Axin2 controls bone remodeling through the β-catenin–BMP signaling pathway in adult mice

The Wnt family consists of a number of small, cysteine-rich, secreted glycoproteins that are involved in the regulation of a variety of cellular activities that are crucial for development (Huelsken and Birchmeier, 2001; Moon et al., 2002; Westendorf et al., 2004). The canonical Wnt signaling pathway is mediated by a protein complex containing β-catenin, through regulation of β-catenin protein levels and its subcellular localization. In the absence of Wnts, β-catenin levels are kept in a steady state. Free β-catenin proteins are ubiquitylated and degraded by the 26S proteasome (Aberle et al., 1997). A multiprotein complex containing the kinases glycogen synthase kinase 3β (GSK-3β) and casein kinase 1 (CK1) along with scaffolding proteins Axis inhibition protein 1 (Axin1), Axin2, adenomatous polyposis coli (APC) and disheveled (Dsh) mediates the degradation of excess β-catenin. This complex phosphorylates specific amino acid residues on β-catenin, creating a docking site for the F-box protein/E3 ligase complex (Behrens et al., 1998; Jiang and Struhl, 1998). Therefore, inhibition of β-catenin phosphorylation prevents degradation of β-catenin, increases its cytoplasmic level, and facilitates its nuclear translocation. Signaling from Wnt proteins makes the complex release β-catenin, and allows it to move to the nucleus, where it interacts with TCF/LEF transcription factors to activate the expression of target genes (Staal and Clevers, 2000; Moon et al., 2002; Cliffe et al., 2003; Westendorf et al., 2004).

At the cell surface, Wnts interact with two classes of proteins: frizzled receptors and low-density-lipoprotein (LDL)-receptor-related protein 5 or 6 (LRP5/6). There are many genes encoding frizzled proteins, including ten in the human genome. Individual frizzled proteins probably have different affinities for various types of Wnt proteins. Wnt proteins can form a complex with both the cysteine-rich domain (CRD) of frizzled and with LRP5/6 simultaneously, leading to the formation of a dual-receptor complex (Bejsovec, 2000; Huelsken and Birchmeier, 2001). The intracellular portion of the receptors communicates this binding information and turns on the pathways that act on β-catenin inside the cell. Following Wnt binding, the intracellular tail of LRP5/6 binds Axin1 (or Axin2) and causes dissociation of β-catenin from its protein complex and activation of β-catenin signaling (Mao et al., 2001).

Axin1 and its homolog Axin2 are negative regulators of canonical Wnt signaling through promotion of β-catenin degradation (Behrens et al., 1998; Kikuchi, 1999; Peifer and Polakis, 2000). Axin1 has interaction sites for proteins involved in Wnt signal transduction, including β-catenin, GSK-3β, CK1, APC, Dsh and LRP5 (Luo and Lin, 2004). Significantly, Axin1 associates with β-catenin and protein kinases GSK-3β and CK1. In this complex, β-catenin is phosphorylated and degraded by the ubiquitin-proteasome pathway. Wnt binding recruits Axin1 to the receptor complex, where it binds the cytoplasmic domain of LRP5 (Cong et al., 2004; Hay et al., 2005; Mao et al., 2001). Axin1 is a central component of the canonical Wnt pathway and acts as a scaffold for the protein complex involved in β-catenin degradation.

Mouse Axin2 cDNA encodes a protein of 840 amino acids (Axin2, also known as conductin). Several important binding domains in Axin2 have been mapped. Axin2 has a β-catenin-binding domain that is located in the central region of Axin2 protein (amino acids 396-465). Axin2 contains a N-terminal RGS (regulator of G protein signaling) domain (amino acids 78-200), a GSK-3β binding domain (amino acids 343-396) and a C-terminal sequence related to the protein Dsh (amino acids 783-833) (Behrens et al., 1998). Axin2 has been shown to mediate APC-induced β-catenin degradation. The Axin2 RGS domain binds the SAMP repeats of APC. A double mutant of β-catenin (W383A/R386A) does not bind APC but this mutant is still degraded by exogenous wild-type (WT) APC, indicating that APC does not need to bind directly to β-catenin to induce β-catenin degradation. A triple mutant β-catenin (W383A/R386A/H260A) does not bind APC or Axin2 and this mutant β-catenin is fully stable in the presence of exogenous WT APC (von Kries et al., 2000). These results demonstrate that Axin2 can link APC to β-catenin and mediate APC-induced β-catenin degradation.

Axin2 has 44% homology with Axin1 and its function is similar (Behrens et al., 1998; Chia and Costantini, 2005), but they have different expression patterns. Axin1 is widely expressed, whereas Axin2 is differentially expressed in various cells and tissues, and during different stages of maturation (Chia and Costantini, 2005; Yu et al., 2005). The deletion of the Axin1 gene results in early embryonic mortality. Mice die at embryonic day 9.5 with forebrain truncation, neural tube defects and axis duplications (Zeng et al., 1997). Knock-in of the Axin2 gene into the deleted Axin1 gene rescues the phenotype of Axin1-deficient mice (Chia and Costantini, 2005). A lacZ knock-in allele that is a null allele of Axin2 (Axin2^–/–) was generated by insertion of the lacZ cDNA into exon2 of the Axin2 gene. The homozygous Axin2 knockout (KO) mice are viable and fertile, but have craniofacial defects and premature closure of the cranial sutures due to increased β-catenin signaling (Yu et al., 2005). Nevertheless, despite some overlapping functions, Axin1 and Axin2 do not have redundant functions, which probably reflects their differential expression patterns.

Although canonical Wnt/β-catenin signaling plays a crucial role in controlling bone development, its role in bone remodeling remains poorly understood because either deletion of the β-catenin gene in a conventional method or tissue-specific manner often leads to embryonic lethality. Using the Axin2 KO mice as a unique mouse model, we have investigated the role of the Axin2/β-catenin signaling pathway in bone remodeling. Our findings demonstrate that Axin2/β-catenin signaling targets Bmp2/4, regulates Osx expression and controls osteoblast differentiation in osteoblast progenitor or precursor cells.

To investigate changes in bone mass and bone microstructure, we analyzed the metaphyseal region of long bones (femora) of 2-, 6- and 12-month-old WT and Axin2 KO mice using micro-CT imaging. No obvious changes in trabecular bone parameters were observed in long bones of 2-month-old Axin2 KO mice compared with WT controls (data not shown). By contrast, a significant increase in bone volume (BV) and bone mineral density (BMD) was observed in 6- and 12-month-old Axin2 KO mice (Fig. 1). No significant difference in bone parameters was observed between heterozygous Axin2 KO mice and WT controls at these three time points (data not shown). These results suggest that Axin2 might play a dominant role in bone formation in older mice.

Trabecular bone structure was significantly denser and more robust in 6-month-old female Axin2 KO mice. Specifically, the bone volume fraction (BVF; BV as a percentage of tissue volume, TV) was 61% higher (Axin2 KO: 12.7±1.8% versus WT: 7.9±1.0%, P<0.05, n=6) (Fig. 1A,B) and bone mineral density (BMD = BMC/TV) was 31% higher (Axin2 KO: 214±12 mg/cm³ versus WT: 164±17 mg/cm³, P<0.05, n=6) in the Axin2 KO mice than WT controls (Fig. 1C). However, the structural model index (SMI) (a measure of the shape of trabeculae; 0 for plates and 3 for cylindrical rods) was significantly decreased in these mice versus WT controls (Fig. 1D), reflecting plate-like morphology of trabeculae in Axin2 KO mice compared to the rod-like morphology in the WT mice. Trabecular number (Tb.N.) and trabecular thickness (Tb.Th) were significantly increased (by 16% and 15%, respectively) in the KO mice (Fig. 1E,F), reflecting increased trabecular volume of the Axin2 KO phenotype. By contrast, trabecular separation (Tb.Sp.) was significantly decreased (17%) in Axin2 KO mice (Fig. 1G) compared to WT controls. To determine the expression patterns of Axin1 and Axin2 in bone cells, we isolated bone marrow stromal (BMS) cells from 2- and 6-month-old wild-type (WT) C57BL/6J mice and examined age-related changes in Axin1 and Axin2 expression. We found that Axin1 expression was reduced 47% and Axin2 expression was increased ninefold in 6-month-old mice compared to that in 2-month-old mice (Fig. 1H,I).

The differences in BV and BMD were much more pronounced and significant by 12 months of age. The BVF was increased 250% (Axin2 KO: 9.8±2.4% versus WT: 2.8±0.5%, P<0.05, n=6) (Fig. 1J,K) and BMD was increased 73% (Axin2 KO: 171±26 mg/cm³ versus WT: 99±8, P<0.05, n=6) in Axin2 KO mice compared with WT controls (Fig. 1L). The SMI was decreased 42% in Axin2 KO mice (Fig. 1M), again reflecting trabeculae that resembled plate-like morphology compared to the rod-like morphology of the WT trabeculae. The Tb.N. and the Tb.Th. were increased by 39 and 21%, respectively, and Tb.Sp. was decreased 24% in Axin2 KO mice (Fig. 1N-P). Similar, but less marked changes were also found in 6- and 12-month-old male Axin2 Ks mice (data not shown).

Long-bone diaphyseal morphology was not found to be different between WT and Axin2 KO femurs nor between tibia, using micro-CT cross-sectional imaging. Diaphyseal cross-sectional area, bending moments of inertia, outer diameter and BMD were not significantly different (Table 1), with the exception that Axin2 KO femoral cortical thickness was 10% greater, whereas the outer medial-lateral diameter was 9% smaller (P<0.05). To determine the effect of Axin2 on bone mechanics, eight femurs from 12-month-old WT and seven femurs from Axin2 KO mice underwent three-point bending mechanical testing. Femurs from Axin2 KO animals were significantly stiffer by 31%, with a maximum bending moment that was 20% greater (P<0.05, Table 2) than for WT controls.

Consistent with the phenotypic changes measured by micro-CT, histological analysis also showed a significant increase in bone volume in 6- and 12-month-old female Axin2 KO mice (Fig. 2A,B). The mineral appositional rates (MAR) and bone formation rates (BFR) were increased 64 and 86%, respectively, in 6-month-old female Axin2 KO mice, as demonstrated by histomorphometric analysis (Fig. 2C,D). The micro-CT and histomorphometric results demonstrated a dramatic increase in bone mass in 6- and 12-month-old Axin2 KO mice and suggest that Axin2 is a negative regulator in the maintenance of bone mass in adult mice. Because BFR is increased in Axin2 KO mice, the data also suggest that osteoblast function is elevated in Axin2 KO mice.

To determine whether increased bone mass in Axin2 KO mice is also due to a decrease in osteoclast formation, we also analyzed changes in osteoclast numbers. We found that osteoclast surfaces and osteoclast numbers were significantly reduced in 6-month-old Axin2 KO mice (Fig. 2E,F). We also examined the ability of BMS cells to form osteoclasts in response to M-CSF (macrophage-colony stimulating factor) and RANKL (receptor activator for nuclear factor-kB ligand). The results showed that osteoclast formation was significantly reduced (n=4) in the cells derived from 6-month-old Axin2 KO mice compared with formation in the cells obtained from WT controls under the same culture conditions (Fig. 2G,H). In addition, we also found that osteoclastic bone resorption was also significantly reduced in bone marrow cells derived from Axin2 KO mice (Fig. 2I). The mRNA expression of osteoprotegerin (Opg) in BMS cells was significantly increased in 6-month-old Axin2 KO mice (Fig. 2J). By contrast, Rankl expression was slightly reduced in Axin2 KO mice (Fig. 2K). In addition, the expression of cathepsin K was significantly reduced in Axin2 KO mice (Fig. 2L). Consistent with the increased Opg mRNA expression, serum OPG protein levels were also significantly increased in Axin2 KO mice (Fig. 2M). These results suggest that reduction in osteoclast formation and bone resorption might also contribute to the high-bone-mass phenotype observed in Axin2 KO mice.

To determine the role of Axin2 in bone loss induced by estrogen deficiency, we performed an ovariectomy experiment in 6-month-old female Axin2 KO mice and WT controls. The changes in bone volume were monitored by micro-CT analysis. There was 48% bone loss (BV/TV) in WT mice 4 weeks after ovariectomy compared with sham-operated mice (n=10). By contrast, only 36% bone loss was found in Axin2 KO mice 4 weeks after ovariectomy (n=10) (supplementary material Fig. S1). These findings demonstrated that Axin2 KO mice are partially resistant to overiectomy-induced bone loss and that this might be related to the increased Opg expression in Axin2 KO mice.

To investigate changes in cellular function, we isolated BMS cells from 6-month-old Axin2 KO mice and WT controls and cultured these cells for different periods of time. We first examined cell proliferation by immunostaining using an anti-Ki-67 antibody and found a significant increase in Ki-67-positive cells (5.7-fold increase) in BMS cells from Axin2 KO mice (Fig. 3A) versus WT controls. It has been reported that β-catenin directly activates cyclin D1 gene transcription in tumor cells (Tetsu and McCormick, 1999). We analyzed changes in cyclin D1 expression and found that the expression of cyclin D1 mRNA and protein was significantly increased in BMS cells derived from Axin2 KO mice compared with that in WT BMS cells (Fig. 3B,C).

We determined changes in osteoblast differentiation by examining a panel of osteoblast maker genes expressed at various developmental stages using enzyme activity assay and real-time PCR assays. The alkaline phosphatase (ALP) activity (Fig. 3D,E) and the expression of osteoblast-specific transcription factors, Runx2 and osterix (Osx), were significantly increased in BMS cells derived from 6-month-old Axin2 KO mice (Fig. 3F,G). As a consequence, the expression of early osteoblast differentiation markers, type I collagen (Col1) and osteopontin (Opn) was also significantly increased in Axin2 KO BMS cells (Fig. 3H,I). The expression levels of later-stage marker genes such as bone sialoprotein (Bsp) and osteocalcin (Oc) were also significantly increased in Axin2 KO cells (Fig. 3J,K). These data suggest that osteoblast differentiation was largely advanced in the Axin2 KO BMS cells. In contrast to the findings observed in BMS cells derived from 6-month-old Axin2 KO mice, no significant changes in cell proliferation, ALP activity or the expression of osteoblast marker genes were found in BMS cells derived from 2-month-old Axin2 KO mice (data not shown). Osteoblast apoptosis was examined by TUNEL staining and no changes in cell apoptosis were found between the Axin2 KO and WT BMS cells from either 2- or 6-month-old mice (data not shown). These in vitro studies demonstrate that the process of osteoblast proliferation and differentiation is accelerated in the Axin2 KO mice.

It has been reported that β-catenin promotes mesenchymal progenitor cells and osteoblast precursor cells to differentiate into osteoblasts (Hill et al., 2005; Day et al., 2005; Rodda and McMahon, 2006) and that it stimulates Opg expression in mature osteoblasts (Glass et al., 2005; Holmen et al., 2005). Because the phenotype found in Axin2 KO mice is consistent with the phenotypes observed in mice with β-catenin gene deletions at the different stages of osteoblast lineages, we examined cytoplasmic and nuclear β-catenin protein levels to determine changes in β-catenin signaling in Axin2 KO mice. Although the protein levels of β-catenin in cytoplasm were slightly increased, the protein levels of β-catenin (total and non-phosphorylated active form) in the nucleus were significantly increased in BMS cells derived from 6-month-old Axin2 KO mice (Fig. 4A,B). These results suggest that the bone phenotype observed in Axin2 KO mice could be due to the activation of β-catenin in osteoblasts.

To confirm the role of β-catenin activation in the bone phenotype of Axin2 KO mice, we crossed Axin2 KO mice with β-catenin^fx/fx mice and generated Axin2^–/–;β-catenin^fx/fx mice. BMS cells were then isolated from 6-month-old WT, Axin2^–/–, β-catenin^fx/fx and Axin2^–/–;β-catenin^fx/fx mice and infected with lentivirus expressing Cre recombinase (Lenti-Cre). ALP activity and the expression of osteoblast marker genes were examined in these cells. The results showed that the increased ALP activity and increased Bmp4 and Oc expression observed in Axin2 KO mice were completely inhibited in β-catenin-deleted BMS cells under the Axin2 KO background (Fig. 4C-E). In addition, the increased Bmp2 expression found in Axin2 KO mice was also significantly inhibited in β-catenin-deleted BMS cells (Fig. 4F). In vitro deletion of the β-catenin gene in BMS cells derived from 6-month-old Axin2^–/–;β-catenin^fx/fx mice was confirmed by western blotting (Fig. 4G). These results strongly suggest that increased osteoblast function in Axin2 KO mice is due to the activation of β-catenin in BMS cells.

The synergistic effect between β-catenin and bone morphogenetic protein (BMP) signaling in bone cells has been reported (Fischer et al., 2002; Mbalaviele et al., 2005; Nakashima et al., 2005; Chen et al., 2007). However, the mechanism remains unknown. To investigate the interaction between BMPs and Axin2 and/or β-catenin, we examined changes in expression levels of Bmp. We analyzed the mRNA levels of Bmp in BMS cells by real-time PCR. The results demonstrated that mRNA levels of Bmp2 and Bmp4 were significantly increased in Axin2 KO BMS cells (Fig. 5A,B). By contrast, the mRNA levels of Bmp6 and Bmp7 was not significantly changed in Axin2 KO mice (data not shown). To determine whether signaling proteins downstream of BMP are activated in Axin2 KO mice we measured changes in protein levels of phosphorylated Smad1/5, the key signaling molecules in BMP signaling. The basal levels of phosphorylated Smad1/5 (two BMP signaling Smads expressed in osteoblasts) were dramatically increased in the cells derived from 6-month-old Axin2 KO mice. There was no change in the protein levels of total Smad1 (Fig. 5C), suggesting that the increased phosphorylated Smad1/5 level is not due to changes in protein degradation. In addition, BMP-2-induced formation of mineralized bone nodules was significantly enhanced in BMS cells derived from Axin2 KO mice (13-fold increase for Axin2 KO cells versus sixfold increase for WT cells) (Fig. 5D,E). Similar results were also obtained when BMS cells were incubated with BMP4 (data not shown). In contrast to the changes in BMP signaling, PTH-induced ALP activity was similar in the BMS cells from Axin2 KO and WT mice (data not shown).

To further determine the regulatory mechanism of Bmp expression by Axin2 and β-catenin, we utilized the chemical compound BIO (6-bromoindirubin-3′-oxime), a GSK-3β inhibitor. BIO activated β-catenin signaling, stimulated ALP activity (Fig. 6A-D) and increased new bone formation in the periosteal bone formation assay in WT mice (Fig. 6E,F). BIO stimulated the expression of Bmp2 and Bmp4 genes over a 24-hour time period in primary osteoblasts (Fig. 7A,B). To determine whether BIO-induced osteoblast differentiation is dependent on Bmp2/4, primary calvarial osteoblasts were isolated from Bmp2^fx/fx;Bmp4^fx/fx mice and infected with adenovirus-expressing Cre recombinase (Ad-Cre). Infection of Ad-Cre in osteoblasts efficiently deleted Bmp2 and Bmp4 genes in these cells (Fig. 7C,D). BIO stimulated Runx2 expression in both WT and Bmp2/4-deleted cells (Fig. 7E), suggesting that BIO-induced Runx2 expression is Bmp2/4-independent. By contrast, BIO-induced Osx expression was significantly inhibited and BIO-induced Col1, Alp and Oc expression was completely abolished in Bmp2/4-deleted osteoblasts (Fig. 7F-I), suggesting that BIO-induced Osx and subsequent osteoblast marker gene expression are Bmp2/4-dependent.

To further determine the role of Osx in BIO-induced osteoblast differentiation, Osx RNAi was transfected into primary osteoblasts, which were subsequently treated with BIO. The results showed that BIO-induced expression of Alp, Col1 and Oc was significantly or completely inhibited when Osx gene expression was knocked down by Osx siRNA (Fig. 7J-M). These results demonstrated that BIO (or activation of β-catenin signaling) activates Osx expression and subsequently regulates the expression of other osteoblast marker genes.

Osteoporosis is a major health threat for the elderly, particularly for postmenopausal women. It is characterized by low bone mass, induced by ineffective bone remodeling, which leads to an increased incidence of fractures and to higher morbidity and mortality in affected individuals. Bone remodeling, a dynamic process involving bone resorption followed by bone formation, takes place throughout life to maintain bone homeostasis. In osteoporotic patients, progressive bone loss with age could be due to increased bone resorption or inadequate bone formation, or to a combination of both. A better understanding of the mechanisms through which bone formation and resorption are controlled is therefore crucial in determining the pathogenesis of osteoporosis and identifying potential molecular targets for the development of new therapies for this debilitating disease.

In the present studies, we discovered that bone mass is significantly increased in 6- and 12-month-old adult Axin2 KO mice in which lacZ cDNA was inserted into exon2 of the Axin2 gene and thereby effectively prevented Axin2 expression in all cells of the mice (Yu et al., 2005). The present studies demonstrate for the first time that Axin2 plays a crucial role in the maintenance of bone mass in adult mice. Our in vivo results are complemented by cell-culture studies that show that osteoblast function is increased in vitro in cells from Axin2 KO mice. We also found that mRNA expression of Bmp2 and Bmp4 and phosphorylated Smad1/5 protein levels are increased in Axin2 KO BMS cells. These findings suggest an important cross-talk between the Axin2/β-catenin and BMP signaling pathways.

Axin2 is a scaffolding protein and a negative regulator of β-catenin function (Behrens et al., 1998). It has been reported that the N-terminal RGS domain of Axin2 binds to APC and mediates APC-induced β-catenin degradation (von Kries et al., 2000). In the present studies, we found that activation of β-catenin signaling is the major event in BMS cells of Axin2 KO mice. β-catenin plays a crucial role in the commitment and differentiation of mesenchymal cells into the osteoblast lineage. Conditional deletion of the β-catenin gene in mesenchymal progenitor cells (targeted by Prx1-Cre or Dermo1-Cre mice) (Hill et al., 2005; Day et al., 2005) leads to the inhibition of osteoblast differentiation, suggesting that β-catenin promotes osteoblast differentiation in osteoblast progenitor cells. By contrast, continued activation of β-catenin signaling in Osx1⁺ cells (targeted by Osx1-Cre mice) promotes cell differentiation in Runx2⁺/Osx1⁺ osteoblast precursor cells but arrests osteoblasts in a low Oc expressing stage (Rodda and McMahon, 2006). In the present studies, we found that nuclear β-catenin protein levels are increased and osteoblast proliferation and differentiation are accelerated in Axin2 KO mice. The results from the rescue experiments further demonstrate that activation of β-catenin signaling is the primary cause for the bone phenotype of the Axin2 KO mice. These results suggest that Axin2 regulates osteoblast function through β-catenin signaling in BMS cells, mainly through osteoblast precursor cells.

In the present studies, we found that Opg expression is increased in Axin2 KO mice. By contrast, the expression of Rankl is not significantly altered. As a consequence, osteoclast formation is reduced in Axin2 KO mice. Although it seems that changes in osteoblast function are more pronounced than those in osteoclast function, the decreased osteoclast formation might also contribute to the bone phenotype observed in Axin2 KO mice. In mature osteoblasts, β-catenin specifically activates Opg expression and inhibits osteoclast formation (Holmen et al., 2005; Glass et al., 2005). In the present studies, we observed a significant increase in Opg expression and a decrease in osteoclast formation in BMS cells of Axin2 KO mice. To further investigate whether Axin2 directly targets mature osteoblasts to regulate osteoclast formation and bone mass, we have recently generated 2.3Col1-Axin2 transgenic mice in which the expression of an Axin2 transgene is driven by the 2.3 kb mouse type I collagen promoter. The bone phenotype of these mice (two independent lines) was analyzed by micro-CT and histological methods. We found that there were no significant changes in bone mass, bone microstructure and bone mechanical properties in 3-month-old 2.3Col1-Axin2 transgenic mice compared with the WT littermates (supplementary material Fig. S2). These results suggest that Axin2 might be mainly targeting osteoblast progenitor and precursor cells, instead of mature osteoblasts.

In these studies, we found that mRNA expression of Bmp2 and Bmp4 and levels of phosphorylated Smad1/5 proteins are significantly increased and addition of noggin, an antagonist of BMP-2 and BMP-4, completely blocks elevated phosphorylated Smad1/5 levels in Axin2 KO BMS cells. Our in vitro studies further demonstrate that the GSK-3β inhibitor BIO induces osteoblast-specific transcription factor Osx and osteoblast marker genes in a Bmp2/4-dependent manner. These findings suggest that Bmp2 and Bmp4 are downstream mediators of Axin2/β-catenin signaling in osteoblast precursor cells and that the accelerated osteoblast differentiation observed in Axin2 KO mice might be the result of the activation of BMP signaling in osteoblast precursor cells. Consistent with these findings, in a separate study, we found that Wnt3a induces the expression of hypertrophic marker gene type X collagen and this effect is completely blocked by addition of noggin to the cell cultures of primary chondrocytes (Chen et al., 2008), suggesting that the BMPs are downstream mediators of canonical Wnt signaling in chondrocytes. Previous studies have shown that Axin2 KO mice have cranial bone defects, a phenotype resembling craniosynostosis in humans (Yu et al., 2005). Further, BMP signaling is activated in calvarial osteoblasts of Axin2 KO mice (Liu et al., 2007). However, it remains to be determined whether osteoblast differentiation induced by Axin2 deletion (or activation of β-catenin) is totally dependent on the activation of BMP signaling. Our evidence that deletion of the Bmp2/4 genes abrogates the BIO-induced Osx and subsequent osteoblast marker gene expression suggests that Axin2 deletion might trigger the β-catenin–Bmp2/4–osterix signaling pathway in osteoblast precursor cells.

It has been reported that β-catenin activates the bone-specific transcription factor Runx2 in osteoblasts and chondrocytes (Gaur et al., 2005; Dong et al., 2006). In the present studies, we found that both Runx2 and Osx expression were increased in Axin2 KO mice and that treatment with BIO in calvarial osteoblasts enhanced the expression of Runx2, Osx and other osteoblast marker genes. To determine whether BIO-induced osteoblast marker gene expression is Bmp2/4-dependent, primary calvarial osteoblasts were isolated from Bmp2^fx/fx;Bmp4^fx/fx mice and infected with Ad-Cre and treated with BIO. We found that BIO-induced Runx2 expression was not affected by deletion of the Bmp2/4 genes. By contrast, deletion of the Bmp2/4 genes significantly inhibited the BIO-induced Osx expression and completely abolished BIO-induced expression of other osteoblast marker genes. We further demonstrated that Osx siRNA also inhibited BIO-induced osteoblast marker gene expression. These observations suggest that BIO (or activation of β-catenin signaling) might stimulate osteoblast differentiation through a Bmp2/4-Osx pathway.

In the present studies, we found a significant increase in the levels of phosphorylated Smad1/5 in BMS cells derived from Axin2 KO mice. A possible mechanism for this could be increased Smad1/5 phosphorylation or decreased Smad1/5 dephosphorylation. Three different types of phosphatases have recently been identified to specifically induce Smad1/5 dephosphorylation. These are pyruvate dehydrogenase phosphatase 1 and 2 (PDP1 and PDP2), small C-terminal domain phosphatase (SCP) and magnesium-dependent protein phosphatase 1α (PPM1α) (Chen et al., 2006; Knockaert et al., 2006; Duan et al., 2006). We have obtained expression plasmids of these phosphatases from different collaborators and tested the effects of Axin2 on Smad1/5 dephosphorylation induced by these phosphatases. We found that PPM1A is the most potent phosphatase for inducing Smad1/5 dephosphorylation in BMS cells. However, Axin2 (or Axin1) could not prevent or delay Smad1/5 dephosphorylation in BMS cells (supplementary material Fig. S3). It has been recently reported that GSK-3β can phosphorylate the linker domain of Smad1, leading to Smad1 polyubiquitylation (Fuentealba et al., 2007). We have looked at whether Axin2 regulates Smad1 levels through affecting the phosphorylation of the linker domain of Smad1 and found that Axin2 has no effect on Smad1 linker domain phosphorylation in 2T3 cells (supplementary material Fig. S4). In addition, the total Smad1 protein levels were not changed in Axin2 KO BMS cells. These results demonstrate that the increase in Smad1 phosphorylation level is due to increased Bmp expression but not to a decrease in Smad1 dephosphorylation or linker domain phosphorylation of Smad1.

In summary, our findings demonstrate for the first time that Axin2 is a crucial negative regulator of bone remodeling in adult mice. Deletion of the Axin2 gene causes an age-dependent high-bone-mass phenotype. In Axin2 KO mice, β-catenin signaling is activated; osteoblast proliferation and differentiation are accelerated; and Opg expression is increased and leads to decreased osteoclast formation. We also discovered that BMPs are downstream mediators of Axin2/β-catenin signaling in osteoblast precursor cells, suggesting an important cross-talk between β-catenin and BMP signaling in the regulation of osteoblast differentiation. Our findings suggest that Axin2 might be an important therapeutic target for modulation of bone homeostasis.

The Axin2^–/– mice and genotyping method were reported previously (Yu et al., 2005). A pair of upper (5′-AGTCCATCTTCATTCCGCCTAGC-3′) and lower (5′-TGGTAATGCTGCAGTGGCTTG-3′) primers were used to detect WT Axin2 and another pair of upper (5′-AGTCCATCTTCATTCCGCCTAGC-3′) and lower (5′-AAGCTGCGTCGGATACTTGCGA-3′) primers were used to detect the Axin2^–/– allele. The Catnb^tm2Kem/J strain (Brault et al., 2001) carrying a floxed allele of β-catenin was obtained from Jackson Laboratory.

Quantitative analysis was performed in bone samples of femora from WT and Axin2 KO mice (6- and 12-month-old) at 10-μM resolution on a micro-CT Scanner (VivaCT 40; Scanco Medical AG, Bassersdorf, Switzerland) (Hildebrand et al., 1999). Briefly, scanning in the femora began approximately at the lower growth plate and extended proximally for 350 slices (each 10-μm thick). Morphometric analysis was performed on 150 slices extending proximally, beginning with the first slice in which the femoral condyles have fully merged. The trabecular bone is segmented from the cortical shell manually on key slices using a contouring tool, and the contours are morphed automatically to segment the trabecular bone on all slices. The three-dimensional structure and morphometry was constructed and analyzed. These parameters include BV/TV (%), BMD (mg HA/mm³), Tb.N. (mm^–1), Tb.Th. (mm) and Tb.Sp. (mm) and SMI. Tb.Th., Tb.N. and Tb.Sp., are computed using algorithms that do not rely on assumptions about the underlying trabecular structure. Micro CT imaging was also performed in the mid diaphysis of the femur. Midshaft evaluation of 100 slices (1.0 mm) was performed to quantify the cross-sectional area, maximum and minimum bending moments of inertia, cortical thickness and maximum and minimum external radii, and cortical bone mineral density.

Changes in mechanical properties of the femoral shaft were determined using three-point bending mechanical testing. Testing was performed using an Instron Dynamite 8841 servo-hydraulic materials testing device equipped with a 1 kN load cell, with control and data collection using Bluehill software version 1.8 (Instron, Norwood, MA). Prior to testing, each femur was harvested and cleaned of soft-tissue and soaked in PBS for 3 hours to ensure hydration (Broz et al., 1993). The anterior surface of each femur was placed on the two support beams, which were separated by an 8-mm span. The femora were bent about the medial-lateral axis at a rate of 5 mm/minute (Jamsa et al., 1998). Force versus deformation data were collected and analyzed to calculate the bending stiffness (N/mm), yield bending moment (N mm), maximum bending moment (N mm) and toughness to maximum bending moment (N mm).

Detail methods for histological and histomorphometric analyses have been described previously (Zhao et al., 2002; Zhao et al., 2004). Histomorphometric analyses were performed on proximal tibial metaphyses in 6-month-old mice. The trabecular bone volume was quantified in the entire area of the bone marrow cavity 0.4–1.4 mm underneath the growth plates and expressed as a percentage of tissue volume. Histomorphometric analysis was performed using the OsteoMeasure system (OsteoMetrics, Atlanta, GA).

Bone marrow cells were isolated from femur and tibia of 6-month-old Axin2^–/– mice and WT littermates and cultured with minimum essential medium (α-MEM) supplemented with 10% fetal bovine serum (FBS) in six-well culture plates and treated with M-CSF (30 ng/ml) (R&D) for 3 days and then cultured in 96-well culture plates at the density of 3×10⁵/ml (200 μl per well). Cells were then treated with M-CSF (30 ng/ml) and RANKL (30 ng/ml) (R&D) for additional 4 days. Tartrate-resistant acid phosphatase (TRAP) staining was performed and the numbers of TRAP-positive multi-nuclear (≥3 nucleus) osteoclasts were counted.

BMS cells isolated from 6-month-old WT and Axin2^–/– mice were plated in a four-well chamber slide at a density of 1×10⁶ cells per well and cultured with α-MEM supplemented with 10% FBS at 37°C for 6 days. The cells were fixed with cold methanol for 5 minutes, endogenous peroxidase quenched with 3% hydrogen peroxide for 20 minutes to block the non-specific binding site and then blocked with 1:20 normal goat serum for 20 minutes. The slides were incubated with 1:20 dilution of rabbit anti-Ki-67 antibody (Lab Vision Corp.) overnight at 4°C and then with 1:200 goat anti-rabbit IgG (Cell Signaling) for 30 minutes, and with 1:250 horseradish peroxidase streptavidin for 30 minutes. The slides were treated with deionized water mixed with 1 drop of Tween 20, and the color reaction with Romulin AEC Chromagen detected for 3-5 minutes. The hematoxylin counterstain was then performed. The nuclei of Ki-67-positive cells showed a dark brown color.

BMS cells or primary calvarial osteoblasts were fixed in 10% formalin. ALP staining was performed by 0.1% fast red salt (Sigma-Aldrich) (buffered with 0.05% Napthol AS-MX phosphate disodium salt, 2.8% N,N-dimethyl formamide, 0.1 M Tris maleate buffer, pH 8.4) at 37°C for 30 minutes. The assays for ALP activity and nodule formation were described in previous studies (Zhao et al., 2002; Zhao et al., 2004).

Details of the method for real-time PCR assay were previously described (Chen et al., 2008; Zhu et al., 2008; Zhang et al., 2009). β-actin was amplified as an internal control. Relative amounts of mRNA were normalized to β-actin and calculated using the software program Microsoft Excel. The primers were: Runx2 (5′-TCCTGTAGATCCGAGCACCA-3′ and 5′-CTGCTGCTGTTGTTGCTGTT-3′), osteopontin (OPN) (5′-CATTGCCTCCTCCCTCCCGGTG-3′ and 5′-GTCATCACCTCGGCCGTTGGGG-3′), type I Collagen (Col-1) (5′-CCTGGTAAAGATGGGCC-3′ and 5′-CACCAGGTTCACCTTTCGCACC-3′), ALP (5′-TGACCTTCTCTCCTCCATCC-3′ and 5′-CTTCCTGGGAGTCTCATCCT-3′), osteocalcin (Oc) (5′-CTTGAAGACCGCCTACAAAC-3′ and 5′-GCTGCTGTGACATCCATAC-3′), bone sialoprotein (BSP) (5′-GAGCCAGGACTGCCGAAAGGAA-3′ and 5′-CCGTTGTCTCCTCCGCTGCTGC-3′), Bmp2 (5′-ACTTTTCTCGTTTGTGGAGC-3′ and 5′-GAACCCAGGTGTCTCCAAGA-3′), Bmp4 (5′-TGAGCCTTTCCAGCAAGTTT-3′ and 5′-CTTCCCGGTCTCAGGTATCA-3′) and β-actin (5′-TGTTACCAACTGGGACGACGACA-3′ and 5′-CTGGGTCATCTTTTCACGGT-3′).

Details of the method for western blotting were described in previous studies (Shen et al., 2006a; Shen et al., 2006b; Zhang et al., 2009). Cyclin D1 mouse monoclonal antibodies (1:1000) (Cell Signaling), anti-β-catenin mouse monoclonal antibody (clone E-5) (1:1000) (Santa Cruz Biotechnology, Santa Cruz, CA), anti-active-β-catenin (anti-ABC) mouse monoclonal antibody (clone 8E7) (1:500) (Upstate Cell Signaling Solutions, Lake Placid, NY), anti-p-Smad1(Ser463/465)/Smad5 (Ser463/465)/Smad8 (Ser426/428) monoclonal antibody (1:1000) (Cells Signaling), anti-lamin-B1 mouse monoclonal antibody (1:400) (Santa Cruz Biotechnology) and anti-β-actin mouse monoclonal antibody (1:3000) (Sigma-Aldrich) were used as primary antibodies.

HFPW-Cre and lentivirus packaging vectors were kindly provided by Eric Brown (University of Pennsylvania, Philadelphia, PA). Lenti-Cre virus was produced in the HEK 293FT cell line with the Virapower Trex Lenti-viral Expression System (Invitrogen) according to the manufacturer's protocol. Lenti-Cre condition media were kept at –80°C until used. BMS cells were infected with Lenti-Cre condition media for 7 days and then cultured with α-MEM containing Zeocin (Invitrogen, 20 μg/ml). At the end of the cell culture, mRNA was extracted and expression of osteoblast marker genes was examined by real-time PCR.

BIO and fibroblast growth factor (FGF-1) were obtained from Calbiochem. FGF-1 was dissolved in PBS containing 0.1% BSA and was used as a positive control. BIO was dissolved in Corn Oil (Sigma) at 1 mg/ml. BIO was injected subcutaneously over the parietal bone of the calvariae of one-month-old C57BL/6J WT mice. Mice received either vehicle (corn oil alone) or BIO (25 and 50 μg/mouse) and FGF-1 (2 μg/mouse) daily for 5 days. Calcein (20 mg/kg i.p.) labeling was performed at days 7 and 14 and mice were sacrificed 2 days after the second label. At the end of the experiments, calvarial bones were removed, fixed in 75% ethanol, and embedded in plastic. Transverse sections were cut at 3-μm thickness and unstained sections were viewed under fluorescent microscope. Mineral appositional rates (MAR) were measured using the OsteoMeasure system (OsteoMetrics, Atlanta, GA).

For comparisons between two data groups the unpaired Student's t-test was used. For multiple comparisons between more than two groups, data was analyzed by one-way analysis of variance (ANOVA), followed by Dunnett's test. Differences were considered significant when a P value was less than 0.05.