Msx2 Exerts Bone Anabolism via Canonical Wnt Signaling*

Msx2 is a homeodomain transcription factor first identified in craniofacial bone and human femoral osteoblasts. We hypothesized that Msx2 might activate skeletal Wnt signaling. Therefore, we analyzed the effects of CMV-Msx2 transgene (Msx2Tg) expression on skeletal physiology and composition. Skeletal Msx2 expression was increased 2-3-fold by Msx2Tg, with expanded protein accumulation in marrow, secondary ossification centers, and periosteum. Microcomputed tomography established increased bone volume in Msx2Tg mice, with increased numbers of plate-like trabeculae. Histomorphometry revealed increased bone formation in Msx2Tg mice versus non-Tg siblings, arising from increased osteoblast numbers. While decreasing adipogenesis, Msx2Tg increased osteogenic differentiation via mechanisms inhibited by Dkk1, an antagonist of Wnt receptors LRP5 and LRP6. Bone from Msx2Tg mice elaborated higher levels of Wnt7 canonical agonists, with diminished Dkk1, changes that augment canonical signaling. Analysis of non-Tg and Msx2Tg siblings possessing the TOPGAL reporter confirmed this; Msx2Tg up-regulated skeletal β-galactosidase expression (p ≤ 0.01), along with Wnt7a and Wnt7b, and reduced circulating Dkk1. To better understand molecular mechanisms, we studied C3H10T1/2 osteoprogenitor cells. As in bone, Msx2 increased Wnt7 genes and down-regulated Dkk1, while inducing the osteoblast gene alkaline phosphatase. Msx2-directed RNA interference increased Dkk1 expression and promoter activity, while reducing Wnt7a, Wnt7b, and alkaline phosphatase. Moreover, Msx2 inhibited Dkk1 promoter activity and reduced RNA polymerase association with Dkk1 chromatin. RNA interference-mediated knockdown of Wnt7a, Wnt7b, and LRP6 significantly reduced Msx2-induced alkaline phosphatase. Msx2 exerts bone anabolism in part by reducing Dkk1 expression and enhancing Wnt signaling, thus promoting osteogenic differentiation of skeletal progenitors.

Msx2, also known as Hox-8, is a homeodomain transcription factor first characterized by Sharpe and co-workers (1) as a calcitriol-regulated transcript in osteoprogenitors isolated from human femur, and subsequently shown to be highly expressed in the murine embryonic craniofacial skeleton (2). Soon thereafter, Msx2 was identified as a transcriptional repressor of the osteoblast-specific osteocalcin (OC) 3 promoter (3,4). Studies of Msx2-regulated gene expression in bone have emphasized its role as a transcriptional repressor of the late osteogenic phenotype (5)(6)(7). For example, in the developing tooth, where stage-specific osteogenic gene expression profiles are spatially resolved, Msx2 and OC exhibit reciprocal patterns of mRNA accumulation (8). Msx2 suppresses OC gene expression in a cell-autonomous fashion, mediated via antagonistic protein-protein interactions between Msx2 and Runx2-containing complexes that support OC promoter activity (6,7).
However, elegant genetic studies have demonstrated that Msx2 also promotes craniofacial bone mineralization and is necessary for robust trabecular and cortical bone formation (9,10). Msx2 Ϫ/Ϫ mice exhibit parietal foramina, characterized by reduced mineralization in calvarial fields that give rise to membranous bone (9). Moreover, Msx2 Ϫ/Ϫ mice exhibit a global, low turnover osteopenia (9). Simultaneous deletion of both Msx2 and Msx1, a homologous Msx gene family member, results in the complete absence of craniofacial bone (9,11). Furthermore, a nonsynonymous CCC to CAC mutation at the 7th codon of the Msx2 homeodomain causes Boston-type autosomal dominant craniosynostosis, characterized by precocious mineralization and differentiation of osteoblasts in the calvarial suture (12). In vivo and in vitro data indicate that this Msx2(P148H) variant is most likely a gain-of-function variant; Msx2(P148H) exhibits increased DNA binding to the Msx CTGAATTRG binding cognate (13,14). Intriguingly, however, the frequency of transgene-induced craniosynostotic bone is three time more prevalent in wild-type Msx2 transgenic mice than Msx2(P148H) transgenic mice (71 versus 27%, respec-tively) (13). This suggests that in vivo the P148H mutation perturbs other important Msx2 functions (13). Indeed, the region of the homeodomain N-terminal arm altered by the P148H substitution participates in regulatory protein-protein interactions with Dlx5 (5,15) and other transcription factors (5)(6)(7). Consistent with the Msx2 gain-of-function model arising from CMV-Msx2 transgenic mice, distal trisomy of chromosome 5q, the physical region encompassing the Msx2 gene, has been identified in six patients with craniosynostosis (16); this demonstrates the exquisite sensitivity of the developing human craniofacial skeleton to Msx2 gene dosage (10,17). Although details have emerged as to how Msx2 inhibits osteoblast terminal differentiation via cell-autonomous actions (5)(6)(7), little is known of the mechanisms whereby Msx2 promotes osteoblastmediated bone formation. The mechanistic underpinnings of the low turnover osteopenia demonstrated in Msx2 Ϫ/Ϫ long bone during postnatal growth have yet to be determined (9). In seminal studies of calvarial bone formation, Maxson and coworkers (10,11) showed that Msx2 promotes the proliferative expansion and survival of neural crest-derived osteoblasts.
To better understand the mechanisms whereby Msx2 augments orthotopic long bone formation, we have evaluated the skeletal phenotype of adult CMV-Msx2 (Msx2Tg) transgenic mice (19). We show that augmenting Msx2 expression in long bone directs bone mesenchymal progenitors to osteogenic lineage, enhancing skeletal osteoblast numbers, mineralizing surface, trabecular number, and trabecular bone formation via canonical Wnt signals.
Assays of Osteogenic and Adipogenic Potential of Primary Mouse Bone Cell Cultures-Calvarial osteoblasts were prepared from neonatal mouse calvaria by timed collagenase digestion using methods described previously (26). Bone mesenchymal cells were isolated from the marrow compartment of long bones of 4-month-old Msx2Tg and WT mice using methods described previously (27) and then tested for the capacity to undergo osteogenic or adipogenic differentiation. For osteogenic differentiation, cells were seeded in 96-well tissue culture plates at a density of 1 ϫ 10 5 cells/well, left undisturbed for 7 days, and then maintained in mineralization medium (modified Eagle's medium containing 10% FBS, 50 g/ml ascorbic acid, and 10 mM ␤-glycerophosphate) for 3 additional weeks. Cultures were subsequently stained for calcium deposition using Alizarin red. Wells containing red nodules with 15 or more cells were considered positive. The osteogenic potential was determined as the percent of positive wells in total wells. For adipogenic differentiation assays, bone mesenchymal cells were seeded at 1 ϫ 10 7 cells per 10-cm culture dish. Seven days later, cells were treated with adipogenic medium containing insulin (5 g/ml), dexamethasone (10 Ϫ7 M), and indomethacin (50 M) for 9 additional days. Cultures were stained with the lipotrophic dye Oil Red O as detailed previously (28,29). The numbers of adipocyte colonies were enumerated, and data were expressed as the percent of total cell colonies that stained positive with Oil Red O.
Immunohistochemistry and Western Blot Analysis-Multiple commercially available immunoreagents were tested for sensitivity and specificity, and only those that proved sensitive and specific are reported here. Prior to immunohistochemistry, the specificity of the antibody reagents was first gauged by Western blot analysis as detailed previously (30), using protein extracts prepared from the following: (a) C3H10T1/2 cells transduced with either control SFG-LacZ virus or with SFG-Msx2 and then (b) mouse primary calvarial osteoblast cell extracts from either WT or Msx2Tg mice. Only those reagents that generated single immunoreactive bands of the appropriate relative mass following SDS-PAGE and Western blot analysis were utilized for immunoperoxidase staining. For neonatal long bones, following 48 h of fixation in 10% neutral buffered formalin, samples were embedded in paraffin without decalcification, and 5-m longitudinal sections were prepared for Msx2, Wnt7a/b, and Wnt1 immunohistochemistry. For adult long bones, ossicles were fixed for 48 -72 h in 10% neutral buffered formalin and then decalcified for 3 weeks at 4°C in 0.375 M EDTA, pH 8, with constant gentle stirring. Decalcified adult long bones were embedded in paraffin, and 5-m sections were cut for subsequent immunohistochemistry using methods described previously (19). Briefly, after deparaffinization of the sections with xylene and rehydration via graded aqueous ethanolic baths, endogenous peroxidases were quenched by treatment with 1% H 2 O 2 in methanol for 20 min, rinsed with water for 1 min, followed by 10 more min of rinsing in PBS, and then incubated in blocking solution (10% normal serum cognate to the secondary antibody and 5% bovine serum albumin in PBS) for 30 min at 25°C. Sections were then incubated with 2 g/ml primary antibody in blocking solution overnight (Ͼ18 h) at 4°C. The next morning, after washing three times for 10 min with PBS at room temperature, antigen-antibody complexes were tagged with VECTASTAIN Elite ABC peroxidase kits directed toward either goat (PK6105) or rabbit (PK6101) primary antibodies as appropriate (see above), by incubating sections with 1:200 dilution of the biotinylated secondary antibody for 30 min. After washing and incubation with ABC anti-biotin reagent, specimens were again washed two times for 10 min with sterile PBS, and complexes were visualized by diaminobenzidine staining (SK-4100), with stain intensity monitored by microscopy. Reactions were stopped by rinsing with tap water. Equal time of stain development was utilized for both WT and MsxTg tissue sections monitored in parallel. After dehydration in graded ethanol to xylene, samples were mounted using VectaMount permanent mounting media and digital images were captured using a Leica 4100 DM microscope coupled with a DFC 420 digital camera, operated with FW 4000 software (Leica Microsystems, Bannockburn, IL) as detailed previously (19).
Histochemistry and Histomorphometric Methods-Histochemistry for LacZ activity in Msx2Tg;TOPGAL and TOPGAL mice was carried out essentially as detailed previously (19). Long bones from Msx2Tg;TOPGAL and TOPGAL siblings were excised, rinsed twice in PBS, and then embedded in OCT (Sakura Tissue-Tek). Subsequently, 25-m-thick longitudinal frozen sections of nondecalcified specimens were analyzed by histochemistry for ␤-galactosidase (LacZ) enzyme activity as detailed previously for mouse aortas (19), comparing LacZ staining of proximal tibiae in Msx2Tg;TOPGAL versus TOP-GAL animals.
For dynamic and static histomorphometric studies, 6-month-old male mice (7 Msx2Tg,11 WT siblings) were sequentially dosed at 8 and 3 days prior to sacrifice, first with subcutaneous tetracycline (1 mg/0.1 ml per 30-g body weight) and then with subcutaneous calcein (1 mg/0.1 ml per 30-g body weight), respectively. Following euthanasia via exsanguination under ketamine-xylazine anesthesia, right hindlimbs were dissected en bloc, and the combined femur, genu, and tibia were dissected free of adherent muscle and fascia. Carefully liberated right femurs were fixed in 10% neutral buffered formalin at 4°C for 2 days, washed twice for 15 min in PBS, and then stored in 70% ethanol until embedded in Osteo-Bed Resin for preparation of longitudinal plastic sections. Dynamic and static histomorphometry was performed under a fee-for-service contract with MDS Pharma Services, formerly SkeleTech, Inc. (Bothell, WA), using Osteomeasure software (33), and following nomenclature and recommendations of the American Society for Bone and Mineral Research Histomorphometry Committee (34). For bone formation rate (BFR), the tissue volume referent (TV), i.e. BFR/TV (%), was chosen to facilitate comparisons with contemporary reports of Wnt-regulated (35), Dkk1-regulated (36), and other anabolic responses (37)(38)(39)(40)(41) in the mouse skeleton.
Conebeam Micro-CT Imaging-The left tibiae obtained from 6-month-old mice described above were used to assess trabecular bones using Scanco micro-CT 40 (Scanco USA Inc., Southeastern, PA) at an energy setting of 55 kV, current setting of 145 mA, and with an integration time of 300 ms. A fixed threshold of 270 Hounsfield units was used to separate bone from background. A total of 300 -350 transverse CT slices was obtained, and three-dimensional analysis was performed on trabecular bones in the 50 slices (16 m thick/slice) starting at about 0.1 mm below the lowest point of the growth plate. Trabecular morphometric parameters, including the percentage of bone volume per total volume (%BV/TV), trabecular thickness (Tb Th, m), trabecular number (per mm 3 ), trabecular spacing (m), structure model index, and connectivity density (connections per mm 3 ) were acquired by direct method of calculation (42).
Serum ELISAs for Tartrate-resistant Acid Phosphatase (TRAP5c), CTX, and Dkk1-After overnight fasting, mice were anesthetized and blood was withdrawn from the inferior vena cava in heparinized syringes. Blood samples were layered on top of Microtainer Serum Separator tubes (365956; BD Biosciences) and incubated overnight at 4°C. Sera were obtained after centrifugation at 2,000 ϫ g for 20 min at room temperature. Serum concentration of osteoclast-derived tartrate-resistant acid phosphatase form 5b was determined by using MouseTRAP assay kit (Code TR103; IDS Inc., Fountain Hills, AZ). Commercial ELISAs for CTX (RatLaps, 1RTL4000, Nordic Bioscience Diagnostics via IDS Inc.) and Dkk1 (DuoSet DY1765; R & D Systems) were carried out per the manufacturer's instructions.
Transfection and Retroviral Transduction-The pseudotyped retrovirus SFG-LacZ (43) was the kind gift of Dr. Dan Ory (Washington University, St. Louis). SFG retroviruses express-ing Dkk1, Msx2, Msx2(P148H), and Msx2(T147A) were generated using methods detailed previously (28). Transduction of C3H10T1/2 cells and primary mouse mesenchymal cells was performed as detailed (28). Cells transduced with SFG-LacZ encoding ␤-galactosidase were used as a negative control. The expression of Msx2 and Dkk1 in SFG-Msx2-and SFG-Dkk1transduced cells was verified via real time RT-qPCR analysis of mRNA (see below).
Gene Expression by Fluorescence RT-qPCR-Two-step RT-qPCR was used to quantify gene expression in total RNA extracted from either mouse bone (44) or cultured cells (28) using the respective methods detailed previously. A mixture of random hexamers and oligo(dT) was used to prime cDNA synthesis, so that relative mRNA accumulation could be normalized to the signal obtained for 18 S ribosomal RNA in parallel. Amplimer pairs used for quantitative RT-qPCR were as follows: Knockdown of Msx2-Wnt Signaling Components by siRNA in C3H10T1/2 Mouse Mesenchymal Osteoprogenitors-Commercially available siRNAs directed to murine Wnt1 (sc-36840), Wnt5a (sc-41113), Wnt7a (sc-41115), Wnt7b (sc-41117), Wnt10b (sc-37186), and control siRNA (sc-44230) were purchased from Santa Cruz Biotechnology. Sequences for these reagents can be obtained upon request from the manufacturer. Custom siRNAs used for knockdown of LRP5, LRP6, and Msx2 were ordered from Qiagen (Valencia, CA) as high performance purity-grade reagents. The siRNA designed for Msx2 knockdown was derived from the viral RNAi strategy previously described by Cossu and co-workers (45). Custom sequences for mouse LRP5 and LRP6 knockdown were designed using Easy siRNA software (Protein Lounge, San Diego). Specific reagents used for Msx2, LRP5, and LRP6 knockdown were as follows: Msx2, 5Ј-r(CAG UAC CUG UCC AUA GCA G)dTdT-3Ј and 5Ј-r(CUG CUA UGG ACA GGU ACU G)dTdT-3Ј; LRP5, 5Ј-r(GGA GAU CCU UAG UGC UCU G)dTdT-3Ј and 5Јr(CAG AGC ACU AAG GAU CUC C)dTdT-3Ј; LRP6, 5Ј-r(GAG AAU GCA ACG AUU GUA G)dTdT-3Ј and 5Ј-r(CUA CAA UCG UUG CAU UCU C)dTdT-3Ј and ordered from Qiagen (Valencia, CA).
Targeted mRNA degradation using siRNA technology was performed essentially as described by Brazas and Hagstrom (46), according to the protocol provided by Santa Cruz Biotechnology. Briefly, C3H10T1/2 cells were seeded at 2 ϫ 10 5 /well in 6-well cluster plates (35-mm diameter wells) in DMEM containing 10% FBS the day before lipofection. To prepare lipid-siRNA complexes, 80 pmol of the indicate siRNA duplex in 100 l of Transfection Medium (sc-36868) and 6 l of siRNA Transfection Reagent (sc-29528) in 100 l of Transfection Medium were combined, incubated for 30 min at 25°C, and then diluted with 800 l of pre-warmed Transfection Medium. Cell were rinsed once with serum-free DMEM, and 1000 l of lipid-siRNA admixture described above was applied per well. After incubation for 6 h at 37°C in a humidified 5% CO 2 cell culture chamber, an additional 1 ml of 20% FBS in DMEM was added per well, and lipofection was allowed to continued overnight. The next morning, the lipofection media was aspirated, and transfected monolayers cells re-fed with fresh C3H10T1/2 growth media (10% FBS in DMEM). Twenty four hours later, total cellular RNA was harvested as detailed previously (19) using Ambion Turbo DNase treatment and removal kit (AM1907; Applied Biosystems) to remove all traces of genomic DNA. Quantitative RT-qPCR was carried out in triplicate using methods described previously (28). Efficiency of target mRNA knockdown of Ͼ30% (range 30 -50%) was observed for all siRNA reagents presented, using RT-qPCR for targeted mRNA accumulation as an assay. For RNAi using primary bone marrow cell cultures from Msx2Tg mice, the above protocol was also implemented, except that TransIT-TKO transfection reagent (Mirus, Madison, WI) was used to introduce the siRNA reagents.
Statistics-For statistical testing of dynamic histomorphometry, micro-CT, and gene expression results, analyses were performed using Student's t test as described previously (44). All data are presented as the means Ϯ S.E. of measurement to facilitate meaningful comparisons (95% confidence interval Ϯ 1.96 S.E.).

CMV-Msx2 Transgenic Mice Exhibit Higher Bone Volume with Greater Numbers of Plate-like Trabeculae as Compared with Wild-type Littermates-We demonstrated previously that
Msx2 stimulated osteogenic differentiation but inhibited adipogenesis in cultured C3H10T1/2 mesenchymal cells (28,29). We wished to better understand the physiological relevance of osteogenic-adipogenic lineage allocation by Msx2 on postnatal bone homeostasis. Therefore, we extended our studies to the evaluation of bone and body composition in our CMV-Msx2 transgenic (Msx2Tg) mice (19), a murine model previously demonstrated by Maxson and co-workers (13) to efficiently recapitulate key features of Msx2 actions in craniosynostosis.
Msx2 mRNA was significantly increased in calvarial osteoblasts (2.57 Ϯ 0.07-fold) and bone marrow cell cultures (1.81 Ϯ 0.17fold) in Msx2Tg mice as compared with WT littermates (p Ͻ 0.05 for both). Western blot and digital image analysis of osteoblast extracts isolated from long bone and calvaria confirmed that Msx2 protein levels paralleled the transgene-mediated increases in mRNA accumulation (Fig. 1A). We localized the spatial patterns of Msx2 expression in long bone from Msx2Tg mice by immunoperoxidase staining in 3-day-old mice. Overall signal intensity was increased in Msx2Tg versus WT siblings, because of expanded Msx2 immunoreactivity in the bone marrow compartment, the periosteum, and secondary ossification centers ( Fig. 1B and supplemental material). Secondary (epiphyseal) ossification centers begin to form in mice in the first postnatal week (47,48); in long bone sections where secondary ossification centers were visualized, Msx2 expression in associated hypertrophic chondrocytes was also expanded in Msx2Tg mice (Fig. 1B). The recently described accumulation of Msx2 protein in a subset of early hypertrophic zone chondrocytes was also noted (49) and was enhanced in Msx2Tg versus WT sibs (Fig. 1B). No immunoreactivity was observed in the absence of primary antibody (Fig. 1B, right panel). In the epiphyseal ossification center, Msx2 appeared to localize primarily to chondrocyte nuclei (Fig. 1C). However, in marrow and periosteal venues, both nuclear and cytoplasmic Msx2 immunoreactivity were clearly noted (Fig. 1D), presumably reflecting stage-specific re-localization of Msx2 in bone mesenchymal progenitors as described recently (50).
To examine the effects of Msx2 on bone content and structure, we performed conebeam micro-CT analysis of the proximal tibia (51), comparing Msx2Tg versus WT siblings. Consistent with increased BMD (see below), Msx2Tg mice exhibited significant increases in trabecular bone volume (Tb.BV/TV; 40% increase), trabecular number (Tb.N; 20%), and trabecular connectivity density (80% increase; Fig. 2, A-C). The structure model index (SMI), an index that diminishes as trabeculae become more sturdy and plate-like (52), was 40% lower in Msx2Tg mice ( Fig. 2D; p ϭ 0.04); thus, trabeculae in Msx2Tg mice were more plate-like than those of their WT sibs (52). Although Msx2Tg-dependent increases in tibial trabecular thickness did not reach significance by micro-CT (not shown), static morphometry of femoral trabeculae did establish a significant 15% increase in femoral trabecular thickness (Tb.Th) and confirmed Msx2-dependent increases in Tb.N and Tb.BV/TV (Fig. 3). Long bone cortical thickness was also significantly increased by 33% in Msx2Tg versus WT siblings (p Ͻ 0.05). Thus, augmenting Msx2 expression in bone significantly increases trabecular bone mass accrual.
Msx2Tg Mice Exhibit Greater Osteoblast Numbers and Increased Trabecular Bone Formation-Detailed static and dynamic histomorphometry (34,53) of secondary spongiosa in the distal femur revealed the cellular mechanisms of Msx2 long bone anabolism. Significant increases in trabecular bone volume/total volume (Tb.BV/TV; Fig. 3A), Tb.N (Fig. 3B), and Tb.Th (Fig. 3C) in Msx2Tg mice (n ϭ 7) versus nontransgenic siblings (n ϭ 11) were observed at 6 months of age. Importantly, the trabecular bone formation rate/tissue volume (BFR/TV; tissue volume referent) was increased 76% (p ϭ 0.03; Fig. 3D). A similar change was noted when bone surface referent was used for BFR (WT ϭ 52.5 Ϯ 9.1 m 3 /m 2 /year versus Msx2Tg ϭ 73.7 Ϯ 8.4 m 3 /m 2 /year, p ϭ 0.065 for the 40% BFR/BS increase). A 2.7-fold increase in osteoid surface/bone surface was seen with no change in osteoid thickness (not shown), indicating normal and unchanged matrix mineralization. Because mineralizing surface (MS/BS; Fig. 3E) was increased without a change in matrix apposition rate (MAR; Fig. 3F), this strongly suggested Msx2Tg-induced increases in BFR/TV were because of increased numbers of osteoblasts. Osteoblast numbers were indeed increased; the osteoblast perimeter/bone perimeter (Ob.Pm/B.Pm) was 50% greater in Msx2Tg versus non-Tg sibs (p ϭ 0.001; Fig. 3G). By contrast, osteoclast numbers were not increased (Oc.Pm/B.Pm; Fig. 3H). Increased bone formation without significant changes in osteoclast-mediated bone resorption was confirmed by serum marker analysis (54); serum TRAP5c (tartrate-resistant acid phosphatase; WT ϭ 1.09 Ϯ 0.11 IU/ml versus Msx2Tg ϭ 1.01 Ϯ 0.17 IU/ml; p ϭ 0.68) and CTX (collagen C-terminal telopeptide; WT ϭ 29.7 Ϯ 4.2 ng/ml versus Msx2Tg ϭ 23.0 Ϯ 3.6 ng/ml; p ϭ 0.28) did not differ between Msx2Tg and non-Tg sibs. Thus, Msx2Tg increases long bone mass and trabecular bone formation by increasing the total number of trabecular bone-forming osteoblasts, without significant changes in osteoclast-mediated bone resorption. Mice transgenic for Msx2 expression enhanced by the CMV promoter were generated as detailed previously (19). Cellular extracts prepared from primary cell cultures of neonatal calvarial and long bone osteoblasts were analyzed for Msx2 protein accumulation by Western blot analysis. Expression levels were referenced to that of actin expression. A, note that Msx2 protein accumulation in both calvarial and long bone mesenchymal cell cultures is enhanced 2-fold by Msx2Tg. Levels of Msx2 mRNA accumulation were equivalently increased (p Ͻ 0.05, see text). B, immunoperoxidase staining for Msx2 expression was more intense in long bones from Msx2Tg versus WT siblings, localized to early hypertrophic zone (HZ) chondrocytes, and periosteal and marrow compartments of the metaphysis (MET) and diaphysis (DIA). Expanded Msx2 protein accumulation was also observed in secondary ossification centers of Msx2Tg mice. No signal was observed in the absence of primary antibody. C, higher power magnification of chondrocytes in secondary ossification centers suggested only nuclear localization of Msx2 in both WT (shown) and Msx2Tg mice (not shown). D, in marrow spongiosa (left panel) and periosteal (right panel) skeletal compartments, both nuclear (white arrowheads) and non-nuclear (black arrowheads) Msx2 protein accumulation was observed. Aby, antibody.  (Fig. 4A). Again, no difference in overall body weight was detected between obese Msx2Tg and non-Tg littermates (p ϭ 0.43; Fig. 4A). Serum leptin, an adipokine marker of body fat content (55), was also reduced in  5C). Intriguingly, no change was observed in the relative accumulation of Runx2 message (Fig. 5C). Accumulation of the nuclear transcriptional co-adapter, ␤-catenin, has been shown to be critical to the elaboration of osteoblast development downstream of osteogenic Wnt signaling (56). Therefore, cellular extracts of bone mesenchymal cell culture from WT and Msx2Tg mice were prepared, and ␤-catenin protein accumulation was quantified by ELISA. Total cellular ␤-catenin protein was elevated 1.6-fold in Msx2Tg versus non-Tg long bone cell cultures (Fig. 5D, p ϭ 0.02) and 2-fold in calvarial osteoblasts (data not shown). Thus, the ex vivo osteogenic potential of mesenchymal progenitors isolated from Msx2Tg mice is enhanced compared with non-Tg WT sibs, consistent with the increased osteoblast numbers observed in Msx2Tg mice long bone as quantified by histomorphometry.

Msx2Tg Activation of Osteogenic Differentiation Is Antagonized by Dkk1, an Inhibitor of Canonical Wnt Signaling-
The reciprocal control of bone mesenchymal cell osteogenesis and adipogenesis by the Msx2 transgene was highly reminiscent of the effects of canonical Wnt signaling (22,23). Wnt10b has been shown to promote osteogenesis and suppress adipogenic differentiation from multipotent marrow osteoprogenitors (22,23). If paracrine canonical Wnt signals contribute to the effects of Msx2Tg on bone marrow osteogenesis, then Dkk1, a highly specific inhibitor of LRP5/6 Wnt receptor signaling complexes (36), should antagonize Msx2Tg actions. We thus generated a retroviral expression construct, SFG-Dkk1, to augment Dkk1 production by bone mesenchymal cells. We first tested the effects of SGF-Dkk1 transduction on C3H10T1/2 cell cultures previously transduced with SFG-Msx2; in this validated cell culture model, Msx2 actions are antagonized by exogenous recombinant purified Dkk1 (19). As shown in Fig. 6A, SFG-Dkk1 completely abrogated the effects of SFG-Msx2 on ALP induction in C3H10T1/2 cells. We next evaluated the effects of SFG-Dkk1 transduction on osteogenic differentiation using calvarial osteoblasts isolated from Msx2Tg versus non-Tg siblings. As compared with control virus (SFG-LacZ), transduction with SFG-Dkk1 profoundly inhibited the mineralized nodule formation up-regulated by the Msx2 transgene (Fig. 6B). Thus, Msx2Tg induction of osteogenic development in transgenic osteoblasts is dependent upon endogenous paracrine canonical Wnt signaling cascades.

The Msx2 Transgene Up-regulates Wnt7 but Suppresses Dkk1 Protein Accumulation in Calvarial Osteoblasts and Long Bone
Mesenchymal Cells-Our studies of arterial calcification have revealed that Msx2 controls osteogenic lineage allocation of aortic vascular progenitors (28) via paracrine Wnt signals (19). The body composition changes observed in response to Msx2Tg resembled those of the canonical ligand Wnt10b (20,22,23). Moreover, because SFG-Dkk1 inhibited Msx2Tg actions, this strongly suggested that modulation of Wnt and Dkk1 protein biogenesis contributes to Msx2 bone anabolism (see above and Ref. 28). To test this notion, we analyzed protein accumulation of Wnt1, Wnt3a, Wnt7 (antibody recognizes both Wnt7a and Wnt7b), Wnt10b, and Dkk1-prominent genomic targets of Msx2 first identified in aortic adventitial myofibroblasts transduced with SFG-Msx2 (19). As compared with non-Tg sibs, Msx2Tg up-regulated Wnt1 and Wnt7 in calvarial osteoblasts ϳ3-fold, but suppressed Dkk1 protein FIGURE 4. Msx2Tg mice exhibit globally increased bone mass and reduced fat mass than non-Tg siblings. Beginning at 1 month of age, male (shown) and female (not shown) Msx2Tg and non-Tg (WT) sibling cohorts were fed high fat diet composition typical of western societies for 16 weeks. Subsequently, animals were weighed, and body composition was determined by dual electron x-ray absorptiometry. A, no differences in overall body weight were observed between genotypes; however, whole body bone mineral density (BMD) was increased in Msx2Tg mice (n ϭ 22) versus non-Tg WT (n ϭ 34) siblings. By contrast, fat mass percentage was decreased in Msx2Tg animals. B, decreases in fat mass in Msx2Tg mice were paralleled by even greater reductions in serum leptin (WT, n ϭ 11; Msx2Tg, n ϭ 7). C, histologic analysis of WT and Msx2Tg siblings demonstrated that marrow fat was also diminished in Msx2Tg animals versus non-Tg sibs when fed westernized diets. The histology shown is representative of results obtained from eight animals analyzed (WT, n ϭ 4; Msx2Tg, n ϭ 4; see supplemental material). D, adipocytic Oil Red O-positive colonies were quantified in primary bone marrow mesenchymal cell cultures obtained from WT (n ϭ 4) and Msx2Tg (n ϭ 7) mice that had been fed the high fat diet for 4 months (see "Experimental Procedures"). Note the relative reduction in adipogenic potential by the Msx2 transgene. accumulation (Fig. 7A, expression normalized to eIF2␣ protein signal intensity). No change was observed in Wnt3a or Wnt10b protein accumulation. In long bone mesenchymal cells, little if any change occurred in Wnt1 and Wnt10b protein levels (Fig.  7A). However, Wnt7 was once again increased, and Dkk1 decreased, by the Msx2 transgene; accumulation of Wnt7 protein was increased ϳ10-fold in Msx2Tg animals versus nontransgenic sibs (Fig. 7A). As compared with calvarial osteoblasts, Wnt3a protein levels were lower in long bone mesenchymal cells but were increased 2-fold by the Msx2 transgene.
Because Wnt7 protein accumulation was consistently and dynamically regulated by the Msx2Tg in primary cell culture, immunohistochemistry was used to spatially resolve the expression of Wnt7 in Msx2Tg long bone. In neonatal mice, Wnt7 accumulated in the endosteal compartment of the mineralizing periosteum and along the trabeculae of the spongiosa (Fig. 7B). Wnt7 expression was also observed in early hypertrophic zone chondrocytes (Fig. 7B), thus overlapping the pattern of Msx2 accumulation in neonatal Msx2Tg mice (Fig. 1, B and C). In adult mouse long bone, Wnt7 immunoreactivity was still prominent, detected in bone marrow vascular structures, mineralizing spongiosa, and osteocytes (Fig. 7C). No signal was observed in the absence of primary Wnt7 antibody (Fig. 7D).
Higher power magnification more clearly revealed vascular sinusoid Wnt7 staining pattern in marrow (see supplemental material), as well as in Haversian bone vasculature (Fig. 7E, asterisk and stout arrow) and in osteocytes (Fig. 7E, long thin  arrows). Of note, the pattern of Wnt7 protein accumulation was completely distinct from that of Wnt1 immunoperoxidase staining, which clearly localized to marrow megakaryocytic cells (data not shown). Thus, Msx2Tg controls canonical Wnt ligand protein expression, consistently augmenting the accumulation of Wnt7 while reducing Dkk1 protein levels in osteoblasts derived from calvaria and long bone.
Msx2Tg Enhances Canonical Wnt Signaling in Long Bone in Vivo-To confirm that the Msx2Tg enhances endogenous canonical Wnt signaling in bone, we crossed Msx2Tg mice with TOPGAL reporter mice (32) and characterized ␤-galactosidase (LacZ) staining in long bones of neonatal mice. As compared with TOPGAL siblings, trabecular LacZ staining in the trabecular primary spongiosa and the endosteal bone envelope was much more prominent in Msx2Tg versus non-Tg sibs (Fig. 8A). Analysis of total long bone RNA from older 2-monthold cohorts confirmed the histochemical results; RT-qPCR demonstrated higher levels of both lacZ reporter and Wnt7a gene expression in Msx2Tgϩ;TOPGAL versus TOPGAL sibs (Fig. 8B). By contrast, no significant up-regulation of Wnt1, Wnt3a, or Wnt10b mRNAs was observed in long bone of Msx2Tg mice (Fig. 8B, left panel). Wnt7b mRNA also accumulated to 3-fold higher levels in long bone from Msx2Tg versus non-Tg animals (Fig. 8B, right panel). Moreover, as predicted from our Western blot analyses, serum levels of Dkk1 were significantly reduced by 43% in Msx2Tg mice versus non-Tg siblings (Fig. 8C).
We next studied LacZ staining of cultured bone mesenchymal cells from TOPGAL mice (32) to provide an additional index of canonical Wnt/␤-catenin signaling. As compared with TOPGAL siblings, long bone mesenchymal cells from Msx2Tg; TOPGAL mice exhibit a 5-fold increase in the numbers of LacZ-stained cells (Fig. 8D; p Ͻ 0.0001). Similar results were obtained with primary calvarial osteoblasts (data not shown). Thus, histochemical, quantitative RT-qPCR, and biochemical data converge to demonstrate that the Msx2Tg augments skeletal canonical Wnt signaling cascades.
Canonical Wnt7-LRP Signaling Contributes to Msx2-dependent Induction of Osteogenesis in C3H10T1/2 Mesenchy- FIGURE 5. Primary bone mesenchymal cell cultures from Msx2Tg mice exhibit greater osteogenic differentiation as compared with non-Tg sibs. Primary cultures of hind limb long bone mesenchymal cells were prepared and cultured under pro-osteogenic conditions. Numbers of osteogenic colonies were scored by Alizarin Red staining as described previously (28). A, note that numbers of osteogenic colonies were increased by ϳ50% in Msx2Tg (n ϭ 5) versus non-Tg sibs (n ϭ 5). B and C, Msx2Tg-enhanced osteogenic differentiation was confirmed by 2-fold increases in both alkaline phosphatase enzyme activity of primary calvarial osteoblast cultures (B) and expression of the osteogenic genes, ALP, BSP, and Osx (C). For both WT and Msx2Tg osteoblast cultures, n ϭ 4. D, total cellular ␤-catenin levels measured by ELISA were also increased in cultures derived from Msx2Tg mice (n ϭ 3) versus non-Tg siblings (n ϭ 3).
mal Cells-Because Wnt7 protein levels were significantly affected by the Msx2Tg, and both Wnt7a and Wnt7b are agonists for canonical Wnt signaling (21,(57)(58)(59), Wnt7 family members are potential paracrine mediators of Msx2 action. Moreover, as observed in other cell systems (21), transient coexpression of Wnt7a and Wnt7b eukaryotic expression plasmids significantly activate canonical TCF/LEF-dependent transcription in C3H10T1/2 cells (see supplemental material). The C3H10T1/2 multipotent mesenchymal cell line faithfully recapitulates the osteogenic versus adipogenic differentiation fate choices of bone marrow mesenchyme (23,28). Thus, to further establish the role of paracrine canonical Wnt tone to the actions of Msx2, we examined the effects of depleting key inhibitors (Dkk1) and activators (Wnt1, Wnt7a, and Wnt7b) of this pathway on the actions of Msx2 in the C3H10T1/2 model. We first established that the receptors for Dkk1 and canonical Wnt signaling, LRP5 and LRP6, participated in Msx2 induction of ALP. As shown in Fig. 9A, siRNA directed to LRP6, and to a lesser extent LRP5, inhibited Msx2 induction of bone ALP in  (19). A, SFG-Dkk1 virus abrogated SFG-Msx2 induction of ALP activity, functionally validating the bioactivity of the Dkk1 expression vector. For each group, n ϭ 3. B, as compared with the SFG-LacZ control virus (upper panels), SFG-Dkk1 transduction (lower panels) markedly diminished osteogenic nodule formation in both WT (left panels) and Msx2Tg (right panels) bone mesenchymal cell cultures. ϫ400 magnification, scale bar, 0.1 mm.

FIGURE 7. The Msx2Tg increases Wnt7 and decreases Dkk1 protein accumulation in calvarial osteoblasts and long bone mesenchymal cells.
A, primary calvarial osteoblast and long bone mesenchymal cell cultures were prepared from WT and Msx2Tg siblings and extracts analyzed for Msx2-responsive canonical Wnt regulators by Western blot. Relative protein accumulation was normalized to that of the housekeeping protein, eIF2␣, and expressed as fold difference of the non-Tg control. Wnt7 (antibody recognizes both Wnt7a and Wnt7b) was increased by the Msx2Tg in calvarial and long bone cultures, whereas Dkk1 protein was reciprocally decreased. Wnt1 and Wnt3a were also regulated, increased in calvarial and long bone, respectively. Although expressed, Wnt10b protein was not induced by the Msx2Tg. B, in neonatal long bone, Wnt7 immunolocalized to endosteum (large arrow), trabecular spongiosa (small arrow), and early hypertrophic zone (HZ) chondrocytes (brackets). Wnt7 protein accumulation was also detected in secondary ossification centers of long bone (not shown). C, striking Wnt7 expression (arrows) was detected in marrow vascular sinusoids (white arrows), and the mineralizing growth plate (black arrows), as well as in osteocytes (fine black arrows) of adult long bone. D, no signal was observed in the absence of Wnt7 primary antibody. E, vascular cells lining the Haversian canal (asterisk and large arrow) also accumulate Wnt7 protein, as do osteocytes (fine black arrows). C3H10T1/2 cells. RT-qPCR analysis demonstrated that siRNA-targeted mRNA degradation following transient transfection reduced targeted cognate mRNA accumulation by 20 -40% (Fig. 9B). Moreover, Western blot analysis confirmed efficient reduction of LRP5 (Fig.  9C), demonstrating that the relatively minor impact of LRP5 siRNA was not because of inefficient reduction in LRP5 protein. We next examined the effects of targeting specific Wnt ligands on Msx2dependent ALP induction. Consistent with our prior results (19), Wnt7a and Wnt7b expression was dependent upon Msx2 (Fig. 9D) and necessary for full ALP induction in response to Msx2 (Fig. 9E). Targeting Wnt1, another canonical Wnt ligand regulated by Msx2, also reduced ALP expression. Targeting the noncanonical ligand Wnt5a had no impact upon Msx2-induced ALP expression (see supplemental material). Moreover, as observed in C3H10T1/2 cells, siRNA directed to Msx2, Wnt7a, and Wnt7b also reduced ALP mRNA accumulation in primary bone mesenchymal cell cultures from Msx2Tg mice (Fig. 9F). Thus, in both C3H10T1/2 mesenchymal cells and primary bone marrow mesenchymal cells, Wnt7a and Wnt7b participate in Msx2-induction of osteogenic differentiation. In C3H10T1/2 cells, Msx2-Wnt signaling is dependent upon LRP6 and, to a lesser extent, LRP5.
Msx2 Inhibits Dkk1 Promoter Activity and Selectively Displaces RNA Polymerase II from the Dkk1 Gene-To better understand the molecular mechanisms whereby Msx2 gives rise to net enhancement of Msx2-Wnt signaling, we examined the effects of Msx2 on Dkk1 promoter activity. As observed in vivo (Fig. 7), Msx2 suppresses Dkk1 mRNA accumulation (19) and Dkk1 protein secretion (data not shown) in C3H10T1/2 mesenchymal cells. Thus, we studied the effects of Msx2 on Dkk1 transcription in this multipotent mesenchymal cell line. The proximal 1.1 kb (nucleotides Ϫ1141  (n ϭ 4). Although Wnt1, Wnt3a, and Wnt10b were not induced, long bone Wnt7a (B, left) and Wnt7b (B, right) were significantly up-regulated by the Msx2Tg. Data for Wnt7b expression in long bone were obtained in parallel from an independent set of animals (n ϭ 3 per genotype). C, circulating levels of Dkk1 were also reduced in Msx2Tg mice (n ϭ 14) versus WT non-Tg sibs (n ϭ 16), consistent with changes in bone formation (Fig. 3) and with changes observed in primary cell culture (Fig. 7A). D, number of LacZ-staining cells in long bone cultures is significantly increased in Msx2Tg(ϩ);TOPGAL(ϩ) mice (n ϭ 7) versus TOPGAL(ϩ) siblings (n ϭ 6). See text for details. to Ϫ1) of the mouse Dkk1 promoter supports transcription in C3H10T1/2 cells, and co-expression of Msx2 significantly reduces Dkk1 promoter activity (Fig. 10A). Sequential 5Ј-deletion analysis revealed that basal activity was dependent upon regulatory elements encoded within the first 170 bp of the Dkk1 promoter (Fig. 10A). As first noted for OC promoter regulation by Msx2 (30), transcriptional suppression of Dkk1 activity was independent of intrinsic Msx2 DNA binding activity (Fig. 10B). C3H10T1/2 cells expressing Msx2, Msx2(P148H), or Msx2(T147A), variants that possess normal, enhanced, or absent DNA binding activity, respectively (30), exhibited equivalent suppression of 170 DKK1LUC activity (Fig. 10B). Moreover, as observed for the endogenous Dkk1 gene (Fig. 10C), siRNA targeting Msx2 significantly increases the activity of the proximal Dkk1 promoter in C3H10T1/2 cells, without altering RSVLUC activity (Fig. 10D).
The specific transcriptional regulators supporting Dkk1 transcription and targeted by Msx2 suppression via protein-protein interactions have yet to be determined. However, the basal transcriptional machinery via TFIIF and RNA pol II is necessary for these actions (30). To confirm the actions of Msx2 on Dkk1 transcriptional complexes, we performed ChIP assays assessing the association of total RNA pol II (N20 antibody) and phosphorylated transcriptionally active pol II (H5 antibody) with the Dkk1 gene. As shown in Fig. 10E, C3H10T1/2 cells transduced with SFG-Msx2 exhibit reduced levels of pol II association with the Dkk1 gene, as compared with control cells transduced with SFG-LacZ. By contrast, Msx2 expression had no impact upon pol II association with the osteopontin gene (Fig. 10F), a gene that is not suppressed by Msx2 expression (8). Thus, Msx2 inhibits Dkk1 gene transcription.

DISCUSSION
The osteoblast homeodomain proteins Msx1 and Msx2 are absolutely required for craniofacial skeletogenesis; in the absence of these two transcription factors, no bone is formed in the developing vertebrate skull (9,11). Importantly, Msx2 deficiency alone gives rise to globally severe hypomineralization, amelogenesis imperfecta, dentinogenesis imperfecta, and reduced long bone skeletal mass characterized as a low turnover osteopenia (9,60). The pathobiology of compromised skeletal mineral homeostasis with Msx2 deficiency is poorly understood. Maxson and coworkers (10,11) first demonstrated that Msx2 controls neural crest-derived osteoblast proliferative expansion as a necessary  and co-workers (22,23) who first established the important roles for canonical Wnt10b signaling in the control of reciprocal osteoblast-adipocyte lineage allocation in marrow. Our studies extend these observations by demonstrating that Msx2 reciprocally controls both the Wnt7 agonists and Dkk1 antagonist of LRP6 signaling, thus utilizing this novel paracrine Wnt signaling mechanism to augment bone mass via canonical ␤-catenin-mediated pathways. Because current data indicate the absence of significant Wnt10b-Msx2 (23) and Msx2-Wnt10b (this work) signaling interactions in bone, unique canonical Wnt ligand regulatory networks make function in parallel to fine-tune the control of mesenchymal cell lineage allocation and osteogenic differentiation.
We recently identified that Msx2 participates in the pro-calcific signals of diabetic aortic calcification, enhancing vascular expression of Wnt3a and Wnt7a while suppressing expression of Dkk1 (19). Importantly, as a universal inhibitor of canonical Wnt agonists for LRP5 and LRP6, Dkk1 inhibits osteoblast lineage allocation and differentiation (36) and bone formation (61,62). Strategies that reduce Dkk1 expression by either genetic or pharmacologic means increase bone mass accrual (63); loss of even a single Dkk1 allele is sufficient to result in increased bone mineral density (62). Conversely, mice expressing a Dkk1 transgene in bone develop osteopenia, with reductions in BFR/TV (36,61). Hence, the rodent skeleton is exquisitely sensitive to changes in Dkk1 tone.
Of note, both serum and skeletal levels of Dkk1 protein were significantly reduced in Msx2Tg mice. This is consistent with Msx2Tg-induced increases in the net activation of marrow Wnt/␤-catenin signaling and skeletal mineral accumulation. Msx2Tg activation of skeletal Wnt signaling was clearly demonstrated by the following: (a) trabecular TOPGAL staining; (b) long bone LacZ reporter expression; (c) increased LacZ expression of bone marrow cell numbers; and (d) increased long bone osteoblast ␤-catenin protein accumulation. Moreover, restoration of Dkk1 expression by transduction of cultured Msx2Tg bone marrow cells with SFG-Dkk1 inhibited osteogenesis and restored adipogenic potential. We propose that a major mechanism whereby Msx2 enhances canonical Wnt signaling is by suppression of the LRP5 and LRP6 antagonist Dkk1, thus augmenting the impact of all paracrine canonical Wnt ligands on mesenchymal cell fate and physiology.
The induction of multiple canonical Wnt ligands by Msx2, with concomitant suppression of Dkk1, represents an additional mechanism whereby Msx2 enhances net canonical Wnt signaling in bone. The relative contributions of Dkk1 suppression versus Wnt agonist induction have yet to be determined. Our observation that depletion of Wnt7a, Wnt7b, and Wnt1 expression in the C3H10T1/2 mesenchymal cell model suggests that Wnt1 and Wnt7 family members substantially contribute to the skeletal osteogenic activity of Msx2. However, this remains to be proven in vivo.
The molecular details of Wnt gene induction by Msx2 are completely unknown; however, given the important role of Msx2 as a transcriptional suppressor, it is tempting to speculate that Msx2 up-regulates target Wnt gene expression indirectly by inhibiting the expression of a Wnt gene inhibitor. Now that activation of Wnt signaling by Msx2 has been identified to con-tribute to mineralization in both orthotopic and heterotopic venues, the mechanisms of Wnt gene induction by Msx2 will become the subject of intense investigation. Msx2Tg regulation of the Wnt expression profile differed somewhat in calvarial versus long bone osteoblasts. Studies detailing Msx2 regulation of genomic targets in neural crest versus paraxial mesoderm osteoprogenitors may provide new insights into the transcriptional underpinnings of membranous versus endochondral bone formation.
Over the past decade, our group (6) and others (4,15,64,65) have characterized the molecular mechanisms whereby Msx2 suppresses gene transcription. Antagonistic protein-protein interactions, independent of Msx2 DNA binding activity, are conveyed by the Msx2 homeodomain N-terminal arm and extension. Targets of Msx2 inhibition include transcriptional activators such as Dlx5 (65), Dlx3 (65), Runx2 (6,15), and C/EBP-␣ (28, 66). As we described previously in the osteocalcin gene (30) and peroxisome proliferator-activated receptor-␥ genes (28), Msx2 suppression of the Dkk1 gene is independent of intrinsic Msx2 DNA binding activity. Moreover, ChIP assays confirm that Msx2 selectively down-regulates pol II association with the Dkk1 gene but not the osteopontin gene, the latter a gene unaffected by Msx2 expression (8). Thus, the mechanism whereby Msx2 suppresses Dkk1 transcription will also be conveyed via antagonistic protein-protein interactions with regulators that support Dkk1 promoter activity. The specific transcriptional activators of the Dkk1 gene targeted by Msx2 are as yet unknown but are in the process of being identified.
Lian and co-workers (67) clearly demonstrated the important stage-specific role of Runx2/Cbfa1 isoforms for osteoblast development in response to canonical Wnt signaling (68). Wnt3a stimulation depends upon ␤-catenin and a TCF cognate encoded by the Runx2 P1 promoter, and we have confirmed their seminal observations (data not shown). We were therefore surprised that Runx2 mRNA accumulation was not enhanced by the Msx2 transgene. However, because Msx2 dimerizes with and antagonizes Dlx5 and Dlx3, factors that support Runx2 gene expression (65,69), initiation of canonical Wnt signals in the setting of constitutive Msx2 expression may blunt the upregulation of other important targets such as Runx2. Indeed, the magnitude of bone formation we observed with Msx2 is less that that observed with lithium, a powerful stimulus for ␤-catenin induced bone formation (35). Whether Msx2 inhibits Wnt induction of Runx2 via Dlx or other transcriptional activators remains to be determined.
Although possessing ϳ77% protein sequence identity, Wnt7a and Wnt7b provide distinct signals in skeletal homeostasis. Wnt7a is a dorsalizing signal in vertebrate limb development that is also required for anterior-posterior forelimb patterning (57,58). Signaling cascades have been incompletely characterized, but both canonical and noncanonical Wnt signals are elaborated by Wnt7a via LRP6 (57,58). Wnt7b was recently identified as controlling endochondral bone formation via noncanonical Wnt signaling pathways (59). Wnt-activated protein kinase C-␦ (PKC-␦) pathway signaling, insensitive to Dkk1 yet still dependent upon fully intact Dishevelled platform protein functions (70), was shown to mediate bone anabolism (59). In our studies, neither rotterlin nor Ro-31-8220, antago-nists of PKC-␦ activity, inhibits Msx2 induction of ALP, and phospho-myristoylated alanine-rich protein kinase C substrate protein accumulation is not affected by Msx2 expression. 4 This indicates that the noncanonical Wnt/PKC-␦ pathway does not mediate the pro-osteogenic activities of Msx2. Rather, our data demonstrate that Msx2 activates canonical Wnt/␤-catenin signaling in the skeleton, conveyed in part via Wnt7a and Wnt7b. Wnt7 expression in Msx2Tg bone exhibits a pattern overlapping that of Msx2 protein accumulation, and persists throughout skeletal maturation to include the osteocytes of the Haversian system. The marrow vascular sinusoid Wnt7 staining pattern is highly reminiscent of that for the CD146-expressing subendothelial marrow stromal cell recently described by Bianco and co-workers (71). Based upon histoanatomic considerations, this important mural cell population is contiguous with the roof of the bone remodeling compartment (72). Future studies will address the relative contributions of Wnt7a and Wnt7b to Msx2 skeletal anabolism in long bone and calvaria and clarify the potential roles of perivascular, osteoblast, and osteocyte Wnt7 expression to postnatal bone homeostasis.
Of note, siRNA directed to the Wnt co-receptor LRP6 was consistently capable of reducing the effects of Msx2 in C3H10T1/2 cells; LRP5-directed siRNA exerted much less effect on Msx2 induction of ALP. Meisler and co-workers (57,58) first demonstrated the interactions between Wnt7a and Dkk1 that are critically important for local dorsal and posterior patterning of the developing mouse limb via LRP6. Elegant murine and human genetic studies have now demonstrated globally important roles for LRP6 in skeletal homeostasis. The human LRP6(R611C) variant is a hypomorphic allele that impairs Wnt ligand dose response and conveys risk for osteoporosis and metabolic syndrome pathophysiology (73). Williams and co-workers (74) recently demonstrated that reduction of one LRP6 allele in mice completely lacking LRP5 also impairs bone mass accrual. Based upon those recent observations, we propose that Wnt7a and Wnt7b convey canonical signals important for Msx2-mediated osteogenesis via LRP6 Ͼ LRP5, receptors antagonized by Dkk1 (see supplemental material for working model). However, the relative contributions of LRP6 and LRP5 to Msx2-dependent osteogenic signaling in calvaria and long bone have yet to be determined in vivo. Moreover, although the down-regulation of Dkk1 by Msx2 potentiates canonical Wnt signaling, the panoply of LRP5 and LRP6 agonists mediating bone formation in vivo have yet to be identified. In the craniofacial skeleton, Wnt1 is also likely to be a critical signal downstream of Msx1 and Msx2 (9,75). Interestingly, the LRP6 Ϫ/Ϫ mouse (76) exhibits absent craniofacial bone and exencephaly observed in Msx2 Ϫ/Ϫ ;Msx1 Ϫ/Ϫ double knock-out mice (9). We speculate that gene dose interactions between LRP6 and Msx2 may yield progressively severe abnormalities in craniofacial bone formation.
Although not quite reaching statistical significance (p ϭ 0.08), the nonsignificant downward trend in osteoclast numbers is also consistent with Msx2 activation of canonical Wnt signals. Several laboratories have demonstrated Wnt-dependent suppression of osteoclast formation via induction of osteoprotegerin (77) and suppression of RANKL production (77) by bone mesenchymal cells. However, Msx2 exerts direct actions in craniofacial osteoclasts as well (60); cell autonomous contributions of Msx2 signaling in long bone osteoclasts have yet to be determined.
Recently, a high throughput screening assay identified TAK-778, a compound that enhances alkaline phosphatase activity via Msx2 (78). Sustained local delivery of TAK-778 in biodegradable capsules enhances repair of critical defects in a low turnover models of fracture non-union (79). Based upon our results, we speculate that TAK-778 will also activate Wnt/␤catenin signaling via paracrine signals provided at least in part via Wnt7 family members. However, the Msx2-Wnt pathway also contributes to arterial calcification in murine diabetic disease models (19,80) and in humans (81,82). Thus, therapeutics that globally enhance Msx2-Wnt signaling may exhibit mechanism-based toxicities in the arterial tree (19).
There are, of course, limitations to our study. First, whereas mesenchymal cells of periosteum and subsets of hypertrophic chondrocytes exhibit expanded Msx2 protein accumulation, the precise cell populations with augmented Msx2 expression in the bone marrow envelope are as yet undefined. In vivo, the human CMV promoter drives expression of transgenes in mice with much more selectivity than occurs in vitro, generally restricted vascular, neural, epithelial, and mesenchymal tissues that are sites of congenital human CMV infection (83). Bone mesenchymal cells with osteogenic potential are clearly identified in primary cell culture, but the full panoply of cells in the marrow environment that elaborate Msx2-regulated genomic programs is as yet unknown. Stage-specific elaboration of Msx2-Wnt signaling during osteoblast lineage maturation may exert differential bone anabolic responses. Moreover, our observations are made in mice expressing the Msx2 transgene during stages of skeletal growth and development (84). Thus, the magnitude of Msx2 bone anabolic actions may differ in the growing and modeling skeleton (this study) versus the remodeling skeleton of maturity. Of note, we chose to utilize immunohistochemistry rather than in situ hybridization for this study; Msx and Dlx family members (5,85) are prodigiously regulated via post-translational protein-protein interactions that control complex stability via Praja1, a ubiquitin E3 ligase (86). Msx2 also binds Miz1/PIASx␤ (87), a small ubiquitin-like molecule E3 ligase that promotes osteogenic mineralization via Osx-NFATc1 complexes without regulating Runx2 (88). Some effects of Msx2 may be mediated via regulation of these novel, mechanically sensitive ubiquitination and sumoylation systems; we have not studied the role of osteogenic E3 ligases in Msx2 actions. Intriguingly, in adult bone the pattern of Wnt7 protein accumulation appears increasingly dissociated from declining levels of Msx2 protein accumulation, 4 even though Msx2 mRNA accumulation remains elevated in MsxTg mice. Thus, mechanisms of durable genetic and epigenetic control of Wnt7 gene expression are likely coupled to Msx2-regulated protein biochemistry. In addition, although siRNA to Wnt7a and Wnt7b clearly reduced ALP and Osx induction in vitro, other Wnt family members will likely contribute to Msx2 actions in vivo; Wnt1 may be particularly important in the craniofacial skeleton (75). Our systematic analysis, however, reveals specificity; siRNAs directed to Wnt5a were without significant effect, even though equivalent knockdown of cognate Wnt mRNA accumulation was achieved. The observation that siRNA to LRP6 inhibits Msx2-induced ALP induction to a greater extent than LRP5 siRNA suggests that the Wnt ligands induced by Msx2 may preferentially activate LRP6. Although redundancy exists between LRP5 and LPR6, human and murine genetic studies demonstrate functionally important distinct roles for LRP6 in skeletal development and postnatal bone homeostasis (see above). The preferential use of LRP6 to transduce Msx2-Wnt paracrine signals remains to be established in vivo and may differ dependent upon cell type. Finally, given the significant functional redundancy of Msx1 and Msx2 (9), Msx1 is also likely to regulate Wnt gene expression. This possibility has not been directly assessed, but the important recent development of conditional alleles for both Msx1 and Msx2 will now permit detailed analysis of endogenous Msx gene contributions to skeletal Wnt signaling (89). Nonetheless, our data demonstrate that, as suggested by studies of ectopic arterial calcification (19,28), Msx2 exerts bone anabolism by augmenting canonical Wnt signals, signals that reciprocally promote osteogenic versus adipogenic differentiation of skeletal mesenchymal progenitors. Although certainly most germane to skeletal physiology, these mechanistic insights have implications for the roles of Msx2-expressing cells in Wnt-regulated breast and central nervous system biology (9,90,91).