Essential Role of β-Catenin in Postnatal Bone Acquisition*

Mutations in the Wnt co-receptor LRP5 alter bone mass in humans, but the mechanisms responsible for Wnts actions in bone are unclear. To investigate the role of the classical Wnt signaling pathway in osteogenesis, we generated mice lacking the β-catenin or adenomatous polyposis coli (Apc) genes in osteoblasts. Loss of β-catenin produced severe osteopenia with striking increases in osteoclasts, whereas constitutive activation of β-catenin in the conditional Apc mutants resulted in dramatically increased bone deposition and a disappearance of osteoclasts. In vitro, osteoblasts lacking the β-catenin gene exhibited impaired maturation and mineralization with elevated expression of the osteoclast differentiation factor, receptor activated by nuclear factor-κB ligand (RANKL), and diminished expression of the RANKL decoy receptor, osteoprotegerin. By contrast, Apc-deficient osteoblasts matured normally but demonstrated decreased expression of RANKL and increased osteoprotegerin. These findings suggest that Wnt/β-catenin signaling in osteoblasts coordinates postnatal bone acquisition by controlling the differentiation and activity of both osteoblasts and osteoclasts.

The Wnt signaling pathway is implicated in the regulation of bone mineral density (1)(2)(3)(4)(5). Wnt ligands bind and activate a specific cellular receptor complex composed of a member of the frizzled family of seven transmembrane-spanning proteins and either LRP5 or LRP6 (6). LRP5 and LRP6 are members of a distinct subfamily of the low density lipoprotein receptor proteins (6). Loss of LRP5 causes osteoporosis pseudoglioma syndrome in humans, which is characterized by low bone density at an early age (7,8), whereas point mutations in LRP5 are associated with extremely high bone density in an autosomal dominant inheritance pattern in humans and mice (2,4,5,9).
Binding of cognate receptors by Wnt ligands can initiate several downstream signaling cascades (6). 1 Activation of the "canonical" pathway involves the stabilization of cytoplasmic levels of ␤-catenin. The main function of the adenomatous polyposis coli (APC) 2 protein appears to be the normal degradation of ␤-catenin. In the absence of APC, ␤-catenin levels are elevated, which ultimately contributes to increased proliferation (11). This pathway requires either LRP5 or LRP6 for activity (6), but LRP5 and LRP6 may have other cellular functions that are not directly related to regulation of Wnt signaling. For example, mice lacking LRP5 display defects in cholesterol and glucose metabolism (12). It remains unclear whether the bone defects seen in humans and mice lacking the Lrp5 gene are due solely to defects in Wnt signaling or if other downstream signaling cascades are involved.
In this work, we have shown that osteoblast-specific deletion of the ␤-catenin gene leads to early onset, severe osteoporosis and is associated with defective osteoblast differentiation in vitro. In contrast, loss of Apc leads to early onset, severe osteopetrosis leading to lethality early in life. Interestingly, we show that these osteoblast-specific deletions are associated with alterations in the regulation of osteoprotegerin (OPG) and receptor activated by nuclear factor-B ligand (RANKL) and with altered osteoclastogenesis in vivo. Thus, this work provides the first genetic evidence that dysregulation of ␤-catenin in osteoblasts leads to defects in bone development and supplies the first link between Wnt/␤-catenin signaling in osteoblasts and functional alterations of the osteoclast regulated by the OPG/ RANKL signaling axis.

EXPERIMENTAL PROCEDURES
Mouse Crosses-OC-cre mice (20) were mated with homozygous conditional mutants carrying modified Apc (19) or ␤-catenin (18) alleles to generate OC-cre/Apc flox/ϩ and OC-cre/␤-catenin flox/ϩ progeny, which were used in subsequent matings. All experiments performed were in compliance with the guiding principles of the "Care and Use of Animals" available at www.nap.edu/books/0309053773/html and approved prior to use by the Van Andel Research Institute Institutional Animal Care and Use Committee.
Genotype Analysis-DNA was prepared from tail biopsies using an AutoGenprep 960 automated DNA isolation system. PCR-based strategies were then used to genotype these mice (details available upon request).
Demineralized Bone Histology-Tissue samples were fixed in formalin overnight, decalcified in Immunocal decalcifying agent (Decal, Baltimore, MD) overnight, and then dehydrated through a graded alcohol series in a Ventana Renaissance processor (Ventana Medical Systems, Tucson, AZ). Tissues were paraffin-embedded, and 5-m sections were adhered to glass slides. Slides were de-paraffinized and stained with hematoxylon and eosin or left unstained for immunohistochemistry.
Mineralized Bone Histology-Femurs were fixed in ethanol at room * This research was supported in part by the Van Andel Research Institute, the National Osteoporosis Foundation, and National Institutes of Health Grant AR44410. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ temperature, dehydrated, and embedded in methylmethacrylate. 3-m sections were cut with a Microm microtome and stained with modified Mason-Goldner trichrome stain. The number of osteoblasts and osteoclasts per bone perimeter were measured at standardized sites under the growth plate using a semiautomatic method (Osteoplan II; Kontron, Munich, Germany) at a magnification of ϫ200. These parameters comply with guidelines of the nomenclature committee of the American Society of Bone and Mineral Research (13).
Osteoblast Isolation and Culture-Osteoblasts were isolated from calvaria of newborn mice by serial digestion in ␣-minimal essential medium (Mediatech, Herndon, VA) containing 10% bovine serum albumin, 25 mM HEPES, pH 7.4, 0.2 mg/ml collagenase type I (Worthington, Lakewood, NJ), 0.7 mg/ml collagenase type 2 (Worthington), and 1 mM CaCl 2 . Calvaria were digested for 15 min at 37°C with constant agitation. The digestion solution was collected, washed with fresh medium, and digested an additional five times. Digestions 3-6 (containing the osteoblasts) were centrifuged, washed with ␣-minimal essential medium containing 10% fetal bovine serum, 1% pen/strep, and plated overnight at 37°C. The next day, the cells were trypsinized and 1.1 ϫ 10 5 cells were plated on 6-cm dishes. The medium was supplemented with 5 mM ␤-glycerophosphate (Sigma) and 100 g/ml ascorbic acid (Sigma) (mineralization medium), which was replaced every other day.
Immunohistochemistry-␤-Catenin was detected in cells using a mouse monoclonal antibody (BD Transduction Laboratories) at a 1:100 dilution. The signal was detected with the Vectastain ABC kit (Vector Labs, Burlingame, CA) and visualized with 3,3Ј-diaminobenzidine tetrahydrochloride (DAB; Vector Labs). Slides were counterstained with hematoxylin.
Von Kossa Staining-Cultures were maintained in differentiation medium for the days indicated and fixed in 10% neutral buffered formalin. The cells were washed with water, dehydrated, and allowed to air dry. Silver nitrate (2%) was added to the cells for 20 min. The cells were washed with water and then incubated with 5% sodium carbonate for 3 min.
Microcomputed Tomography (CT)-High resolution images of the femur were acquired using a desktop microtomographic imaging system (MicroCT40; Scanco Medical AG, Basserdorf, Switzerland). The femur was scanned at 45 keV with an isotropic voxel size of 6 m, and the resulting two-dimensional cross-sectional images are shown in gray scale. Scanning was started in the mid-epiphysis and extended proximally for ϳ3.6 mm (600 CT slices/specimen).
OPG Serum Enzyme-linked Immunosorbent Assay-Serum was collected from animals, and serum OPG levels were quantitated using the mouse OPG/TNFSRSF11B immunoassay according to the manufacturer's specifications (R&D Systems, Minneapolis, MN).

Mice with Osteoblast-specific Deletions of ␤-Catenin or Apc
Die by 4 Weeks of Age-Because global inactivation of either the ␤-catenin gene (14,15) or Apc (16,17) results in early embryonic death, we crossed mice containing conditionally inactivatable alleles of ␤-catenin (␤-catenin-flox) (18) or Apc (Apc-flox) (19) to mice expressing cre under the control of the osteocalcin (OC) promoter (20) to disrupt these genes in osteoblasts. Previous analysis of the OC-cre strain via crossing it to strains carrying cre reporter transgenes indicated that expression of cre recombinase was specific to cells of the osteoblast lineage with no detectable expression or function in other cell lineages (20). As expected, deletion of the ␤-catenin gene led to loss of ␤-catenin protein (Fig. 1, a and b). Consistent with its role in mediating ␤-catenin degradation (11), deletion of Apc was associated with significantly elevated levels of ␤-catenin as assessed by immunohistochemistry (Fig. 1, b and c). By 1 week of age mice homozygous for either mutation could be identified by their reduced size (Fig. 1, d-g). OC-cre;␤-catenin-flox/flox (⌬␤catenin, or ⌬␤-cat) mice died within 5 weeks (Fig. 1h). Growth retardation was even more severe in the OC-cre;Apc-flox/flox (⌬APC) mice (Fig. 1g), and these mice generally succumbed within 2 weeks (Fig. 1i). At weaning, only 75% of the expected number of ⌬␤-catenin mice and only 10% of the expected number of ⌬APC mice were identified (Fig. 1, h and i). No defects in the incisors were observed either grossly or histologically, and the teeth erupted normally in both mutants (data not shown). The cause of the postnatal lethality in these mutant mice is currently unclear but is certainly not because of aberrant expression of the osteocalcin promoter in extraosseous tissues (20). Importantly, mice carrying osteoblast-specific deletions of both the Apc and ␤-catenin genes (⌬APC/⌬␤-catenin) display growth and survival characteristics similar to those lacking only the ␤-catenin gene (data not shown), suggesting that the severe phenotype induced by loss of Apc is due to dysregulation of ␤-catenin signaling.
⌬␤-Catenin and ⌬APC Mice Have Dramatic Defects in Bone Development-Analysis of femurs from ⌬␤-catenin mice by CT revealed striking reductions in both the trabecular and cortical bone compartments (Fig. 2, a and b). In contrast, analysis of ⌬APC mice revealed a significant accumulation of bone matrix in the femur, to the point where the marrow space was almost completely filled (Fig. 2, d and e). Bone in the metaphyseal region was poorly mineralized, whereas osteoid in the diaphysis was more completely mineralized and entirely filled the marrow cavity. CT images of femurs from ⌬APC/⌬␤-catenin mice were similar to those seen in ⌬␤-catenin mice, again suggesting that loss of APC induced phenotypes in a ␤-catenindependent manner (Fig. 2c).
Examination of undecalcified sections from tibia of the ⌬␤catenin mice disclosed a dramatic reduction in mineralized cortical and trabecular bone (Fig. 3, a-d), consistent with changes observed by CT. By contrast, the ⌬APC mice had dramatically increased bone deposition associated with disturbances in bone architecture and composition (Fig. 3, e-h). For example, the growth plate of ⌬APC mice lacked a secondary ossification center and was misshapen (Fig. 3g), possibly because of the rapid rate of bone formation and the lack of osteoclasts (see below) that would normally function to shape the ends of the bone. Further analysis of decalcified bone sections from 4-week-old ⌬␤-catenin and 2-week-old ⌬APC mice showed marked abnormalities in all bone examined, including vertebrae, long bones, and calvaria (Fig. 4).
Effects of Dysregulation of ␤-Catenin on Osteoblast Differentiation in Vitro-To examine the cellular mechanisms responsible for these disturbances in the mutant mice, we studied the effect of conditional deletion of the ␤-catenin and Apc genes in calvarial osteoblasts in vitro. Cells derived from mice carrying the floxed alleles were infected with adenovirus expressing the cre recombinase (Creϩ) or a control adenovirus directing the expression of green fluorescent protein (CreϪ) and then differentiated in the presence of ␤-glycerol phosphate and ascorbate (mineralizing medium). No obvious qualitative differences in proliferation rates or osteoblast density were observed in osteoblasts mutant for either gene (data not shown). However, consistent with the osteopenic phenotype of the ⌬␤-catenin mice, osteoblasts deficient in ␤-catenin showed delayed and diminished expression of osteocalcin as well as a marked reduction in calcified nodule formation (von Kossa staining). In -FIG. 1. Comparison of mutant and control mice. a-c, immunohistochemical detection of ␤-catenin protein expression in cultured osteoblasts. a, ⌬␤-catenin osteoblasts, 40ϫ; b, cre-/flox/flox osteoblasts; 40ϫ; c, ⌬APC osteoblasts, 40ϫ. d, appearance of ⌬␤-catenin mice at postnatal day 10; e, ⌬APC mice at postnatal day 11. Growth curve of ⌬␤-catenin (f) and ⌬APC (g) mice compared with control littermates. Survival curve for ⌬␤-catenin (h) and ⌬APC (i) mice. terestingly, Runx-2 expression levels were similar in control and ␤-catenin osteoblasts (Fig. 5a), suggesting that early events in the osteoblast differentiation program (21) do not depend on signaling through ␤-catenin. In contrast, ⌬APC os-teoblasts appeared to mature and mineralize normally in vitro, although osteocalcin expression levels increased somewhat prematurely as compared with controls (Fig. 5b). It appears, therefore, that excess ␤-catenin signaling does not severely impact the ability of osteoblasts to differentiate, at least in vitro. Thus, the more dramatic effects of disruption of the ␤-catenin gene observed in vivo may be due to non-cell autonomous functions of APC in osteoblasts.
Dysregulation of ␤-Catenin in Osteoblasts Is Associated with Abnormal Osteoclastogenesis-Initial histological analysis (Fig. 3) suggested a disturbance in osteoclastogenesis in both mutants. Indeed, quantitation of osteoclasts in representative long bone sections showed that osteoclast numbers were dramatically increased in the ⌬␤-catenin mutants but entirely absent in the ⌬APC mice (Fig. 6a). In addition, osteoblasts were dramatically decreased in ⌬␤-catenin mutants at 4 weeks of age and were absent in ⌬APC mice at 2 weeks of age (data not shown). However, the dramatic deposition of osteoid material in ⌬APC mice suggests they were present at an earlier age. These findings suggested that dysregulated ␤-catenin signaling in osteoblasts not only causes cell-autonomous osteoblast defects but can also impact bone resorption by altering the numbers of osteoclasts. To explore this possibility further, we measured the expression of RANKL, the major osteoclast differentiation factor, and the osteoclast inhibitory factor, OPG (22). In ⌬␤-catenin osteoblasts, OPG mRNA expression was decreased compared with controls after 5 and 10 days of differentiation in mineralization medium, whereas RANKL expression was increased (Fig. 6b). The reverse pattern was observed in the ⌬APC osteoblasts (Fig. 6c). In addition, levels of serum OPG in the ⌬APC mice were 3-fold higher than controls (Fig.  6d). Serum OPG levels were similar to controls in both ⌬␤catenin and ⌬APC/⌬␤-catenin mice (data not shown). Taken together, these observations suggest that alterations in ␤-catenin signaling in osteoblasts brought about by each mutation lead to marked disturbances in osteoclast differentiation. DISCUSSION In this study we used conditional mutagenesis to disrupt selectively the genes encoding ␤-catenin and Apc in mouse   FIG. 4. Histological analysis of decalcified bone from ⌬␤-catenin and ⌬APC mice. a, d, and g, femur; b, e, and h, vertebrae; c, f, and i, skull from 32-day-old ⌬␤-catenin (a-c) and ⌬APC (g-i) mice as compared with age-matched (d-f) control littermates. The skull sections were taken from equivalent positions immediately dorsal to the hippocampus. Calibration marks are indicated.
osteoblasts. In both models, osteoblast-specific disruption of the canonical Wnt signaling pathway led to postnatal death. In the case of the ⌬APC mice, we speculate that deficiencies in hematopoiesis may be responsible. As shown, ⌬APC mice develop bone in which the vast majority of the marrow component is absent. In other situations where osteopetrosis is observed in humans and mice, it is often accompanied with hepatosplenomegaly associated with extramedullary hematopoiesis. Interestingly, despite severe osteopetrosis, we have not observed any evidence of extramedullary hematopoiesis in these mice. Importantly, mice carrying osteoblast-specific deletions of both the Apc and the ␤-catenin genes are phenotypically similar to those lacking only the ␤-catenin gene (Fig. 2), suggesting that the majority of the phenotype induced by loss of APC is because of dysregulated ␤-catenin signaling. In the case of ⌬␤-catenin mice, we occasionally observe animals with paralysis, consistent with osteoporotic-related fractures, accounting for some portion of the lethality observed. However, we find that many of these animals die very suddenly between 26 and 30 days of age ( Fig. 1) without prior evidence of paralysis. One explanation is that catastrophic fractures may occur leading to full paralysis and sudden death. Alternatively, defects in hematopoietic cell regulation or other systemic defects may play a role.
The skeletal phenotypes observed in these mice allow two important conclusions regarding the role of Wnt signaling in skeletal development. First, Wnt signaling controls postnatal bone acquisition by determining ␤-catenin levels in osteoblasts, and the previously identified alterations in bone mass in humans carrying mutations in LRP5 (1,2,4,9) result from aberrant ␤-catenin signaling. However, it is important to note that the bone phenotypes resulting from inactivation of osteoblast ␤-catenin signaling, although consistent with the alterations seen in patients carrying mutations in LRP5, were more severe. One explanation for this difference is the potential role of LRP6 in ␤-catenin regulation in osteoblasts. We speculate that the continued presence of LRP6 in the context of LRP5 deficiency allows residual Wnt signaling through ␤-catenin. In agreement with this idea, we (23) and others (24) have shown that mice carrying mutations in both Lrp5 and Lrp6 show synergistic defects in bone development. In an alternative model, functions of ␤-catenin independent of Wnt signaling contribute to the observed phenotype. For example, ␤-catenin also plays a key role in mediating cell-cell adhesion through its interaction with E-cadherin (25). E-cadherin is a single-pass transmembrane protein that forms homotypic dimers with E-  Fig. 4). c, assays identical to those in panel b were performed on Apc-deficient (Creϩ) and control osteoblasts (CreϪ). d, the levels of OPG in the serum of 2-3-week-old ⌬APC mice (n ϭ 3) are shown relative to control littermates (n ϭ 3). OPG levels in ⌬APC mice are significantly increased (p Ͻ0.01).
cadherin proteins expressed on adjacent cells (25). The intracellular portion of E-cadherin binds directly to either ␤-catenin or the related plakoglobin protein, which then associates with ␣-catenin. The association of ␣-catenin with the actin cytoskeleton completes the coupling of cell-cell adhesion to the cytoskeleton. Although some studies have suggested that plakoglobin and ␤-catenin are interchangeable in mediating these cell adhesion complexes (26), it is possible that loss of ␤-catenin could also affect cell-cell adhesion in osteoblasts. However, the fact that ⌬APC mice exhibit the converse phenotype suggests that the functions of ␤-catenin in canonical Wnt signaling underlie the observed phenotypes. Also, our observations that mice carrying osteoblast-specific mutations in both Apc and ␤-catenin genes have phenotypes similar to those seen in mice mutant for only the ␤-catenin gene suggests that the osteopetrosis seen in ⌬APC animals is due to dysregulation of ␤-catenin signaling. A more definitive picture of the role of cadherin-mediated adhesion in osteoblast differentiation may be provided by creating mice carrying osteoblast-specific knockouts of the ␣-catenin gene. This will disrupt cadherin-based adhesion in osteoblasts without directly affecting the canonical Wnt signaling pathway.
The second conclusion from our studies is that ␤-catenin signaling in the mature osteoblast controls postnatal bone acquisition through both cell-autonomous and non-autonomous mechanisms. Thus, ␤-catenin signaling may not be required for the initial commitment of cells to the osteoblast lineage, based on Runx2 expression (3,21), but appears to be essential for the performance of the more mature osteoblast (e.g. osteocalcin expression and mineralization) (27,28). However, the marked alterations in osteoclast numbers and reciprocal patterns of expression of OPG and RANKL in the context of dysregulated ␤-catenin signaling suggest that non-cell autonomous effects on osteoclast progenitors also impact overall bone acquisition. Further support for this latter effect has come from a recent report indicating that the OPG gene may be a direct transcriptional target for complexes containing the ␤-catenin protein (10).
These combined effects of Wnt signaling through the canonical pathway in mature osteoblasts are therefore critical in the control of normal bone acquisition, and this pathway represents a plausible pharmaceutical target for treating osteoporosis and other bone disorders.