The WNT molecules belong to a family of secreted proteins consisting of 19 members in mammals and play important roles in embryonic development and tissue homeostasis.1 WNT proteins activate multiple downstream signaling cascades depending on the cell context, one of which is the β-catenin–dependent pathway. In this pathway, binding of WNT to a Frizzled (Fz) receptor and a low-density lipoprotein receptor-related protein 5/6 (LRP5/6) coreceptor leads to stabilization of β-catenin, which subsequently translocates to the nucleus where it interacts with members of the T-cell factor/lymphoid enhancer factor (TCF/LEF) transcription factors to activate transcription of downstream target genes.2
Genetic studies, by conditionally deleting β-catenin at different stages of osteoblast development during embryogenesis, have implicated WNT/β-catenin signaling in regulating multiple steps of osteoblast differentiation. Specifically, deletion of both LRP5 and LRP6, or β-catenin in mesenchymal progenitors results in a failure to initiate or maintain Osterix (Osx)-positive preosteoblasts.3–6 Deletion of β-catenin in Osx-positive cells abolishes subsequent differentiation to mature osteoblasts.7 Interestingly, deletion of β-catenin in mature osteoblasts or osteocytes indirectly increases osteoclast number and activity without a notable effect on osteoblasts.8–10 Thus, WNT/β-catenin signaling plays multiple and stage-specific roles in the osteoblast lineage.
Other genetic studies have implicated WNT signaling in promoting postnatal bone formation. In humans, loss-of-function mutations in LRP5 cause osteoporosis-pseudoglioma syndrome, a form of early-onset osteoporosis,11 whereas gain-of-function point mutations in LRP5 result in high bone mass.12, 13 Moreover, deficiency in either expression or function of sclerostin (SOST), a secreted protein that inhibits the binding of WNT molecules to LRP5/6, results in Van Buchem disease or sclerosteosis, both exhibiting high bone mass.14, 15 Mouse genetic studies have demonstrated that LRP5 promotes postnatal bone mass accrual by increasing osteoblast number and function,16, 17 with LRP6 performing a redundant role.18 In addition, mice lacking SOST or missing one DKK1 allele exhibit higher bone mass predominantly resulting from increased bone formation.19, 20 Thus, multiple lines of genetic evidence support an important role for WNT-LRP5/6 signaling in stimulating postnatal bone formation, but an alternative mechanism for LRP5 function has also been proposed.21 Because embryonic deletion of β-catenin in early-stage osteoblast-lineage cells leads to perinatal lethality whereas deletion in the more mature cells did not affect osteoblasts per se, it remains an open question whether WNT-LRP5/6 signaling regulates postnatal osteoblast number or function in a β-catenin–dependent manner.
Here we use the Osx-CreERT2 mouse that expresses tamoxifen (TM)-inducible Cre from the regulatory sequences of Osx (official symbol Sp7), both to examine the contribution of postnatal Osx-expressing cells and to determine the role of β-catenin in these cells. Osx is a zinc-finger transcription factor whose expression initiates in preosteoblasts, and it is indispensable for osteoblast differentiation.22, 23 During endochondral bone development in the mouse embryo, Osx-expressing cells migrate to the nascent bone marrow cavity and produce trabecular osteoblasts, osteocytes, and stromal cells.24 Here we report that in postnatal mice, Osx-expressing cells include both osteoblast-lineage cells and transient progenitors that give rise to bone marrow stromal cells and adipocytes. We further provide genetic evidence that β-catenin in Osx-expressing cells is necessary for osteoblast function in postnatal bones.
Subjects and Methods
The β-catenin conditional mouse (Ctnnb1c/c), R26-mT/mG reporter mice, and Osx-CreERT2 mice have been reported.24–26 The Animal Studies Committee at Washington University reviewed and approved all mouse procedures used in this study.
Two-month-old Osx-CreERT2;R26-mT/mG, Osx-CreERT2;Ctnnb1c/c, or Ctnnb1c/c mice were subjected to either corn oil (vehicle) or TM administration (80 µg per gram of body weight) by oral gavage once daily for 5 consecutive days. The mice were harvested at 8, 21, or 49 days after the first treatment for analyses. All data presented were derived from male mice.
Cryojane section and immunofluorescence
Mice were perfused with 4% paraformaldehyde (PFA) according to a standard protocol. After perfusion, mouse femurs were isolated and then fixed in 4% PFA overnight. After 3 days of decalcification by 14% EDTA, femurs were snap-frozen in optimal cutting temperature (OCT) embedding medium and then sectioned at 8 µm using a Leica cryostat equipped with Cryojane (Leica, IL). For detection of green fluorescent protein (GFP), perilipin, or osteocalcin immunostaining was performed on frozen sections using a chicken polyclonal GFP antibody (Abcam, Cambridge, MA, USA), rabbit monoclonal perilipin antibody (Cell Signaling Technology, Danvers, MA, USA), or rabbit polyclonal osteocalcin antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA).
X-ray radiography and micro–computed tomography
Radiographic images of either the whole body or the hindlimbs were acquired using a Faxitron X-ray system (Faxitron X-ray Corp) at 25 kv for 20 seconds. The femurs were scanned by a micro–computed tomography (µCT) system (µCT 40; Scanco Medical AG) following the recommendations by American Society for Bone and Mineral Research.27 For quantifying trabecular bone parameters, 100 µCT slices (1.6 mm total) immediately below the growth plate of femurs were analyzed (threshold set at 300). For quantifying cortical bone parameters, 50 µCT slices (0.8 mm total) starting from 6.8 mm below the articular surface of femurs were analyzed.
Histology and histomorphometry
For histological analysis, femurs were isolated from mice after perfusion with 4% PFA, and fixed in 10% buffered formalin overnight at room temperature, followed by decalcification in 14% EDTA for 2 weeks. After decalcification, femurs were processed for paraffin embedding and then sectioned at 6 µm thickness. Hematoxylin and eosin (H&E), Alcian blue/picrosirius red staining, and tartrate-resistant acid phosphatase (TRAP) staining were performed on paraffin sections following the standard protocols. For dynamic histomorphometry, calcein (Sigma, St. Louis, MO, USA) was injected intraperitoneally at 20 mg/kg on days 7 and 2 prior to euthanasia, and bones were fixed in 70% ethanol and embedded in methyl-methacrylate for sectioning. Both static and dynamic bone histomorphometry were performed with the computer software Bioquant II.
RNA extraction and qPCR
For RNA extraction, femurs and tibias were cleanly dissected. After removal of the epiphysis and the bone marrow (by centrifugation), the bones were rinsed twice in the cold PBS, and cut into small pieces. Then 1 mL Trizol (Invitrogen) was added to extract RNA. One microgram (1 µg) RNA was used to synthesize cDNA using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA). qPCR was performed with SYBR-Green Supermix (Bio-Rad). The primers used were as follows: Ctnnb1 (5′-CCTCCCAAGTCCTTTATGAATGG-3′ and 5′-CCGTCAATATCAGCTACTTGCTCTT-3′); osteoprotegerin (OPG) (5′-CCGAGGACCACAATGAACAAGT-3′ and 5′-CTGGGTTGTCCATTCAATGATG-3′); receptor activator of NF-κB ligand (Rankl) (5′-CTGGGCCAAGATCTCTAACATGA-3′ and 5′-GGTACGCTTCCCGATGTTTC-3′); and ribosomal RNA (18S) (5′-CGGCTACCACATCCAAGGAA-3′ and 5′-GCTGGAATTACCGCGGCT-3′).
Serum cross-linked C-telopeptide and N-terminal propeptide of type I collagen assays
For serum cross-linked C-telopeptide (CTX) and amino-terminal propeptide of type I procollagen (P1NP) assays, serum was collected from mice after 6 hours of fasting. CTX and P1NP assays were performed with the RatLaps ELISA kit and the Rat/Mouse P1NP EIA kit (both from Immunodiagnostic Systems, Ltd.), respectively.
Postnatal Osx-expressing cells include transient progenitors for bone marrow stromal cells and adipocytes
We first examined what cells are targeted by Osx-CreERT2 upon tamoxifen (TM) administration in 2-month-old mice. To this end, we generated Osx-CreERT2;R26-mT/mG animals so that GFP expression could be activated by TM only in Osx-expressing cells and their progeny. The mice were administered either vehicle or TM once daily for 5 consecutive days, and then harvested at different time points for GFP detection (Fig. 1A). To detect the cells that were initially targeted by this strategy, we harvested the mice at 8 days after the first treatment (D8). As expected, bones from the vehicle-treated animals did not show any GFP (Fig. 1B). In contrast, bones from the TM-treated mice exhibited many GFP-positive cells (Fig. 1C). Most positive cells were detected at the chondro-osseous junction of the primary ossification center (Fig. 1C1), the trabecular bone surface (Fig. 1C2) and the endocortical surface (Fig. 1C3). Quantitative analyses revealed that GFP-positive cells covered approximately 65% of the trabecular bone surface and 85% of the endocortical surface in the femur. GFP was also detected at the secondary ossification center, albeit at a lower intensity (Fig. 1C, asterisk), but none within the growth plate (Fig. 1C, bracket) or at the periosteum (Fig. 1C3, red arrow). Relatively few osteocytes expressed GFP (Fig. 1C3, white arrow), and no GFP was detected in the bone marrow (Fig. 1I). Double immunostaining for GFP and osteocalcin (OC) confirmed that a majority of the GFP-positive cells on the bone surface express OC (Fig. 1D). Thus, Osx-CreERT2 initially targets mainly osteoblast-lineage cells on the bone surface.
To track the fate of the cells initially targeted by Osx-CreERT2, we analyzed GFP expression in animals harvested at 21 days after the first TM treatment (D21). Here, 55% of the trabecular bone surface remained GFP-positive, slightly but statistically significantly reduced from the coverage at D8 (Fig. 1F, H). However, many GFP-positive cells appeared within the bone marrow (Fig. 1F). These cells exhibited a typical reticular morphology of stromal cells, and were often present in clusters (Fig. 1J, arrow). Because they were not detected at D8 (Fig. 1I), the GFP-positive bone marrow stromal cells were most likely derived from Osx-expressing progenitors initially targeted by Osx-CreERT2.
To examine the fate of the targeted cells for a longer time, we analyzed the mice at D49. GFP-positive cells were still detected on both trabecular and endocortical bone surfaces, but their coverage of the total surfaces was greatly reduced (down to 25% for trabecular bone) (Fig. 1G, H; and data not shown). Similarly, the number of GFP-positive stromal cell clusters and the size of each cluster were dramatically reduced when compared to D21 (Fig. 1K, L). Thus, both bone-surface osteoblast-lineage cells, and bone marrow stromal cells experienced appreciable turnover between D21 and D49, and the Osx-expressing progenitors had only a limited capacity to replenish these populations.
Because adipocytes are an important constituent of the bone marrow in adult mice, we next asked whether Osx-expressing cells normally contribute to adipocytes. We immunostained the bone sections from the Osx-CreERT2;R26-mT/mG mice for both GFP and the adipocyte-specific marker perilipin. At D8, the TM-treated mice, like the vehicle-treated controls, exhibited no colocalization of GFP and perilipin in the 261 adipocytes counted, indicating that adipocytes were not directly targeted by Osx-CreERT2 (Fig. 2A, B). However, at D21, GFP and perilipin were found to colocalize in 3% of the 436 adipocytes counted (Fig. 2C). The colocalization was specific to bone marrow adipocytes, because it was not found in other adipose tissue (data not shown). At D49, colocalization was observed in 1% of 637 adipocytes counted (Fig. 2D). Thus, an Osx-expressing progenitor normally contributes to adipocytes in the bone marrow of postnatal mice at a low rate.
Deletion of β-catenin in postnatal Osx-lineage cells results in high-turnover osteopenia and increased bone marrow adiposity
Having established the cell populations targeted by Osx-CreERT2, we proceeded with using this approach to delete β-catenin (Ctnnb1) in the postnatal skeleton. We administered TM to 2-month-old Osx-CreERT2;Ctnnb1c/c mice and their littermate controls (ie, Ctnnb1c/+ or Ctnnb1c/c) once daily for 5 consecutive days. These mice were harvested at 8, 21, or 49 days after the first TM administration (hereafter KO-D8, KO-D21, and KO-D49, respectively). X-ray radiography detected no obvious change in the KO-D8 mice, but severe osteopenia throughout the skeleton in the KO-D21 and KO-D49 animals (Fig. 3A, B; and data not shown). Because a similar phenotype was observed with both male and female mice, we have focused all subsequent analyses on males. µCT analyses of the trabecular bone in the femur confirmed that bone mass (BV/TV) in the KO-D8 mice was relatively normal, but that in the KO-D21 and KO-49 mice was progressively reduced when compared to their respective littermate controls (Fig. 3C–E). The decrease in bone mass in the KO-D21 and KO-49 mice was associated with reduced trabecular number and increased trabecular spacing, with little change in trabecular thickness (Fig. 3D, E). In keeping with the µCT results, picrosirius red staining of the femur revealed progressive worsening of the trabecular bone loss in the KO-D21 and KO-D49 mice, although the KO-D8 mice were relatively normal (Fig. 3F). This staining also showed a loss of the cortical bone at the metaphyseal region in the KO-D49 mice (Fig. 3F, arrows). Moreover, µCT analyses of the diaphyseal cortical bone showed a smaller diameter and reduced cortical thickness in the KO-49, but not the KO-21 mice when compared to their respective littermate controls (Fig. 3G–I). Overall, TM-induced deletion of β-catenin by Osx-CreERT2 in 2-month-old mice results in progressive general osteopenia.
We next sought to understand the cellular basis for the osteopenia caused by β-catenin deletion. We assessed total bone resorption activity in the body by measuring serum CTX levels. There was no change in the KO-D8 mice, but CTX was increased by 118% and 149% over the controls in the KO-D21 and KO-D49 mice, respectively (Fig. 4A, E, I). TRAP staining on femur sections revealed an increase in both osteoclast number per bone perimeter (No.OC/BS) and the spreading of individual osteoclasts (µm/OC) in the KO-D21 and KO-D49 mice, which together led to a significant increase in the percentage of bone surface covered by osteoclasts (OC.S./BS) (Fig. 4F–H, J–L). In contrast, these osteoclast parameters were not altered in the KO-D8 mice (Fig. 4B–D). To probe the molecular basis for the enhanced osteogenesis, we measured RANKL and OPG levels in the serum. However, these assays did not detect any significant difference between any of the mutant mice and their respective controls (Fig. 4M–R). On the other hand, qPCR experiments with RNA extracted from the bone surface cells detected a decrease in Opg but no change in Rankl mRNA in the KO-D8 versus control mice (Fig. 4S, T). These experiments also revealed that β-catenin (official name Ctnnb1) mRNA was reduced by ∼40% in the KO-D8 mice (Fig. 4U). Whether or not the observed changes in Opg can account for the increased osteoclastogenesis remains to be tested. Overall, deletion of β-catenin in Osx-expressing cells induces excessive osteoclastogenesis and bone resorption.
We next examined total bone formation activity by measuring serum P1NP. Although the KO-D8 mice did not show an overall reduction in bone mass, they exhibited a marked reduction in P1NP (Fig. 4V). In contrast, the KO-21 and KO-49 mice that were severely osteopenic showed a much higher level of P1NP than their control littermates (Fig. 4W, X). Thus, β-catenin deletion in the postnatal Osx-expressing cells initially decreases bone formation activity without overtly reducing bone mass, but subsequently exuberant bone resorption causes high turnover osteopenia.
In analyzing the bone phenotype, we observed a marked increase in bone marrow adiposity in the KO-D49 mice (Fig. 5A, B). The increase was localized specifically to the metaphyseal regions of the long bones (Fig. 5A1–A3, B1–B3). On H&E sections of the femur, we found a 14-fold increase in the number of adipocytes in the KO-D49 mice over their littermate controls (Fig. 5C), and that the average size of the adipocytes in the mutant mice also increased (Fig. 5A1, B1). To investigate the possibility that Osx-positive progenitors might differentiate preferentially to adipocytes upon β-catenin deletion, we generated mice with the genotype of Osx-CreERT2;R26-mT/mG;Ctnnb1c/c (KO-GFP), and compared their percentage of GFP-positive adipocytes among total adipocytes with that in the Osx-CreERT2;R26-mT/mG mice at D21. Although we observed an increase in 1 out of 3 KO-GFP mice, a similar change was not seen in the other 2 (Fig. 5D). We conclude that fate switch of the Osx-positive progenitors may not be the driving force for the excessive adipogenesis in the KO-D49 mice.
Loss of β-catenin impairs osteoblast activity and increases osteoblast turnover
We next sought to determine the basis for the impaired bone formation in the KO-D8 mice. Because Osx-CreERT2 mainly targets bone-surface osteoblast-lineage cells at D8, we reasoned that deletion of β-catenin might directly affect osteoblast number and/or function. Histomorphometry showed that osteoblast number normalized to bone surface was not changed in the KO-D8 mice, although it was slightly increased in the KO-D21 and KO-D49 mice (Fig. 6A–C). However, the mineral apposition rate (MAR), an indicator of osteoblast function, was greatly reduced in the KO-D8 mice (Fig. 6D, E). Moreover, both the double-labeled surface normalized to total bone surface (Dls/BS) and the mineralized surface normalized total bone surface (MS/BS) were markedly decreased in the KO-D8 mice (Fig. 6F; and data not shown). Consequently, the bone formation rate (BFR) was much reduced in these mice (Fig. 6G). Because of the extremely low bone mass, we could not obtain reliable data for osteoblast activity in the KO-21 or KO-49 mice. Thus, β-catenin deletion in osteoblast-lineage cells greatly suppresses osteoblast activity in postnatal mice.
We next sought to explain the unexpected observation that osteoblast numbers and bone formation activity were increased in the KO-D21 and KO-49 mice. We hypothesized that osteoblasts initially targeted for β-catenin deletion might be replaced by newly produced wild-type osteoblasts by D21. To investigate this possibility, we monitored the presence of the β-catenin–deficient and GFP-positive osteoblasts in the KO-GFP mice at D21. Indeed, there were notably fewer GFP-positive osteoblast-lineage cells associated with bone surfaces in the KO-GFP mice than the Osx-CreERT2;R26-mT/mG control mice at D21 (Fig. 7A, B). Quantification revealed that only 28% of the trabecular bone surfaces were covered by GFP-positive cells in the KO-GFP mice at D21, compared to the Osx-CreERT2;R26-mT/mG mice with 55% at D21 and 25% at D49. Thus, loss of β-catenin greatly accelerates the turnover of osteoblast-lineage cells at the bone surface.
We investigated the fate of postnatal Osx-expressing cells, and the role of β-catenin in the Osx-lineage cells in postnatal bone homeostasis. We found that Osx-expressing cells in postnatal mice include not only osteoblast-lineage cells, but also transient progenitors that contribute to stromal cells and adipocytes in the bone marrow. Deletion of β-catenin in the Osx-lineage cells initially greatly reduces bone formation, but later markedly increases bone resorption and marrow adiposity. These results provide, to our knowledge, the first genetic evidence that β-catenin critically regulates osteoblast behavior in postnatal bones.
This study provides new insight about the turnover of osteoblast-linage cells. In a normal mouse, the trabecular bone surface covered by labeled osteoblast-lineage cells decreased from 55% to 25% within 4 weeks (D21 to D49). Assuming no major replenishment of the labeled cells, we estimate that approximately one-half of the cells turn over within this time, and that the average life span of bone surface osteoblast-lineage cells is about 2 months. This estimation is in agreement with a previous report about calvarial osteoblasts.28 The apparent turnover rate between D8 and D21 was lower than expected (from 65% to 55%), but this could be explained if a labeled progenitor population transiently replenished the bone-surface cells during this time. Alternatively, this could simply reflect the time delay required for sufficient removal of β-catenin from the targeted cells. Importantly, deletion of β-catenin in the Osx-lineage cells markedly accelerated the apparent turnover rate of bone-surface cells by D21. Because the high apparent turnover rate exceeded what could be expected from a normal osteoblast lifespan even if the replenishment rate was reduced to zero, we conclude that loss of β-catenin shortens the life span of osteoblasts. It is not clear at present whether the role of β-catenin in osteoblast life span reflects a similar role for LRP5. Studies to date have only reported on osteoblast apoptosis in the calvaria of LRP5 mouse models. Although one study showed that overexpression of a high-bone-mass variant of LRP5 reduced apoptosis,29 another study detected no change in osteoblast apoptosis in the LRP5–/– mice.16
After the initial suppression of osteoblast activity, deletion of β-catenin in postnatal Osx-expressing cells subsequently caused excessive osteoclastogenesis. The increased osteoclastogenesis is consistent with previous reports about deletion of β-catenin in mature osteoblasts or osteocytes.8–10 These previous studies proposed downregulation of OPG expression by osteoblasts or osteocytes as the primary mechanism for the increased osteoclastogenesis. Although we detected a modest decrease in OPG mRNA in bone surface cells in the KO-D8 mice (3 days after last TM administration), we did not observe a change in OPG protein concentrations in the serum at any time point. In addition, we did not detect any changes in the serum RANKL protein level in our mice, even though it was reported to be higher than normal when β-catenin was deleted specifically in osteocytes.10 It is possible that a change in the local RANKL/OPG ratio within the bone microenvironment was responsible for the drastic increase in osteoclastogenesis, but this possibility needs to be examined further in the future. Finally, because Osx-CreERT2 also targeted a subset of osteocytes in response to TM, deletion of β-catenin in this population might have contributed to the increased osteoclastogenesis, especially at the later time points when the targeted osteoblasts had been replaced largely by normal osteoblasts.
Concurrent with the increased bone resorption, bone formation also increased in the KO-D21 and KO-D49 mice. The increased bone formation correlated with an increased number of osteoblasts per bone surface area. This increase in osteoblast number and total bone formation activity is unexpected, but is explained by the finding that a large majority of the osteoblasts at the later time points following TM treatment were not targeted by our gene deletion strategy. The mechanism for the large increase in wild-type osteoblasts is not understood at present, but may be secondary to the excessive bone resorption, which may stimulate osteoblastogenesis through the release of growth factors from the bone matrix.30, 31
Through lineage-tracing experiments, we show that in postnatal mice, there are Osx-expressing cells that give rise to bone marrow stromal cells and adipocytes. This result extends the previous finding that during embryogenesis, Osx-expressing progenitors within the perichondrium contribute to bone marrow stromal cells.24 Because the postnatal Osx-expressing cells cannot sustain a stable population of stromal cells or adipocytes, they are most likely transient progenitors, not self-renewing stem cells. It is not known at present whether stromal cells, adipocytes, and osteoblasts share the same progenitor, or simply the same marker Osx among the different lineage-restricted progenitors.
Finally, it is not fully understood at present what caused the marked increase in bone marrow adiposity following β-catenin deletion in postnatal Osx-expressing cells. While this article was in preparation, others reported a similar observation when β-catenin was deleted with the Tet-off Osx-Cre by keeping the mice off Dox for 2 to 4 months beginning at 2 months of age.32 Through both in vitro cell culture and in vivo lineage tracing experiments, the authors concluded that loss of β-catenin directly shifted the differentiation potential of Osx-positive preosteoblasts to adipocytes. Here we show that Osx-positive progenitors normally contribute to a small percentage of bone marrow adipocytes in our experimental setting, but deletion of β-catenin did not cause a consistent increase in the contribution at D21. It remains possible that a shift in differentiation to adipocytes occurred at a later stage to cause the marked increase in bone marrow adiposity. On the other hand, because the increase in adipocyte number was not obvious until long after TM treatment (D49), loss of β-catenin in the Osx-positive progenitors may not be the sole driver for the excessive adipogenesis. Instead, changes in the bone marrow microenvironment caused by excessive bone resorption may be an important factor. Overall, the present work demonstrates that β-catenin in postnatal Osx-expressing cells plays multiple roles in maintaining bone homeostasis. In particular, the role of β-catenin in regulating postnatal osteoblast function and life span lends support to the view that WNT/LRP5 signaling may regulate postnatal bone accrual in part through β-catenin.
All authors state that they have no conflicts of interest.
This work was supported by NIH grants AR060456 and AR055923 (to FL), and P30 AR057235 (Washington University Musculoskeletal Research Center). We thank Dr. Henry M. Kronenberg (Massachusetts General Hospital, Harvard Medical School) for his generous donation of the Osx-CreERT2 mouse strain.
Authors' roles: FL directed the project, JC conducted the experiments, JC wrote a draft, and FL wrote the final version of the paper.