Secondary chondrocyte-derived Ihh stimulates proliferation of periosteal cells during chick development

Intramembranous, or direct, ossification is distinguished from endochondral ossification by the absence of a cartilage scaffold in the former. The significance of this difference is highlighted as we learn more about the key processes of endochondral ossification, although knowledge regarding direct ossification remains scarce. A linear pathway can be discerned for the endochondral maturation of chondrocytes, during which molecular and histological markers of differentiation accrue along the axis of the element from epiphysis to diaphysis. The endpoint for this series is the hypertrophic chondrocyte, a post-proliferative, enlarged cell that possesses characteristics of both chondrocyte and osteoblast(Gerstenfeld and Shapiro,1996). It is reached via an intermediate, the pre-hypertrophic chondrocyte, which is characterised by exit from the cell cycle, but not expressing the full hypertrophic phenotype. Coincident with the formation of pre-hypertrophic and hypertrophic chondrocytes, changes in the cells surrounding the cartilage element occur. The perichondrium overlying the prehypertrophic and hypertrophic zone begins to secrete mineralised matrix; by loosely followed convention this is designated the periosteum(Lacroix, 1951). The pathway of chondrocyte maturation, and its effects on the attendant endochondral ossification, can therefore be distilled into two processes: cell cycle regulation and differentiation. The periosteum comprises pre-osteoblasts and osteoblasts, which undergo a parallel but less overt differentiation pathway than do chondrocytes (Aubin and Liu,1996).

Skeletogenic tissues are highly responsive to applied forces, and hence are uniquely shaped by epigenetic influences (which we define as cellular responses to macroscopic influences); an example in development is the evocation of secondary cartilage. Secondary cartilage arises after membrane bone formation at a variety of sites in birds and mammals (reviewed by Fang and Hall, 1997; Beresford, 1981). In the chick,secondary cartilage has been shown to occur at articulations between membrane bones and cartilage elements of the head(Murray, 1963). Muscle action across these joints leads to the induction of cartilage within the bounds of the periosteum (Murray, 1963),and this tissue is believed to be the source of the chondrocytes. This evocation process can be mimicked in explant culture by mechanical articulation (Hall, 1967), or prevented through in ovo paralysis (Murray and Smiles, 1965). In the mouse temporo-mandibular joint (TMJ),immunohistochemical data indicate a rapid hypertrophy of secondary chondrocytes (Silbermann et al.,1990; Shibata et al.,1997). Despite the unusual epigenesis, the lifetime of a secondary chondrocyte is short, and follows the normal course for a chondrocyte:resorption of matrix by chondroclasts and replacement with endosteal bone(Hall, 1972).

The identification and study of two transcription factors, Sox9 and Cbfa1/Runx2, has greatly enhanced our understanding of the skeletal tissues. Sox9 is recognised as the key regulator of chondrocyte differentiation: it is obligate for chondrocyte specification (Bi et al., 1999), regulates cartilage specific genes(Bell et al., 1997; Lefebvre et al., 1997; Sekiya et al., 2000; Xie et al., 1999) and can induce ectopic chondrogenesis when misexpressed(Bell et al., 1997; Healy et al., 1999). Its transcriptional regulation is consistent with these roles(Wright et al., 1995; Zhao et al., 1997), and excludes any putative role in osteoblast development.

The runt-box-containing transcription factor Cbfa1/Runx2is similarly essential for osteoblast maturation(Ducy et al., 1997; Komori et al., 1997);haploinsufficiency leads to the human skeletal disorder cleidocranial dysplasia (Mundlos et al.,1997). Hypertrophy of chondrocytes in the Cbfa1-knockout mouse is also defective in some long 'bone' elements(Inada et al., 1999; Kim et al., 1999), which is consistent with upregulation of Cbfa1 expression in chondrocytes preceding hypertrophy (Kim et al.,1999). The requirement for Cbfa1 in chondrocyte maturation has been further investigated using the Col2 promoter to drive Cbfa1 expression in immature chondrocytes and in those that would not normally hypertrophy (Takeda et al., 2001). Cbfa1 is sufficient in these cells to promote hypertrophy and to rescue the chondrocyte phenotype of Cbfa1-deficient mice. Moreover, these studies identify Cbfa1 as the first transcription factor to regulate both chondrocyte and osteoblast differentiation (Takeda et al.,2001).

As mentioned above, Cbfa1 plays an important role in the attainment of hypertrophy by some chondrocytes; an example of the regulatory integration between bone and cartilage. The co-dependence of these tissues has also been illustrated by elucidation of the role of Indian hedgehog (Ihh) in skeletogenesis. Overexpression of Ihh in the chick limb leads to suppressed hypertrophy and an increase in bone collar formation(Vortkamp et al., 1996). The former effect results from a feedback loop involving Pthrpupregulation. Ihh expression in the pre-hypertrophic chondrocytes coincides with osteogenesis in the adjacent perichondrium; a link shown to be causal by the failure of osteogenesis in the long bones of Ihh^-/- mice(St-Jacques et al., 1999). Thus, in conjunction with its role in chondrocyte proliferation, achieved directly (Karp et al., 2000)and indirectly (Vortkamp et al.,1996), Ihh couples cartilage differentiation with differentiation of adjacent osteoblasts (Chung et al.,2001). It can be seen, therefore, that a crucial step in osteoblast induction is chondrocyte exit from the cell cycle leading to pre-hypertrophy and Ihh upregulation. The tight choreography of chondrocyte differentiation and osteogenesis in the perichondrium/periosteum cedes the primary role to cartilage. It is appropriate therefore that in the membrane bones of Ihh^-/- mice, where cartilage forms secondarily, osteogenesis is normal(St-Jacques et al., 1999).

Owing to the likely periosteal origin of secondary chondrocytes, one may expect their genesis to challenge the established bifurcation between Cbfa1-specified osteoblasts and Sox9-specified chondrocytes. The regulation of these pivotal genes has not previously been examined in secondary chondrogenesis, but the importance of epigenetic influences on skull development and evolution is widely acknowledged(Thorogood, 1993). We have studied the evocation and function of secondary cartilage in the chick, and found that the integration of epigenetic, and more familiar genetic, pathways underlies its role as a signalling centre. Our study also indicates that Sox9, like Cbfa1 and chondrocyte differentiation, can divert pre-osteoblasts to chondrogenesis: the hypertrophic chondrocyte being the common endpoint, reached via the functionally crucial Ihh-expressing pre-hypertrophic chondrocyte.

Radioactive in situ hybridisation was performed as described(Francis et al., 1994). In brief, embryos were fixed in 4% paraformaldehyde and processed through to wax,sectioned at 8 μm and hybridised with [³⁵S]-labelled riboprobes at 55°C. Chick probes used were as follows: Cbfa1/Runx2 - XhoI(T3; gift from Dr J. Helms), Sox9(Kent et al., 1996), Col2 (Devlin et al.,1988), Col10 (Kwan et al., 1989), Ihh(Vortkamp et al., 1996), Ptc2 (Pearse et al.,2001), Bmp7 (Houston et al., 1994), Frzb(Ladher et al., 2000), and Osteocalcin and Osteopontin PCR fragments (gift from Dr V. Church). Sections were viewed on a Zeiss Axioskop and photographed using a Sony DSC75 digital camera. Montages were arranged using Adobe Photoshop. In some cases bright field images were overlayed with inverted dark field images.

Fertilised wild-type chicken eggs (Ross White) were obtained from Henry Stewart and Co., Lincolnshire, UK. Eggs were incubated at ∼38°C for the stated number of days. Embryos at stage e14 were used for the ex vivo experiments, as the joint was less liable to dislocation when manipulated. The quadratojugal/quadrate (QJ/Q) was dissected from either side of the head and stripped of most attendant connective tissue, except that holding the two elements together. The Q and QJ were trimmed and then placed on a filter(Millipore; 0.2 μM) upon a metal grid at the air/medium interface. Explants were cultured in DMEM (Gibco) plus 15% chick serum (Sigma), at 37°C/5%CO₂. Articulation regimes involved operating the joint through its normal movement (10 times on the hour) in order to mimic the sporadic movements in ovo, for the specified time (following Hall, 1967).

Anti-hedgehog antibody, 5E1 (Ericson et al., 1996), was obtained from the DSHB, University of Iowa, IA. Partially purified antibody was obtained and used at a dilution of 1:10.

Explants were labelled in 10×BrdU diluted in culture medium (as above), according to the manufacturer's protocol (Roche), immediately following dissection (for 0 hour timepoint) or at the specified time post-dissection for 90 minutes. Explants were then fixed in 4%paraformaldehyde, processed through to wax and sectioned. BrdU incorporation was revealed using a biotin-conjugated anti-BrdU antibody (MD-5215; Caltag,USA). ABC kit from Vectastain (Vector Labs, Peterborough, UK) was then used,followed by DAB (Vector Labs) for the colour reaction. Sections were counterstained with Toluidine Blue (0.02%) for cell counting, or Alcian Blue(prior to immunochemistry) and Chlorantine Fast Red (post-immunochemistry) to assess differentiation. For pulse-chase experiments the QJ/Q joint was rinsed in medium, placed on a fresh filter in fresh medium and re-incubated. Parathyroid hormone (PTH) was used at a concentration of 10 nM.

High-density micromass culture was carried out essentially as described by Fang and Hall (Fang and Hall,1996), based on the sequential digestion method for calvaria(Wong and Cohn, 1974). 2-3 dozen eggs were harvested to provide e13 QJs and Qs that were then dissected free from muscle and connective tissue, and pooled in PBS. The hook,containing the secondary chondrocytes, was removed and the shafts were digested in 0.5% trypsin/collagenase type 1A (200 units/ml; Sigma) in PBS,sequentially for 30 minutes, 30 minutes and 45 minutes. Between each digestion the solution was removed and replaced. The solution containing the cells was then spun down and rinsed in serum-containing medium. For establishing the micromass cultures we used fractions 2 and 3. These were pooled, passed through a cell strainer (Falcon), spun down and resuspended in 200 μl of medium. A 10 μl drop was added to the centre of a well of a multi-well plate, allowed to settle for 2 hours before flooding with medium (3/4 Ham's F12+1/4 BGJb, containing 10% FBS, L-glutamine, ascorbate 50 μg/ml,penicillin/streptomycin and β-glycerophosphate). Proliferation was assessed by adding BrdU for 30 minutes on the fourth day, before fixing and processing as outlined above.

Infection with retroviruses was achieved by adding filtered conditioned medium at the flooding stage. Alkaline phosphatase activity was detected using Napthol AS MX-PO₄ (Sigma) and Red Violet (Sigma), and mineralised matrix was detected using 2.5% silver nitrate enhanced by carbonate formaldehyde (von Kossa). The RCAS retroviral construct used in this study, RCAS-Ihh, has been previously described(Vortkamp et al., 1996); RCAS-Gfp was constructed as described by Church et al.(Church et al., 2002).

Previous studies in the chick by Murray, and by Fang and Hall, have demonstrated that secondary chondrocytes arise in the quadratojugal (QJ)between embryonic days (e) 10 and 11(Murray, 1963; Fang and Hall, 1997). In order to characterise the formation of secondary chondrocytes in this location, we studied initially the expression of Cbfa1 and Sox9. Using radioactive in situ hybridisation of adjacent sections at e10(Fig. 1A), we found Cbfa1 (Fig. 1B) and Sox9 (Fig. 1C) to be expressed in exclusive tissues: Cbfa1 was expressed in membrane bone periostea and osteoblasts, and Sox9 was expressed in the primary cartilaginous elements (e.g. the quadrate; Q). Cbfa1 was expressed throughout the periosteum (Fig. 1E), but at the tip of the QJ at e10, where the secondary chondrocytes will arise, there was neither Alcian Blue staining(Fig. 1D), nor expression of Sox9 (Fig. 1C), type II collagen (Col2; Fig. 1G) or type X collagen (Col10; data not shown). Analysis of Gdf5expression, a potent pro-chondrogenic growth factor of the TGFβsuperfamily (Francis-West et al.,1999), revealed it to be expressed only in the perichondrium of the Q (Fig. 1F).

Alcian Blue staining can first be detected within the boundary of the periosteum of the membrane bone at e11, on the inner face of the QJ-hook(Fig. 1H), the earliest histological evidence of secondary chondrocytes (SCs). Observed simultaneously with this change, (though necessarily slightly in advance of it; data not shown) Sox9 transcripts could be detected within the same region of the QJ (Fig. 1J). The Sox9 expression domain was confined to the inner region of the QJ tip(the secondary chondrocytes) and did not extend to the adjacent periosteum,which, from hereon, we shall call the germinal region(Hall, 1968). Cbfa1continued to be expressed unaltered in the periosteum and germinal region, and was also expressed in the forming chondrocytes(Fig. 1I). Detection of the same genes at e13 (Fig. 1K,L,M)show clearly that Cbfa1 was expressed in secondary chondrocytes(Fig. 1L), the periosteum and the germinal region. Sox9 was also expressed throughout the SCs at this time point, but was excluded from the germinal region(Fig. 1M).

We further characterised the secondary chondrocytes by examining the expression of Col2, an early chondrocyte matrix constituent; Ihh, expressed by pre-hypertrophic chondrocytes(Vortkamp et al., 1996); Col10, a marker for terminally differentiated chondrocytes(Devlin et al., 1988); and Frzb, expressed by condensing mesenchyme and early chondrocytes(Ladher et al., 2000).

Col2 was expressed in a domain coincident with that of Alcian Blue staining (Fig. 2A,B); in this example, both in anterior and posterior SC domains of the tip of the QJ, but not in the germinal region. Col10 expression on an adjacent section(Fig. 2C) revealed that although the domain of Col10 was contained within that of Col2, it had been rapidly upregulated by the secondary chondrocytes. This apparent rapid hypertrophy was borne out by the co-expression of Ihh (Fig. 2E) and Col10 (Fig. 2F), and was consistent with the enlarged appearance of the secondary chondrocytes. Additionally, a negative regulator of chondrocyte differentiation, Pthrp, was undetectable at e11 and at e14 (data not shown). As secondary chondrocytes are believed to arise from the germinal region, we examined the expression of known perichondrial/periosteal markers to see whether changes in the expression of these genes presaged or reflected the formation of SCs. We found that Bmp7 was expressed in the periosteum of the membrane bones (Fig. 2H), including around the SCs. Gdf5 expression was not detected in the germinal region (Fig. 2I), even after SC formation, nor was it expressed in the remaining periosteum, but it could be detected in the quadrate perichondrium. We also looked at further markers of chondrocytes, for example, the pre-chondrogenic marker Frzb (Fig. 2L), which was expressed coincident with Sox9upregulation (Fig. 2K) and in advance of Alcian Blue staining. For markers of definitive osteoblasts and late stage hypertrophic chondrocytes we also compared the expression of osteocalcin (ocn) and osteopontin (opn) in the quadratojugal. Neither gene was expressed in SCs up to e13(Fig. 2M,N,O; data not shown). In the membrane bone, opn was expressed by the osteocytes of the trabecular bone (Fig. 2N),whereas ocn was expressed earlier, in the osteoblasts adjacent to the periosteum (Fig. 2O). Opn was upregulated in SCs at e14(Fig. 2P,Q), coinciding with the erosion of secondary cartilage by chondroclasts. By contrast, ocnexpression remained osteoblast specific(Fig. 2R).

Exit from the cell cycle is a key event in the maturation of chondrocytes,occurring at the transition to prehypertrophic chondrocyte. To determine how,or whether, this transition occurred in secondary chondrocytes, we explanted the quadratojugal/quadrate joints (QJ/Q), from e14 embryos, onto filters and labelled them with bromo-2′-deoxyuridine (BrdU) for 90 minutes whilst culturing at the air/medium interface. Explants were then fixed and processed for detection of BrdU. Labelling revealed striking differences in proliferation between the SCs, the germinal region and the periosteum - very few SCs were found to be in S-phase. Comparison of the percentage of BrdU labelled cells in the three regions confirmed the clear variation in proliferation rates (Fig. 3):periosteum (5.2%; 114/2175), germinal region (15.9%; 563/3531) and secondary chondrocytes (0.51%; 10/1978) (nine sections from three QJs were assessed for each result). Histologically, these results represented a marked cut- off in proliferation at the boundary between the SCs and germinal region(Fig. 4A), with only occasional cells at the periphery of the SC region being labelled(Fig. 4C; n=6, 36 sections).

Given the distinction between proliferation and non-proliferation, and the parallel contrast between undifferentiated germinal region and differentiated secondary chondrocytes (as assayed by genetic markers), we realised that it would be possible to use BrdU incorporation to label the germinal cells and then follow their fate. The immediate labelling by BrdU of S-phase cells would therefore assess commitment to the chondrocytic pathway under the in vivo conditions of articulation. Subsequently, however, in the absence of mechanical stimulation ex vivo, the germinal region should start to give rise to osteoblasts, not chondrocytes (Hall,1968) (see also Fig. 8). All the following experiments were carried out in the absence of mechanical stimulation, unless otherwise stated. In order to follow the fate of proliferating cells, paired QJ/Q joints were incubated in BrdU, as described above, and one was immediately fixed (time 0). The contralateral side was rinsed of BrdU, moved to a new well with a fresh filter/medium and re-incubated for 24 hours (the chase) prior to fixation.

By following the pulse/chase protocol, we found that pairs of BrdU-labelled nuclei could be detected within the chondrocyte zone(Fig. 4B,D). The daughter cells shared a lacuna within the hypertrophic-like domain of the secondary chondrocytes, where at time 0 BrdU data showed proliferation to be very low or undetectable. To confirm that self-renewing pre-chondrocytes were not present,we incubated the explanted QJ/Q joint for 16 hours prior to labelling with BrdU, and followed this by a 24-hour chase in the absence of mechanical stimulation. If there was an amplification zone of pre-chondrocytes, these would be labelled at the second division and labelled chondrocytes would be observed. We found that pulsing at 16 hours, followed by immediate fixation,labelled the germinal region (Fig. 4E), but the chase of the contralateral side failed to label chondrocytes (Fig. 4F;labelling osteoblasts instead), even though conditions were clearly permissive for chondrocyte proliferation as judged by that detected in the quadrate.

To explore the role of cell division in differentiation, we assessed the expression of Col2 and Col10 in explanted QJ/Q joints labelled with BrdU, on adjacent sections. As at e11 and e13 (see Fig. 2B,C; data not shown), Col2 expression (Fig. 4H) slightly preceded that of Col10(Fig. 4H), although BrdU labelling within the Col2-expressing area was rare(Fig. 4G). When we compared this to the situation after a 24 hour chase(Fig. 4I, BrdU labelling at 0 hours), we found that the Col2 and Col10 expression domains at the germinal region/SC boundary were now coincident(Fig. 4J). Also, cell doublets could now be observed within the Col10 expression domain(Fig. 4I). This observation was consistent with a cell committing to the chondrocyte lineage, dividing once,and then upregulating Col10 (n=3). The rapid hypertrophy indicated by this result, led us to assess whether this was an intrinsic property of SCs, or was related to the absence of Pthrp expression. We attempted to delay hypertrophy by adding PTH peptide to cultured QJ/Q explants. We found no significant difference between PTH-treated and untreated explants when we compared Col2 with Col10 expression on adjacent sections after 24 hours culture (n=4; data not shown).

The proliferation rate in the germinal region was almost three times that of the periosteum in other regions, indicating that the germinal region was subject to influences unique to this location. In order to assess whether hedgehog (Hh) signalling from SCs was responsible, we analysed the expression of Ptc2 in relation to Ihh. As an inducible transcriptional target, Ptc2 is a sensitive marker for receipt of Hh signal(Pearse et al., 2001; Hartmann and Tabin, 2001). At e11, although Ihh was readily detectable(Fig. 2E, Fig. 5B), Ptc2expression was at the limit of detection(Fig. 5C). However, at e12 and e14 Ptc2 expression was evident within the germinal layer(Fig. 5F,I), adjacent to the Ihh domain in the SCs (Fig. 5E,H). Ptc2 was not expressed in the periosteum surrounding the membrane bone, and we were unable to detect Ptc1expression at either e11 or e14 (data not shown).

We addressed the functional significance of the Ihh domain by using the Hh blocking antibody 5E1 to ablate Ihh function in our explant system (Ericson et al., 1996; Wu et al., 2001). We then assessed proliferation after 16 hours of incubation by BrdU labelling as before. We found that proliferation in the contralateral joint in 5E1-treated sections, as measured by the number of cells in S-phase, was dramatically reduced when compared with non-Hh antibody treated, or no antibody treated controls (Fig. 6A,D). t-test analyses on scores for four pairs of QJ/Qs revealed significant differences in each case (P<0.05): on average the total number of BrdU-labelled cells in the germinal region of 5E1-treated samples was reduced to a third of that in the controls. To ensure the effect was specific, we analysed adjacent sections for Ihh and Ptc2expression: Ihh was unaffected(Fig. 6B,E), whereas Ptc2 was downregulated in the treated side relative to the control(Fig. 6C,F).

Our data indicated that Ihh signalling to the periosteal cells was having a proliferative effect. We assessed whether this was a general property of membrane bone periosteal cells by using high-density micromass cultures with QJ-derived periosteal cells that excluded the germinal region(Fang and Hall, 1996). Such cultures are alkaline phosphatase (AP) positive, express osteocalcinand osteopontin, do not produce Alcian Blue staining matrix (data not shown) or type II collagen (Fang and Hall,1996), and go on to form bone after 14 days(Fang and Hall, 1996). We assayed proliferation by BrdU incorporation 4 days after seeding and infection with RCAS-Ihh or RCASGFP. We found that Ihh almost doubled the proliferation rate (1.8×, s.d. ±0.15, three experiments in triplicate; Fig. 7A). In longer-term cultures (day 11), we observed polygonal cells in both control(Fig. 7B) and RCAS-Ihhinfected cultures (Fig. 7C),but observed widespread bone nodule formation only in the latter(Fig. 7D). Both control and RCAS-Ihh-infected cultures were AP positive(Fig. 7E,F), but the pronounced nodules present in Ihh-infected cultures also reacted strongly for von Kossa stain, demonstrating precocious formation of mineralised matrix(Fig. 7E,F; n=6). After 14 days in culture, we found greatly enhanced AP activity around the periphery of the Ihh-infected micromasses compared with the control(Fig. 7G,H). The core of the micromass did not express AP activity in either control or Ihh-infected cultures (osteocytes are AP negative)(Aubin and Liu, 1996), but the AP-negative core of Ihh-infected culture was larger. Visualisation of mineralised bone matrix in the same samples confirmed that osteocytes had formed in the centre of each micromass(Fig. 7I,J; n=8). Thus, Ihh-infected cultures exhibited increased proliferation of pre-osteoblasts at 4 days, and at the later timepoints showed enhanced progression to osteocytes. Together these findings indicated an osteogenic effect of Ihh on periosteal cells.

To explore the role of mechanical cues in the genesis of SCs, we used the transcriptional regulation of osteocalcin as an assay; this would also provide evidence on whether the Cbfa1-expressing nature of the germinal cells was truly indicative of pre-osteoblastic commitment. We utilised an hourly articulation regime, based on the methods of Hall(Hall, 1968), in which the joint is manually flexed whilst in explant culture to recapitulate the effects of jaw opening and closing as occurs in ovo. We coupled analysis of ocn expression with BrdU labelling in order to determine potential effects on the cell cycle simultaneously. Pulsing at 0 hours, followed by hourly articulation for 9 hours and analysis at 24 hours, revealed labelling of chondrocyte pairs (Fig. 8A)and an absence of ocn expression(Fig. 8B). The control joint possessed similar labelling of chondrocyte pairs(Fig. 8C) after 24 hours, but in contrast to the articulated joint, ocn expression was readily detected at the periphery of the SC domain(Fig. 8D; 0/4 in articulated joints; 3/4 in non-articulated joints). Thus, in the absence of mechanical stress, the cells of the germinal region rapidly give rise to definitive osteoblasts. Having established that mechanical stress resulting from the articulation of the QJ/Q joint was sufficient to repress osteoblast differentiation we wanted to assess whether it was also promoting chondrogenic differentiation. Using the same regime but with a delayed labelling strategy incorporating a chase (i.e. pulse at 24 hours, chase for 24hours), we could follow the fate of cycling cells. By doing so we found that, as before, in the absence of articulation chondrocytes were not labelled(Fig. 8E), but after the articulation regime labelled chondrocytes could be detected(Fig. 8F; n=3). This indicated that germinal region cells had been committed to the chondrocyte pathway.

The rapid shift in cell-type differentiation, dependent upon the presence or absence of mechanical stimulation, indicated that the germinal region was extremely plastic, with the fate decision being a late event in germinal region expansion. To determine whether this was the case, we labelled e11 explanted QJ/Qs at 0 hours followed by a 48 hour chase (n=3). The resultant BrdU-labelling of both chondrocytes and osteoblasts(Fig. 8G) indicated that precursor cells capable of giving rise to osteoblasts, as well as cells committed to giving rise to chondrocytes, had both been proliferating under the influence of mechanical stimulation(Fig. 8H).

Our study of the events underlying secondary chondrogenesis has revealed the details and potential function of a unique differentiation pathway: the formation of hypertrophic-like chondrocytes from osteoblast precursors. In essence, this transition reflects the transcriptional upregulation of the Sox9 gene within a precursor population that already expresses Cbfa1. Subsequently, a raft of genetic markers, normally associated with the hypertrophic chondrocyte phenotype, become rapidly expressed within these cells: Ihh, Col10, Wnt11 (data not shown) and, later, osteopontin. Some stratification within the secondary chondrocytes is apparent; Col2 expression precedes Ihh and Col10expression, and osteopontin expression reflects the last phase of hypertrophy. Thus the hypertrophic character that Murray observed(Murray, 1963) and used to distinguish secondary chondrocytes from their primary counterparts is found to be bona fide, in common with mammalian secondary chondrocytes(Shibata et al., 1997). In itself, the evidence that Sox9 expression is upregulated is unsurprising given its pivotal role in chondrocyte specification and gene regulation (reviewed by de Crombrugghe et al., 2000; Bi et al.,1999); what is unusual is the succession of gene upregulation.

It is well known that immature chondrocytes of the growth plate gradually acquire the range of markers outlined above, but start from a population of precursors expressing Sox9. Misexpression of Cbfa1 in immature, Sox9-expressing chondrocytes has shown that it is sufficient to promote hypertrophy and exit from the cell cycle(Enomoto et al., 2000; Takeda et al., 2001). What our study of SCs shows is that the reverse is also true: mechanical evocation of Sox9 in ovo, in a population expressing Cbfa1, leads to the same array of gene targets becoming upregulated. Consequently, we have shown that the key motivator of chondrogenesis on the one hand (Sox9) and osteogenesis on the other (Cbfa1) appear to affect the same outcome(pre-hypertrophic chondrocytes) when co-expressed, irrespective of their primacy. Consequently, a bifurcating pathway of endochondral ossification and secondary chondrogenesis can be described in which early lineage determination from a mesenchymal precursor, arguably governed by Sox9 and Cbfa1 activity,much later rejoins with the differentiation of the hypertrophic chondrocyte(Fig. 9).

This reversibility of succession is evidence for a simple mechanism underlying hypertrophy and the control of chondrocyte maturation. It is also noteworthy that although the endpoint for pre-osteoblasts and chondrocytes can be the same, the genes expressed by (pre-) hypertrophic chondrocytes are not merely the sum of Sox9 targets (de Crombrugghe et al., 2000) plus Cbfa1 targets(Ducy, 2000). We characterise the germinal cells as pre-osteoblastic because they express Cbfa1,and, if not diverted into chondrogenesis, they will upregulate osteocalcin and produce matrix that binds Chlorantine Fast Red when they differentiate. Furthermore, prior to the evocation of secondary chondrocytes, the periosteum at the tip of the QJ gives rise to osteoblasts and no changes in the expression of peripheral markers occurred coincident with the genesis of secondary chondrocytes. However, we must note that SCs do derive from these Cbfa1-expressing cells, and this can be interpreted as evidence that Cbfa1 does not dictate commitment to the osteoblast lineage,but rather confers bi-potentiality.

The transition from perichondrium to periosteum during endochondral ossification follows exit from the cell cycle and pre-hypertrophy in the underlying chondrocytes. This process of chondrocyte maturation is meticulously regulated and deficiency of any of a number of genes that affect proliferation or cell cycle control causes its perturbation, for example, Pthrp (Karaplis et al.,1994), Pthrp-r(Lanske et al., 1996), Ihh (St-Jacques et al.,1999), p57 (Zhang et al., 1997) and Cyclin D1(Beier et al., 2001). When we consider secondary chondrogenesis, the periosteum (germinal region) is the source of the chondrocytes as well as the responding tissue. However, the chondrocytes formed rapidly exit the cell cycle and undergo hypertrophy. Thus,commitment to the chondrocyte lineage appears not to be accompanied by the formation of a self-renewing proliferative layer; this is in contrast to the growth plate of the long bone, where chondrocyte precursors are maintained independent of movement of the joint. This conclusion is also supported by in ovo experiments that show that, following paralysis of the embryo,proliferating secondary chondrocyte-committed cells are not found(Hall, 1979).

This achieves one end - rapid hypertrophy - but necessitates continued mechanical articulation of the joint to re-commit precursors to the chondrocyte lineage (Murray and Smiles,1965; Hall, 1967; Hall, 1968; Hall, 1979) (this paper). Our finding that Col10 expression expanded to coincide with the Col2 domain after 24 hours in culture indicated that commitment,differentiation and exit from the cell cycle are intimately correlated in SCs. Thus the initial spur to chondrogenesis - compression and upregulation of Sox9 (Takahashi et al.,1998) - appears to be followed by a single round of division and hypertrophy.

The mouse chondrodysplasias mentioned above result in either delayed or accelerated hypertrophy. These changes then feed into ossification via Ihh signalling, resulting in reduced or increased bone formation,respectively. As in endochondral ossification, accelerated exit from the cell cycle in SCs is accompanied by Ihh upregulation, which then signals to the germinal region but not to the chondrocytes. As a result, cycling cells in the germinal region express Ptc2, but, as they become chondrocytes, the cells exit the cell cycle, downregulate Ptc2 and upregulate Ihh. Hence, the cells progressed from Hh-signal receiving to Hh-signal producing at the germinal region/SC boundary, superficially resembling an extreme form of the growth plate of the Pthrp-rknockout mouse (Lanske et al.,1996).

The threefold enhanced proliferation evident in the germinal region around the SCs was quenched by blocking Ihh function with the 5E1 antibody(Ericson et al., 1996; Wu et al., 2001). This finding was consistent with the demonstrated role of Ihh signalling to immature chondrocytes to drive proliferation (Long et al., 2001), but was transposed to the germinal region. Given the pivotal roles of Ihh in chondrogenesis and osteogenesis of the long bones,it might also be expected to affect differentiation of the germinal cells. However, we did not find precocious or delayed expression of the osteogenic marker osteocalcin in 5E1 treated samples (data not shown). It was also apparent from the osteocalcin expression data at 24 and 48 hours that Ihh alone, in the absence of mechanical stimulation, was not sufficient to cause chondrogenic differentiation in germinal region descendants. This is consistent with the fact that Ihh is expressed only after SCs have formed, excluding it from any role in the initiation of SCs themselves. Thus the principal, and perhaps sole, function of Ihh signalling in secondary chondrogenesis is to drive proliferation of the germinal region and thereby increase indirectly recruitment to the chondrogenic fate. In turn, exit from the cell cycle appears to be a prerequisite for the synthesis of Ihh.

The mitogenic effect of Ihh in micromass culture was not directly comparable with the role of Ihh in SC genesis owing to the different fates of the responding cells. However, it was consistent with the conclusion above. Our in vitro micromass data additionally demonstrate that Ihh promotes osteogenesis, as measured by nodule formation and deposition of mineralised matrix. This conclusion is comparable to that from the phenotype of the Ihh^-/- mice, that Ihh is necessary for osteogenesis in cartilage-replacement bones(St-Jacques et al., 1999). Our findings are significant not only for understanding secondary cartilage but also because they extend the similarities between periostea of membrane bones and long bones. This implies that it is ontogenetic differences that separate these periostea, not genetic ones(St-Jacques et al., 1999).

One of the most striking features of the SCs described is that their formation is dependent upon articulation of the QJ/Q joint. In ovo this is achieved by muscular contraction that opens and closes the beak, which can be mimicked in explant by the manual operation of the joint(Hall, 1968) (this paper). Whereas evocation of chondrogenesis has certainly been described, we found that the discreet nature of the germinal region as proliferative centre allowed us to explore the short-term response to mechanical articulation. The ability of mechanical articulation to prevent the default osteogenic pathway,as determined by ocn upregulation, indicates a direct role for mechanical stimulation in the regulation of gene expression. Ocn is upregulated by osteoblasts as they exit the cell cycle(Aubin and Liu, 1996), and in the absence of mechanical stimulation the germinal cells follow this route. Thus, the cells of the germinal region that give rise to secondary chondrocytes would normally exit the cell cycle at this juncture, but the acquisition of chondrogenic fate diverts them - a limited form of bi-potentiality (Fang and Hall,1996). The fact that no self-renewing chondrocyte-committed precursors appear to be established, as evidenced by gene expression and BrdU labelling (see also Hall,1979), accentuates this as a defining feature of secondary chondrogenesis in the chick.

The unusual differentiation undergone by SCs, and their transient nature,alludes to a role for SCs that is not structural, but is rather as a growth and signalling centre; our study demonstrates that a molecular basis for this function is Ihh production. The clear inference is that the pre-hypertrophic chondrocyte is of such utility as a source of extracellular signals for bone morphogenesis that, even when no cartilage scaffold exists, pre-hypertrophic chondrocytes are evoked. The finding that Ihh stimulates proliferation of the germinal region highlights the reciprocity between epigenetic and genetic pathways that characterise the development of the cranial tissues, and that underpin their evolution. Such a scenario resonates with other phenomena, such as fracture repair, where a short-lived, Ihh-synthesising soft callus forms from the periosteum under the influence of movement(Vortkamp et al., 1998; Ferguson et al., 1999). Conversely, the common characteristics that we have found between endochondral and secondary chondrogenesis indicate that the subordinate role of bone to cartilage [and other epigenetic influences(Herring, 1993)] is a skeleton-wide phenomenon. Moreover, although requiring confirmation in mammals, our findings provide additional mechanistic evidence in support of Scott's view of cartilage as a pacemaker of cranial bone growth(Scott, 1954).

We would like to thank Stalin Kariyawasam for help with histology, and all those who provided reagents for this study: C. Devlin, B. Houston, A. Kwan, C. Healy, J. Helms, A. Sinclair, C. Tabin and A. Vortkamp. This work was funded by the Arthritis Research Campaign.