preparing the sections of chick and mouse embryos, Professor Linda Sandell for her gift of the anti-type IIA antibody, and the late Jim Bee for his assistance in preparing images of differentiating cartilage in Figure 2. We are very grateful to Profs. Cheryl Tickle, Brian Johnstone, and Philippa Francis-West for their astute and critical comments regarding this manuscript. We also thank Drs. Imelda McGonnell and Steve Allen (Royal Veterinary College) for their critical reading and clear developmental insights. Finally, we are indebted to Professor John Fallon for his constructive advice and encouragement during the preparation of this review and for allowing us to reproduce the images of developing embryos stained with Victoria Blue that he produced together with Joseph Lancman.

  • 1

    Researchers from different backgrounds are likely to address these events from disparate perspectives. Some, for example, may concentrate on the intricacies of the changing extracellular matrices and related cell behaviour, while others focus on the underpinning changes in cellular gene expression. Confusion regarding joint specification and formation events may be partly due to imprecision in terminology. Thus, a cartilaginous “continuity” denotes an uninterrupted extracellular cartilaginous matrix for some, but uniformity in cellular gene expression for others (see Introduction section).

  • 2

    At stage 22, the peripheral mesenchyme is separated from the central prechondrogenic mesenchyme by the “capillary net.” The matrix of the central mesenchyme contains less collagen fibrils and proteoglycan granules than the peripheral mesenchyme (Singley and Solursh,1981). Condensation within the pre-chondrogenic mesenchyme involves the removal of hyaluronan leading to the close packing of the cells prior to chondrogenic differentiation (Toole,1972).

  • 3

    Types IIA and IIB collagens differ only in that type IIA contains a 689–amino acid cysteine-rich region in the NH2 propeptide. The antibody used here is specific to this region and hence does not recognise type IIB collagen (Organesian et al.,1997). In contrast, all antibodies to type IIB collagen will also recognise type IIA.