Indian hedgehog synchronizes skeletal angiogenesis and perichondrial maturation with cartilage development

The ontogeny of the skeleton, like that of so many embryonic tissues,begins when loose mesenchymal cells aggregate to form a condensation, and continues as a reiterative series of interactions between the condensed cells and their surrounding environment. In the case of the endochondral skeleton,cells in the center of the condensation differentiate into chondrocytes while the surrounding cells eventually differentiate into bone. Both programs of cell differentiation absolutely require signals originating from the other tissue (Lanske et al., 1996; Long and Linsenmayer, 1998; Vortkamp et al., 1996).

The vasculature is another tissue essential for the program of skeletogenesis. The initiation of skeletal tissue formation requires that blood vessels vacate the area of the future condensation(Yin and Pacifici, 2001), but the differentiation of cells into osteoblasts requires vascular invasion at a later stage (Gerber et al.,1999; Vu et al.,1998). This co-dependency among tissues that comprise the skeleton suggests why it is difficult to distinguish the role cartilage plays in the orchestration of vascular and perichondrial development, relative to its own developmental progression. We set out to determine the extent to which indian hedgehog (IHH) functions as a `molecular coordinator' of chondrocyte differentiation, perichondrial development and vascular remodeling during the process of fetal skeletogenesis.

Hedgehog proteins are prime candidates for coordinating cell differentiation and thus orchestrating tissue formation(Bumcrot and McMahon, 1995). In the appendicular skeleton, IHH is secreted by pre-hypertrophic and early hypertrophic chondrocytes (Chung et al.,2001; Kobayashi et al.,2002; Long et al.,2004; Vortkamp et al.,1996). IHH is involved in osteoblast differentiation, as illustrated by the fact that Ihh^-/- appendicular skeletal elements do not ossify (Chung et al.,2001; St-Jacques et al.,1999). Although some data suggest that the absence of ossification results from a primary defect in osteoblast differentiation(Chung et al., 2001; Long et al., 2004), secondary defects might also contribute to the phenotype. For example, the Ihh^-/- phenotype may be the result of a failure to specify appendicular perichondrial cells to an osteoblast lineage, rather than an actual failure of specified osteoblasts to differentiate(Long et al., 2004; Naski et al., 1998). In either of these two scenarios, IHH would act directly on the osteoblast. Alternatively, IHH may only secondarily block ossification because of its primary role in regulating chondrocyte differentiation(Long et al., 2001; Vortkamp et al., 1996). Another possibility is that IHH influences vascular development(Byrd et al., 2002; Dyer et al., 2001; Pola et al., 2001), which, in turn, affects osteoblast differentiation. To understand the extent to which IHH controls osteogenesis and angiogenesis in addition to chondrogenesis, we took a closer look at the Ihh^-/- phenotype for new clues regarding these tissue interactions. A series of embryonic tissue manipulations shed new light on the function of this morphogen during fetal skeletogenesis.

Ihh^+/- mice in the C57B6 background carrying one null allele of the Ihh gene were mated with each other to generate homozygous Ihh^-/-embryos. The tails of the embryos were genotyped by PCR as previously described(St-Jacques et al., 1999). Ihh^+/- mice were crossed with PtchlacZ mice that contain an insertion of the lacZ transgene in one patched 1(Ptch1) locus (Goodrich et al.,1997) in order to generate Ihh^+/-;PtchlacZ mice. Ihh^+/-;PtchlacZ mice were mated with Ihh^+/- mice to obtain Ihh^-/-;PtchlacZ embryos. Ihh^-/-;Rosa26 embryos were generated by crossing Ihh^+/- mice with Rosa26 mice, which carry the lacZ transgene expressed ubiquitously (Jackson Laboratory, Maine,USA). To detect the presence of the PtchlacZ or Rosa26 transgene,adult tails or embryonic heads were collected and processed for whole-mount X-gal staining after fixation in 0.4% paraformaldehyde in PBS overnight at 4°C or in 0.2% glutaraldehyde (0.2% glutaraldehyde, 5 mM EDTA and 2 mM MgCl₂ in 1× PBS) for 30 minutes at room temperature as described previously (Colnot et al.,2004).

Stylopod cartilage anlagen from E14.0 wild-type, PtchlacZ,Ihh^-/-, Ihh^-/-;Rosa26, wild type;Rosa26,or Ihh^-/-;PtchlacZ embryos were dissected and transplanted underneath the renal capsule of adult wild-type or Rosa26 mice,as previously described (Colnot et al.,2004). Mice used as hosts and donors for the renal capsule transplantations were both bred in the C57B6 background to avoid immune rejection. All procedures followed standard Stanford and UCSF CAR/LARC protocols. Skeletal elements were collected after 24 hours, 48 hours and 4-14 days and processed for cellular and molecular analyses.

To determine if a hedgehog (HH) signal was provided by the kidney environment, we grafted E14.0 PtchlacZ and Ihh^-/-;PtchlacZ cartilage elements under the renal capsule of wild-type hosts in the absence or presence of exogenous SHH-N protein (Curis, Inc.). As Ptch is a downstream target of HH signaling, up-regulation of the PtchlacZ transgene is indicative of activation of the HH signaling pathway in vivo. The SHH-N protein was added as a positive control to demonstrate that the tissue was responsive to a hedgehog signal. Affi-Gel Blue beads (BioRad) were soaked in a solution containing 400μg/ml of the recombinant SHH-N protein in PBS with 0.1% bovine serum albumin. Two beads were placed on each side of the cartilage elements at the time of transplantation into the renal capsule.

Embryonic limbs and skeletal elements collected at various time points after transplantation in the renal capsule were fixed in 4% paraformaldehyde at room temperature for 1 hour or at 4°C overnight, decalcified at 4°C in 19% EDTA, pH 7.4, for 1 to 7 days, dehydrated and embedded in paraffin. Five micron-thick sections were collected on superfrost-plus slides and analyzed for histology using Safranin-O/Fast green (SO) and trichrome (TC)staining. Tartrate-resistant acid phosphatase (TRAP) staining, MMP9 and PECAM immunohistochemistry were performed as previously described(Colnot et al., 2003). Smooth muscle actin (SMA) and COL4 immunohistochemistry was performed according to established protocols (Marcucio and Noden,1999) at dilutions of 1:200 (mouse anti-αSMA, clone 1A4;Sigma) and 1:400 (rabbit anti-collagen type IV; cat. no. c7510-51, US Biological). SMA primary antibody was detected using the Vectastain ABC mouse IgG detection kit (Vector Labs, Inc.), and COL4 was detected using the Vectastain ABC rabbit IgG detection kit (Vector Labs). Since the mouse anti-αSMA was used on mouse tissue, we compared this staining to negative control samples that were exposed to the Vectastain mouse IgG detection kit alone (absent primary SMA antibody). For detection of lacZ or PtchlacZ expression, whole mount X-gal staining was performed as described above. Samples were then embedded, sectioned and counterstained with Eosin.

³⁵S-labelled antisense riboprobes corresponding to cDNAs for collagen type I (Col1), collagen type IIa1 (Col2),osteopontin (Op; Spp1 - Mouse Genome Informatics), matrix metalloproteinase 13 (Mmp13), indian hedgehog (Ihh), patched 1 (Ptch) and vascular endothelial growth factor (Vegf) were hybridized on tissue sections and the in situ hybridization signal was visualized as described previously(Albrecht et al., 1997; Ferguson et al., 1999).

To distinguish cells derived from the graft and cells derived from the host, E14.0 wild type;Rosa26 or Ihh^-/-;Rosa26 embryos were transplanted into wild-type hosts (Jackson Laboratory, Maine, USA). They were processed for X-gal staining as described above and by Colnot et al.(Colnot et al., 2004).

We analyzed in greater detail the Ihh^-/- skeletal phenotype for new evidence as to how IHH synchronizes chondrogenesis,osteogenesis and angiogenesis. We began by examining the program of cartilage differentiation. Previous reports indicate that Ihh^-/-cartilage is slow to hypertrophy(St-Jacques et al., 1999; Vortkamp et al., 1996). We reasoned that a delay such as this, or an actual halt in chondrocyte differentiation, might leave Ihh^-/- skeletal elements in an anti-angiogenic state, which in turn would be unfavorable for vascular invasion and ossification.

Our molecular analyses, however, belied this theory. Although cartilage maturation was definitely delayed, most Ihh^-/-chondrocytes still progressed to the point of expressing markers of late hypertrophy such as matrix metalloproteinase 13 (Mmp13) and osteopontin (Op; Fig. 1A-C,E-G). In addition, Ihh^-/- chondrocytes expressed vascular endothelial growth factor (Vegf), indicating that regardless of the fact that it was disorganized, Ihh^-/-hypertrophic cartilage still switched from an anti-angiogenic to an angiogenic state (Fig. 1D,H). We therefore concentrated our efforts on the condition of the vasculature surrounding the mutant appendicular skeleton.

During normal fetal mouse development, the humeri are invaded by blood vessels at E15.0 (Colnot and Helms,2001). In Ihh^-/- humeri, vascular invasion was delayed until E17.5 (Fig. 2)and instead of endothelial cells invading the center of the hypertrophic cartilage domain and spreading distally and proximally to create a marrow cavity, Ihh^-/- endothelial cells were restricted to a few small islands in the disorganized hypertrophic cartilage(Fig. 2A,B,E,F, arrows). We further characterized these invading cells using immunohistochemical markers,and found that they were positive for TRAP and MMP9 (osteoclast markers), and PECAM (a marker of endothelial cells), indicating that mutant blood vessels can penetrate Ihh^-/- hypertrophic cartilage(Fig. 2C,D,G,H, arrows). One day later, at E18.5, the PECAM- and TRAP-positive cells had disappeared, and the cartilage matrix was degraded. In addition, we noted a large number of condensing, terminally differentiated chondrocytes(Farnum and Wilsman, 1989; Roach and Clarke, 2000) in that region and only weak MMP9 expression remained in the disintegrated extracellular matrix (Fig. 2I,J,K, arrows). These data led us to conclude that Ihh^-/- hypertrophic chondrocytes undergo terminal differentiation and that Ihh^-/- blood vessels invade the mutant hypertrophic cartilage. Having done so, however, they soon disappear. Why do Ihh^-/- blood vessels not persist? One possibility is that the Ihh^-/- hypertrophic cartilage is inhospitable to blood vessels and they collapse soon after invasion. To address this possibility we turned to an ex vivo system where we could directly test whether the mutant cartilage anlagen could support wild-type blood vessels.

When wild-type skeletal elements are transplanted underneath the renal capsule, they become vascularized and progress through the same molecular,cellular and histological phases of endochondral ossification that take place when the elements develop in situ (Colnot et al., 2004). Using this system, we produced a chimeric situation where Ihh^-/- skeletal anlage were grown in the presence of wild-type endothelial cells. By transplanting Ihh^-/-;Rosa26 elements into wild-type host mice we could unambiguously identify the sources of cells participating in the development of the ex vivo skeletal elements because all graft-derived cells turned blue after X-gal staining.

Using this system we found that the Ihh^-/- skeletal anlage were indeed angiogenic: in an ex vivo environment, these elements became vascularized and formed a marrow cavity(Fig. 3A,B,F,G; n=80 wild type and n=80 Ihh^-/-). This finding is in keeping with the fact that Vegf expression persists in Ihh^-/- hypertrophic cartilage in situ(Fig. 1).

We next turned our attention to the status of the Ihh^-/- blood vessels in the ex vivo setting. Initially, we speculated that Ihh^-/- endothelial cells might not actually participate in the vascular invasion process. Examination of samples collected after 4 days ex vivo revealed, however, that Ihh^-/- vessels invaded Ihh^-/-cartilage anlage, just as wild-type blood vessels invaded the wild-type cartilage anlage (Fig. 3C,H).

On day 10 these same lineage analyses demonstrated that the vasculature had become chimeric. While most of the blood vessels were still derived from the graft, some were derived from the host, at least in the case of wild-type tissues (Fig. 3D, and data not shown). When we examined the Ihh^-/- grafts we found that the majority of vessels were derived from the wild-type host and only a small percentage of the total vasculature was composed of Ihh^-/-vessels (Fig. 3I, and data not shown). In addition, the Ihh^-/- vessels had an abnormal morphology (Fig. 3I). For example, normally wild-type vessels line the bone trabeculae and are positioned perpendicular to the surface of the chondro-vascular junction(Fig. 3D) but Ihh^-/- blood vessels had a large lumen and were arranged parallel to the chondro-vascular junction(Fig. 3I).

By day 14 wild-type grafts continued to contain graft-derived endothelial cells in significant numbers (Fig. 3E) but in mutant grafts Ihh^-/- endothelial cells had disappeared entirely from the mutant ossification center and had left behind only host-derived vessels that sustained development of the mutant anlage (Fig. 3J). Collectively,these data taken from three time points indicate that Ihh^-/- endothelial cells were capable of penetrating the mutant hypertrophic cartilage, but were unable to persist in the Ihh^-/- hypertrophic cartilage. These two observations were identical to those we made in the Ihh^-/- in vivo setting. In the ex vivo environment, however, wild-type blood vessels were able to replace the mutant blood vessels, and thereby sustained the Ihh^-/- anlage in a manner that was not possible in vivo.

We drew three conclusions from this part of the study. First, the ex vivo results confirmed that Ihh^-/- appendicular cartilage can undergo an angiogenic switch. Second, terminal differentiation of Ihh^-/- hypertrophic chondrocytes can be uncoupled from vascular invasion and expansion, since chondrocyte terminal differentiation continued to progress even though vascular invasion came to an abrupt halt. Third, Ihh^-/- blood vessels retained their invasive potential but were unable to persist, despite the presence of wild-type vessels and blood-born growth factors. This last finding suggested that Ihh^-/- blood vessels themselves might be defective.

In order to further investigate the disappearance of Ihh^-/- vessels both in vivo and ex vivo, we performed an immunohistochemical comparison of vascular maturity in wild-type and mutant vessels. At E15.0, endothelial cells at the initial site of vascular invasion express PECAM and COL4; these same markers are evident at E17.5, at the chondrovascular junction (Fig. 4A,C, and data not shown). These same vessels had yet to be covered by SMA-positive pericytes (Fig. 4B, and data not shown). Vessels in the marrow cavity of E17.5 samples, however, were SMA positive (Fig. 4D,E, red arrows). Using these markers of endothelial cell maturity, we examined Ihh^-/- samples from various stages and found that endothelial cells which formed the vascular islands were similar in maturation state to those found at the wild-type chondrovascular junction, in that they were COL4 positive but SMA negative(Fig. 4H-J). In order to examine the later stages of ossification in the Ihh^-/-anlage, we turned again to the renal capsule transplant system. Ihh^-/-;Rosa26 elements transplanted at E14.0 into wild-type hosts and collected after 10 days contained some vessels that resembled those found in the marrow cavity of E17.0 wild-type elements;vessels that were both COL4 and SMA positive(Fig. 4K-M). Because the transplants were done with Rosa26 grafts, we were able to identify the origin of these vessels and found that some of the mature, SMA-positive vessels were derived from the Ihh^-/- graft(Fig. 4L,M, red arrows).

To test whether vascular cells were a target of hedgehog signaling, we examined blood vessels in the appendages of heterozygous PtchlacZembryos (Milenkovic et al.,1999). Cells and tissues from PtchlacZ embryos have served as a functional readout of hedgehog activation in a wide variety of assays (Byrd et al., 2002; Milenkovic et al., 1999; Pola et al., 2001). While we were unable to confirm or deny the expression of Ptch in endothelial cells because of the diffuse nature of the β-gal staining (data not shown), we did find that pericytes adjacent to skeletal elements express Ptch strongly, indicating that they were responsive to hedgehog signaling (Fig. 4F,G). In some developmental contexts, endothelial cell survival is dependent upon pericyte coverage (Benjamin et al.,1998). In this case, however, the only pericyte-covered endothelial cells were found either in the soft tissue adjacent to the developing element (Fig. 4F,G)or in fully formed marrow cavities (Fig. 4D,E). Thus, while we demonstrated that pericytes can and do respond to hedgehog signaling, the blood vessels in and near the site of cartilage hypertrophy and initial vascular invasion were not associated with SMA-positive pericytes (Fig. 4B,D-G, and data not shown). Therefore the disappearance of vessels subsequent to the formation of vascular islands in Ihh^-/- animals is unlikely to be due to a pericyte defect.

In addition to revealing the Ihh^-/- vascular anomaly,the ex vivo experiments yielded another unexpected finding: all Ihh^-/- elements ossified in the renal capsule(Fig. 5A,B,D,E; n=80 wild type and n=80 Ihh^-/-). Because these experiments were performed using Rosa26 donor embryos, we could unequivocally determine that the osteoblasts were Ihh^-/- cells(Fig. 5C,F). This finding was unexpected for two reasons. First, these same Ihh^-/-elements never ossify in vivo (St-Jacques et al., 1999) and second, IHH is thought to be required for osteoblast differentiation in the limb(Long et al., 2004). That osteoblasts were potentially differentiating in the absence of hedgehog was an intriguing possibility.

In situ hybridization analyses performed on Ihh^-/-embryos confirmed that markers of early osteoblast differentiation such as collagen type 1 (Col1) and neuropilin 2 (Nrp2) were expressed in Ihh^-/- perichondria, just as they are in wild-type bone trabeculae (Fig. 6B,C,F,G). Trichrome staining revealed a small amount of osteoid-like matrix in the mutant perichondrium as well(Fig. 6A,E). In contrast, only wild-type perichondrial cells expressed osteocalcin (Oc; Fig. 6D,H), a marker of mature osteoblasts (Aubin et al.,1995). At no time did we detect Oc or Runx2expression in Ihh^-/- perichondria in vivo(St-Jacques et al., 1999).

We wondered if these Col1-, Nrp2-expressing cells in the Ihh^-/- perichondrium were able to differentiate ex vivo simply because they were provided with sufficient time to reach maturity(recall that Ihh^-/- embryos die at birth). This did not appear to be the sole explanation however, since the Ihh^-/- elements began to ossify after only 4 days ex vivo(data not shown), whereas, given four more days to develop in vivo (reaching E18.0), the Ihh^-/- elements had still not ossified(Fig. 2). Another possibility was that these Col1-, Nrp2-expressing cells in the Ihh^-/- perichondrium differentiated into osteoblasts in the renal capsule because this ex vivo environment provided a growth factor or a cell population that was missing in the mutant environment. The most obvious prospect was that the renal capsule supplied exogenous hedgehog (HH)protein.

We set about determining if the renal capsule provided exogenous hedgehog protein to the transplanted Ihh^-/- elements, and discovered that Ihh RNA was detectable by RT-PCR in the cortex of the kidney (data not shown). This finding raised the possibility that the protein might diffuse far enough and penetrate deep enough to act on Ihh^-/- perichondrial cells. We took advantage of the fact that cells and tissues from PtchlacZ embryos can serve as a functional readout of hedgehog activation(Byrd et al., 2002; Milenkovic et al., 1999; Pola et al., 2001) and devised a functional assay to test whether exogenous hedgehog protein reached the Ihh^-/- elements in the renal capsule. We first tested the feasibility of using this reporter assay by transplanting E14.0 wild-type PtchlacZ skeletal elements to a syngenic wild-type renal capsule and evaluating the elements after 48 hours. As anticipated, there was abundant X-gal staining in the cartilage and perichondrium, indicating that Ptch expression was maintained in the wild-type transplanted element(Fig. 7A,B; n=6). When we examined Ihh^-/-;PtchlacZ elements we did not detect any X-gal staining, either in whole mounts or sections(Fig. 7F,G; n=4). The most plausible explanation for this finding was that the reporter was not activated because neither the renal capsule environment nor the skeletal environment itself (i.e. Ihh^-/- chondrocytes) provided functional hedgehog protein. Alternatively, one might argue that Ihh^-/-;PtchlacZ elements were incapable of responding to a hedgehog signal, even if it were present in the environment. To confirm that the Ihh^-/-;PtchlacZ cells could respond, and to ensure that our assay system was sensitive enough to detect a small amount of hedgehog protein in the environment, we conducted another series of experiments in which we placed beads soaked in SHH-N protein next to the Ihh^-/-;PtchlacZ elements. After 48 hours in the renal capsule, Ptch expression was strongly up regulated in both wild-type and mutant elements in response to the exogenous hedgehog protein(Fig. 7C,D,H,I; n=6 wild type and n=4 Ihh^-/-).

Even after 7 or 14 days in the renal capsule we did not detect any up regulation in lacZ expression in spite of the fact that the Ihh^-/-;PtchlacZ elements had undergone vascular invasion and ossification (Fig. 7E,J; n=4 wild type and n=4 Ihh^-/- and data not shown). Thus we concluded that the ex vivo environment does not provide functional levels of exogenous hedgehog protein, and that the renal capsule either supplies a cell population or a blood-born factor that circumvents the hedgehog pathway and permits the differentiation of Ihh^-/- osteoblast precursors. Since Ptch is normally down-regulated in wild-type mature osteoblasts and osteocytes both in vivo and ex vivo (Fig. 7E, and data not shown), these findings collectively suggest that hedgehog signaling is no longer required once a bony matrix has been deposited.

We next focused our attention on how the loss of IHH affected morphogenesis of the perichondrium. During endochondral ossification, the majority of osteoblasts are derived from the perichondrium(Colnot et al., 2004) and although osteoblast precursors were present in Ihh^-/-perichondrium (Fig. 6), the organization of the tissue was clearly affected. The first reported evidence of an Ihh^-/- skeletal defect is at E13.5(St-Jacques et al., 1999), but we decided to re-examine the mutant skeletal elements beginning at the stage when Ihh is first expressed, at E11.5 in the central chondrogenic condensation (Bitgood and McMahon,1995). Within 24 hours of the initial induction of Ihhtranscription (i.e. E12.5) we found that in mutants, Ihh^-/- perichondrial cells failed to align and condense around the cartilage core like their wild-type counterparts(Fig. 8A,F). The reason for this became clearer when we mapped the expression of Ptch and Gli1 to cells that surrounded the wild-type condensation, in the region of the putative perichondrium (Fig. 8B, and data not shown). These two hedgehog target genes were absent from Ihh^-/- limbs(Fig. 8G, and data not shown),correlating with the thin and disorganized perichondrium(Fig. 8F). Thus, an Ihh-dependent perichondrial defect was present well before the advent of the pre-hypertrophic and hypertrophic chondrocyte defects that have been reported previously (St-Jacques et al.,1999).

By E14.0, the Ihh^-/- perichondrial defect was fully manifest. Wild-type perichondria at this stage consisted of organized,flattened cells several layers thick whereas the Ihh^-/-perichondria were still very thin and disorganized(Fig. 8C,H). this morphological defect was accompanied by molecular defects in gene expression: for example, Ptch and Gli1 were no longer expressed by the Ihh^-/- putative perichondrial cells and Col1transcripts were not restricted to cells immediately adjacent to the Col2-expressing chondrogenic condensation as they were in wild-type elements (Fig. 8D,I). Instead,the diffuse Col1 pattern suggested that one defect in the Ihh^-/- mouse may be an inability to recruit Col1-expressing cells to a perichondrial fate.

The Ihh^-/- vasculature was also affected at this very early stage. Vessels surrounding the chondrogenic condensation normally reside within a Ptch-, Gli1-positive domain(Fig. 8B, and data not shown)but in Ihh^-/- elements neither of these hedgehog target genes was expressed, which markedly changed the local environment of the Ihh^-/- endothelial cells(Fig. 8G). So although endothelial cells still surrounded the mutant cartilage condensations, the lack of hedgehog signaling adversely affected their subsequent development. For example, while PECAM-positive endothelial cells normally reside within the inner layer of the perichondria (Fig. 8E, arrows) the same cells were restricted to regions outside the Ihh^-/- perichondria(Fig. 8J, arrows). Although blood vessels eventually infiltrated Ihh^-/- perichondria(data not shown) and reached the hypertrophic cartilage, they did not persist in the putative ossification center (Fig. 3). We therefore concluded that the inability of Ihh^-/- endothelial cells to respond properly to angiogenic signals in and around the ossification center at E14.5 was probably due to improper programming at E11.5, when Ihh is first expressed and its target genes are expressed in this cell type(Fig. 8B, and data not shown).

IHH is required for the ossification of long bones(St-Jacques et al., 1999), and earlier reports suggested that this is because IHH has a direct effect on osteoblast differentiation (Jemtland et al., 2003). This claim is supported by experiments in a chimeric system, in which Ihh-expressing cells induced premature differentiation of adjacent osteoblasts(Chung et al., 2001). Also supporting a direct role for IHH in osteoblast differentiation are conditional mutations in the hedgehog mediator smoothened (Smo), where Smo^-/- perichondrial cells do not participate in the formation of the periosteum or bone collar(Long et al., 2004). In our ex vivo system, we were able to show that perichondrial osteoblasts are capable of differentiating in the absence of an IHH signal (Figs 5, 6, 7), suggesting that IHH may not be required for the differentiation of all osteoblasts and that other factors can overcome the lack of IHH ex vivo. Supporting this conclusion is the fact that dermal bones of the skull vault ossify in the absence of IHH(St-Jacques et al., 1999).

Our results also indicate a role for IHH beyond its potential function as a regulator of osteoblast differentiation. Long before osteoblasts are called upon to differentiate, the Ihh^-/- phenotype is evident. Ihh^-/- cells surrounding the mutant cartilage condensation are disorganized by E12.5 (Fig. 8). Consequently, Ihh^-/- perichondria are thinner and since this tissue is the primary source of osteoblasts during endochondral ossification (Colnot et al.,2004), this provides a probable explanation for the lack of a bone collar in the Ihh^-/- appendicular skeleton.

There is another, earlier defect that contributes to the lack of a bony collar and that is in the recruitment of cells to the perichondrium. The hedgehog-responding domain (marked by Ptch and Gli1expression) initially extends some distance from the edge of the cartilage condensation; with time, the domain becomes restricted to the perichondrium proper (Fig. 8, and data not shown). The consolidation of hedgehog-responsive cells corresponds to the consolidation of Col1, from a broad expression domain to a tightly localized one, surrounding the cartilage condensation(Fig. 8, and data not shown). In Ihh^-/- elements the hedgehog-responding domain is lost and the Col1 expression boundary fails to sharpen in a timely manner(Figs 6 and 8). Thus, the Ihh^-/- osteoblast phenotype is exacerbated because IHH is probably required for the proper segregation of cells to a perichondrial lineage, and later, for osteoblast differentiation.

The Ihh^-/- vascular defect also arises in the early stages of skeletogenesis. First, Ihh^-/- blood vessels do not position themselves properly with respect to the perichondrium(Fig. 8), and second, they do not persist after an initial invasion of hypertrophic cartilage(Fig. 2). These skeletal-associated blood vessels normally develop in a field of active IHH signaling, and since Ihh^-/- vessels develop in the absence of this morphogen, we theorized that the cells might be less mature than wild-type vessels. In comparing the vessels within the Ihh^-/- vascular islands to those present at the time of vascular invasion in wild-type elements, however, we could find no clear difference in their maturity. Both sets of vessels were positive for markers of early (PECAM) and late (COL4) endothelial cell maturation and devoid of SMA-positive pericytes (Fig. 4). Therefore, since SMA-positive pericytes are not present at the time of vascular invasion in wild-type tissues, it is unlikely that a smooth muscle defect contributes to the Ihh^-/- vascular defect(i.e. failure of vessels to expand into a true ossification center and their subsequent disappearance). This is in spite of the fact that pericytes are plainly responsive to hedgehog proteins in the marrow cavity (data not shown)and in adjacent tissues (Fig. 4F,G). Although we were unable to identify a difference in the expression of vascular markers between wild-type and Ihh^-/- blood vessels, these results still leave open the possibility that IHH may act on local vessels in more subtle ways than were detectable by our immunological assays.

Some hints as to these kinds of vascular defects came from the ex vivo environment, where we found that even though Ihh^-/-endothelial cells can invade (Fig. 3), and become associated with pericytes(Fig. 4L), they eventually disappear and leave behind vessels composed entirely of wild-type endothelial cells (Fig. 3). Because wild-type and Ihh^-/- endothelial cells in the ex vivo setting are exposed to the same extracellular environment, the eventual disappearance of Ihh^-/- endothelial cells cannot be attributed to an absence of some trophic signal in the surrounding environment. Neither can the disappearance of mutant endothelial cells be explained by evoking their inability to produce IHH, since wild-type endothelial cells do not normally produce it. They do, however, respond to the morphogen. So while one report refutes the idea that hedgehog proteins act directly on endothelial cells (Pola et al., 2001), others provide evidence that hedgehog proteins directly induce endothelial cells to form capillaries(Kanda et al., 2003; Vokes et al., 2004). The altered functionality of Ihh^-/- endothelial cells associated with the endochondral skeleton would not necessarily manifest itself until the cells were challenged to perform a specific task, such as persisting and expanding within the ossification center (see Fig. 2). Neither would an altered functionality necessarily register on in vitro assays (e.g. migration,proliferation, or differentiation assays), as their function may be context dependent. A subtle effect of hedgehogs on endothelial cell identity and function is not without precedent, as sonic hedgehog is required to impart arterial over venous identity in the zebrafish(Lawson et al., 2002). One enticing possibility then is that IHH modulates the ability of endothelial cells to respond to local environmental cues that direct organ-specific angiogenesis.

Overall, the data shown here and those of others indicate that IHH is involved in multiple steps and acts through multiple mechanisms in the program of skeletogenesis, ranging from a direct role in cartilage and perichondrial differentiation to a role in determining osteoblast and endothelial cell fate. While previous studies have demonstrated that IHH regulates chondrocyte proliferation and maturation (Chung et al.,2001; St-Jacques et al.,1999; Vortkamp et al.,1996), we have here begun to uncover the earliest manifestations of the Ihh^-/- phenotype and to understand how these early defects may contribute to the full-blown defects characteristic of late-stage Ihh^-/- embryos.

The authors thank Angelo Kaplan for technical assistance and Brian Eames for helpful discussions and comments on the manuscript. Thiswork was supported by funds from R01 DE31497 (J.A.H.), P60 DE13058 (J.A.H. and Z. Werb) and R01 AR46238 (Z. Werb and J.A.H.), and funding from the Oaks Foundation to J.A.H. B.S.J. is supported by funds from the Shriners Hospitals of North America.