Rib Fractures and Death from Deletion of Osteoblast βcatenin in Adult Mice Is Rescued by Corticosteroids

Ribs are primarily made of cortical bone and are necessary for chest expansion and ventilation. Rib fractures represent the most common type of non-traumatic fractures in the elderly yet few studies have focused on the biology of rib fragility. Here, we show that deletion of βcatenin in Col1a2 expressing osteoblasts of adult mice leads to aggressive osteoclastogenesis with increased serum levels of the osteoclastogenic cytokine RANKL, extensive rib resorption, multiple spontaneous rib fractures and chest wall deformities. Within days of osteoblast specific βcatenin deletion, animals die from respiratory failure with a vanishing rib cage that is unable to sustain ventilation. Increased bone resorption is also observed in the vertebrae and femur. Treatment with the bisphosphonate pamidronate delayed but did not prevent death or associated rib fractures. In contrast, administration of the glucocorticoid dexamethasone decreased serum RANKL and slowed osteoclastogenesis. Dexamethasone preserved rib structure, prevented respiratory compromise and strikingly increased survival. Our findings provide a novel model of accelerated osteoclastogenesis, where deletion of osteoblast βcatenin in adults leads to rapid development of destructive rib fractures. We demonstrate the role of βcatenin dependent mechanisms in rib fractures and suggest that glucocorticoids, by suppressing RANKL, may have a role in treating bone loss due to aggressive osteoclastogenesis.


Introduction
bcatenin plays a critical role in commitment and differentiation of mesenchymal progenitors into different lineages and promotes osteoblast differentiation [1,2,3,4]. Osteoblasts not only form new bone but also control bone resorption by modulating osteoclast formation. Osteoblasts express the osteoclastogenic cytokine RANKL (receptor activator of nuclear factor kappa B ligand) that binds to RANK receptor present on osteoclast progenitors and induces osteoclast formation. RANKL is a critical determinant of osteoclastogenesis and close coupling of osteoblast and osteoclast activity regulates bone mass in the adult mammal [5]. In age related osteoporosis or bone loss related to excessive osteoclastogenesis, the skeleton does not lose bone in a uniform distribution, rather some skeletal sites exhibiting a far greater degree of bone loss compared to other regions. The biological basis of such skeletal site specific osteoclastogenesis remains unclear. bcatenin activity in bone is affected by physical loading of the skeleton and is important for skeletal site specific response to load [6,7]. Osteoblast bcatenin modulates RANKL activity and osteoclastogenesis [8] but whether site specific modulation of osteoclastogenesis by bcatenin occurs remains unknown.
Although osteoporotic studies concentrate on fractures of the hip and spine, rib fractures represent the most common nontraumatic fractures in the elderly and a large fraction of rib fractures occur in the absence of osteopenia [9,10]. Rib fractures and deformity leading to cardiopulmonary insufficiency are the leading cause of death in osteogenesis imperfecta [11]. Site specific osteoclastogenesis predominating in ribs also occurs in metastatic cancer and myeloma [12] but few studies have focused on the biology of site specific skeletal fragility or queried a possible contribution of local bcatenin expression in regulating osteoclastogenesis and fractures in ribs. Here, we present a model of bcatenin regulated skeletal site specific osteoclastogenesis with increased RANKL levels, aggressive rib resorption and fractures, and demonstrate the potential use of corticosteroids in suppressing RANKL to interrupt rapid osteoclastogenesis.

Ethics Statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. This protocol was approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill. Analgesics were appropriately used to minimize suffering.

Generation of Col1a2CreERT/bcatenin fl/fl Mice
Col1a2CreERT transgenic mice carry a tamoxifen inducible Cre recombinase element under the control of a promoter sequence of pro a2(I) collagen gene [13]. Cre transgenic mice (C57Bl/6) were crossed with mice having both bcatenin alleles floxed (bcatenin fl/fl ) (C57Bl/6) to generate mice heterozygous for both alleles. Mice heterozygous for both alleles were backcrossed with bcatenin fl/fl mice to yield Col1a2CreERT/bcatenin fl/fl mice. Tamoxifen dissolved in corn oil (10mg/mL) was administered intra-peritoneally daily for 11 days in 8 week old Col1a2CreERT/bcatenin fl/fl mice (23 male and 22 female) to delete bcatenin in Cre expressing cells. Analysis of survival was performed following completion of 10 days of tamoxifen administration. For lineage reporter analysis, Col1a2CreERT mice or Cola2CreERT/bcatenin fl/fl mice were crossed to Rosa26R lacZ transgenic mice to create Col1a2CreERT/ Rosa26R lacZ or Cola2CreERT/bcatenin fl/fl /Rosa26R lacZ mice. Pamidronate (5mg/kg) was injected intra-peritoneally 5 days prior to starting tamoxifen injections and continued twice/weekly for the duration of analysis. Dexamethasone phosphate (equivalent to 1mg/kg dexamethasone) was administered subcutaneously thrice/ weekly, started concomitantly with tamoxifen, or after stopping tamoxifen and continued for the duration of the analysis. This dose of dexamethasone is consistent with that used in several murine studies investigating the effects of this agent on bone loss and is similar to that used in humans for treating inflammatory states [14,15,16].

Pulmonary Function Testing and Estimation of Oxygen Saturation
Mice were anesthetized with sodium pentobarbital, subjected to tracheostomy and mechanically ventilated with a computer controlled small animal ventilator and pulmonary function testing was done with invasive testing as described [17]. Briefly, mice were paralyzed with pancuronium bromide and airway pressure, volume and flow were measured using a precisely controlled piston during a single inspiration and expiration, with custom designed software (Flexivent, Scireq). Forced oscillation techniques were used to calculate airway impedance. For measurement of peripheral blood oxygen saturation, mice were anesthetized with a 3 minute isoflurane challenge and a mouse-OX (Starr Life Science) oximeter probe was placed on the scruff of the neck to determine the oxygen saturation.

CT Imaging for Lung and Bone Volumes
Tomographic images were acquired with a field emission x-ray cone beam micro CT following image acquisition techniques as described [18]. Image acquisition was physiologically gated to eliminate motion blur, synchronizing x-ray exposures with the inhalation phase of respiration. Respiration monitoring was performed with a novel contactless fiber optic displacement sensor. The sensor was positioned adjacent to the animal's rib cage and was able to detect minute changes in chest position during respiration. Each CT scan required approximately 8-15 minutes for completion, dependent upon the subject's respiration rate. Images were reconstructed into DICOM format for analysis. Bone samples were scanned by SCANCO mCT 40 for image acquisition and assessment of bone density. Region of Interest (ROI) analysis was performed with Image J software. Trabecular bone analysis was performed by drawing an ROI in the center of the vertebral body (no greater than 1/3 of the diameter of the vertebral body) and in the femur (just proximal to the distal femoral metaphysis). This region was found to be most repeated identifiable across the animals. Care was taken to draw the ROIs in consistent anatomic locations with ROI volumes of at least 1000 voxels, across at least 3 slices each as described [18]. Cortical bone volume in the ribs was measured based on an ROI circling the rib on an image in tangential cross section (smallest local diameter). The outside of the ROI encircled the cortex, with the inner medullary portion excluded with a second ROI. Bone volume was then calculated based on percentage of pixels exceeding a consistent threshold defined for each bone type, and all imaging for each bone type was performed under consistent imaging parameters. Image analysis was performed with Image J. In vivo pulmonary images were also acquired with a field emission X-ray cone beam micro-CT. 3 D reconstructions created with appropriate software (Volview, www.kitware.com) [19].

Alkaline Phosphatase and TRAP Staining
Commercially available kits were used for determining alkaline phosphatase and TRAP activity (Sigma). For alkaline phosphatase staining, ribs were collected and frozen in O.C.T. compound immediately followed by staining according to the manufacturer's instructions. For TRAP staining, bone was initially fixed in 4%PFA for 24 hours at 4uC, subjected to decalcification for 36 bcatenin-CKO group and 5-6 animals for control groups, *p,0.001) (D) Tidal volume (mean6S.E.M.; n = 6 animals/group, *p,0.05 versus control groups) and (E) lung histology, 10 days post tamoxifen (F) CT scan and (G) 3-dimensional lung reconstruction at Day 0 and Day 10 post tamoxifen (n = 6 animals/group). (White arrowheads indicate chest wall deformity and black arrowheads loss of lung volume. Oil injection refers to Cre+/ bcatenin fl/fl animals injected with oil and Cre neg control refers to tamoxifen injected bcatenin fl/fl animals). (Scale bar: 100 mm). doi:10.1371/journal.pone.0055757.g001 hours in 14% EDTA at 4uC and then kept in 70% ethanol prior to paraffin embedding. Sections were subsequently deparaffinized and TRAP activity determined by following manufacturer's instructions.

Complete Blood Counts and Serum Chemistry
Blood was collected in micro tubes containing EDTA (BD Biosciences) following submandibular venesection in mice. Peripheral blood counts were analyzed by Heska's animal blood counter. For blood chemistries, plasma was collected using vacuum tubes (BD Biosciences) and samples analyzed with an automated chemical analyzer (Johnson and Johnson VT350).
Statistics p values were computed with student's t test and one way Anova with Bonferroni's post hoc tests as appropriate. All statistical analysis was completed using Graph Pad Prism software.

Deletion of Osteoblast bcatenin Leads to Respiratory Failure and Death from Rib Fractures and a Flail Chest Wall
While studying the role of bcatenin in regulating lineage and function of cells of mesenchymal origin, we observed that deletion of bcatenin in osteoblasts of adult mice led to rapid development of spontaneous rib fractures, flail chest and death from respiratory insufficiency. bcatenin was deleted in osteoblasts by crossing    [20,21]. Tamoxifen was administered daily for 11 days to 8 week old Col1a2CreERT:bcatenin fl/fl mice to excise bcatenin in Col1a2 expressing cells (bcatenin-conditional knock out or bcatenin-CKO) ( Figure 1A). Lineage analysis using the Col1a2CreERT:R26R lacZ reporter mice demonstrated that in addition to osteoblasts in bone, smooth muscle cells and fibroblasts in the heart, lung, liver, spleen, skeletal muscle and kidney expressed Col1a2 ( Figure S1). Within several days of the last tamoxifen dose, bcatenin-CKO animals developed labored breathing with median survival of 11 days and 100% lethality by day 25 (Figure 1B). Not a single death occurred in the vehicle (oil) injected animals or tamoxifen injected bcatenin fl/fl or Col1a2CreERT:R26R lacZ animals ( Figure 1B).
To determine why deletion of bcatenin in Col1a2 expressing cells of adult mice led to labored breathing and rapid demise, we subjected the mice to a pulmonary challenge ten days after the last dose of tamoxifen. Following inhalation of low dose isoflurane for 3 minutes, bcatenin-CKO animals exhibited a profound peripheral oxygen desaturation of 26% (9860.36% vs 7267%, mean6-S.E.M., p,0.001) ( Figure 1C); 20% of the animals died during isoflurane challenge. Mice in the control groups had minimal changes (,5%, p.0.05) in their peripheral oxygen saturation ( Figure 1C). Invasive measurements demonstrated 50% reduction in tidal volume and significantly abnormal ventilatory parameters in the bcatenin-CKO animals ( Figure 1D and Figure S2). Computer assisted tomography (CT) scans and 3D reconstruction of the lungs of bcatenin-CKO animals demonstrated chest wall asymmetry and lung collapse ( Figure 1F,G). However, lung histology did not reveal any abnormalities ( Figure 1E). Cardiac function, blood biochemical tests and peripheral blood counts were also normal except for mild neutropenia (Table S1, S2, S3 and Figure S3). The combination of lung collapse and asymmetry of the chest wall associated with histologically normal lungs suggested defects in the thoracic skeleton as the likely etiology of impaired ventilation and respiratory insufficiency.

Rapid and Extensive Bone Loss Secondary to Osteoclastogenesis Leads to Rib Fractures
To determine potential defects in the chest wall micro CT scans of the entire thoracic cage were performed. Within 10 days of completing tamoxifen dosing, ribs in the bcatenin-CKO animals were overtly osteopenic with deformities due to multiple fractures (Figure 2A,B and Video S1, S2). High resolution CT scan of the ribs confirmed gross bone destruction with a decrease in rib bone volume (bone volume/total volume, BV/TV) by 6663% compared to control groups (p,0.0001) ( Figure 2C,D). To evaluate expression of Col1a2 in the ribs of the affected animals, bcatenin-CKO animals were crossed with the lineage reporter R26R lacZ mice to generate bcatenin-CKO lineage reporter animals (Col1a2CreERT:bcatenin fl/fl :R26R lacZ ). X gal staining of the rib cage in situ demonstrated uniform lacZ expression across the surface of the ribs in the Col1a2 lineage reporter (Col1a2-CreERT:R26R lacZ ) animals ( Figure 2E). In bcatenin-CKO lineage reporter animals, lacZ expression was intensified in areas of callus formation at multiple fracture sites ( Figure 2F). Xgal staining of rib sections of bcatenin-CKO:R26R lacZ animals demonstrated Col1a2 expressing cells scattered throughout the area of bony destruction while in control animals they were prominently present in the endosteal and periosteal surfaces as well as in cortical bone ( Figure 2G,H).
Using alkaline phosphatase to identify osteoblasts, we found that Col1a2 expressing cells in the ribs were cells of osteoblast lineage and that bcatenin protein was not detectable in 80% of these cells ten days post tamoxifen ( Figure S4A,B). Hematoxylin-eosin staining of sections of bcatenin-CKO ribs demonstrated extensive destruction of bone associated with a dense cellular infiltrate surrounding the areas of bone loss compared to control mice ( Figure 2I,J). Consecutive ribs were affected and multiple ribs demonstrated areas of near complete rib resorption ( Figure 2K,L). The cells comprising the infiltrate in the region of rib destruction were largely alkaline phosphatase positive, likely representing coupled bone formation in response to bone loss ( Figure S5A,B). Masson trichrome staining (identifies collagen) demonstrated significant decrease in rib osteoid ( Figure S6A-C) and Von Kossa staining (identifies calcium deposits) showed a similar decrease in calcium content of the affected ribs ( Figure  S6D-F).
We next investigated whether rapid bone loss in the ribs was secondary to osteoclastogenesis. Tartrate resistant acid phosphatase (TRAP) staining to identify osteoclasts showed a significant increase in the number of osteoclasts in bcatenin-CKO ribs ( Figure 2M,N). Blood alkaline phosphatase was significantly elevated by 2 fold in bcatenin-CKO groups, consistent with extensive bone remodeling ( Figure 2O). These data thus suggest that rib destruction was secondary to extensive osteoclast recruitment following bcatenin deletion in rib osteoblasts expressing Col1a2.

bcatenin-CKO Animals Exhibit Less Dramatic Bone Resorption at other Skeletal Sites
We next investigated changes in bone density elsewhere in the skeleton. Micro CT of the vertebrae showed extensive loss of vertebral bone with likely loss of both cortical and trabecular bone ( Figure 3A,B). Affected animals demonstrated a 40% significant decrease in vertebral trabecular bone volume (BV/TV), compared to control animals ( Figure 3C). Consistent with observations made on CT scan, hematoxylin-eosin staining showed loss of bony trabeculae ( Figure 3D,E). Xgal staining in situ of long bones such as the femur of the Col1a2Cre lineage reporter mice showed expression at the trabecular ends of the long bones ( Figure 4A-C) in contrast to the uniform expression across the entire surface of the ribs. Consistent with this expression pattern, there was ,10% decrease (p.0.05) in bony volume (BV/TV) of the femoral metaphysis ( Figure 4D) in the bcatenin CKO mice. Micro CT of the femur did not reveal any fractures ( Figure 4E) and TRAP staining demonstrated no significant differences in the number of osteoclasts in the trabecular metaphysis between the bcatenin-CKO animals and control groups ( Figure 4F-H). Histomorphometry of the femoral diaphysis also did not show any significant change in cortical bone ( Figure 4I-K). Increased expression of Col1a2 in the ribs compared to the femur could potentially explain the relative sparing of bone in the femur. volume) (E) Micro CT of femurs and (F, G) TRAP staining of metaphyseal region of femurs of (F) control and (G) bcatenin-CKO animals (arrowheads show TRAP positive cells) and (H) number of osteoclasts/mm 2 of femoral metaphysis (n = 3 femurs/group, n.s = not significant) (I, J) Hematoxylineosin staining of femoral diaphysis sections of (I) control and (J) bcatenin-CKO animals and (K) histomorphometric assessment of cortical bone area of femoral diaphysis (n = 9 femurs/group, n.s = not significant) (Scale bar: 100 mm). doi:10.1371/journal.pone.0055757.g004

Rib Fractures and Death are Ameliorated by Dexamethasone
We investigated whether inhibition of osteoclast activity would prevent bone resorption and rescue the bcatenin-CKO phenotype. The bisphosphonate, pamidronate (5mg/kg) [22] was administered concurrently with tamoxifen and continued throughout the remaining duration of the experiment. This dose of pamidronate is considered to be a ''high dose'' and can be effectively used in rapid bone loss such as hypercalcemia of malignancy [23]. Pamidronate at this dose delayed but did not prevent death in the large majority of the animals with only 20% of the animals surviving 65 days after the last dose of tamoxifen ( Figure 5A). This suggested that osteoclastogenesis arising from bcatenin deletion in rib osteoblasts was significantly more aggressive than that seen in bisphosphonate amenable osteoporosis. Osteoblasts regulate osteoclast formation and express the osteoclastogenic cytokine RANKL that leads to increased osteoclast recruitment [8]. Bisphosphonates do not work through regulation of RANKL [23,24] but glucocorticoids have been shown to decrease RANKL expressed by inflammatory cells in immune mediated arthritis [25]. Although the affected ribs did not contain activated T, B cells or macrophages as shown by the absence of CD8, B220 and F4/80 antigens ( Figure S5C-G), given the rapidity of the process we elected to treat the bcatenin-CKO mice with this immune modulator. Dexamethasone (dexamethasone phosphate equivalent to 1mg/kg of dexamethasone) was started concurrently with tamoxifen injections and continued thrice weekly throughout the duration of survival analysis (65 days post cessation of tamoxifen). Strikingly, 75% of the dexamethasone treated bcatenin-CKO animals survived at 65 days compared to 0% survival at day 25 in the tamoxifen only group (p,0.0001) ( Figure 5A). No gender specific differences were observed with equal numbers of males and females surviving. Hematoxylin-eosin staining of rib sections harvested 10 days following cessation of tamoxifen demonstrated greater amount of bone in the dexamethasone treated animals ( Figure 5B,C). Consistent with histological assessment, high resolution micro CT scan of the ribs at 10 days post tamoxifen showed 46% increase in rib BV/TV of the ribs of dexamethasone treated animals ( Figure 5D-G) (p,0.05, n = 3 animals/group with 4 ribs examined/animal). At 65 days post tamoxifen, the ribs in the surviving animals showed extensive bony remodeling ( Figure 5H). Hematoxylin-eosin staining of ribs 65 days post tamoxifen in dexamethasone treated animals showed evidence of healing by endochondral ossification with cartilage formation at the surface of the ribs and partial restoration of the bony circumference ( Figure 5I,J). Importantly, CT scan of the chest in dexamethasone treated bcatenin-CKO animals 65 days posttamoxifen demonstrated bilaterally inflated lungs without evidence of lung collapse ( Figure 5K).
TRAP staining performed 10 days post tamoxifen showed 8 fold increase in osteoclasts in the ribs of bcatenin-CKO but 60% reduction in the number of osteoclasts in ribs of animals treated with dexamethasone ( Figure 5L-N), demonstrating that dexa-methasone was significantly attenuating osteoclastogenesis. Measurement of serum RANKL 8 days post tamoxifen dosing demonstrated a significant elevation in bcatenin-CKO animals compared to control mice (Tamoxifen injected Cre negative) (443634 pg/mL vs 246642 pg/mL, mean6S.E.M, p,0.05) ( Figure 5O). In animals treated with dexamethasone, serum RANKL levels (199639 pg/mL) were indistinguishable from control animals ( Figure 5O). Immunohistochemistry for RANKL on frozen sections of ribs showed abundant RANKL expression associated with the rib osteoblast infiltrate in the bcatenin-CKO animals ( Figure S7A,B). In contrast, concurrent dexamethasone decreased RANKL expression mirroring the decrease in serum RANKL ( Figure S7C).
We confirmed that the bone sparing effects of dexamethasone were not a result of dexamethasone induced suppression of Col1a2 driven Cre-recombinase. In bcatenin-CKO:R26R lacZ animals, where Cre expressing cells are identified by lacZ expression, dexamethasone treatment did not decrease the number of lacZ expressing cells at 11 days following initiation of dexamethasone as compared to bcatenin-CKO:R26R lacZ animals which were not treated with dexamethasone ( Figure S8). Furthermore, initiating dexamethasone after completing tamoxifen treatment still led to significantly higher survival rates, albeit less than when dexamethasone was started concurrent with tamoxifen injections ( Figure  S9). Taken together, these data suggest that dexamethasone, particularly if initiated early has remarkable bone preserving effects during a state of rapid and aggressive osteoclastogenesis.

Discussion
Our study demonstrates a remarkable phenotype of aggressive osteoclastogenesis of the thoracic cage within days of osteoblast specific deletion of bcatenin that results in a disappearing thoracic skeleton incapable of supporting ventilation and life. Our findings highlight the role of Col1a2 osteoblast bcatenin in determining rib fragility. Previous studies with embryonic deletion of bcatenin in Col1a1 expressing osteoblasts did not cause such a dramatic phenotype and there was no note of rib lesions leading to increased mortality [8]. These differences may be secondary to different cell specific promoters driving Cre recombinase or other compensatory pathways preventing or rescuing such phenotypes during embryonic development. The rate of cortical bone turnover is highly sitespecific and our data suggests that rib cortical bone has high turnover rates. Site specific bone resorption might depend on pathological RANKL expression by local osteoblasts, rather than by the secretion of RANKL by matrix embedded bone osteocytes thought to be predominant during unloading-associated osteoclastogenesis in trabecular bone [26] or in remodeling associated with normal skeletal development [27]. Indeed, our results show that bcatenin deletion in rib cortical osteoblasts leads to unusually high local expression of RANKL demonstrated histologically. This might suggest that RANKL is expressed by osteoblasts under pathological conditions, and lead to site specific bone resorption, similar to the expression of RANKL by T-lymphocytes after ovariectomy [28] or from multiple sources in inflamed joints [29,30,31].
Our study suggests that corticosteroids might, used judiciously, limit pathological osteoclastogenesis. This is surprising, as glucocorticoids have been shown to enhance the RANKL axis in osteoblasts in vitro and chronic glucocorticoid therapy is associated with bone loss and fractures [16,32]. However, glucocorticoid injection into joints decreases RANKL expression by synovial fluid lymphocytes, and, in TNF primed osteoblasts, downregulates RANKL/osteoprotegerin expression [25]. Moreover, when inflammation is ameliorated in inflammatory arthritis, bone formation is induced, in part through improvement in Wnt antagonism [33]. It is possible that the intense osteoclastogenesis consequent to bcatenin deletion in osteoblasts in adults in our model is accelerated by inflammatory signals. Although we did not find evidence of lymphocytic or macrophage infiltration in resorbing ribs, it is possible that osteoblasts themselves might be sources of corticosteroid-responsive cytokines.
Aggressive osteoclastogenesis in the ribs and at other skeletal sites can be observed in metastatic cancers and multiple myeloma [12,34]. Future studies can capitalize on our findings to uncover interactions that regulate bone destruction, in these conditions, and may be improved by stimulating bcatenin signaling. Further, the model of accelerated osteoclastogenesis and aggressive bone turnover reported here presents a vehicle for rapid pharmacologic testing of anti-resorptive and other bone sparing agents. In summary, we present here a novel model of osteoclastogenesis with a predominant rib phenotype and demonstrate that glucocorticoids, through decreasing RANKL and limiting osteoclastogenesis, may have a role in the treatment of bone loss from rapid osteoclastogenesis. Table S1 Echocardiographic parameters of cardiac function in bcatenin-CKO and control mice before injection and 10 days post cessation of tamoxifen or oil injection. LVEDD, left ventricular end-diastolic dimension; LVESD, left ventricular end-systolic dimension; FS, fractional shortening, calculated as (LVEDD-LVESD)/LVEDD6100; EF%, ejection fraction calculated as (End diastolic volume-End systolic volume)/End diastolic volume6100. Data expressed as mean6S.D. None of the cardiac parameters were statistically significantly different between the bcatenin-CKO mice and other groups either prior to post injection (2 ways Anova). (TIF)