Wnt/β-catenin signaling interacts differentially with Ihh signaling in controlling endochondral bone and synovial joint formation

Vertebrate skeletal development is mainly controlled by endochondral ossification, during which mesenchymal progenitor cells first differentiate into chondrocytes that form cartilage shafts for the future bone. In the developing cartilage, chondrocytes sequentially go through a tightly controlled program of proliferation and differentiation. Eventually,proliferative chondrocytes exit the cell cycle and differentiate into hypertrophic chondrocytes. In the meantime, osteoblasts differentiate in the perichondrium. These events are followed by the invasion of blood vessels and osteoblasts to the hypertrophic cartilaginous region to replace it with trabecular bone. Chondrocyte hypertrophy is required for ossification because hypertrophic chondrocytes secrete factors such as Indian hedgehog (Ihh) and vascular endothelial growth factors (Vegfs), which are required for osteoblast differentiation (reviewed by de Crombrugghe et al., 2001; Karsenty and Wagner, 2002; Kronenberg,2003; Zelzer and Olsen,2003). Wnt and hedgehog signaling pathways control cell proliferation, differentiation, survival and migration in both developing embryos and adult animals. In the developing skeletal systems, several aspects of endochondral bone formation are critically regulated by both Wnt/β-catenin and Ihh signaling pathways. Inactivation of β-catenin or Ihh signaling leads to ectopic chondrocyte differentiation at the expense of bone formation (Day et al.,2005; Hill et al.,2005; Hu et al.,2005; Long et al.,2004). Wnt/β-catenin and Ihh signaling pathways also regulate chondrocyte hypertrophy and maturation(Akiyama et al., 2004; Day et al., 2005; St-Jacques et al., 1999; Vortkamp et al., 1996). In the developing cartilage, Ihh is expressed in prehypertrophic chondrocytes. It controls osteoblast differentiation in the perichondrium by controlling the expression of Runx2, which encodes a transcription factor required for osteoblast differentiation(St-Jacques et al., 1999). Ihh is also a major driving force for chondrocyte proliferation(Karp et al., 2000), and it sets the pace for chondrocyte hypertrophy by activating the expression of parathyroid hormone related peptide (Pthrp; Pthlh - Mouse Genome Informatics), which inhibits chondrocyte hypertrophy(Lanske et al., 1996; St-Jacques et al., 1999; Vortkamp et al., 1996). In contrast to what is known about the functional mechanism of Ihh signaling, the molecular mechanism underlying the Wnt/β-catenin functions in endochondral skeletal development remains to be elucidated. Deciphering the genetic relationship between Wnt/β-catenin and Ihh signaling pathways will provide significant new insight into the molecular mechanisms by which both signaling pathways control endochondral bone formation.

It has been proposed that Ihh may signal upstream of Wnt signaling because the expression of Wnt7b and Tcf1 in the perichondrium is lost in the Ihh mutant (Hu et al., 2005). However, instead of being regulated by Ihhdirectly, the expression of Wnt7b and Tcf1 may be simply associated with osteoblast differentiation in the perichondrium, which is abolished in the Ihh mutant. Consistent with this, Tcf1 is not required for osteoblast differentiation in mice(Glass et al., 2005) and there is no evidence to support that Wnt7b is required for osteoblast differentiation in vivo. Another unsettled issue is the role of Wnt/β-catenin signaling in regulating the expression of the transcription factor osterix (Osx; Sp7 - Mouse Genome Informatics) that acts downstream of Runx2 in controlling osteoblast differentiation(Nakashima et al., 2002). Although Osx expression was clearly detected in the perichondrium/periosteum when β-catenin was inactivated by Cre-mediated recombination (Day et al.,2005), its expression was missing in another similar study(Hu et al., 2005). Therefore,the control of Osx expression by the Wnt/β-catenin signaling,and the genetic interactions between Wnt/β-catenin and Ihh signaling await further investigation by rigorous genetic tests.

Another important aspect of endochondral skeletal development is synovial joint formation, which is controlled by Wnt/β-catenin signaling(Guo et al., 2004; Hartmann and Tabin, 2001). Although joint formation is not disrupted in Ihh mutant mice(St-Jacques et al., 1999), it is likely that Ihh signals to the joint region, and it is possible that Ihh signaling needs to be kept at a low level to allow normal joint formation.

All mammalian hedgehogs (Hhs), including Ihh, transduce their signals through two multipass transmembrane proteins, smoothened (Smo) and patched 1(Ptch1) (reviewed by Huangfu and Anderson,2006; Lum and Beachy,2004). The current model proposes that, in the absence of the Hh ligand, the signaling activity of Smo is inhibited by Ptch1. Hh binding to Ptch1 relieves its inhibition of Smo, allowing Smo to transduce the Hh signal to the intracellular components. It is predicted that Hh signaling can be maximally activated in a cell-autonomous manner by inactivating Ptch1. The Wnt/β-catenin signaling is the canonical Wnt signaling transduced through stabilizing β-catenin (reviewed by Logan and Nusse, 2004; Moon, 2005). In the absence of Wnt ligands, β-catenin is phosphorylated by glycogen synthase kinase 3(Gsk3) and then targeted for degradation in the proteasome. When Wnt ligands bind to their receptors frizzled and low-density lipoprotein receptor-related protein 5/6 (Lrp5/6), β-catenin phosphorylation is inhibited. This causes the cytoplasmic accumulation and nuclear translocation of β-catenin,which activates downstream gene transcription by binding to Lef/Tcf factors. Therefore, the strength of the canonical Wnt signal can be manipulated by altering the protein levels of β-catenin. Canonical Wnt signaling has been blocked cell autonomously by tissue-specific inactivation ofβ-catenin in many developmental processes.

Here, we have generated a floxed allele of Ptch1 and activated Hh signaling cell autonomously in osteochondral progenitor cells during endochondral bone formation. We found that upregulation of Hh signaling uncoupled chondrocyte hypertrophy and osteoblast differentiation. Endochondral ossification occurred in the absence of chondrocyte hypertrophy. In addition,upregulation of Hh signaling led to mild joint fusion. Furthermore, when canonical Wnt signaling was blocked at the same time that Ihh signaling was upregulated, we observed cell-context-dependent genetic interactions between the canonical Wnt and Ihh signaling pathways. In osteoblast differentiation,β-catenin acts, not only downstream of the Ihh pathway, but also downstream of Osx. In chondrocyte survival, Wnt/β-catenin is upstream of Ihh signaling when inhibiting chondrocyte apoptosis. Finally,during chondrocyte hypertrophy and synovial joint formation,Wnt/β-catenin and Ihh signaling pathways exert opposite activities and there is a synergistic interaction between them during synovial joint formation.

Generation of the conditional Ptch1 mutant mice is described in Fig. 1. Genotyping was done by PCR using genomic DNA prepared from embryonic liver. Oligonucleotides Ptch1b(5′-GCAAGTTTTTGGTTGTGGGTCTCC-3′) and PtclacZ2(5′-GCGATTAAGTTGGGTAACGCC-3′) were used to genotype the Ptch1 floxed allele (348 bp band size). Oligonucleotides Ptc2b(5′-CCTTCCCGCGAGCTGGATGTG-3′) and PtclacZ2 were used to genotype the Ptch1^ΔloxP allele (333 bp band size). Oligonucleotides Ptch1b and Ptch1f(5′-GCTACAAGGAGGCTCTAGGTGC-3′) were used to genotype the Ptch1 wild-type allele (200 bp band size). β-catenin alleles were genotyped as described (Guo et al.,2004).

Embryos were dissected in PBS. The embryos were then skinned, eviscerated,and fixed in 95% ethanol. Skeletal preparations were performed as described(McLeod, 1980).

Embryos were fixed in 4% paraformaldehyde at 4°C overnight. Fixed samples were embedded in paraffin and sectioned at 5 μm thickness. Histological analysis, BrdU labeling, immunohistochemistry and radioactive ³⁵S-RNA in situ hybridization were performed as described(Yang et al., 2003). Primary antibodies included anti-p21 mouse monoclonal IgG (Santa Cruz) at 1:40,anti-p27 mouse monoclonal IgG (Santa Cruz) at 1:40, anti-p57 rabbit polyclonal IgG at 1:160 (Santa Cruz) and anti-cyclin D1 mouse monoclonal IgG (Santa Cruz)at 1:80. Signals were detected using biotin-conjugated secondary antibodies and the ABC Kit (Vectorstain). Cell death was detected as described(Barrow et al., 2003). Probes for in situ hybridization have been described previously: ColX, ColII,Ihh and Runx2 (Yang et al.,2003); Osx (Nakashima et al., 2002); Mmp13(Day et al., 2005); Gli1 (Lewis et al.,2001); and Hip1(Chuang and McMahon, 1999). An EST clone (Clone ID 5346064, from Invitrogen) was used to generate Pthrp probes.

Micromass cultures and Cre-Adenovirus infection were performed as previously described (Guo et al.,2004). Recombinant mouse Bmp2 and noggin were purchased from R&D Systems. Bmp2 (0.5 μg/ml) and noggin (1.5 μg/ml) were added to the micromass culture 24 hours after the cells were plated at high density. In the limb culture, the BGJB (Invitrogen) medium, containing noggin (1.5μg/ml), was changed every 2 days.

As both Ihh and canonical Wnt signaling control multiple aspects of chondrocyte and osteoblast development, testing the genetic epistasis of the Ihh and the canonical Wnt signaling pathways specifically in the developing skeletal tissue is essential for gaining more insight into the molecular regulatory mechanisms of both signaling pathways in vivo. To achieve this would require the upregulation of Ihh signaling and the inactivation of canonical Wnt signaling simultaneously in the same cells. For this purpose, we generated a floxed allele of Ptch1 in mice(Fig. 1) that allowed cell autonomous activation of Hh signaling upon Cre-mediated recombination. Ptch1 was inactivated using the Col2a1-Cre line in which Cre-mediated recombination was observed in all chondrocytes and many osteoblasts (Day et al., 2005). We found that in the Ptch1^c/-; Col2a1-Cre embryos, long bones in the limb were shorter, bone mineralization was enhanced and the posterior skull failed to form (Fig. 2L-N). These results are consistent with previous findings when Ihh was overexpressed during endochondral bone formation in mice(Long et al., 2004; Long et al., 2001). It appears that failure of posterior skull formation in the Ptch1^c/-;Col2a1-Cre embryos may be due to enhanced antichondrogenic Wnt/β-catenin signaling in the skull mesenchyme, which resulted from abnormal brain development. In support of this, loss of β-catenin completely rescued skull formation in the Ptch1^c/-;Catnb^c/-; Col2a1-Cre double mutant embryos(Fig. 3A). In the limb sections of the Ptch1^c/-; Col2a1-Cre embryos, we found that chondrocyte hypertrophy was inhibited (Fig. 2A,B) and Pthrp expression was upregulated(Fig. 2C,D), suggesting that Ihh signaling was activated in the developing cartilage. Indeed, the expression of Hh signaling transcriptional targets such as Gli1 and Hip1 was activated in both chondrocytes and perichondrium, with the strongest sites of activation observed in the perichondrium and joint region(Fig. 3B). These results demonstrate that the Ptch1 allele was inactivated by Cre-mediated recombination, which led to Hh signaling activation in the Ptch1^c/-; Col2a1-Cre embryos. As a result of Hh signaling activation, we found that mineralized periosteum extended ectopically to the joint region, where mineralization was never found in wild-type controls(Fig. 2E,F). Mild joint fusion was also observed in the Ptch1^c/-; Col2a1-Cre embryo(Fig. 2G,H), which is consistent with previous observation that Shh overexpression in the cartilage resulted in joint fusion(Tavella et al., 2004). These results suggest that, to maintain joint integrity, Ihh signaling in the joint needs to be kept sufficiently low not to allow ectopic osteoblast differentiation and cartilage fusions.

Ectopic osteoblast differentiation in the perichondrium of Ptch1^c/-; Col2a1-Cre embryos allowed us to directly test whether Ihh signaling requires canonical Wnt signaling for inducing osteoblast differentiation, by simultaneously blocking canonical Wnt signaling and activating Ihh signaling cell autonomously. We generated Ptch1^c/-; Catnb^c/-; Col2a1-Cre double conditional mutant embryos in which upregulation of Ihh signaling occurred in Col2a1-expressing cells and their descendants, where canonical Wnt signaling was abolished. We found that ossification and mineralization, as indicated by Alizarin red staining, were more severely inhibited in Ptch1^c/-; Catnb^c/-; Col2a1-Cre mutant embryos than in Catnb^c/-; Col2a1-Cre single mutant embryos(Fig. 3A). This was due to the blockage of both osteoblast differentiation and chondrocyte hypertrophy in Ptch1^c/-; Catnb^c/-; Col2a1-Cre double conditional mutant embryos. In the limb sections, mature osteoblasts,indicated by von Kossa staining, were almost completely missing, except for in a very small spot in the Ptch1^c/-; Catnb^c/-;Col2a1-Cre double mutant embryo at 16.5 days post-coitum (dpc; Fig. 4B). However, when the activity of Hh signaling was assessed by examining expression of the hedgehog transcriptional targets Gli1 and Hip1, both were still ectopically activated in the perichondrium and the joint in Ptch1^c/-; Catnb^c/-; Col2a1-Cre double mutant embryos at similar levels to those in Ptch1^c/-; Col2a1-Cresingle mutants, although the perichondrium in the Ptch1^c/-;Catnb^c/-; Col2a1-Cre double mutant was slightly thinner(Fig. 3B). These results indicate that Hh target gene expression in endochondral bone development does not require β-catenin and suggest that β-catenin may be required for perichondrial cell proliferation. Consistent with this, we found that the expression of Pthrp, a cartilage-specific transcriptional target of Ihh signaling, was similarly upregulated(Fig. 3B), and chondrocyte hypertrophy was inhibited, as shown by the lack of ColX expression that marks hypertrophic chondrocytes (Fig. 4A,B), in both Ptch1^c/-; Catnb^c/-;Col2a1-Cre and Ptch1^c/-; Col2a1-Cre mutant embryos. Taken together, our results show that Wnt/β-catenin signaling is required for osteoblast differentiation downstream of, or in parallel to, Ihh signaling.

We found in our previous studies that the expression of Runx2 and Osx, two early osteoblast differentiation markers, was present, but at a much reduced level, in the perichondrium of Catnb^c/-;Col2a1-Cre mouse embryos (Day et al.,2005). It was not clear whether the residual expression of Runx2 and Osx was a direct consequence of residualβ-catenin-positive cells caused by incomplete recombination, or an indirect result of reduced Ihh signaling and thinner perichondrium caused by reduced Ihh expression in the Catnb^c/-;Col2a1-Cre embryo (Day et al.,2005), as Ihh is required for cell proliferation and the expression of Runx2 and Osx(Karp et al., 2000; Long et al., 2004). To clarify this issue, we took advantage of the thicker perichondium and the ligand-independent cell autonomous upregulation of Hh signaling in Ptch1^c/-; Col2a1-Cre mouse embryos and investigated whether the removal of Wnt signaling in these embryos would affect Runx2 and Osx expression. In the Ptch1^c/-;Col2a1-Cre mouse embryo at 14.5 dpc, we found that Runx2expression was significantly upregulated throughout the thickened perichondrium, including the joint region, in a similar pattern to Gli1 and Hip1 expression(Fig. 4A). Importantly, in the Ptch1^c/-; Catnb^c/-; Col2a1-Cre double mutant embryo, we still observed strong ectopic expression of Runx2 and Osx in the perichondrium, although the perichondrium was slightly thinner (Fig. 4A). The persistent and strong expression of Runx2 and Osx cannot be due to β-catenin expression in the perichondrium because ossification,which requires β-catenin, was almost completely inhibited in the Ptch1^c/-; Catnb^c/-; Col2a1-Cre double mutant(Fig. 3A, Fig. 4B). These results indicate that like Gli1 and Hip1 expression, Runx2expression may be directly controlled by Ihh signaling. Furthermore,β-catenin is not required for the expression of Runx2 and Osx, two downstream targets of Ihh signaling, in early osteoblasts in the developing long bones. Our results demonstrate that β-catenin is required downstream of Ihh signaling and Osx expression during early osteoblast differentiation.

Interestingly, osteoblast maturation, indicated by von Kossa staining and the expression of mature osteoblast markers osteopontin (Opn; Spp1 - Mouse Genome Informatics) and matrix metalloproteinase 13(Mmp13), was not detected in the Ptch1^c/-;Col2a1-Cre mouse embryo at 14.5 dpc when ossification had just started in the wild-type embryo, although the expression of Runx2 and Osx was strongly activated (Fig. 4A). Osteoblast maturation caught up later at 16.5 dpc(Fig. 4B). Furthermore, in the humerus of the Ptch1^c/-; Col2a1-Cre embryo, chondrocyte hypertrophy indicated by ColX expression was missing, and yet osteoblast differentiation occurred at 16.5 dpc(Fig. 4B). Therefore, activated Ihh signaling is sufficient to induce osteoblast differentiation and maturation without chondrocyte hypertrophy. Because Ihh is also required for Runx2 expression and osteoblast differentiation(St-Jacques et al., 1999), our results further indicate that during normal endochondral bone formation, the coupling of chondrocyte hypertrophy with osteoblast differentiation and maturation is mediated by Ihh signaling. The slight delay in osteoblast maturation in the Ptch1^c/-; Col2a1-Cre embryo is likely to be due to the severely delayed chondrocyte hypertrophy(Fig. 4A,B), because factors such as Vegf and Mmps are expressed in hypertrophic chondrocytes and are required to promote osteoblast maturation(Inada et al., 2004; Stickens et al., 2004; Zelzer et al., 2002). We also observed that the Mmp9 expression was ectopically activated in both the Ptch1^c/-; Col2a1-Cre and Ptch1^c/-;Catnb^c/-; Col2a1-Cre double mutant mouse embryos in a pattern similar to that of Gli1 and Hip1 expression(Fig. 4A), suggesting that Mmp9, but not Mmp13, expression may also be directly controlled by Hh signaling. In contrast to the short delay in osteoblast maturation in the Ptch1^c/-; Col2a1-Cre mutant, osteoblast differentiation was severely blocked in the Ptch1^c/-;Catnb^c/-; Col2a1-Cre double mutant in which there was almost no von Kossa staining or expression of Opn and Mmp13 at 16.5 dpc, although the expression of Osx was still stronger than that in the Catnb^c/-; Col2a1-Cre mutant(Fig. 4B). These results confirm that β-catenin is required downstream of Osx for the progression of osteoblast maturation.

As Ihh and canonical Wnt signaling both control chondrocyte hypertrophy, we tested whether they act in the same pathway. In contrast to the humerus, where chondrocyte hypertrophy was not detected in the Ptch1^c/-;Col2a1-Cre embryos, chondrocyte hypertrophy did occur in the radius, ulna and tibia, which allowed us to compare the effects Ihh and Wnt/β-catenin signaling in regulating chondrocyte hypertrophy. The differential inhibition of chondrocyte hypertrophy in different long bones in the Ptch1^c/-; Col2a1-Cre embryos was not due to different Cre activities, as Hh target genes were expressed at similar levels in the long bones (data not shown). In the tibia, the expression domains of Ihhand ColX, which mark prehypertrophic and hypertrophic chondrocytes,respectively, were smaller in both Ptch1^c/-; Col2a1-Creand Catnb^c/-; Col2a1-Cre mutants than in the wild-type control (Fig. 5), indicating that chondrocyte hypertrophy was delayed in each single mutant. In the Ptch1^c/-; Catnb^c/-; Col2a1-Cre double mutant, Pthrp expression was similarly upregulated when compared with that in the Ptch1^c/-; Col2a1-Cre mutant, but chondrocyte hypertrophy was more delayed than that in either of the single mutants,because Ihh and ColX expression was completely missing(Fig. 5). These results reveal that Ihh signaling inhibits chondrocyte hypertrophy independently of canonical Wnt signaling.

Because the regulation of chondrocyte hypertrophy is coupled with proliferation in that chondrocyte hypertrophy is delayed when proliferation is prolonged by removing cell cycle inhibitors(Scheijen et al., 2003; Yan et al., 1997; Zhang et al., 1997; Cobrinik et al., 1996), we examined the status of chondrocyte proliferation by BrdU labeling in Ptch1 and β-catenin mutant embryos. In the developing cartilage,chondrocytes are organized into two zones with distinct proliferation rates before hypertrophy (Long et al.,2001; Yang et al.,2003). The resting zone chondrocytes are located at the articular end and proliferate slowly. The proliferative zone is composed of fast proliferating chondrocytes, and is located between the resting zone and the prehypertrophic region where Ihh is expressed(Fig. 6A). We found that, as expected, in the Ptch1^c/-; Col2a1-Cre embryo, chondrocyte proliferation in the presumed resting zone was significantly increased to a level similar to that in the proliferative zone by activated Hh signaling(Fig. 6A,B). In contrast to a previous study (Akiyama et al.,2004), we did not detect a reduction in chondrocyte proliferation in the Catnb^c/-; Col2a1-Cre mouse embryo(Fig. 6B). Interestingly,chondrocyte proliferation in Ptch1^c/-; Catnb^c/-;Col2a1-Cre double mutant embryos resembles that in the Catnb^c/-; Col2a1-Cre mouse embryo(Fig. 6A,B), particularly in the resting zone, indicating that Wnt/β-catenin signaling may be required for the activity of Hh signaling in promoting cell proliferation in resting zone chondrocytes. As both Hh and Wnt/β-catenin signaling can regulate cell proliferation through regulating cyclin D1 expression(Long et al., 2001; Tetsu and McCormick, 1999), we next examined the expression of cyclin D1, and p21, p27 and p57, members of the Cip/Kip family of cyclin-cyclin dependent kinase (CDK) inhibitors, in mutant embryos at 14.5 dpc.

Consistent with the status of cell proliferation revealed by BrdU labeling,cyclin D1 expression is normally expressed highly in proliferative chondrocytes and we found that its expression was greatly upregulated in the presumed resting chondrocytes in the Ptch1^c/-; Col2a1-Cremutant embryo (Fig. 7A). In addition, we did not observe a reduction in cyclin D1 expression in the Catnb^c/-; Col2a1-Cre mutant embryo, and the cyclin D1 expression pattern was similar in Catnb^c/-; Col2a1-Cre and Ptch1^c/-; Catnb^c/-; Col2a1-Cre mutant embryos. These results suggest that although β-catenin is required for cell proliferation and cyclin D1 expression downstream of Ihh signaling in the resting zone, there may be a β-catenin-independent pathway in the proliferative zone to promote cyclin D1 expression and cell proliferation. P27 was expressed in chondrocytes of all zones and its expression was not altered by Hh or Wnt/β-catenin signaling (Fig. 7A). P57 is strongly expressed in prehypertrophic chondrocytes and is required to promote chondrocyte hypertrophy(Yan et al., 1997; Zhang et al., 1997). The expression pattern of p57 was not altered in the Catnb^c/-;Col2a1-Cre mutant embryo. But in the Ptch1^c/-;Col2a1-Cre and especially the Ptch1^c/-;Catnb^c/-; Col2a1-Cre mutant embryo, strong expression of p57 was detected throughout the cartilage even though chondrocyte hypertrophy was significantly delayed (Fig. 7A). Interestingly, p21 was expressed almost exclusively in prehypertrophic chondrocytes in the wild-type embryo. Only very few cells in the proliferative zone expressed p21 at both 14.5 and 18.5 dpc(Fig. 7A,B). In the Ptch1^c/-; Col2a1-Cre double mutant embryo, few chondrocytes expressed p21, similar to what was observed in the proliferative region of the wild-type cartilage at 14.5 and 18.5 dpc(Fig. 7A,B). In the Catnb^c/-; Col2a1-Cre mouse embryo, slightly more cells in the resting and proliferative zones expressed p21 at 14.5 dpc(Fig. 7A). Strikingly, at 18.5 dpc, significantly more cells in the proliferative and resting zones of the Catnb^c/-; Col2a1-Cre embryo expressed p21(Fig. 7B). But in the double mutant embryo, p21 expression was reduced to a level similar to that in the Ptch1^c/-; Col2a1-Cre embryo at both 14.5 and 18.5 dpc(Fig. 7A,B), indicating that p21 expression is inhibited by both the Wnt/β-catenin and Hh signaling pathways in proliferative and resting chondrocytes, and that Hh signaling acts downstream of Wnt/β-catenin signaling. It should be noted here that, in contrast to cyclin D1 expression, changes in p57 and p21 expression did not correlate with those of chondrocyte proliferation in the mutants. p57 or p21 expressing cells were increased in either Ptch1^c/-;Col2a1-Cre or Catnb^c/-; Col2a1-Cre embryos,respectively, but there was no corresponding reduction in cell proliferation(Figs 6, 7). As p21 acts, not only as an inhibitor for cell proliferation, but also as a regulator of cell death with both anti- and pro-apoptotic function depending on the cellular context(Gartel and Radhakrishnan,2005; Tsao et al.,1999; Wang et al.,2005; Yamamoto and Nishioka,2005), we examined cell death by TUNEL assay. In the wild-type embryo, cell death was mostly observed in the joint and the most mature hypertrophic chondrocytes at 18.5 dpc. Cell death was not evident in the proliferative and resting chondrocytes(Fig. 7B). By contrast, in the Catnb^c/-; Col2a1-Cre embryo at 18.5 dpc, extensive cell death was detected in both proliferative and resting chondrocytes, and this cell death phenotype was completely suppressed in the Ptch1^c/-;Catnb^c/-; Col2a1-Cre embryo(Fig. 7B). These data demonstrate that Wnt/β-catenin acts upstream of the Hh signaling pathway in promoting chondrocyte survival, and further suggest that p57 and p21 may act downstream of canonical Wnt and Hh signaling in this process.

The synovial joint fusions in Ptch1^c/-; Col2a1-Creembryos (Fig. 2H) prompted us to investigate whether Ihh signaling is activated during normal synovial joint formation by examining the expression of Hip1, a transcriptional target of Hh signaling, at 14.5 dpc. We found that Hip1 was expressed strongly in the joint region (Fig. 8A), as well as Gli1 and Ptch1 (data not shown). These results indicate that Hh signals to the joint cells directly and that Hh signaling may need to be kept at a low level to allow normal joint formation. The joint cells may be more responsive to Hh signaling, as the Hh signaling transcriptional targets Gli1 and Hip1 were activated more strongly in the joint region than in the neighboring chondrocytes in Ptch1^c/-; Col2a1-Cre embryos(Fig. 3B). Interestingly, we found extensive synovial joint fusions in the Ptch1^c/-;Catnb^c/-; Col2a1-Cre double mutant, which were much more severe than each of the single mutants that only exhibited mild fusions in some joints (Fig. 2H) (see also Guo et al., 2004). For instance, the elbow joint was normal in both Ptch1^c/-;Col2a1-Cre and Catnb^c/-; Col2a1-Cre single mutant embryos, but was completely fused in the double mutant(Fig. 8B). To test whether the joint ever forms in the double mutant, we examined the earliest elbow joint development at 12.5 dpc when the elbow joint interzone has just formed (see Fig. S1 in the supplementary material). We found that the elbow joint interzone failed to form only in the double mutant embryos, based on both morphological and molecular criteria (see Fig. S1 in the supplementary material). These results indicate that the inhibition of canonical Wnt signaling was strongly enhanced by the activation of Ihh signaling, thus causing synovial joint fusion.

Consistent with previous findings in Col2a1-Gal4; UAS-Ihh mouse embryos (Minina et al., 2002),we observed that the expression of Bmp7, and to a lesser extent, Bmp2 and Bmp4, but not the joint marker Gdf5, was activated in chondrocytes, perchondrium and the forming joint region in both Ptch1^c/-; Col2a1-Cre and Ptch1^c/-;Catnb^c/-; Col2a1-Cre mutant embryos(Fig. 8C). Because Bmp signaling acts to inhibit synovial joint formation(Brunet et al., 1998; Tsumaki et al., 2002; Zou et al., 1997), it is likely that the Bmp genes mediate the activity of Ihh signaling in regulating synovial joint formation and that canonical Wnt signaling interacts with Ihh signaling indirectly through Bmps. We tested this possibility directly by culturing the limb buds from 12.5 dpc wild-type and Ptch1^c/-;Col2a1-Cre mutant embryos. Bmp signaling was inhibited in the culture by the addition of noggin, a Bmp inhibitor. The humerus-scapula joint was fused in the Ptch1^c/-; Col2a1-Cre mutant embryos in vivo and in vitro, and cells in the presumptive joint area maintained chondrocyte-specific Safranin O staining. Importantly, in the noggin-treated Ptch1^c/-; Col2a1-Cre embryos (n=3), many cells in the humerus-scapula joint were elongated and had much lighter Safranin O staining, thereby resembling those joint cells found in the wild-type embryos(Fig. 8D). Furthermore, we found that the expression of the joint marker Gdf5 was lost in the humerus-scapula joint of the Ptch1^c/-; Col2a1-Cre mutant,but was significantly rescued by noggin treatment (see Fig. S2 in the supplementary material), demonstrating that joint formation was rescued in the Ptch1^c/-; Col2a1-Cre mutant by blocking Bmp signaling. Noggin did not affect joint formation in the wild-type limbs (n=3; Fig. 8D). These results demonstrate that joint fusion in Ptch1^c/-; Col2a1-Cremutant embryos is caused by upregulated Bmp signaling.

To further understand the synergistic interaction between Bmp signaling and loss of β-catenin in joint fusion, we performed micromass cultures to mimic the in vivo chondrogenesis in vitro, as joint fusion is caused by enhanced chondrocyte differentiation. We found that Bmp2 could further promote cartilage nodule formation already enhanced by inactivating β-catenin(Fig. 8E), indicating that Bmps can promote chondrocyte differentiation independently of canonical Wnt signaling. Furthermore, we found that inactivation of β-catenin was sufficient for extensive chondrocyte differentiation even in the presence of noggin, which almost completely inhibited chondrocyte differentiation in wild-type progenitor cells (Fig. 8E). These results suggest that when the canonical Wnt pathway is inhibited, a much lower level of Bmp signaling is needed for chondrocyte differentiation. It is possible that the canonical Wnt and Bmp signaling pathways act in parallel with opposite activities in regulating the expression of a common target, such as Sox9, that is required for chondrocyte differentiation.

Here, we report that the canonical Wnt and Ihh signaling pathways interact with each other in a cell context-dependent manner in regulating different aspects of endochondral skeletal development, including osteoblast differentiation, chondrocyte proliferation, hypertrophy, survival and synovial joint formation. Different relationships between the two signaling pathways in these developmental processes reveal the unique genetic hierarchy in each developmental process and highlight the distinct contributions of each signaling cascade.

Canonical Wnt signaling is required downstream of Ihh signaling and Osx in regulating osteoblast differentiation during endochondral ossification Previous studies have uncovered the requirement of the Ihh and canonical Wnt signaling pathways in osteoblast differentiation during endochondral bone formation (Day et al.,2005; Hill et al.,2005; Hu et al.,2005; Long et al.,2004). In this study, by activating the Hh signaling pathway while inactivating β-catenin simultaneously in the same cells in Ptch1^c/-; Catnb^c/-; Col2a1-Cre double mutant embryos, we clarified two issues that were not settled before. First, the genetic relationship between Ihh signaling and β-catenin in promoting osteoblast differentiation was not clearly demonstrated in vivo. It has been suggested that β-catenin may act downstream of Ihh signaling based on correlations of gene expression. In particular, expression of Tcf1and Wnt7b was suggested as a downstream target of Ihh signaling in activating the canonical Wnt signaling (Hu et al., 2005). However, the lack of Tcf1 or Wnt7b expression in the perichondrium of Ihh^-/-embryos could be a consequence of a loss of the right cell types in which Tcf1 and Wnt7b are normally expressed. Second, it was not clear whether β-catenin could act downstream of Osx in controlling osteoblast differentiation, as Osx expression was reduced, but not abolished, in Catnb^c/-; Dermo1-Cre and Catnb^c/-; Col2a1-Cre mutant embryos(Day et al., 2005) (this study). The reduced expression of Osx could be a result of reduced Ihh signaling and thinner perichondrium in the mutant embryos. In this study,the thickness of perichondrium, the expression of both Runx2 and Osx, and the final bone formation were greatly increased by activating Hh signaling in the Ptch1^c/-; Col2a1-Cremutant. Yet, further removal of β-catenin in the Ptch1^c/-;Catnb^c/-; Col2a1-Cre double mutant reduced bone formation to the same degree as that in the Catnb^c/-; Col2a1-Cremutant, despite the fact that expression of Runx2 and Osx in the perichondrium of the double mutant was similar to that in the wild type and stronger than that in the Catnb^c/-; Col2a1-Cre mutant. Therefore, Osx-expressing cells fail to give rise to bone-forming mature osteoblasts in the absence of β-catenin. β-catenin is required not only downstream of the Ihh signaling pathway, but further downstream of Osx in osteoblast maturation. Our results, together with previous studies, indicate that Runx2 is a transcriptional target of Ihh signaling. As Osx is downstream of Runx2,β-catenin is required at least three steps downstream of Ihh signaling. Because ectopic chondrocyte differentiation occurred in the periosteum of the developing long bones of the Catnb^c/-; Col2a1-Cre and Osx^-/- embryo (Day et al., 2005; Nakashima et al.,2002), it is likely that loss of β-catenin not only blocked osteoblast differentiation, but also caused Osx-expressing cells to lose Osx expression later and become chondrocytes. Consistent with this, we found that in the absence of β-catenin, Osx expression was much weaker in older Catnb^c/-; Col2a1-Cre embryos than in wild type (Fig. 4A,B),suggesting that β-catenin may even be required for the maintenance of Osx expression.

During normal long bone development, mature ossification is initiated from the middle of the cartilage. This is thought to be a result of the chondrocyte hypertrophy that is required for ossification and which occurs in the middle segment first (reviewed by Kronenberg,2003). In the Ptch1^c/-; Col2a1-Cre mouse embryo, hedgehog signaling was activated cell autonomously in all chondrocytes and throughout the perichondrium independently of Ihh ligand and chondrocyte hypertrophy. However, mature ossification was still initiated from the middle of the long bones, although ectopic ossification did occur throughout the perichondrium, including the joint, eventually. The directional ossification along the longitudinal axis cannot be a consequence of localized chondrocyte hypertrophy as, in the developing humerus of the Ptch1^c/-;Col2a1-Cre mouse embryo, chondrocyte hypertrophy was inhibited by upregulated Pthrp expression (Fig. 4). The directional ossification was not a result of a differential response to Hh or Bmp signaling along the longitudinal axis of the developing long bones either, as we found that the expression of Hh signaling targets, Runx2 and Bmps was upregulated strongly throughout the perichondrium, including the joint region. The directional osteoblast maturation might be caused by some unidentified inhibitor(s) of osteoblast differentiation localized in the joint region. The earliest sign of directional ossification in the Ptch1^c/-; Col2a1-Cremutant is the expression of Osx that was expressed at the strongest level in the middle segment (Fig. 4A). Because osteoblast maturation eventually occurred in the joint, we favor another explanation: that the directional osteoblast maturation may simply reflect the time required by the Runx2expressing cells to proceed to the next stages of osteoblast maturation,including activating the expression of Osx. In the Col2a1-Cre mice, Cre is expressed in early osteochondral progenitor cells and chondrocytes, and not in late perichondral cells in the middle segment of the long bones. However, the effects of Cre-mediated recombination, which is inactivation of Ptch1, are inherited in all descendants of the early Cre expressing cells. As the development of chondrocytes and osteoblasts proceed from the joint to the middle segment of long bones, many osteoblasts in the middle segment are descendants of the earlier Col2a1-Cre expressing cells, and they have been activated by Hh signaling for a longer period of time.

We observed that the phenotype of delayed chondrocyte hypertrophy in the Ptch1^c/-; Col2a1-Cre mutant embryo was enhanced by loss ofβ-catenin. As canonical Wnt signaling promotes chondrocyte hypertrophy,and Ihh signaling regulates chondrocyte hypertrophy mainly indirectly through activating and maintaining the expression of Pthrp, the canonical Wnt signaling pathway may also antagonize Pthrp signaling in controlling chondrocyte hypertrophy. It is likely that the canonical Wnt and Pthrp signaling pathways exert opposite activities in regulating common targets required for the execution of chondrocyte hypertrophy. The common targets may be Runx2 or Sox9, or both, as they are both required to control chondrocyte hypertrophy (Bi et al.,2001; Enomoto et al.,2000; Kim et al.,1999). Consistent with the role of Runx2 in promoting chondrocyte hypertrophy, Runx2 expression was reduced in chondrocytes in Ptch1^c/-; Col2a1-Cre embryos, although its expression was upregulated in the perichondrium (Fig. 4A). But a reduction of Runx2 expression in chondrocytes was not obvious in the Catnb^c/-; Col2a1-Cre mouse(Fig. 4A). It is also possible that the reduction of Runx2 expression in Ptch1^c/-;Col2a1-Cre cartilage is only secondary to the initial delay of chondrocyte maturation caused by a Runx2-independent mechanism. This is supported by a recent finding that Pthrp, a major target of Hh signaling in controlling chondrocyte hypertrophy, regulates chondrocyte hypertrophy through both Runx2-dependent and -independent pathways(Guo et al., 2006). In contrast to Runx2, Sox9 inhibits chondrocyte hypertrophy(Bi et al., 2001). Loss ofβ-catenin increases Sox9 activity in chondrocytes(Akiyama et al., 2004), whereas Pthrp potentiates Sox9 activity by activating PKA, which phosphorylates Sox9(Huang et al., 2001). As Sox9 is expressed in all Col2a1 expressing chondrocytes at all developmental stages, the regulation of Sox9 activity is at least one of the underlying mechanisms for the opposite activities of canonical Wnt and Pthrp signaling in regulating chondrocyte hypertrophy.

Our results indicate that both the canonical Wnt and Hh signaling pathways regulate chondrocyte survival. It has been demonstrated that the oxygen-sensitive hypoxia-inducible factor 1 alpha (Hif1α) and Vegfa are required for chondrocyte survival and the regulation of chondrocyte proliferation in cartilage, which is an avascular tissue(Schipani et al., 2001; Zelzer et al., 2004). The phenotypes of Catnb^c/-; Col2a1-Cre embryos bear some similarities to those of Hif1a and Vegfa cartilage mutants in chondrocyte survival, although they are not as severe. It will be interesting to further investigate whether the canonical Wnt and Hh signaling pathways interact with the Hif1α and Vegfa pathway in controlling chondrocyte survival.

We have previously shown that inactivation of β-catenin promotes chondrocyte differentiation (Day et al.,2005; Guo et al.,2004). Our finding here that β-catenin inactivation also resulted in a dramatically increased cell death in differentiated chondrocytes further indicates that just inhibiting Wnt/β-catenin signaling alone is not sufficient to maintain healthy cartilage. The rescue of chondrocyte apoptosis and the inhibition of chondrocyte hypertrophy by Hh signaling in Ptch1^c/-; Catnb^c/-; Col2a1-Cre mouse embryos is significant from the therapeutic standpoint: to maintain healthy joint cartilage or to treat degenerating joint cartilage, inhibiting canonical Wnt signaling needs to be combined with the manipulation of other signaling pathways, such as activating Hh signaling or perhaps its downstream Pthrp signaling.

We thank members of the Yang lab for stimulating discussion during the work. Work in the Yang lab is supported by the intramural research program of NIH, National Human Genome Research Institute. Work in the Chuang lab was supported by NIH grant HL67822 and NHLBI-funded Program for Genomics Applications ('BayGenomics') HL66600. We also thank Ms Julia Fekecs for help in preparing the figures.