Glucocorticoid-dependent Wnt signaling by mature osteoblasts is a key regulator of cranial skeletal development in mice

The vertebrate skeleton is formed via two distinct mechanisms:intramembranous and endochondral ossification. Endochondral ossification accounts for the formation of the vertebrae and long bones. In this process,mesenchymal cells first condense and then differentiate into cartilage, which provides the template for the subsequent formation of bone by osteoblasts(Olsen et al., 2000). Hence,endochondral ossification proceeds through stages of cartilage maturation and mineralization before the cartilage template is remodelled into bone.

By contrast, the cranial bones are formed through intramembranous ossification, where mesenchymal precursor cells derived from both the neural crest and mesoderm convert directly into osteoblasts without the precondition of a cartilage intermediate (Jiang et al.,2002). Intramembranous ossification begins with centres of condensing mesenchymal cells in which osteoblasts subsequently differentiate. These centres then expand, and adjacent bones form sutures that consist of mesenchymal cells providing a reservoir of stem cells for the further bone formation and growth (the `osteogenic front')(Opperman, 2000). Premature closure of the sutures results in cranial dysmorphisms such as craniosynostosis, one of the most common human craniofacial deformities(Cohen and MacLean, 2000). Genetic and molecular evidence suggests that factors such as bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs) and the Wnt/β-catenin signaling cascade play an important role in controlling cranial bone formation (Byeong S. Yoon,2004; Day et al.,2005; Liu et al.,2002; Nie et al.,2006; Ohbayashi et al.,2002; Spater et al.,2006). However, the regulation of intramembranous skeletal development upstream of these signaling pathways remains unknown.

Although the formation of calvarial bones does not follow the endochondral ossification pattern, calvarial cartilage does exist during skull formation. Embryonic cranial cartilage is prominent at mid-gestation and grows substantially in the latter part of embryogenesis, reaching its most conspicuous extent at E17.5 (Holmbeck et al., 2003). However, unlike in endochondral bone formation, this unmineralized cartilage does not become mineralized but gradually disappears as intramembranous ossification proceeds. There is some evidence that the cranial cartilage is removed, at least in part, through the action of matrix metalloproteinase 14 (Mmp14) (Holmbeck et al., 2003). As the calvarial cartilage is not directly involved in intramembranous bone formation, the processing of calvarial cartilage has often been ignored in studies of cranial bone development. As a result, the mechanisms responsible for the patterning of cartilage processing during skull development remain obscure. Specifically, it is unclear why and when the cranial cartilage appears and disappears, which interactions occur between cranial cartilage and other tissues, and which cellular signals eventually induce cartilage removal.

Glucocorticoids (GC) play an important role in bone cell differentiation and are known to influence both osteoblast and adipocyte lineage commitment(Herbertson and Aubin, 1995; Shalhoub et al., 1992; Zhou et al., 2008). In addition to the action of GC through its cognate receptor, specific enzymes modulate GC metabolism within the cell at the pre-receptor level(Draper and Stewart, 2005; Stewart and Krozowski, 1999). Within certain tissues, two isoforms of 11β-hydroxysteroid-dehydrogenase(11βHSD) vary intracellular GC concentrations independently of circulating GC levels: 11βHSD type 1 (11βHSD1) predominantly converts inactive cortisone to active cortisol to increase intracellular GC concentrations; by contrast, 11βHSD type 2 (11βHSD2)unidirectionally catalyses the conversion of active GC to their inactive metabolites (Stewart and Krozowski,1999). Kream and colleagues generated a Col2.3-11βHSD2 transgenic mouse, in which the rat gene for 11βHSD2 was linked to the 2.3 kb collagen type I (Col2.3) promoter to target transgene expression to mature osteoblasts (Kalajzic et al.,2002b). The Col2.3 promoter has been well characterized and specifically targets gene expression to mature osteoblasts and osteocytes in the bones of transgenic mice, with no expression in bones derived from wild-type littermates (Kalajzic et al.,2002a; Kalajzic et al.,2005). As overexpression of Col2.3-11βHSD2 in mice results in the inactivation of cytoplasmic corticosterone, GC signaling is effectively disrupted in the mature osteoblasts of Col2.3-11βHSD2 transgenic mice(Sher et al., 2006; Sher et al., 2004). Female Col2.3-11βHSD2 transgenic mice exhibit vertebral osteopenia, reduced femoral cortical bone area and thickness, and impaired mineralized nodule formation in primary calvarial cultures(Sher et al., 2004). We have recently shown that cell cultures derived from Col2.3-11βHSD2 transgenic mice show dominant adipogenesis and reduced osteoblastogenesis in vitro. This phenotypic change is due to a failure in Wnt signaling, which normally allows osteoblasts to exert direct control over the lineage commitment of their mesenchymal progenitors (Zhou et al.,2008).

In the present study, we investigate the role of endogenous glucocorticoids on embryonic and postnatal murine skeletal development. Using a tissue-specific approach via osteoblast-targeted transgenic disruption of endogenous GC signaling, we delineate a novel paracrine mechanism in which osteoblasts - under the control of glucocorticoids - orchestrate the process of intramembranous bone formation by directing mesenchymal cell commitment towards osteoblastic differentiation while simultaneously initiating and controlling cartilage dissolution in the postnatal mouse. This pathway involves Wnt and Mmp14 as downstream effectors of GC action, and therefore may have relevance to wider areas of clinical concern, such as understanding the therapeutic and adverse effects glucocorticoid treatment, and the cellular mechanisms involved in disorders such as cancer and inflammatory joint disease.

Col2.3-11βHSD2 transgenic mice (CD-1 outbred background) were generated as described previously (Sher et al., 2004) and were a gift from Dr Barbara Kream (Department of Medicine, University of Connecticut Health Center, Farmington, CT, USA). Col2.3-GFP mice generated in the CD-1 outbred background(Kalajzic et al., 2002b) were provided as a gift by Dr David Rowe (Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, CT, USA). Mice were maintained at the animal facilities of ANZAC Research Institute (Sydney,Australia) in accordance with Institutional Animal Welfare Guidelines and an approved protocol.

Mice were maintained under specific pathogen-free, temperature-controlled conditions throughout this study at the animal facilities of ANZAC Research Institute in accordance with Institutional Animal Welfare Guidelines and an approved protocol.

After sacrifice, mice were eviscerated and skin was removed. Following fixation in 95% ethanol for 24 hours, mice were stained in Alcian Blue solution (150 mg Alcian Blue, 800ml 98% ethanol, 200ml acetic acid) overnight. After rinsing with 95% ethanol for several hours, the specimens were transferred to 2% KOH for 24 hours and then further stained in Alizarin Red solution (50 mg/l Alizarin Red in 2% KOH) for a further 24 hours. Skeletons were kept in 1% KOH/20% glycerol until the skeletons became clearly visible and then were stored in 50% ethanol/50% glycerol.

For histological analysis, tissues were harvested from Col2.3-11βHSD2 transgenic mice and their wild-type littermates and fixed with 4%paraformaldehyde in PBS for 24 hours, decalcified in 10% EDTA for 3-7 days(depending on animal age), embedded in paraffin and sectioned at 5 μm. Sections were then stained with Hematoxylin and Eosin for morphological studies, with Toluidine Blue for detection of cartilage, and with alkaline phosphatase (ALP) for detection of osteoblastic cells.

Micro-CT of mouse heads was performed using a Skyscan 1172 scanner(SkyScan, Kontich, Belgium). Scanning was carried out at 60 kV, 167 μA,with no filter, and exposure set to 1180 ms. In total, 1125 projections were collected at a resolution of 12.1 μm/pixel. Reconstruction of sections was carried out with software associated with the scanner (Nrecon) with beam hardening correction set to 50%. To obtain 3D visualization from reconstructed sections we used VGStudio MAX 1.2 software (Volume Graphics GmbH, Heidelberg,Germany). The sagittal suture area was measured on digitally recorded projections of micro-CT 3D images using interactive image analysis software(ImageJ, NIH).

Paraformaldehyde-fixed tissue sections on slides were deparaffinized and rehydrated into distilled water through a series of decreasing ethanol washes. TUNEL labeling of apoptotic cells was performed using an In Situ Cell Death Detection Kit (Roche Diagnostics), according to the manufacturer's protocol. The proportion of apoptotic cells was quantified by counting two fields of each section and two sections in each animal.

The 986 bp murine Mmp14 and 889 bp murine Wnt9a probes were generated by reverse transcription PCR using RNA derived from mouse calvarial osteoblasts. The resultant fragments (Mmp14, nucleotides 409-1394, GenBank Accession Number X83536; Wnt9a, nucleotides 3-891, GenBank Accession Number AB072311) were cloned into pGEM-T (Promega). The DIG-labeled riboprobes were transcribed with either T7 or SP6 RNA polymerase to generate antisense and sense riboprobes using a RNA labeling kit (Roche Diagnostics) according to the manufacturer's instructions.

In situ hybridization was performed as previously described(Kartsogiannis et al., 1998). To ensure the sections of wild-type and transgenic mice were hybridized under identical detection conditions, the wild-type and transgenic calvarial samples were always sectioned as a pair and mounted on the same slide. After dewaxing,sections were deproteinized with 0.2 M HCl followed by digestion with proteinase K at 2 μg/ml in 0.1 M Tris buffer (pH 8.0)/50 mM EDTA for 30 minutes at 37°C. Tissues were then fixed in 4% paraformaldehyde for 15 minutes at room temperature before hybridization. Hybridization was performed with hybridization buffer containing DIG-labeled antisense or sense probes at a final concentration of 4-8 ng/μl for 16-18 hours. Slides were washed and the hybridized probe was detected with the alkaline phosphatase-coupled anti-DIG antibody (Roche Diagnostics).

Immunolocalization of Mmp14 was performed on 10 μm cryosections using anti-Mmp14 rabbit polyclonal antibody (1:100; Chemicon International).β-catenin immunohistochemical staining was performed on 5 μm paraffin sections using anti-β-catenin rabbit polyclonal antibody (1:50; Cell Signaling Technology) after heat-induced citrate buffer antigen retrieval. The signals were detected using a biotinylated goat anti-rabbit secondary antibody(1:150 dilution; Vector Laboratories) in combination with the ABC kit (Vector Laboratories) and DAB substrate (Vector Laboratories). Wild-type and transgenic calvarial samples were sectioned as a pair and mounted on the same slide so they were incubated under identical detection conditions.

One-day-old Col2.3-11βHSD2 transgenic mice were injected subcutaneously with 100 ng of recombinant Wnt3a protein (Chemicon International) in 0.01% CHAPS/PBS in a volume of 10 μl over their calvaria above the sagittal suture between the parietal bones daily for 2 days. A group of Col2.3-11βHSD2 transgenic mice received 0.01% CHAPS/PBS injections as vehicle control. The mice were sacrificed 3 days after birth and were subjected to micro-CT and histology analysis.

Parietal bones were dissected by cutting skull bones along coronal and lambdoid sutures. Total RNA from parietal bones was isolated by Trizol(Invitrogen). First-strand cDNA was synthesized from 2 μg of total RNA by incubating for 1 hour at 50°C with Superscript III reverse transcriptase(Invitrogen) following oligo (dT) priming. The expression of mRNAs was estimated by RT-PCR or/and real-time RT-PCR (primer sequences are available upon request). Real-time RT-PCR carried out using IQ SYBR Green Supermix(Bio-Rad) according to the manufacturer's instructions, using a Bio-Rad iCycler iQ5 Real-Time PCR Detection System (Bio-Rad). 18S or GFP were used for cDNA normalization.

All data were presented as the mean±s.e.m. and statistical analyses were performed using Student's t-test. A P<0.05 was considered statistically significant.

Microcomputed tomography (micro-CT) and histological analysis of the calvaria of Col2.3-11βHSD2 transgenic mice revealed a distinct phenotype in neonatal or 1- (P1), 3- (P3) and 7- (P7) day old animals(Fig. 1). Compared with wild-type animals, the frontal, parietal and interparietal bones were hypoplastic and the extent of mineralized bone was greatly reduced in P1 and P3 transgenic mice, resulting in an enlargement of all sutures (best seen at the sagittal suture; Fig. 1B,D). On day 7, coronal sutures were still widely separated with clear gaps at all suture areas in transgenic mice, whereas in wild-type mice all suture gaps were narrowing (Fig. 1E,F). In addition, the cranial bones at the anterior apical region were poorly mineralized, as evidenced by their patchy radio-opacity(Fig. 1F). Despite larger suture areas, the overall size of the head was smaller in transgenic animals compared with their wild-type littermates(Fig. 1A-F).

To determine at what developmental stage this skeletal defect first occurred, embryos of Col2.3-11βHSD2 transgenic mice and their wild-type littermates were compared by whole body skeletal staining. In Col2.3-11βHSD2 transgenic embryos, a reduction in the size and mineralization of the frontal and parietal bones became apparent as early as E15.5 (data not shown). At E16.5, the cranial bones show reduced mineralization (Fig. 1H,K,L)and increased bony separation (Fig. 1L). In addition, a marked reduction in the size of the mineralized regions in the interparietal bones was seen(Fig. 1L).

Another striking abnormality apparent in Col2.3-11βHSD2 transgenic mice was the persistence of intact cartilage underneath the calvarial bones(Fig. 2B). These cartilage remnants were present in all P1 Col2.3-11βHSD2 transgenic mice and significantly larger than those seen in their wild-type littermates(Fig. 2A,B). Histology of the parietal bones revealed a chaotic bone matrix and disorganized osteoblasts in Col2.3-11βHSD2 transgenic mice (Fig. 2C and 2D). By contrast, Toluidine Blue staining confirmed the presence of a broad and vital band of cartilage directly underneath the poorly developed parietal bones in all transgenic mice(Fig. 2F), whereas in wild-type mice only small islets of cartilage were seen, and bone was well organized and almost completely formed (Fig. 2E). Alkaline phosphatase (ALP) staining and evidence of mineralization were absent in the remnant cartilage plates(Fig. 2G,H), indicating this cartilage differs from endochondral cartilage, which usually stains strongly for ALP (data not shown). In addition, numerous ALP⁺preosteoblastic cells were seen at the advancing edge of developing calvarial bone above the cartilage plate in wild-type mice(Fig. 2I) but only a few ALP⁺ osteoblastic cells were present at a similar position in transgenic mice (Fig. 2J). These observations suggest that osteoblast differentiation is suppressed during calvaria development in Col2.3-11βHSD2 transgenic mice.

A further abnormality noted in Col2.3-11βHSD2 transgenic mice is the appearance of ectopic cartilage below the sutures. In 7-day-old (P7)Col2.3-11βHSD2 transgenic mice, the region around the sagittal and lambdoid sutures stained clearly with Alcian Blue, indicating the presence of cartilage (Fig. 3B). Histology of the same regions revealed the presence of ectopic cartilage in the sagittal sutures with greater separation of parietal bones in Col2.3-11βHSD2 transgenic mice (Fig. 3D) when compared with wild-type animals (Fig. 3C). These observations correlate well with the results of the micro-CT studies as described above (Fig. 1E,F). In addition, thicker cartilage plates remained lateral to the parietal bone regions, suggesting that the removal of parietal cartilage is further delayed (Fig. 3F). Whereas on day 7, the parietal bones had formed to their full extent, these bones remained thinner, the osteoblasts stayed disorganized and the sub-parietal cartilage remnants were retained(Fig. 3F,H).

By 2 weeks of age, the abnormal cartilage had disappeared, though the calvarial bones were still thinner (see Fig. S1B in the supplementary material) in transgenic mice compared with wild-type mice (see Fig. S1A in the supplementary material). Micro-CT revealed that the cranial bones were still poorly mineralized with patchy radio-opacity and separated by clear gaps at all suture areas in transgenic mice (see Fig. S1D in the supplementary material). The long bones of transgenic mice, which form by endochondral rather than intramembranous bone formation, did not show any defects as assessed by histology (data not shown).

Mmp14 (or MT1-Mmp) has been shown to be essential for calvarial cartilage removal in rodents. Mice genetically deficient in Mmp14 exhibit impaired cranial cartilage degradation and reduced apoptosis of non-hypertrophic chondrocytes (Holmbeck et al.,2003). Interestingly, these Mmp14 knockout mice seem to share a similar calvarial cartilage phenotype to that seen in Col2.3-HSD2 transgenic mice. This led us to hypothesize that the delayed cranial cartilage removal in Col2.3-11βHSD2 transgenic mice may result from impaired Mmp14 expression.

To this aim, we compared cranial bone sections of neonatal (P0)Col2.3-11βHSD2 transgenic mice and their wild-type littermates. At this age, the cranial cartilage has not been fully removed in wild-type animals,allowing for assessment of Mmp14 expression and activity in wild-type and Col2.3-11βHSD2 transgenic animals. In situ hybridization and immunohistochemistry revealed that both mRNA and protein levels of Mmp14 were lower in cranial cartilage chondrocytes of Col2.3-11βHSD2 transgenic(Fig. 4B,D) compared with their wild-type littermates (Fig. 4A,C). Real time RT-PCR demonstrated that mRNA expression for Col2a1, a cartilage marker, was significantly higher (P=0.003) in transgenic than in wild-type mice, whereas mRNA levels of Mmp14 were four times lower (P=0.01) in transgenic parietal bone compared with their wild-type counterparts (Fig. 4K). In wild-type mice, removal of the calvarial cartilage commenced with proteoglycan loss from the cartilage matrix, followed by a gradual dissolution of the matrix itself(Fig. 4E) and the apoptotic demise of chondrocytes (Fig. 4G,L). By contrast, those processes were not seen in Col2.3-11βHSD2 transgenic mice (Fig. 4F,H). TUNEL staining demonstrated near-complete absence of apoptotic chondrocytes in the cartilage of Col2.3-11βHSD2 transgenic mice(Fig. 4H,L).

Taken together, these data suggest that disrupted GC signaling in mature osteoblasts is associated with a suppression of both Mmp14 expression and Mmp14-dependent apoptosis in the neighbouring chondrocytes. However, Mmp14 is a membrane-type I protein and as there is little or no direct cell-cell contact between osteoblasts and chondrocytes, osteoblast-derived Mmp-14 is unlikely to directly act on the calvarial cartilage cells. Rather, the enzyme needs to be expressed by resident chondrocytes to catalyse cartilage degradation. Immunohistochemical staining confirmed the 11βHSD2 transgene to be expressed in mature osteoblasts only, whereas it was undetectable in chondrocytes (Fig. 4J). This indicates that the suppression of Mmp14 in cranial cartilage chondrocytes is not a direct result of altered GC signaling in chondrocytes but due to the action of paracrine upstream molecules secreted by osteoblasts which regulate Mmp14 expression in calvarial cartilage.

Previously, by culturing primary osteoblasts derived from Col2.3-11βHSD2 transgenic and their wild-type littermates, we have discovered that mature osteoblasts provide a GC-dependent paracrine Wnt signaling to control mesenchymal progenitor cell lineage commitment through the active secretion of Wnt7b and Wnt10b proteins(Zhou et al., 2008). Interestingly, Mmp14 is a direct downstream target gene for the canonical Wnt signaling pathway, as the Mmp14 promoter is directly targeted byβ-catenin/Tcf4 complex (Takahashi et al., 2002). Accumulation of β-catenin by recombinant Wnt3a or LiCl treatment resulted in upregulation of Mmp14 expression in human mesenchymal stem cells (Neth et al.,2006). This led us to examine, by immunohistochemical staining,potential changes in the accumulation of β-catenin protein in calvaria of neonatal Col2.3-11βHSD2 transgenic mice and their wild-type littermates. In wild-type animals, high levels of β-catenin protein were found in both calvarial osteoblasts and chondrocytes(Fig. 5A). By contrast, no accumulated β-catenin protein was detected in Col2.3-11βHSD2 transgenic mice in either cranial osteoblasts or chondrocytes, despite intense staining of neuronal cells (Fig. 5B). These findings indicate that Wnt signaling was suppressed in the calvaria of transgenic mice. Interestingly, loss of Wnt9a leads to ectopic differentiation of cartilage in the sagittal suture and at the base of the parietal bones (Spater et al.,2006), a phenotype similar to that of Col2.3-11βHSD2 transgenic mice. We therefore investigated Wnt9a mRNA expression in parietal bones. In situ hybridization revealed that Wnt9a mRNA was localized in osteoblasts of the calvarial bone surfaces of wild-type mice(Fig. 5E). By contrast, Wnt9a mRNA was not detected in transgenic littermates(Fig. 5F). Interestingly, Wnt9a mRNA was not detectable in parietal cartilage of either wild-type or Col2.3-11βHSD2 transgenic mice (Fig. 5E,F), indicating that Wnt9a is mainly expressed in osteoblastic cells. As there was no transgene detected in cranial chondrocytes, the reduced accumulation of β-catenin protein in the chondrocytes of Col2.3-11βHSD2 transgenic mice is likely to be due to reduced stimulation by Wnt proteins secreted by neighbouring osteoblasts, which are known to lack normal intracellular GC signaling.

To investigate Wnt protein expression levels in the same cell population that expresses the transgene in transgenic mice, we crossed mice expressing green fluorescent protein (GFP) under the control of the identical promoter,Col2.3-GFP mice (Kalajzic et al.,2002b), with Col2.3-11βHSD2 transgenic mice to generate Col2.3-11βHSD2-GFP-transgenic and Col2.3-GFP wild-type littermates. Col2.3-11βHSD2-GFP transgenic mice carry the same phenotype as Col2.3-11βHSD2 transgenic mice. We then assessed, in the parietal bones of Col2.3-11βHSD2-GFP transgenic and Col2.3-GFP wild-type mice, the mRNA expression levels for Wnt7b, Wnt10b and Wnt9a relative to GFP mRNA expression levels. The relative expression in the parietal bones of both Wnt10b and Wnt9a mRNA, which are predominantly transcribed in osteoblasts, was significantly lower in Col2.3-11βHSD2 transgenic than in Col2.3-GFP littermates(Fig. 5G). Wnt7b mRNA was expressed at a very low level in Col2.3-GFP-wild-type animals (just detectable after 40 cycles of RT-PCR) and was not significantly lower in Col2.3-11βHSD2 transgenic mice (data not shown).

If endogenous glucocorticoids regulate parietal cartilage removal and suture expansion through Wnt/β-catenin signaling, the phenotype in transgenic mice should be rescued through activation of β-catenin by exogenous Wnt3a treatment. Recombinant Wnt3a (100 ng) was applied daily for 2 days by subcutaneous injection over the calvaria of 1-day-old mice. At day 3,the parietal bone volume was significantly increased and, as a consequence,suture areas were significantly reduced in Wnt3a-treated transgenic mice as shown by micro-CT analysis (Fig. 6C,D). The cranial bones of Wnt3a-treated mice were well mineralized as evidenced by a more solid appearance and reduced patchy radio-opacity (Fig. 6C).

Histological analysis revealed that whereas cartilage was present underneath the poorly developed parietal bones in Col2.3-11βHSD2 transgenic mice (Fig. 6F), this abnormal cartilage had disappeared and bones were well organized and formed in Wnt3a-treated transgenic mice (Fig. 6G). Likewise, the sutures of Wnt3a-treated transgenic mice(Fig. 6J,M) were comparable with those of their wild-type littermates(Fig. 6H,K), with similar microstructure and suture widths.

In the present study we delineate a novel signaling pathway between osteoblasts and chondrocytes, where mature osteoblasts - under the control of endogenous glucocorticoids - function not only as bone forming cells, but also direct suture expansion and postnatal cartilage removal. We demonstrate that cell-specific, osteoblast-targeted disruption of endogenous GC signaling in mice causes a distinct phenotype characterized by calvarial bone hypoplasia,osteopenia and disorganization of osteoblasts; increased suture patency;ectopic differentiation of cartilage in the sagittal suture; and a defect in the postnatal removal of parietal cartilage. Concurrently, mRNA and protein expression of Mmp14, an enzyme essential for calvarial cartilage removal, was markedly reduced in both the parietal bone and cartilage of transgenic animals. When compared with wild-type mice, the expression of Wnt proteins was significantly lower in transgenic osteoblasts, whereas accumulation ofβ-catenin, the upstream regulator of Mmp14 expression, was significantly decreased in transgenic chondrocytes. These results establish a novel molecular pathway by which glucocorticoids, mainly through the mature osteoblast, regulate the cellular mechanisms that govern cranial skeleton development.

During skull formation, embryonic parietal cartilage is formed as primordium at day E16.5, mostly in the area of the parietal and interparietal bones. This cartilage then grows substantially during the later part of embryogenesis and the early neonatal period, but gradually disappears as ossification proceeds concurrent with the degradation of the extracellular matrix of cartilage and chondrocyte apoptosis. This process is arrested in Mmp14-deficient mice (Holmbeck et al.,1999; Holmbeck et al.,2003; Holmbeck et al.,2005). As Mmp14 is a type I membrane-bound protein, it needs to be expressed by resident chondrocytes to catalyse cartilage degradation. In Col2.3-11βHSD2 transgenic mice, in which GC signaling has been rendered dysfunctional in mature osteoblasts only, cranial cartilage removal was delayed concurrent with a reduction in the expression of chondrocytic Mmp14 mRNA and protein, and decreased chondrocyte apoptosis. These observations indicate that there are paracrine upstream molecules secreted by osteoblasts that regulate Mmp14 expression in calvarial cartilage.

Using calvarial cells in an in vitro culture system, we have previously shown that mature osteoblasts direct mesenchymal progenitor cells to differentiate away from the adipogenic towards the osteoblastic lineage by a glucocorticoid-dependent mechanism (Zhou et al., 2008). Dominant adipogenesis and greatly reduced osteoblastogenesis were observed in calvarial cell cultures from Col2.3-11βHSD2 transgenic mice when compared with wild-type mice. This phenotypic shift in mesenchymal progenitor cell commitment coincided with a reduction in Wnt7b and Wnt10b mRNA and β-catenin protein levels in transgenic versus wild-type cultures. In addition, transwell co-culture of transgenic mesenchymal progenitor cells with wild-type osteoblasts restored commitment to the osteoblast lineage, as did treatment of transgenic cultures with exogenous Wnt3a. The ability of wild-type osteoblasts to restore commitment to the osteoblast lineage was blocked by sFRP1, a Wnt inhibitor.

Given that MMP14 is a canonical Wnt target gene(Neth et al., 2006; Takahashi et al., 2002) and mature osteoblasts are a dominant source of Wnt proteins(Zhou et al., 2008), we proceeded to investigate the nature of canonical Wnt signaling pathways between osteoblasts and neighbouring chondrocytes in calvarial bone. Thus, we demonstrate that Wnt10b and Wnt9a mRNA levels were lower in the parietal bones of Col2.3-11βHSD2 transgenic mice compared with their wild-type littermates. Using in situ hybridization, we further show that in wild-type mice, Wnt9a mRNA localizes only to calvarial osteoblasts but not to chondrocytes. By contrast, large amounts of β-catenin protein were found both in mature osteoblasts and nearby chondrocytes of wild-type mice. Taken together, these findings indicate that chondrocytes can be stimulated by osteoblast-derived Wnt proteins to initiate the intracellular canonical Wnt signaling cascade. By contrast, Wnt9a was not detectable in Col2.3-11βHSD2 transgenic mice by in situ hybridization and, accordingly,β-catenin protein was absent from calvarial chondrocytes and present at reduced levels in only a few nearby osteoblasts. These observations suggest that activation of β-catenin in chondrocytes of parietal cartilage is dependent on Wnts produced by the neighbouring osteoblasts under the control of endogenous glucocorticoids. The primary role of these secreted canonical Wnts is to control calvarial bone and cartilage development. The reason why Wnt7b expression was observed in cultured calvarial osteoblasts(Zhou et al., 2008), instead of Wnt 9a as seen in our in vivo experiments, is currently unclear but is probably due to differences in the local regulation of Wnt expression occurring in vivo. Importantly, applying recombinant Wnt3a, a canonical Wnt protein, by supracalvarial injection to transgenic mice resulted in complete cartilage degradation. This finding adds further proof to the concept that Wnt signaling acts as an upstream regulator of Mmp14 expression during murine cranial development.

During skull growth, the sutures serve as growth centres, where mesenchymal cells reside as a reservoir for postnatal osteogenesis and new bone formation. In this process, Wnt/β-catenin signaling is required to suppress chondrogenesis and to allow osteoblasts to form(Day et al., 2005; Hill et al., 2005). It has been shown that either knockout of Wnt9a or inactivation of β-catenin in mesenchymal cells induces ectopic cartilage formation in the developing calvaria, particularly below the sutures(Day et al., 2005; Spater et al., 2006). By contrast, mice with mutant Axin2, a negative regulator of the canonical Wnt pathway that promotes degradation of β-catenin, display premature closure of sutures as a result of excessive β-catenin accumulation(Yu et al., 2005). Thus, it seems clear that canonical Wnt signaling is essential for intramembranous ossification. However, so far the source of Wnt in this signaling cascade has remained obscure. In the present study, we observed that Col2.3-11βHSD2 transgenic mice show a developmental phenotype similar to that of Wnt9a mutant mice (Spater et al., 2006)with calvarial bone hypoplasia and osteopenia, increased suture patency, and ectopic differentiation of cartilage in the sagittal suture. Of note, the phenotype in Col2.3-11βHSD2 transgenic mice was associated with a dramatic reduction in β-catenin protein accumulation in calvarial osteoblasts and progenitor cells located in the sutures, indicating that canonical Wnt signaling was attenuated. Treatment with exogenous Wnt3a by supracalvarial injection rescued the transgenic phenotype to resemble the wild type with reduced suture areas, well mineralized cranial bones and complete cranial cartilage removal. Hence, depending on the cells' local environment,canonical Wnt signaling may act as a molecular switch between osteoblast,chondrocyte and adipocyte cell fates when mesenchymal progenitor cells are differentiating.

The pronounced disorganization of both the resident osteoblasts and the collagenous bone matrix in Col2.3-11βHSD2 transgenic calvaria indicate that endogenous GC, via osteoblastic Wnt signaling, may have a direct effect on bone formation through modulating mature osteoblast function. We have found reduced β-catenin protein accumulation and Mmp14 expression not only in calvarial chondrocytes but also in Col2.3-11βHSD2 transgenic calvarial osteoblasts. In addition, Wnt3a treatment not only rescued the abnormal cartilage phenotype, but also significantly improved parietal bone formation,mineralization and, consequently, suture narrowing. It is interesting to note that Mmp14-deficient mice develop a similar phenotype of disturbed cranial intramembranous ossification, characterized by reduced calvarial bone formation in association with chaotic bone matrix organization and disordered osteoblasts (Holmbeck et al.,1999). These findings indicate that Mmp14 may play an important role in normal bone formation. We therefore hypothesize that osteoblasts are controlled through a GC-dependent autocrine Wnt signaling loop. Therefore, in addition to the lack of paracrine Wnt signaling to mesenchymal progenitor cells, reduced autocrine Wnt signaling in transgenic osteoblasts may impair osteoblastic Mmp14 expression, contributing to the phenotype of disturbed calvarial bone formation as observed in Col2.3-11βHSD2 transgenic mice.

In this model (Fig. 7), we suggest that endogenous glucocorticoids stimulate the expression and secretion of Wnt proteins in mature cranial osteoblasts, in keeping with our previous results (Zhou et al., 2008). The ensuing canonical Wnt signaling cascade induces: (1) mesenchymal progenitor cells to differentiate away from the chondrocyte towards the osteoblast lineage to form bone; (2) osteoblasts to initiate Mmp14-mediated remodeling of the collagenous matrix surrounding osteoblasts; and (3) parietal cartilage chondrocytes to initiate Mmp14-mediated cartilage degradation. These concurrent and tightly interconnected pathways establish a novel role for both glucocorticoids and osteoblasts in the intricate process of intramembranous bone development.

Disruption of GC signaling in mature osteoblasts results in delayed calvarial (i.e. intramembranous) development without affecting the endochondral formation of long bones. Day and colleagues have shown that intramembranous ossification requires high levels of β-catenin, through upregulation by Wnt signaling, to promote osteoblast differentiation. By contrast, during endochondral ossification, β-catenin protein levels are kept low by inhibition of canonical Wnt signaling inside mesenchymal condensations to ensure that only chondrocytes can form initially(Day et al., 2005). This difference may explain why long bone development (i.e. endochondral ossification) is not affected in Col2.3-11βHSD2 transgenic mice.

Our study highlights an important role for glucocorticoids in skeletal function that is mediated through regulation of Wnt signaling with downstream effects on Mmp14, and therefore may have relevance to wider areas such as understanding the therapeutic and adverse effects GC treatment. This study also highlights the importance of the interaction of glucocorticoids and Wnt proteins with Mmp14, the latter being a significant mediator of adverse outcomes in cancer and chronic arthritis. Thus, our findings may open novel avenues to advance research into the mechanism of malignant and inflammatory joint disease.