Ihh controls cartilage development by antagonizing Gli3, but requires additional effectors to regulate osteoblast and vascular development

Indian hedgehog (Ihh) signaling is indispensable for proper development of the endochondral skeleton. In the developing cartilage, Ihh is primarily expressed by prehypertrophic chondrocytes (chondrocytes immediately prior to hypertrophy) as well as in early hypertrophic chondrocytes, and Ihh signals to both immature chondrocytes and the overlying perichondrial cells(St-Jacques et al., 1999; Vortkamp et al., 1996). Ihh^-/- mice exhibited multiple defects in development of the endochondral skeleton, these including a 50% reduction in chondrocyte proliferation, profound dysregulation of chondrocyte maturation, a severe delay in cartilage vascularization and a complete lack of osteoblasts(Long et al., 2001; St-Jacques et al., 1999). Subsequent genetic analyses of smoothened (Smo), which encodes a seven-pass transmembrane protein indispensable for all Hh signaling, revealed that direct Ihh input was required for proper proliferation of chondrocytes and development of the osteoblast lineage(Long et al., 2004; Long et al., 2001). Regulation of chondrocyte maturation, however, did not require direct Ihh signaling in chondrocytes as it appeared to be mediated primarily via a secondary signal parathyroid hormone-related peptide (Pthlh) whose expression was dependent upon Ihh signaling (Karp et al.,2000; Long et al.,2001; St-Jacques et al.,1999). Regulation of cartilage vascularization by Ihh is less well understood. Mouse embryos with Smo conditionally removed by Col2-Cre exhibit normal vascularization(Long et al., 2001),indicating that Ihh controls vascularization independent of its signaling in chondrocytes. In chimeric embryos developed from wild-type and Smo^-/- ES cells, Smo^-/- cells contribute to the endothelium of blood vessels in the bone marrow cavity(Long et al., 2004), arguing against a requirement for direct Ihh signaling in endothelial cells for cartilage vascularization.

The role of vascularization in osteoblast formation during endochondral skeletal development is not well understood. Cartilage anlagen that prefigure the long bones are initially avascular; nascent vascularization coincides with osteoblast differentiation in the perichondrium. Ihh^-/-mice exhibited profound defects in cartilage vascularization, and thus have not been informative in previous analyses for assessing the role of vascularization in osteoblast development in the endochondral skeleton. However, the skull bones did form in the Ihh^-/- embryo(St-Jacques et al., 1999),indicating the presence of an Ihh-independent osteogenic pathway in the intramembranous bones.

Hh signaling is mediated through transcriptional regulation by the zinc-finger transcription factors Gli1, Gli2 and Gli3 in mammals, or Cubitus interruptus (Ci) in Drosophila. Mutant studies in the mouse demonstrated crucial roles for Gli2 and Gli3 in the developing embryo, whereas Gli1 appeared to be dispensable (Bai et al., 2002; Ding et al.,1998; Matise et al.,1998; Mo et al.,1997; Motoyama et al.,1998; Park et al.,2000), although additional phenotypes were noted when Gli1 was deleted in the Gli2^-/- background(Park et al., 2000). In particular, Gli2 appeared to function in vivo predominantly as a transcriptional activator both in the central nervous system(Ding et al., 1998; Komori et al., 1997; Matise et al., 1998; Sasaki et al., 1999) and in the developing hair follicle (Mill et al.,2003). Gli3, however, was shown to function primarily as a transcription repressor of Hh targets. Removal of both Gli3 and Shh in the mouse rescued a multitude of defects in specification of the ventral neuron types, patterning of the limb and lung organogenesis caused by the loss of Shh alone (Li et al., 2004; Litingtung and Chiang, 2000; Litingtung et al., 2002). However, additional studies have also revealed an activator function for Gli3 in both the somite and the spinal cord(Bai et al., 2004; Buttitta et al., 2003), and a repressor activity for Gli2 in the somite(Buttitta et al., 2003). These studies underline the diverse roles of the Gli molecules in mediating Hh signaling in different physiological contexts.

To determine the potential roles of Gli3 in mediating Ihh signaling during skeletal development, we have generated double mutant embryos of Ihh^-/-; Gli3^-/-, and have analyzed the skeletal phenotype. The results demonstrate that derepression of Gli3 accounts for multiple, but not all, aspects of Ihh signaling in the endochondral skeleton.

The Ihh^+/- and Gli3^Xt-J mouse strains are as previously described(Maynard et al., 2002; St-Jacques et al., 1999). The Gli3^Xt-J strain was acquired through the Jackson Laboratory (Bar Harbor, Maine). The wild-type and deletion alleles of Gli3 were detected by PCR as previously described(Maynard et al., 2002). Embryos of Dermo1-Cre; β-catenin^c/c were generated as previously described (Hu et al., 2005).

Whole-mount skeletal staining of embryos was based on McLeod(McLeod, 1980). For analyses on sections, embryonic tissues were collected in PBS, fixed in 10% formalin overnight at room temperature, then processed and embedded in paraffin prior to sectioning at 6 μm. For detection of mineralization, sections were stained with 1% silver nitrate (von Kossa method) and counterstained with Methyl Green. For histology or in situ hybridization on E17.5 and E18.5 embryos, limbs were decalcified in 14% EDTA in PBS (pH 7.4) for 48 hours after fixation and prior to processing. In situ hybridization was performed as described previously by using ³⁵S-labeled riboprobes(Long et al., 2001). All in situ probes were as previously described(Hu et al., 2005; Long et al., 2004; Long et al., 2001), with the exception that the Pthlh probe was derived from a full-length cDNA clone (ATCC, catalog no. 7061403). For BrdU analysis, pregnant females were injected with BrdU at 0.1 mg/g body weight at 2 hours prior to harvest. Embryos were collected in ice-cold PBS, processed and sectioned as above. BrdU detection was performed using a kit from Zymed Laboratories (South San Francisco, CA) as per instructions. Labeling index was scored for a least four sections from various planes of section through the cartilage. Multiple wild-type and mutant pairs of littermates were analyzed and results from a representative pair are reported here.

For western analyses of Gli3, whole-cell lysates were prepared using a RIPA buffer [20 mM Tris (pH 8.0), 1% NP-40, 0.25% sodium deoxycholate, 150 mM NaCl,0.1% SDS, 1 mM EGTA, 1 mM EDTA, protease inhibitor cocktail (Roche),phosphatase inhibitor cocktails 1 and 2 (Sigma, St Louis, MO)] from the fore-and hindlimb cartilage elements of E14.5 embryos. Cartilage was pulverized in liquid nitrogen using a mini-pestle (Clontech) prior to protein extraction. Total cell lysate (100 μg) was separated on a SDS-PAGE gel and a specific antibody (Li et al., 2004) was used to detect Gli3 by the ECL method (Amersham Biosciences). An antibody against α-catenin (Santa Cruz Biotechnology) was used as a loading control.

We generated double mutant embryos of Ihh^-/-;Gli3^-/- (DKO) by mating double heterozygous(Ihh^+/-; Gli3^+/-) mice. The Gli3^+/- strain obtained from the Jackson Laboratory(Gli3^Xt-J) contains a deletion of 51.5 kb at the Gli3 allele, which results in a non-functional fusion transcript that lacks the coding sequence for the DNA binding motif(Buscher et al., 1998; Maynard et al., 2002). Of 358 total embryos harvested between E13.5 and E18.5, 11 embryos were identified as DKO (3%). The reasons for the low recovery rate (expected Mendelian rate 6%)were not further investigated. The DKO embryos that developed to term all died at birth.

Whole-mount skeletal preparations from E18.5 embryos revealed that most skeletal elements of the DKO embryos exhibited an intermediate length between wild-type and Ihh^-/- littermates(Fig. 1). This was most evident with the forelimb skeleton whose overall length was markedly increased in the DKO (Fig. 1A3) over the Ihh^-/- embryo (Fig. 1A5), although the hindlimb skeleton showed a more modest increase in length (Fig. 1, compare B3 with B5). Notably, both limb skeletons of DKO embryos remained considerably smaller than those of either the wild-type(Fig. 1A1,B1) or the Gli3^-/- embryos (Fig. 1A2,B2). The increase over Ihh^-/- mutants in overall size of the skeleton was not evident in embryos of Ihh^-/-; Gli3^+/- (I^-/-;G^+/-) in which one copy of Gli3 remained intact(Fig. 1, compare A4 with A5, B4 with B5). In addition to a slight reduction in the length of long bones(Fig. 1, compare A1 with A2, B1 with B2), the Gli3^-/- embryos exhibited a number of characteristic patterning defects in the limbs, including the absence of the deltoid tuberosity (Fig. 1A1,asterisk) in the humerus and a varying degree of truncation of the distal tibia (Fig. 1B2, big arrowhead). In addition, the Gli3^-/- embryos invariably exhibited bifurcation of phalange 1 (P1) of all digits(Fig. 1B2, small arrowhead, out of plane of focus in A2) as well as polydactyly in both the forelimb(Fig. 1A2, digits 1-7) and the hindlimb (Fig. 1B2, digits 1-6,2^* representing syndactyly at digit 2). Importantly, all defects characteristic of the Gli3^-/- embryo were maintained in the DKO embryo, including polydactyly, bifurcation of P1(Fig. 1A3,B3, small arrowheads)and truncation of the tibia (Fig. 1B3, big arrowhead). These results indicate that Gli3 functions antagonistically with and genetically downstream of Ihh to regulate linear growth of the long bones.

Examination of the whole-mount stained skeletons at E18.5 also revealed that the bone collar was absent in most long bones of the DKO embryo. As in the Ihh^-/- mutant (Fig. 1A5,B5), the metacarpal (`mc'), metatarsal (`mt') and phalanges(`p') of the DKO embryo exhibited little or no mineralization(Fig. 1A3,B3). Also similar to those in the Ihh<sup>-/-</sup> mutant(Fig. 1A5,B5), the radius (`r')(Fig. 1A3) as well as the fibula (`fi') (Fig. 1B3)contained mineralized cartilage (stained dark red) but no bone collar. Interestingly, the ulna and the more proximal elements such as the humerus(`h') (Fig. 1A3) and the femur(`f') (Fig. 1B3) in the DKO embryo contained some abnormal bone (green arrowheads in insets) around the mineralized cartilage at the diaphysis of these elements. Overall, bone formation remained largely dysregulated in the DKO embryo.

The DKO embryo exhibited additional rescue phenotypes in the rest of the skeleton. The rib cage that was characteristically diminished in the Ihh^-/- embryo (Fig. 1C2) was restored in the DKO embryo(Fig. 1C3) to a similar size to wild type (Fig. 1C1). Notably,the sternum of the DKO embryo was not fused (asterisk, Fig. 1C3), a feature characteristic of the Gli3^-/- mutant(Fig. 1C4) but not that of the Ihh^-/- embryo (Fig. 1C2). Moreover, the Ihh^-/- mutant typically exhibited a `dome-shaped' skull, and a shortened snout(Fig. 1D2), and these features were largely corrected in the DKO embryo(Fig. 1D3). Importantly, the chondrocranium of the DKO embryo was noticeably bigger than its Ihh^-/- counterpart but similar to the wild-type size(asterisks, Fig. 1D1-D3). Thus,normalization of cartilage development may be responsible for the rescue phenotype in the skeleton of the DKO embryo.

We next determined whether Ihh antagonizes Gli3 activity by regulating the proteolytic processing of Gli3. Western analyses were performed with whole cell lysates harvested from E14.5 limb cartilage, using a Gli3 antibody. Although the truncated repressor form of Gli3 (Gli3^R, 83 kDa) was detectable in the normal cartilage (Fig. 1E, `WT'), its amount was much increased in the absence of Ihh (Fig. 1E,`Ihh^-/-'). Interestingly, in both the wild type and the Ihh^-/- cartilage, the full-length activator form(Gli3^A, 190 kDa) was a minor form that could be detected only after longer exposure (Fig. 1E,inset). Moreover, the total amount of Gli3 in the Ihh^-/-cartilage appeared to more than that in the wild-type cartilage. Thus, the loss of Ihh likely resulted in an increase in both Gli3 total production as well as the relative levels of Gli3^R. The specificity of the Gli3^R and Gli3^A bands was confirmed by their absence in the Gli3^-/- sample(Fig. 1E,`Gli3^-/-'). A non-specific 80 kDa band also recognized by the antibody was conveniently used as a loading control (asterisk, Fig. 1E). The similar loading across the genotypes was confirmed by using an antibody againstα-catenin, an abundant adherens junction protein(Fig. 1E).

The apparent increase in total Gli3 production in the Ihh^-/- embryo prompted us to examine its expression level in the mutant versus wild-type embryo by in situ hybridization. At E14.5 in a wild-type embryo (Fig. 1H1), Gli3 was expressed by the proliferative chondrocytes (`P'), in the perichondrium (green arrows) as well as at the periarticular region(asterisk). In the Ihh<sup>-/-</sup> embryo, Gli3 levels in all domains appeared to be elevated (Fig. 1H2). On the contrary, Gli1, which was expressed in similar domains as Gli3 in the wild-type embryo, was undetectable in the Ihh<sup>-/-</sup> littermate (compare Fig. 1F1 with F2). Finally, Gli2 was normally expressed in the perichondrium (green arrows) as well as the proliferating chondrocytes (`P')(Fig. 1G1), and maintained a similar expression level in these domains in the Ihh^-/-embryo, although the proliferative zone was much diminished in the mutant humerus (Fig. 1G2). Curiously, Gli2 expression in the periarticular region appeared to be elevated in the Ihh^-/- embryo (asterisk, Fig. 1G2). Thus, all three Gli molecules were expressed in the developing long bones and Gli3 expression was elevated in the absence of Ihh.

Results from the whole-mount skeleton analyses prompted us to examine first whether cartilage development was normalized in the DKO embryo. We first analyzed chondrocyte maturation by histological analyses. At E18.5, the Ihh^-/- embryo exhibited only very rudimentary vascularization at the center of the humerus (asterisk, Fig. 2B) and contained a much reduced growth region (red double-headed arrow, Fig. 2B) as previously described (St-Jacques et al.,1999). By contrast, in the DKO embryo vascularization was more advanced (asterisk, Fig. 2C),and the growth region was markedly increased (red double-headed arrow, Fig. 2C). Interestingly, both the nonhypertrophic (green double-headed arrow) and the hypertrophy zone(black double-headed arrow) were widened in the DKO embryo over the wild-type counterpart (Fig. 2, compare A and C), suggesting likely delays in both the onset and the removal of hypertrophic chondrocytes in the DKO embryo.

Examination at a higher magnification revealed a rescue of the organization of the growth region chondrocytes in the DKO embryo. In the wild-type embryo at E18.5, stacks of flattened cells (Fig. 2D1) arranged in columns (columnar chondrocytes) preceded the hypertrophic chondrocytes (Fig. 2E1) that became fully vacuolated prior to their removal at the chondro-osseous junction. In the Ihh^-/- embryo,chondrocytes were severely disorganized so that no stacks of cells were evident (Fig. 2D2). Moreover,although a majority of chondrocytes have undergone hypertrophy in the Ihh^-/- embryo at E18.5(St-Jacques et al., 1999), the`hypertrophic' chondrocytes here were noticeably smaller than the wild-type counterparts (Fig. 2E2). Remarkably, in the DKO embryo, both columnar chondrocytes(Fig. 2D3) and the hypertrophy phenotype (Fig. 23E) were fully restored. Thus, chondrocyte organization and maturation at the growth region was normalized in the DKO embryo.

Analyses of the histological sections also revealed differences in the perichondrium among the various embryos. In the wild-type embryo, concurring with chondrocyte development, the perichondrium also underwent distinct morphological changes along the epiphyseal-diaphyseal axis. Specifically, in regions flanking the prehypertrophic chondrocytes(Fig. 2F1), the perichondrium(orange double-headed arrows) differentiated into an inner region with more and cuboidal cells (red arrows) contrasting with an outer region with fewer and elongated cells (green arrows). The morphological distinction between the layers became more pronounced in regions flanking the hypertrophy zone(Fig. 2G1) where the inner layer containing cuboidal cells (red arrows) directly lined the bone collar(yellow contour), whereas the outer layer (green contour) became more fibrous and contained fewer and much elongated cells (green arrows). In the Ihh^-/- embryo, the perichondrium was generally hypoplastic(orange double-headed arrow, Fig. 2F2). At the diaphysis where rudimentary vascularization occurred(Fig. 2G2), although the perichondrium thickened with more cuboidal cells (red arrows) in the inner region and more elongated cells (green arrows) in the outer region (green contour), no distinct layers were established. In the DKO embryo, the width of the perichondrium (orange double-headed arrow, Fig. 2F3) was markedly increased over the Ihh^-/- mutant and was similar to the wild-type counterpart. However, in contrast to that in the wild-type embryo,the perichondrium flanking the prehypertrophic chondrocytes in the DKO animal failed to differentiate into the morphologically distinct layers, but instead was uniformly populated with elongated cells (green arrows, Fig. 2F3). Differentiation of the perichondrium eventually occurred around the mid-region of the hypertrophic zone where bone formation began to occur (yellow contour, Fig. 2G3). Here, as in the wild-type animal, the cuboidal cells (red arrows) resided immediately adjacent to the bone collar, whereas the elongated cells (green arrows) populated the outer fibrous layer (green contour). Thus, the perichondrium in the DKO embryo exhibited normal growth but failed to undergo normal differentiation.

To confirm that chondrocyte maturation was rescued in the DKO embryo, we next examined the expression of chondrocyte maturation markers by in situ hybridization. At E18.5, in the wild-type humerus, parathyroid hormone-related peptide receptor (Pthr) was highly expressed in the prehypertrophic chondrocytes (`PH', Fig. 3A1),in addition to expression in the osteoblast lineage as detected in the primary spongiosa (`PS'), the bone collar (`BC') as well as the perichondrium (orange arrows). Type X collagen (Col10a1) was specifically expressed in all hypertrophic chondrocytes (Fig. 3B1). Within the hypertrophic zone, vascular endothelial growth factor A (Vegfa) exhibited a high level expression in the mid-region hypertrophic chondrocytes (green line, Fig. 3C1), whereas matrix metalloproteinase 13 (Mmp13) was largely restricted to the late hypertrophic chondrocytes (green line, Fig. 3D1). Vegfa was also detected at a lower level throughout the core of the immature cartilage(purple asterisk, Fig. 3C1),whereas Mmp13 was also expressed in cells associated with the primary spongiosa (`PS', Fig. 3D1). As previously documented, at E18.5 the Ihh^-/- embryo exhibited precocious maturation of chondrocytes, as evidenced by the close proximity of prehypertrophic (Pthlh positive) and hypertrophic(Col10a1 positive) cells to the epiphysis (note reduced lengths of green double-headed arrows, Fig. 3A2,B2). Accordingly, cells expressing Vegfa(Fig. 3C2) and Mmp13(Fig. 3D2) were also closer to the epiphysis (demarcated by purple arrows, contiguous with radius because of joint fusion). Moreover, the linear progression of chondrocyte maturation from diaphysis to epiphysis was disrupted in the Ihh^-/- embryo,as the maturation markers were expressed in a concentric fashion(Fig. 3A2-D2). By contrast, in the DKO embryo expression of all markers was restored to largely normal domains (Fig. 3A3-D3). Consistent with the histological results (above), both the immature (green double-headed arrows, Fig. 3A3,B3) and the hypertrophic (Col10a1 positive, Fig. 3B3) zones were widened over the wild-type counterparts. Importantly, the domains expressing later hypertrophic makers such as Vegfa (green line, Fig. 3C3) and Mmp13(green line, Fig. 3D3) were markedly expanded in the DKO embryo, consistent with a delay in the removal of late hypertrophic chondrocytes. The widened zone for immature chondrocytes reflected a delay in maturation in the DKO embryo, and this was most evident in the humerus at E14.5, when the DKO embryo showed a clear retardation in Col10a1 expression (Fig. 3, compare E1 with E3), in contrast to the Ihh^-/- embryo where Col10a1 expression was much accelerated (Fig. 3E2). Overall, molecular analyses confirmed that proper chondrocyte maturation was largely restored in the DKO embryo.

Restoration of proper chondrocyte maturation in the DKO embryo prompted us to examine the expression of Pthlh. Pthlh has previously been shown to be a principal regulator of chondrocyte maturation functioning as a secondary signal for Ihh (Karp et al.,2000; Lanske et al.,1996; Vortkamp et al.,1996). At E14.5, the wild-type radius expressed high levels of Pthlh both in the periarticular chondrocytes (yellow asterisk) and in the outer layers of perichondrium flanking the immature chondrocytes (purple arrows, Fig. 3F1). These regions corresponded to a low level of patched 1 (Ptch1) expression(yellow asterisk and purple arrows, Fig. 3G1) but were distinct from domains expressing high levels of Ptch1 (green asterisk and blue arrows, Fig. 3G1). As previously reported, Pthlh expression was abolished in the developing cartilage of Ihh^-/- embryos (Fig. 3F2), similar to the elimination of Ptch1 expression(Fig. 3G2). Remarkably, in the DKO embryo Pthlh expression was fully restored in both the periarticular (yellow asterisk) and the perichondrium region (purple arrows, Fig. 3F3). Similarly, a low level of Ptch1 expression was detected in both the immature chondrocytes (yellow asterisk) and the perichondrium (purple arrows, Fig. 3G3) in the DKO embryo,whereas the high level Ptch1 expression was not restored. In summary,both Pthlh and a low level of Ptch1 expression were rescued in the cartilage of the DKO embryo.

We next examined the status of chondrocyte proliferation in the growth region cartilage. BrdU labeling experiments showed that the proliferation rate of chondrocytes in E14.5 DKO embryos was similar to the wild-type level, in contrast to a 50% reduction in Ihh^-/- embryos(Fig. 3H).

The whole-mount skeletal staining experiment above revealed abnormal bone formation in certain skeletal elements of the DKO embryo. To confirm this finding, we stained tissue sections by the von Kossa method. At E18.5, the wild-type embryo had developed a definitive bone collar (arrows) surrounding the mineralized cartilage (red asterisks) in the long bones such as the ribs(Fig. 4A1), the humerus(Fig. 4B1) and the femur(Fig. 4C1). The Ihh^-/- embryo, however, as previously reported, completely lacked the bone collar (arrows) in all long bones although cartilage mineralization occurred (asterisks, Fig. 4A2-C2). Like the Ihh^-/- embryo, the DKO embryo also lacked a bone collar at the normal position in all long bones(arrows, Fig. 4A3-C3) despite ample mineralization in the cartilage (red asterisks, Fig. 4A3-C3). Interestingly,however, in the humerus (Fig. 4B3 and B4) and the femur (Fig. 4C3 and C4), as well as certain ribs(Fig. 4A3 and A4), a varying amount of bone deposition (arrows, Fig. 4A4-C4) was detected at the diaphysis. A similar observation was made in the ulna (data not shown). The bone deposition was often asymmetrical and not adherent to the underlying cartilage (double-headed arrows, Fig. 4B4,C4). An additional important feature of the ectopic bone was that it invariably correlated with vascularization of the cartilage, as evidenced by the presence of red blood cells (purple asterisk, Fig. 4A3). Moreover, the degree of ossification also correlated with that of vascularization. For example, the humerus(Fig. 4B3) often exhibited more ossification and was also more advanced in vascularization, whereas the femur(Fig. 4C3) lagged in both aspects. Thus, in the DKO embryo, the normal bone collar failed to form but ectopic bone developed at the diaphysis of long bones where vascularization occurred.

Cartilage vascularization in the long bones of the DKO embryo was partially rescued over the Ihh^-/- embryo, although it remained defective compared with the wild-type embryo. This was exemplified in a rib of an E18.5 DKO embryo by the appearance of a rudimentary marrow cavity (red asterisk, Fig. 4D3), whereas no vascularization was noted in any ribs of an Ihh^-/- embryo at same age (Fig. 4D2). The vascularization in the DKO embryo was noticeably delayed compared to the wild-type embryo, resulting in a wider hypertrophic zone(Fig. 4, double-headed arrows,compare D1 with D3).

The lack of normal bone collars in the DKO embryo prompted us to investigate in detail the development of the osteoblast lineage. To this end,we examined the expression of a panel of markers activated at various stages of osteoblast development. At E15, in a wild-type humerus, Runx2 as well as low levels of Col10a1 and alkaline phosphatase (Ap; Akp - Mouse Genome Informatics) were characteristically expressed in the perichondrial cells (red arrows, Fig. 5A1-C1) towards the epiphysis (`e', Fig. 5A1), representing early stages of the osteoblast lineage. Towards the diaphysis (`d', Fig. 5A1) where cells of the lineage became progressively more mature, the levels of Col10a1 and Ap were elevated in the perichondrium (green arrows, Fig. 5A1,C1), whereas Runx2 expression (green arrow, Fig. 5B1) remained relatively constant. Meanwhile, Osx (Sp7 - Mouse Genome Informatics)and subsequently bone sialoprotein (Bsp) were activated in the more mature perichondrium (green arrows, Fig. 5D1,E1). In addition to the expression in the osteoblast lineage,high levels of Runx2, Ap and Bsp were also detected in hypertrophic chondrocytes (green asterisks, Fig. 5B1,C1,E1, respectively),whereas Osx was also in the prehypertrophic cells (green asterisk, Fig. 5D1). Moreover, a lower level of Runx2 was also detected in the immature chondrocytes (red asterisk, Fig. 5B1). In the Ihh^-/- embryo, consistent with previous reports(Hu et al., 2005; St-Jacques et al., 1999), none of the osteoblast markers were expressed in the perichondrium at E15 (red arrows, Fig. 5A2-E2), although a number of molecules were diffusedly expressed in the cartilage, reflecting the dysregulation of chondrocyte maturation (green asterisks, Fig. 5B2-E2). Interestingly in the DKO embryo, although Col10a1 and Runx2 were expressed in the perichondrium (red arrows, Fig. 5A3,B3, respectively), Ap, Osx or Bsp could not be detected there (red arrows, Fig. 5C3-E3, respectively). Moreover, the level of Col10a1expression remained low throughout the perichondrium (red arrows, Fig. 5A3). In the cartilage,expression of Ap and Osx (green asterisks, Fig. 5C3 and D3, respectively)became restricted to discrete domains similar those in the wild-type embryo. Bsp and high levels of Runx2, both characteristic of later hypertrophic chondrocytes, were not yet expressed in the cartilage at this stage (Fig. 5D3 and B3,respectively), consistent with a delay in chondrocyte maturation in the DKO embryo, as noted earlier. However, similar to the wild-type embryo, a low level of Runx2 expression was detected in the immature chondrocytes in the DKO embryo (red asterisk, Fig. 5B3). Thus, development of the osteoblast lineage was initiated but was later stalled in the DKO embryo.

We next further investigated the molecular basis for the arrest of osteoblast development in the DKO embryo. Our previous work identified that Wnt signaling through the canonical pathway was crucial for osteoblast development and that this signaling was abolished in the Ihh^-/- embryo (Hu et al., 2005). We therefore examined whether canonical Wnt signaling was rescued in the DKO embryo. At E14.5, in the wild-type embryo Tcf1and Dkk1, both direct transcriptional targets of Wnt canonical signaling (Chamorro et al.,2005; Gonzalez-Sancho et al.,2005; Niida et al.,2004; Roose et al.,1999), were upregulated in the perichondrium (red arrows, Fig. 5F1,G1). In the DKO embryo, however, as in the Ihh^-/- mutant(Hu et al., 2005), Tcf1 and Dkk1 were not detected in the perichondrial cells(Fig. 5F2,G2, respectively). Importantly, the canonical Wnt pathway remained quiescent even at E18.5, when Tcf1 was normally expressed in the perichondrium flanking the prehypertrophic chondrocytes (red arrows, Fig. 5H1) but was not detectable in a similar region in the DKO embryo (red arrows, Fig. 5H2). Notably, Tcf1 expression was activated in the periosteum overlying the ectopic bone (asterisk, Fig. 1H2),indicating that canonical Wnt signaling was coupled with bone formation at the ectopic site. Thus, in the DKO embryo, osteoblast development at the orthotopic position was arrested prior to the activation of canonical Wnt signaling in the progenitors.

We next investigated whether the defect in orthotopic osteoblastogenesis was merely a temporary delay that was corrected at later stages. At E18.5, in the wild type humerus Ap and Bsp(Fig. 6A1,B1, respectively)were expressed highly in cells of the osteoblast lineage associated with the primary spongiosa (`PS'), the bone collar (`BC') as well as the perichondrium(green arrows) flanking the prehypertrophic chondrocytes (`PH'), in addition to their expression in the hypertrophic chondrocytes (`H'). Similarly, Osx (Fig. 6C1) was expressed in the primary spongiosa (`PS'), the bone collar (`BC') and the perichondrium (arrows). Osteocalcin (Oc; Bglap1 - Mouse Genome Informatics) (Fig. 6D1),a specific marker for the mature osteoblast, was detected in the primary spongiosa (`PS') as well as the bone collar (`BC') that extended to flank the mid-region (green arrows) of the hypertrophic zone (`H'). By contrast, in the DKO humerus (Fig. 6A2-D2) none of the markers were detected in the corresponding regions of the perichondrium(green arrows), indicating a persistent lack of orthotopic bone formation. However, all the markers were expressed in the ectopic bone formed around the diaphysis (green asterisks, Fig. 6A2-D2). Thus, orthotopic osteoblast development was persistently arrested in the DKO embryo, even though ectopic bone formed at the diaphysis.

Our earlier observation that less bone formed at the femur than at the humerus in the DKO embryo (Fig. 4, compare B3 with C3) prompted us to examine the fate of the perichondrial mesenchymal cells at the diaphysis of the femur. Interestingly,at E18.5 in areas of the femur where no overt ossification had occurred,mesenchymal cells around the diaphysis (red asterisks, Fig. 7A,B) accumulated under the perichondrium (`PC', Fig. 7B), forming an asymmetrical `mesenchymal collar'. Remarkably, the mesenchymal cells appeared to have migrated through the late hypertrophic chondrocytes splitting the hypertrophic zone (`H', Fig. 7A,C). Moreover, the mesenchymal cells embedded between the hypertrophic zones appeared to have died, resulting in an empty canal (green asterisk, Fig. 7C). The cell death correlated with the paucity of cartilage vascularization, which was reduced to small areas of the hypertrophic cartilage(Fig. 7D). More importantly,the viable mesenchymal cells beneath the perichondrium expressed markers of the osteoblast lineage, including Ap, Osx(Fig. 7B1-B2, respectively) and Runx2 (data not shown), none of which was detected in the perichondrium (`PC'). The expression of these markers was not associated with hypertrophic chondrocytes as the cells did not express Col10a1(Fig. 7B3). Thus, in the DKO embryo, perichondrial mesenchymal cells accumulated around the diaphysis of the femur developed along the osteoblast lineage.

We next examined whether the perichondrial mesenchymal cells adopted a similar fate in the absence of canonical Wnt signaling. Similar to the DKO embryo, embryos with β-catenin selectively deleted in the skeleton(Dermo1-Cre; β-catenin^c/c) that we previously generated (Hu et al.,2005) also accumulated cells beneath the perichondrium at the diaphysis (Fig. 7E). There, as in the DKO embryo, orthotopic osteoblast development was also arrested prior to Osx activation (Hu et al.,2005). However, distinct from the DKO embryo, the Dermo1-Cre; β-catenin^c/c embryo did not form ectopic bone at the diaphysis of the skeletal elements, despite advanced vascularization of the cartilage at E18.5(Hu et al., 2005). Instead,cells accumulated beneath the perichondrium (`PC') exhibited the morphology of chondrocytes (red asterisks, Fig. 7E,F). Accordingly, the majority of the cells expressed high-levels of type II collagen (Col2a1)(Fig. 7F1) and the cells closer to the marrow cavity activated Col10a1 expression(Fig. 7F2). Thus, in the absence of β-catenin, the perichondrial mesenchymal cells failed to form osteoblasts but instead developed into chondrocytes.

Finally, we examined the fate of the perichondrial cells in the E18.5 Ihh^-/- embryo. As noted earlier(Fig. 2), the perichondrium in the Ihh^-/- embryo was generally hypoplastic but slightly thickened at the diaphysis of the humerus (boxed region, Fig. 7G), where rudimentary vascularization occurred (asterisk, Fig. 7G). Moreover, the perichondrial cells within this region underwent morphological differentiation(Fig. 7G1), as indicated by the cuboidal cells (red arrows) within the inner region (`1') versus the elongated cells (yellow arrows) at the outer region (`2'). Importantly, the inner cells expressed Runx2, Osx and Ap(Fig. 7H1-J1, respectively) but not Col10a1 (data not shown). Thus, in the Ihh^-/-embryo, rudimentary vascularization correlated with early stages of osteoblastogenesis from the perichondrium.

We have identified Gli3 as a key effector for Ihh signaling in the developing endochondral skeleton (Fig. 8A). Removal of Gli3 bypassed the requirement for Ihh in the proper proliferation and maturation of chondrocytes, indicating that Ihh normally regulates the progression of the cell cycle as well as expression of Pthlh by antagonizing the repressor activity of Gli3 (Gli3^R). The delay in chondrocyte maturation in the DKO embryo represents an`overcorrection' phenotype and may reflect residual Gli3^Ractivities that escape Ihh antagonism in a normal embryo. However, orthotopic osteoblast differentiation, as well as vascularization of the hypertrophic cartilage, was only partially rescued in the DKO embryo. Thus, an additional effector(s), most probably one or more of the activator forms of the Gli molecules (Gli^A), is required for mediating the role of Ihh signaling in later development of osteoblasts and in specific aspects of cartilage vascularization.

The identity of Gli^A is presently unknown. Although Gli2^A has been shown to possess activator activities necessary for Shh signaling in both the central nervous system(Ding et al., 1998; Komori et al., 1997; Matise et al., 1998; Sasaki et al., 1999) and the developing hair follicle (Mill et al.,2003), Gli2^-/- embryos exhibited only quantitative defects in bone formation (Miao et al.,2004). Thus, multiple Gli activators may function in combination to control progression of osteogenesis.

Notably, despite the rescue of cartilage proliferation and maturation, the limb skeletal elements of the DKO embryos remained considerably shorter than their wild-type counterparts. This finding underscores the important contribution of a normal bone collar (which failed to develop in the DKO embryo) to the overall linear growth of skeletal elements. Similarly, our previous studies showed that the Col2-Cre; Smo^n/c mice, which had a similar proliferation defect as the Ihh^-/- mutant but maintained a largely normal bone collar, exhibited a much milder growth defect than the Ihh^-/- embryo (Long et al.,2001). Thus, development of the bone collar, together with growth of the cartilage, contributes to the linear length of a long bone.

The recovery of a low level of Ptch1 in the DKO embryo is consistent with a dual mechanism by which Ihh controls Ptch1 transcription: whereas a low level expression can be achieved by derepression of Gli3^R, the full activation of Ptch1 expression may require the activator forms of Gli molecules in response to robust Ihh signaling. Ptch1 was recently shown to be a direct transcription target of Gli proteins(Agren et al., 2004).

Despite the defect in cartilage vascularization in the DKO embryo,perichondrial mesenchymal cells independently invaded the hypertrophic cartilage. Thus, invasion of the mesenchymal cells can be uncoupled from cartilage vascularization. The molecular basis underlying the mesenchymal invasion is presently unknown but appears to be distinct from that for the blood vessel invasion, as Mmp9(Vu et al., 1998) was not expressed by the invading mesenchyme (data not shown).

Ihh signals via distinct transcriptional effectors at different stages of osteoblast development (Fig. 8B). Initially, Ihh activates expression of Runx2 and low levels of Col1a1 by counteracting the repressor function of Gli3 (Gli3^R)(`1'). Further progression of the lineage, however, requires additional Ihh signaling, most probably transcriptional activation via Gli<sup>A</sup> (`2'). Previously, we showed that Ihh signaling activated canonical Wnt signaling which was in turn required for the activation of Osx expression(Hu et al., 2005). Thus,Gli^A is likely required to activate the canonical Wnt signaling,which remained silent in the DKO embryo (`3'). Alternatively, defective Wnt signaling could be secondary to earlier defects in the osteoblast lineage(e.g. absence of Ap^low-positive cells) in the DKO embryo.

The perichondrial mesenchymal cells are bi-potential for either osteoblasts or chondrocytes (Fig. 8B, `4'). In embryos of either Dermo1Cre; β-catenin<sup>c/c</sup>or DKO, the cells accumulated beneath the perichondrium at the diaphysis,forming a characteristic `wedge-shape' mesenchyme. In the absence ofβ-catenin, the cells failed to express Osx and subsequently developed into chondrocytes, in a manner similar to that in the Osx^-/- embryo(Nakashima et al., 2002). In the DKO embryo, Osx expression was activated in these cells,presumably by a vasculature-derived signal, and osteogenesis ensued. Thus, Osx appears to function as a molecular switch between the osteoblast and chondrocyte fates.

Chondrogenesis is unlikely a mere default fate for the perichondrial progenitors, as no chondrogenesis was observed in the perichondrium of the Ihh^-/- embryo even though osteogenesis was impaired. Here,the perichondrium was generally hypoplastic, and no `wedge-shape' mesenchyme was present at the diaphysis. In agreement with the absence of Runx2expression in the perichondrial cells prior to vascularization, the perichondrium phenotype here is similar to that in the Runx2^-/- embryo(Komori et al., 1997; Otto et al., 1997),implicating Runx2 in the proper growth of the perichondrium. Similarly, in the Runx2^-/- embryo, the perichondrial cells failed to undergo chondrogenesis, despite the defect in osteoblast development. These findings could indicate that Runx2 activity is required for the perichondrial cells to achieve their capacity for chondrocyte differentiation. Alternatively, the failure in chondrogenesis could be secondary to the perichondrium hypoplasia associated with Runx2 deficiency. Regardless of the mechanism, the absence of chondrogenesis in the Ihh^-/- perichondrium seemed to be at odds with our previous finding that Smo^-/- perichondrial cells (nonresponsive to Ihh) developed into chondrocytes(Long et al., 2004). However,the difference could be because the Ihh-nonresponsive cells in the previous study resided among wild-type cells in which Ihh signaling occurred. It is possible that secondary factors from the normal neighboring cells steered the Ihh-nonresponsive cells towards chondrogenesis.

Our results also indicate an osteogenic role for the vasculature(Fig. 8B). In this model, a signal from the vasculature could promote bone formation from Col1a1^low- and Runx2-postive cells (the `stalled'cells in the DKO embryo) by activating Ap expression (`6') and the canonical Wnt/β-catenin pathway (`5'). Moreover, a vascularization-dependent signal could initiate osteogenesis independent of Ihh signaling (as in the Ihh<sup>-/-</sup> embryo) (`7'). It should be noted that we could not discern at present which mechanism contributed to the `ectopic' bone in the DKO embryo. In addition, it is presently unknown whether mature osteoblasts can eventually develop in the Ihh^-/- embryo as neonatal lethality of the mutant precludes analyses at later stages. However, a recent study reported that an Ihh^-/- cartilage element, when implanted into the kidney capsule, became vascularized and developed bone, apparently without activating hedgehog signaling (Colnot et al.,2005). The nature of the vascularization-dependent signal is presently unknown but previous studies have implicated signals for the endothelium in development of the liver(Matsumoto et al., 2001) and the pancreas (Lammert et al.,2001). A definitive link between bone formation and endothelial signaling, however, awaits further studies.

The concept that vascularization functions as an osteogenic signal independent of Ihh has interesting implications. It follows that normally two distinct processes (Ihh dependent versus vascularization dependent) contribute to bone formation in the endochondral skeleton. Disruption of both in the Ihh^-/- embryo resulted in a complete loss of bone. This notion could also explain the relatively normal intramembranous bones in the skull of the Ihh^-/- embryo. As the skull bones are evolved earlier, does vascularization represent a more primitive and universal mechanism for bone formation, whereas Ihh dependence could be a recent invention to accommodate bone formation from a cartilage template?

We thank Dr Chin Chang (Vanderbilt University) for kindly providing the Gli3 antibody. We thank Drs Raphael Kopan and David Ornitz (Washington University Medical School) for critical reading of the manuscript. This work was supported by a NIH grant DK065789 (F.L.) and a NIH postdoctoral training grant 5T32AR07033 (M.J.H.).