Microfibril-associated Glycoprotein-1, an Extracellular Matrix Regulator of Bone Remodeling*

MAGP1 is an extracellular matrix protein that, in vertebrates, is a ubiquitous component of fibrillin-rich microfibrils. We previously reported that aged MAGP1-deficient mice (MAGP1Δ) develop lesions that are the consequence of spontaneous bone fracture. We now present a more defined bone phenotype found in MAGP1Δ mice. A longitudinal DEXA study demonstrated age-associated osteopenia in MAGP1Δ animals and μCT confirmed reduced bone mineral density in the trabecular and cortical bone. Further, MAGP1Δ mice have significantly less trabecular bone, the trabecular microarchitecture is more fragmented, and the diaphyseal cross-sectional area is significantly reduced. The remodeling defect seen in MAGP1Δ mice is likely not due to an osteoblast defect, because MAGP1Δ bone marrow stromal cells undergo osteoblastogenesis and form mineralized nodules. In vivo, MAGP1Δ mice exhibit normal osteoblast number, mineralized bone surface, and bone formation rate. Instead, our findings suggest increased bone resorption is responsible for the osteopenia. The number of osteoclasts derived from MAGP1Δ bone marrow macrophage cells is increased relative to the wild type, and osteoclast differentiation markers are expressed at earlier time points in MAGP1Δ cells. In vivo, MAGP1Δ mice have more osteoclasts lining the bone surface. RANKL (receptor activator of NF-κB ligand) expression is significantly higher in MAGP1Δ bone, and likely contributes to enhanced osteoclastogenesis. However, bone marrow macrophage cells from MAGP1Δ mice show a higher propensity than do wild-type cells to differentiate to osteoclasts in response to RANKL, suggesting that they are also primed to respond to osteoclast-promoting signals. Together, our findings suggest that MAGP1 is a regulator of bone remodeling, and its absence results in osteopenia associated with an increase in osteoclast number.

Originally thought to serve a purely structural role, the extracellular matrix (ECM) 2 is now known to be an important regu-lator of tissue development and homeostasis. Microfibrils are an abundant component of the ECM and can be found alone as microfibril bundles or associated with elastin in elastic fibers. Three fibrillins (FBN1, -2, and -3) provide the major structural components of these 10 nm diameter fibrils (1)(2)(3), although numerous microfibril-associated proteins interact with fibrillin and contribute to microfibril function. Fibrillin expression is widespread throughout development and is a product of most mesenchymal/interstitial cells (4). In the developing skeletal system, fibrillin expression has been documented in limb bud development, as well as in the adult bone (1,5). Within the bone, fibrillin microfibrils can be found in the periosteal matrix, surrounding osteocytes, chondrocytes, and osteons, on the endochondral surface, and within the trabecular matrix (7).
Microfibrils have clinical significance as mutations in the fibrillin genes give rise to a number of heritable connective tissue disorders. Mutations in the gene for fibrillin-1, for example, are linked to Marfan syndrome, dominant ectopia lentis, and the autosomal-dominant form of Weill-Marchesani syndrome (8 -11). In addition to eye defects, these diseases can include abnormalities of the axial and appendicular skeleton, craniofacial and cardiovascular defects, and altered muscularity and adiposity (reviewed in Ref. 12). Mutations in the gene for fibrillin-2 have been genetically linked to congenital contractural arachnodactyly, a rare disorder that shares some of the skeletal manifestations of Marfan syndrome (9,13).
Fibrillin-1 mutation or deficiency in mice results in a skeletal phenotype similar to that seen in humans with Marfan syndrome (14,15). Disruption of fibrillin-1 expression leads to rib and long bone overgrowth and kyphosis. Significant loss in cortical and trabecular bone volume, reduced femoral bone mineral density, altered trabecular microarchitecture, and a defect in osteoblastogenesis were also reported in fibrillin-1-mutant tight skin (Tsk) mice (15). Fibrillin-2 deficiency in mice mimics the human congenital contractural arachnodactyly phenotype, and this mouse model implicates fibrillin-2 as a regulator of limb patterning (16). Providing a potential mechanism for dis-ease pathogenesis, recent studies with these animal models directly connect fibrillin mutations with dysfunction of TGF␤ and BMP (16 -19). The large latent complex of TGF␤ localizes to microfibrils in the ECM of tissues and cells (20), and latent TGF-␤ binding protein-1 was shown to interact with a sequence near the amino terminus of fibrillin-1 (21,22). Several BMPs through their pro-domain also interact with fibrillin-1 at the same region (23). Antagonism of TGF␤ has been shown to rescue phenotypes associated with fibrillin mutation in vivo (17). Further, the skeletal abnormalities seen in Marfan syndrome are similar to those seen in Loeys-Dietz syndrome-I and -II, Shprintzen-Goldberg syndrome, and other diseases associated with TGF␤ receptor-I and -II mutation (24,25). Consequently, microfibrils are believed to regulate the bioavailability of growth factors through their sequestration in the ECM away from the cell (26).
The microfibril-associated glycoproteins (MAGPs) are a common, if not constituent, component of fibrillin microfibrils in vertebrates (27,28). The two members of the MAGP gene family, MAGP1 and MAGP2, are small, secreted glycoproteins (ϳ20 kDa) that, like the fibrillins, exhibit widespread expression during development (29 -31). MAGP2 exhibits a more restricted tissue and developmental pattern of distribution, suggesting that it has a more specialized role than MAGP1. These proteins have no significant similarity to any other proteins in the databases and hence form a small unique ECM protein family. No known human diseases have been linked directly to either MAGP gene, although it is relevant to the findings in this report that the gene for MAGP1 lies within a quantitative trait locus for bone mineral density (32,33).
Like other ECM proteins, MAGP1 is organized into multiple functional domains. The N-terminal half of the protein contains a cluster of acidic and sulfated tyrosine residues that define a binding domain for cationic proteins, such as growth factors like TGF␤ and BMPs. In contrast to fibrillin-1, which binds latent forms of the TGF␤ growth factor family, MAGP1 binds active TGF␤ and BMPs (34). The back half of MAGP1 is rich in cysteines and contains the protein's matrix binding domain, which facilitates interactions with fibrillin, notch, and other EGF-containing molecules (35).
Protein binding studies demonstrate that MAGP1 interacts with fibrillin at a sequence near fibrillin's N terminus and at a second site in the middle of the molecule (36 -40). The location of these binding sites provides insight into MAGP1 function. The N-terminal site of fibrillin is the interaction domain that binds latent TGF␤ superfamily members. Consequently, MAGP1's presence or absence may impact the ability of fibrillin to bind these growth factors. The second MAGP1 binding site is the region where fibrillin mutations result in the most severe phenotypes seen in Marfan syndrome and where almost all congenital contractural arachnodactyly mutations cluster. It is possible that these mutations alter the ability of fibrillin to bind MAGP1 and that this contributes to disease severity. Thus, when considering the role of microfibrils in tissue development and homeostasis, it is necessary to evaluate how the MAGPs and fibrillins together impart microfibril function.
MAGP1 was first isolated from tissues rich in elastic fibers and was proposed to be important for elastic fiber assembly.
However, targeted inactivation of the MAGP1 gene (Mfap2) failed to reveal a defect in elastin-rich tissues such as blood vessels and the lung. Instead, MAGP1-deficient mice showed increased fat deposition, a wound-healing defect, bleeding diathesis, and a skeletal phenotype characterized by lesions on their lower limbs suggestive of spontaneous bone fracture (34,41). The goal of this current study was to better characterize the bone phenotype in MAGP1-null animals and to determine the cellular basis of the suspected bone fragility. MAGP1-null mice are osteopenic relative to WT controls, have significantly less trabecular and cortical bone, have altered trabecular microarchitecture, and have an increased number of osteoclasts. Together, these findings suggest that MAGP1 plays a role in bone remodeling and is a negative regulator of osteoclast number.

EXPERIMENTAL PROCEDURES
Nomenclature-The nomenclature for the MAGPs and their genes is somewhat confusing. The gene name for MAGP1 is Mfap2, whereas the gene name for MAGP2 is Mfap5. To avoid confusing MAGP1 and MAGP2 when referring to knock-out mice, we will refer to the MAGP1 knock-out genotype (Mfap2 Ϫ/Ϫ ) as MAGP1⌬.
Statistical Analysis-A paired t test was used to determine the statistical significance between genotypes. Values were considered significantly different when p values were Ͻ0.05.
Animals-Generation and genotyping of the MAGP1⌬ colony has been described (34). Male mice were utilized for all studies and ranged in age from 4 weeks to 6 months as indicated. All mice were in the Black Swiss background (BkSw, Taconic, Hudson, NY) and were housed in a pathogen-free animal facility, fed standard chow ad libitum, and treated following animal protocols approved by the Washington University Animal Studies Committee.
BMSC Isolation and Culture-Femurs and tibias were removed from 3-month-old WT and MAGP1⌬ mice and cleaned of soft tissue, and the distal end of the bone was removed. Marrow cells were collected by brief centrifugation of the bone followed by incubation in red blood cell lysis buffer for 10 min at room temperature. The lysis buffer was removed, and the remaining cells were passed through a 70-m cell filter, then cultured in ␣-MEM containing antibiotics and 10% fetal bovine serum (OB growth media). For osteogenic differentiation, cells were cultured in osteogenic media (OB growth media supplemented with 50 g/ml ascorbic acid and 10 mM ␤-glycerolphosphate). The medium was replaced biweekly. Cells were lysed for RNA analysis or fixed for alkaline phosphatase activity at day 7. Cells for mineralization assays were fixed and stained using Alizarin Red on day 28. Alkaline phosphatase staining was performed using a Sigma-Aldrich kit (85L1) according to the manufacturer's instructions.
BMM Isolation and Culture-Femurs and tibias were removed from 4-to 6-week-old WT and MAGP1⌬ mice, cleaned of soft tissue, and flushed with ␣-MEM to recover the bone marrow. Marrow cells were cultured on plastic dishes in ␣-MEM containing penicillin/streptomycin, 10% inactivated fetal bovine serum, and 1:10 CMG (conditioned media containing macrophage-colony stimulating factor (43)). After 4 days in culture, the adherent bone marrow macrophage cells (BMMs) were trypsinized and re-plated into tissue culture-treated plastic dishes containing OC differentiation media (␣-MEM, penicillin/streptomycin, 10% inactivated fetal bovine serum, 1:50 CMG, and 50 ng/ml recombinant glutathione S-transferase-RANKL). The cells were grown for an additional 4 -5 days under these conditions then stained for TRAP (Sigma-Aldrich) or lysed for protein analysis. TRAP staining was performed as previously described (44).
Calvaria Osteoblast Isolation and Culture-Whole calvaria were extracted from 4-day-old pups and cleaned of soft tissue. Calvaria from 4 pups per genotype were pooled. Osteoblasts were liberated by serial collagenase treatment. BMMs from 4-week-old mice were harvested on the same day (described above). Osteoblasts and BMMs were cultured for 4 days prior to being lifted and cultured together for 5-6 days in ␣-MEM media containing penicillin/streptomycin, 10% fetal bovine serum, 10 nM VitD3.
RNA Extraction and Quantitative RT-PCR-RNA from cultured cells was isolated and purified using a Qiagen (Valencia, CA) RNeasy RNA purification kit; RLT lysis buffer was supplemented with ␤-mercaptoethanol (1%). Purified RNA was treated with DNase I (Invitrogen) prior to reverse transcription (RT). RT was performed using SuperScript III (Invitrogen) and quantitative PCR (qPCR) was performed using Power SYBR green master mix (Applied Biosystems, Foster City, CA) and gene-specific primers. The qPCR reaction was run on ABI Prism 7000 (Applied Biosystem). Transcript levels (relative units, RU), normalized to cyclophilin, were determined by the equation, 1/(2∧⌬CT)*10,000. qPCR was performed on whole bone tissue as described above, but with the following modifications. Marrow-flushed tibias from 6-month-old mice, or calvaria from 6-week-old mice, were cleaned of soft tissue, and snap frozen. Bones were pulverized using a Braun Mikrodismembrator. Pulverized tissue was collected in TRIzol, and RNA was extracted with chloroform and further purified using the Qiagen RNeasy kit described above.
Whole Body Bone Mass (DEXA)-WT and MAGP1⌬ mice were anesthetized with ketamine/xylazine (86.98 mg/kg and 13.4 mg/kg, respectively; delivered intraperitoneally), and whole body measurements, excluding the head, were made by dual energy x-ray absorptiometry (DEXA, PIXImus Lunar-GE, Madison, WI) at 4, 8, 12, 16, 20, and 24 weeks. Body measurements included bone mineral density (BMD), bone mineral content, and bone area. The DEXA machine was calibrated daily before use, one person performed all scans, and mice were always in the prostrate position on the imaging-positioning tray.
Microcomputed Tomography-Microcomputed tomography (CT) was completed on tibias of 24-week-old WT and MAGP1⌬ mice. Frozen tibias were thawed in phosphate-buffered saline for 30 min prior to embedding in 2% agarose gel. The tibia, from the proximal epiphysis through the tibia-fibula junction, was scanned at 16-m voxel resolution using a Scanco CT 40 (Scanco Medical AG, Zurich, Switzerland). Measurements of both trabecular and cortical bone were made. For trabecular bone, 30 sections (16 m/section) below the growth plate were contoured to exclude the cortical bone. For these trabecular regions the following three-dimensional measurements were obtained: bone volume/tissue volume, connectivity density, structure model index, trabecular separation, trabecular thickness, trabecular number, and BMD. For cortical bone, the sections analyzed were at the mid-diaphysis, 5 mm proximal to the tibio-fibular junction. Tissue area, marrow area, cortical bone area, and cortical bone width were calculated from the contours of three sections per mouse. Cortical bone width was determined from the average of four positions per slice and three slices per bone. For cortical mineral density (BMD), the contours of 10 sections per mouse were averaged. For all measurements a threshold of 220 (on a 0 -1000 scale) was maintained.
Tissue Preparation-Tibias were harvested, cleaned of all soft tissue, and fixed in 10% buffered formalin. For non-decalcified sections, tibias were rinsed in water then dehydrated through incubation in an ethanol gradient (20%, 30%, 50%, and 70% ETOH, for 0.5 h each). Bones stored in 70% ETOH were then embedded in methylmethacrylate and cut into 5-m longitudinal sections. Non-decalcified sections were left unstained. For decalcified sections, following fixation, tibias were rinsed in water then incubated in 14% EDTA (pH 7.2) for 14 days with solution changes every 3 days. Following decalcification, tibias were rinsed in water then incubated in the ETOH gradient described above. Bones stored in 70% ETOH were then embedded in paraffin, cut into 4-m longitudinal sections, and stained with hematoxylin and eosin or for tartrate-resistant acid phosphatase (TRAP).
Dynamic Bone Formation Analysis (Calcein-Alizarin Labeling)-17-week-old WT and MAGP1⌬ mice were injected with 7.5 mg/kg calcein (Sigma-Aldrich). Seven days later mice were injected with 30 mg/kg alizarin 3-methyl iminodiacetic acid, and tibias were harvested after an additional 2 days for histologic analysis. Dynamic assessment of trabecular bone formation was determined on non-decalcified MMA sections based on single and dual calcein-alizarin labeling using BioQuant software (BioQuant, Nashville, TN). This software provided measures of bone surface, percent single-and double-labeled bone surface (BS), mineralizing surface, mineral apposition rate, and bone formation rate. Trabecular bone measurements were taken from a region encompassing a 500-m-long field, across the width of the bone (located 50 -550 m below the growth plate). For each animal, two serial sections were analyzed, and the measurements were averaged (n ϭ 12 for WTs and 8 for MAGP1⌬). Pictures were obtained using an Olympus 1X51 fluorescent microscope fitted with a DP70 camera (Olympus, Center Valley, PA).
Quantitative Histomorphometry-TRAP and hematoxylin and eosin-stained tibial sections were used to visualize osteoclasts and osteoblasts, respectively, using the Olympus BX51 light microscope and camera described above (Olympus). Trabecular osteoblasts and osteoclasts were measured in a 500-m field across the bone in an area located 50 m distal to the growth plate using BioQuant software. Measurements included osteoclast number/BS and osteoblast number/BS.
Western Blotting-Cultured BMM cells were lysed in buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 1 mM NaF 1ϫ protease inhibitor (Roche Applied Science). Lysates were cleared of cellular debris by centrifugation, and protein concentration was determined. Lysate was run on 10% SDS-PAGE gels then transferred to nitrocellulose membranes. Membranes were blocked in 1% casein then incubated overnight with the indicated primary antibodies. Proteins were visualized through use of fluorescence-labeled secondary antibodies and the Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE).
Bone Mechanical Testing (3-Point Bend)-Femurs were thawed in phosphate-buffered saline for 1 h prior to use, and testing was carried out at room temperature using a servohydraulic testing machine (8841 Dynamite, Instron, Norwood, MA). Femurs were positioned on two supports 7 mm apart, and the central loading point was mid-diaphysis. Displacement was applied transverse to the long axis of the bone at a rate of 0.03 mm/s until failure. Force-displacement data were recorded at 60 Hz and analyzed to determine measures of stiffness (rigidity) and whole bone strength (ultimate moment) (45).
Bone-resorptive Pit Staining-BMMs were cultured on bovine bone slices in the presence of macrophage-colony stimulating factor and RANKL for 6 days. Mechanical agitation was used to remove the cells from the bone slices. The bone slices were then incubated with peroxidase-conjugated wheat germ agglutinin (Sigma) then 3,3-diaminobenzidine (Sigma). Resorption lacunae size was determined from the average of five fields from four chips per mouse, and three mice per genotype.

MAGP1-deficient Animals Have Reduced Bone Strength and
Become Osteopenic with Age-We (Fig. 1A) and others (46) have demonstrated MAGP1 to be abundantly expressed in bone. MAGP1-deficient mice, however, appear to have normal skeletal development, and tibia length measurements demonstrate only a slight overgrowth in the MAGP1⌬ animals (Fig.  1B). We previously reported that a percentage of aged MAGP1deficient mice develop lesions on their lower limbs consistent with bone fracture (34). Using the 3-point bend test, we were able to confirm that femurs from MAGP1⌬ mice have reduced whole bone strength and rigidity (Fig. 1C).
To better understand why MAGP1⌬ bones are more susceptible to failure, we performed a longitudinal DEXA scan study following mice from 4 to 24 weeks old. The findings show that MAGP1⌬ mice were unable to achieve the BMD or bone mineral content levels seen in WTs (Fig. 2, A-C), thus becoming relatively osteopenic with age. Osteopenia in the MAGP1⌬ mice was easily visualized by x-ray of the tibias (Fig. 2D) and appeared most prominent in the trabecular bone (see arrows). CT analysis confirmed that the trabecular bone was more affected by MAGP1 deficiency than the cortical bone. Comparison of WT and MAGP1⌬ mice showed that trabecular BMD  JULY 30, 2010 • VOLUME 285 • NUMBER 31

MAGP1 Regulates Bone Resorption
was reduced 40% in MAGP1⌬ animals compared with only a 5% reduction in cortical BMD (Fig. 2, E and F). MAGP1⌬ Mice Have Significantly Less Trabecular Bone and Disrupted Microarchitecture-Histological and CT analyses confirmed that the trabecular bone was significantly affected by MAGP1 deficiency. Histology on longitudinal tibia sections revealed a significant reduction in trabecular bone (Fig. 3A). Quantification showed that MAGP1⌬ mice have 56% less trabecular bone surface compared with WT (Fig. 3B). Based on CT analysis of the tibial epiphyses, trabecular BMD was reduced 40% in MAGP1⌬ animals (Fig. 2E). Consistent with the reduction in volumetric BMD, three-dimensional images clearly demonstrate significantly less trabecular bone mass in the MAGP1⌬ mice (Fig. 3C). Quantitatively, there is a 58% reduction in the ratio of bone volume to tissue volume (BV/TV, Fig. 3D). Furthermore, measures of trabecular microarchitecture were altered in the MAGP1⌬ mice. The trabecular connectivity density (Conn.Dens.) was reduced 76% compared with WT (Fig. 4A). The structure model index was increased from 1.6 to 2.6 in the bones of MAGP1⌬ animals, indicating an increase in the ratio of rod-like to plate-like trabeculae (Fig. 4B). Trabecular separation (Tb.Sp., Fig. 4C) was significantly increased in the knockouts. There was also a slight reduction in trabecular thickness (Tb.Th., Fig. 4D), but no difference in the trabecular number (Tb.N., Fig. 4E). These data indicate that MAGP1-deficient mice have diminished trabecular bone mass and that the trabecular network is fragmented and with thinner trabeculae relative to WT. These findings demonstrate that MAGP1 is a regulator of bone remodeling.

MAGP1⌬ Mice Have Altered Cortical Bone Morphology and
Density-CT of the tibial diaphysis revealed altered cortical morphology (Fig. 5A). Cortical mineral density (BMD) was reduced slightly (Ϫ5%) but significantly (Fig. 4F). Analysis of diaphyseal cross-sections revealed that cortical bone area was reduced ϳ9% (Ct.Ar., Fig.  5B) and cortical bone width was reduced ϳ2% (Ct. Wi., Fig. 5C). These modest changes in the amount of cortical bone were accompanied by a significant reduction of 33% in the medullary area of MAGP1⌬ bones (Ma.Ar., Fig. 5D). Accordingly, tissue area was reduced 16% (T.Ar., Fig. 5E). The large reduction in cross-sectional bone size is consistent with reduced whole bone strength reported in the bending studies (Fig. 1C). Together, the morphologic analyses demonstrate that MAGP1 is involved in skeletal remodeling and maintenance.
Osteoblasts Function Normally in the Absence of MAGP1-To delineate the mechanism underlying reduced bone volume and BMD, bone marrow stromal cells (BMSCs) were isolated from WT and MAGP1-deficient mice to determine their propensity to differentiate into osteoblasts. Following 7 days of incubation in osteogenic medium, differentiated MAGP1⌬ cells expressed osteoblast gene markers at levels similar to (RUNX2, OC, and OSX), or sometimes greater (bone sialoprotein and Col1a1) than, that of the WT cells (Fig. 6A). MAGP1⌬ BMSCs also stained positive for alkaline phosphatase activity at levels comparable to WT BMSCs (Fig. 6B). Further, following 28 days in osteogenic media there was no difference in the mineralization potential of MAGP1⌬ cells (Fig. 6C).
Analysis of histologic sections found no significant difference in the number of osteoblasts lining the trabecular bone surface (Figs. 6D). Dynamic indices of bone formation were measured in WT and MAGP1⌬ mice using dual calcein-alizarin labeling. There was no difference in the ratio of mineralizing surface to bone surface (MS/BS, Fig. 6E). There was only a slight, but not significant, reduction in both mineral apposition rate and the bone formation rate to bone surface ratio (data not shown). These findings imply that the osteopenia of MAGP1-deficient mice does not result from a defect in either osteoblastogenesis or osteoblast function.  that increased bone resorption might be the cause of the osteopenic phenotype. Indeed, BMMs isolated from MAGP1⌬ mice produced a significantly higher number of osteoclasts after 4-day exposure to RANKL (50 ng/ml) relative to WT BMMs (Fig. 7A). Enhanced osteoclastogenesis was also evaluated by monitoring osteoclast differentiation markers over time. ␤3-integrin and c-Src, markers of commitment to the osteoclast lineage, were visualized via Western blot, and both proteins were found to be expressed at higher levels and at earlier time points in the MAGP1-deficient cells (Fig. 7B).
To determine if the resorptive capacity of individual mutant polykaryons is enhanced, we generated WT and MAGP1⌬ osteoclasts on bone and measured the area of pits they excavated. In contrast to osteoclast number, the average pit area per cell is indistinguishable between genotypes (Fig. 7C).
To determine whether an increase in osteoclast number is also present in MAGP1-deficient mice in vivo, quantitative histomorphometry was performed on TRAP-stained tibia sections. The results confirmed that MAGP1⌬ mice have 57% more osteoclasts lining the trabecular surface when compared with WT mice (Fig. 7D). These findings suggest that MAGP1 functions to inhibit osteoclast differentiation such that, when absent, there is an increase in the ratio of osteoclasts to osteoblasts.
MAGP1⌬ BMMs Are Primed for Osteoclastogenesis-MAGP1 is abundantly expressed in the bone. Utilizing RT-qPCR it was determined that committed osteoblasts, not cells of the osteoclast lineage, are the major producers of MAGP1 (Figs. 8, A and B). Thus, it was important to determine whether culturing WT osteoblasts with MAGP1⌬ BMM could reverse the enhanced osteoclastogenesis associated with MAGP1⌬ BMM (Fig. 7A). Calvaria osteoblasts from 4-day-old WT and MAGP1⌬ pups were cultured with either WT or MAGP1⌬ BMMs. After 5-6 days of co-culture, osteoblasts were removed, and the remaining osteoclasts were visualized by TRAP stain. As seen in Fig. 9A    a RANKL decoy receptor, blocks RANKL-RANK binding and thus bone turnover (47). The RANKL/OPG ratio is considered an important predictor of bone mass (48). We therefore evaluated whether MAGP1 deficiency augmented the RANKL/OPG system. RANKL and OPG transcript levels were determined from cultured calvaria osteoblasts (Fig. 9C) and whole calvaria tissue (Fig. 9D). We found cultured calvaria osteoblasts from MAGP1⌬ mice expressed 14-fold more RANKL relative to WT cells. We were also able to detect a 3-fold increase in RANKL in MAGP1⌬ whole calvaria. In both experiments, there was no difference in OPG expression between genotypes.

DISCUSSION
One of the most intriguing phenotypes in aged MAGP1⌬ mice is the appearance of lesions on their lower limbs that are indicative of spontaneous bone fracture (34). Femoral mechanical testing reported in this current study confirmed that bones from MAGP1⌬ mice are more susceptible to failure relative to WT. CT analysis revealed MAGP1⌬ cortical bone to have reduced mineralization and cross-sectional size, which is consistent with their reduced whole bone stiffness and strength determined by bending test.
Trabecular bone remodels at a faster rate than cortical bone. Consequently, if MAGP1 deficiency causes dysregulation of bone turnover, it will be most apparent in the trabeculae. Interruptions in the trabecular network result in a reduced surface for osteoblasts to replace resorbed bone, and disturbing the trabecular network can result in reduced load bearing capacity thereby increasing susceptibility to fracture. Individuals with osteoporosis, for example, are most likely to fracture where  Osteoclastogenesis ex vivo and in vivo is altered in MAGP1⌬ mice. BMMs were isolated from 5-week-old WT (ϩ/ϩ) and MAGP1⌬ (Ϫ/Ϫ) mice. Osteoclast differentiation was first assessed by trap staining of BMMs exposed to RANKL (50 ng/ml) for 4 days. Three WT and MAGP1⌬ animal pairs were evaluated, and cells were counted as OC if they stained positive for TRAP and possessed more than three nuclei. Data are presented in A as the mean of four wells (ϮS.D.) per animal (W ϭ ϩ/ϩ, K ϭ Ϫ/Ϫ); *, p Ͻ 0.05. Osteoclastogenesis was then evaluated by visualizing the c-Src and ␤3-integrin differentiation markers via Western blot (B). Osteoclast-resorptive capacity was normal in MAGP1⌬ cells (C). BMM were isolated and cultured as above, except they were grown on bovine bone slices. After 6 days, cells were lifted, resultant pits were stained, and pit area per cell was determined (n ϭ 3 mice per genotype). D, quantitative histomorphometry of osteoclast numbers from tibia trabeculae. Decalcified tibia sections from 18-week-old mice were stained with TRAP. Osteoclast number per bone surface (N.Oc/BS) was determined from a 500-m field spanning the whole bone width, 50 m distal to the growth plate (mean Ϯ S.D.; *, p Ͻ 0.05; n ϭ 6 -11 mice per group). trabecular bone dominates (49). We found MAGP1⌬ mice to have abnormal trabecular microarchitecture consistent with osteopenia. Trabecular BMD and bone volume in these animals are significantly reduced. Further, the trabeculae have fewer interconnections and have a more rod-like morphology as opposed to a plate-like morphology. These changes in both the trabecular and cortical bone are consistent with the bone fragility seen previously in MAGP1⌬ mice (34).
Osteoblasts in MAGP1⌬ Mice Are Normal in Number and Function-The balance of bone formation by osteoblasts and bone resorption by osteoclasts is key to maintaining bone integrity. When BMSC were induced to differentiate in vivo, alkaline phosphatase activity, mineralized nodule formation, and expression of osteoblast differentiation markers were essentially equivalent in cells from MAGP1⌬ and WT BMSCs animals. When the number of osteoblasts was normalized to bone surface size there was no significant difference between the two genotypes. Similarly, mineralized surface, mineral apposition rate, and bone formation rates were unaffected by MAGP1 deficiency. Together, these data support normal osteoblast differentiation, number, and bone forming ability in the absence of MAGP1. MAGP1 not playing a role in osteoblast differentiation is consistent with it being produced by mature osteoblasts after differentiation has occurred (Fig. 8A).
Elevated Number of Normal Osteoclasts in MAGP1⌬ Bone-The presence of normal osteoblasts in MAGP1⌬-deficient bone suggests that enhanced bone loss is due to a shift in the balance of bone remodeling toward osteoclasts and bone resorption. Indeed, our studies documented significantly more osteoclasts lining the trabecular bone surface in MAGP1-deficient animals, which correlated with a significant loss of trabecular bone volume and thickness, and greater trabecular bone separation. An explanation for the increased osteoclast number was suggested by in vitro experiments wherein more osteoclasts were derived from MAGP1⌬ BMMs than WT BMMs in response to stimulation with RANKL. Further, MAGP1⌬ BMMs expressed osteoclast differentiation markers earlier in response to RANKL stimulation. The response seen in these BMM cultures suggests that MAGP1⌬ BMMs are sensitized to RANKL and, hence, are primed for osteoclastogenesis. Pit forming assays on whole bone showed that the osteoclasts from MAGP1-null animals were functionally normal. Because BMMs express insignificant amounts of MAGP1 relative to osteoblasts, the MAGP1 effects on osteoclastogenesis are likely derived from microfibrils produced by mature osteoblasts (Fig.  8, A and B). This hypothesis was confirmed by demonstrating that cocultures of WT OB and MAGP1⌬ BMM could partially reverse the increase in osteoclast cell number observed in FIGURE 8. MAGP1 expression in BMSC and BMM. A, BMSCs isolated from WT (ϩ/ϩ) and MAGP1⌬ (Ϫ/Ϫ) animals were lysed for RNA following 0 or 7 days (d0 and d7) in osteogenic medium. RT-qPCR was performed using primers specific to MAGP1 or alkaline phosphatase (ALKPhos). B, BMMs isolated from WT and MAGP1⌬ animals were lysed for RNA following 0-or 3-day (d0 and d3) exposure to 50 ng/ml RANKL. RT-qPCR was performed using primers specific to MAGP1 or cathepsin K (CathK). Replicate experiments were performed with consistent results (mean Ϯ S.D.; *, p Ͻ 0.05 relative to d0; † , p Ͻ 0.05 relative to ϩ/ϩ). Transcript levels (relative units (RU)) were determined by RT-qPCR utilizing MAGP1-, RANKL-, or OPG-specific primers (top panels) normalized to cyclophilin. The ratio of RANKL to OPG transcript is given in the bottom panels (mean Ϯ S.D.; *, p Ͻ 0.05 relative to WT). Microfibrils and Growth Factor Binding-The evolving model of microfibril function is that they contribute to tissue development and homeostasis by interacting directly with integrins and other cell surface receptors and through their ability to concentrate and sequester growth factors in the ECM (26,50). The fibrillins have been shown to bind latent TGF␤ and pro-BMP. MAGP1, in contrast, binds active forms of these growth factors (34) such that the presence or absence of MAGP1 could determine which growth factors bind microfibrils. For instance, in the absence of MAGP1 there could be a preference for the latent growth factors binding to microfibrils as opposed to the active forms when MAGP1 is present. This raises the interesting possibility that MAGP1 and fibrillin, as two components of the intact microfibril, have distinct but complementary functions.

MAGP1 Regulates Bone Resorption
Clinically, mutation of both the type-I and -II TGF␤ receptors results in skeletal abnormalities similar to Marfan syndrome (i.e. fibrillin mutation) (51). Members of the TGF␤ superfamily have a well documented role in skeletogenesis and bone remodeling. Osteoblasts produce and deposit these growth factors into the bone matrix, which can be liberated and activated during osteoclast-mediated resorption. BMPs are responsible for coordinating cell condensation, chondrogenesis, and osteogenesis. Consequently, inhibition of this pathway results in severe defects in bone formation (52). Despite being one of the most abundant cytokines in the bone matrix, the exact function of TGF␤ is still unclear. TGF␤1 is likely dispensable for skeletogenesis, having a larger role in postnatal bone remodeling. Osteoporosis is observed in mice with increased TGF␤ signaling as well as in smad3-deficient animals (53,54), whereas reduced TGF␤ signaling results in increased bone mass in other models (55)(56)(57). Recent studies by Tang et al. demonstrate TGF␤ acts as a coupling molecule for osteoblastmediated bone formation and osteoclast-mediated resorption (58).
TGF␤ induces RANKL expression in numerous cell types (6, 59 -61). It was therefore interesting that we found enhanced RANKL expression in MAGP1⌬ osteoblasts and bone tissue. Due to the known role of microfibrils in sequestering TGF␤, and the ability of MAGP1 to selectively bind TGF␤, we hypothesize that MAGP1 deficiency results in increased free TGF␤. An increase in free TGF␤ could then act directly on the macrophage driving osteoclastogenesis or indirectly by enhancing RANKL expression. Studies evaluating TGF␤ activity in MAGP1⌬ tissue are currently underway.
Phenotypic Overlap in MAGP1 and Fibrillin Mutations-Given that MAGP1 is a binding partner of fibrillin in the microfibril, it is reasonable to speculate that mutations in fibrillin might have negative effects on MAGP1 binding and function. The consequences would be phenotypes that are similar to the MAGP1 loss of function phenotypes observed in the MAGP-null mouse. Indeed, mice carrying the autosomal dominant tight skin mutation (Tsk/ϩ) have a genomic duplication within the fibrillin-1 gene that results in an in-frame duplication of exons 17-40, which includes a region encompassing one of fibrillin's MAGP1 binding sites (42). The mutant protein is produced in normal amounts and when combined with fibrillin from the WT allele forms a microfibril of altered structure and function. Although there are phenotypic differences between Tsk/ϩ and MAGP1⌬ mice, the skeletal phenotype is strikingly similar in the two animals (15), suggesting that alterations in MAGP1 binding might contribute to the phenotype.
Summary-We have identified MAGP1 to be an important regulator of bone remodeling. In the absence of MAGP1, mice develop osteopenia, have significantly less trabecular bone and altered cortical bone modeling relative to age-matched controls. Our data indicate that the reduced mineralization and bone volume associated with MAGP1 deficiency is the consequence of an uncoupling in bone formation and resorption. More specifically, osteoblast-mediated bone formation remains normal, whereas an increased osteoclast number can lead to enhanced bone resorption. These findings support a role for microfibrils (fibrillin plus MAGP1) in bone remodeling and homeostasis.