Extracellular Matrix Proteoglycans Control the Fate of Bone Marrow Stromal Cells*

Extracellular matrix glycoproteins and proteoglycans bind a variety of growth factors and cytokines thereby regulating matrix assembly as well as bone formation. However, little is known about the mechanisms by which extracellular matrix molecules modulate osteogenic stem cells and bone formation. Using mice deficient in two members of the small leucine-rich proteoglycans, biglycan and decorin, we uncovered a role for these two extracellular matrix proteoglycans in modulating bone formation from bone marrow stromal cells. Our studies showed that the absence of the critical transforming growth factor-β (TGF-β)-binding proteoglycans, biglycan and decorin, prevents TGF-β from proper sequestration within the extracellular matrix. The excess TGF-β directly binds to its receptors on bone marrow stromal cells and overactivates its signaling transduction pathway. Overall, the predominant effect of the increased TGF-β signaling in bgn/dcn-deficient bone marrow stromal cells is a “switch in fate” from growth to apoptosis, leading to decreased numbers of osteoprogenitor cells and subsequently reduced bone formation. Thus, biglycan and decorin appear to be essential for maintaining an appropriate number of mature osteoblasts by modulating the proliferation and survival of bone marrow stromal cells. These findings underscore the importance of the micro-environment in controlling the fate of adult stem cells and reveal a novel cellular and molecular basis for the physiological and pathological control of bone mass.

The extracellular matrix (ECM) 1 provides structural strength to tissues, maintains the shape of organs, and is often involved directly or indirectly in regulating cell proliferation and differentiation (1)(2)(3). ECM components modulate the bioactivities of growth factors and cytokines, such as TGF-␤, tumor necrosis factor-␣, and platelet-derived growth factor, by 1) activating them by proteolytic processing (4,5), 2) inactivating them by sequestering and preventing binding to their respective receptors (6 -9), or 3) directly binding to cytokine receptors, such as the epidermal growth factor receptor (10,11).
Proteoglycans, which are characterized by a core protein with at least one glycosaminoglycan chain attached, commonly mediate the interactions of ECM components with growth factors and cytokines (12). Small leucine-rich proteoglycans (SLRPs) are some of the major non-collagen components of the ECM (13). The core proteins of the SLRPs consist of leucinerich repeats flanked by two cysteine-rich clusters. The size of the core proteins (ϳ40 kDa) is relatively small compared with aggrecan and versican (Ͼ200 kDa) (1,10,14). The SLRP superfamily currently consists of 13 known members that can be divided into 3 distinct subfamilies based on the genomic organization, structure, and similarity of their amino acid sequences (13). SLRPs are involved in skeletal growth (15)(16)(17), craniofacial structure (15), dentin formation (18), and collagen fibrillogenesis (17,19,20). However, to date, little is known about the precise mechanism through which SLRPs regulate the formation and maintenance of the skeletal system. Biglycan (Bgn) and decorin (Dcn) are class I type SLRPs that are expressed in various tissues (7,15,21), including bones and teeth. Mice with targeted disruption of the bgn gene have impaired postnatal bone formation, which leads to the early onset of osteoporosis (16). Although Dcn-deficient mice do not show histological or macroscopical changes in bone (17,20), they do contain collagen fibrils of abnormal size distribution, density, and shape in mineralized tissue (15,17). Because of the similarity of their primary sequences as well as their overlapping expression patterns in skeletal tissues, it is not surprising that Bgn and Dcn function redundantly and synergistically. Mice deficient in both bgn and dcn display more severe defects than mice deficient in either bgn or dcn, in tissues, such as bone and skin, where Bgn and Dcn are coexpressed (17). These results indicate that Bgn and Dcn may share some common functions in bone and skin and compensate, at least in part, for each other when one is deleted. To eliminate the potential compensation, it is, therefore, necessary to use mice deficient in multiple proteoglycans. Our specific approach was to use bgn/dcn-deficient mice to study the cellular and molecular mechanisms of ECM proteoglycans in regulating osteogenesis.
Bone marrow stromal cells (BMSCs), which contain skeletal stem cells (also known as mesenchymal stem cells), are derived from clonogenic adherent bone marrow cells, which are colonyforming units-fibroblastic. Under certain conditions, BMSCs can give rise to osteoblasts, chondrocytes, adipocytes, myelosupportive stroma, and even muscle and nerve tissue (22)(23)(24). These cells have been used for both in vivo and in vitro studies to understand the mechanisms that modulate skeletal formation and pathological changes in bone disease. In vivo xenogenic transplantation of BMSCs mimics bone formation where transplanted BMSCs can form a miniature bone organ that contains bone, adipocytes, and hematopoiesis-supporting stroma. They can form cartilage in micromass cultures in vitro (23,25). BMSCs have also been used to examine the ability of osteogenic stem cells to proliferate, differentiate, and form mineral nodules in vitro (25,26). BMSCs, therefore, permit in-depth analysis of the signal transduction pathways that regulate proliferation, survival, and differentiation of osteogenic stem cells.
In the current studies, we utilized BMSCs from bgn/dcndeficient mice to examine the combined effects of Bgn and Dcn on bone formation. We found that BMSCs deficient in both SLRPs have increased TGF-␤ signaling due to the inability to sequester TGF-␤ in the ECM by Bgn and Dcn, resulting in switched fate from growth to apoptosis. The premature death of osteogenic stem cells and osteoblast precursors led to a decreased number of mature osteoblasts thereby contributing to decreased osteogenesis and an osteoporosis-like phenotype in bgn/dcn-deficient mice. Our studies demonstrate, for the first time, an unsuspected role for these two SLRPs in controlling bone mass by regulating TGF-␤ activity, which in turn modulated the proliferation and survival of osteogenic stem cells. This provides a new paradigm in understanding the molecular basis of osteoporosis caused by low bone formation.
Bone marrow cells (10 -20 ϫ 10 6 ) in culture medium containing 55 M 2-mercaptoethanol (Invitrogen) were plated into a 100-mm tissue culture dish. The cells were incubated until ϳ80% confluence (12-14 days) then washed with PBS to remove non-adherent cells. The adherent cells were detached with 200 units/ml collagenase, and plated for the experiments as described below for each assay. Colony Forming Efficiency-Colony forming efficiency assays were performed as previously described (27,28). Colonies containing more than 50 cells were counted using a dissecting microscope. Because each colony is theoretically formed from a single stem cell, this permits a direct estimate of the stem cell colony forming efficiency by counting the number of colonies formed by 1 ϫ 10 6 marrow cells originally plated.
In Vitro Calcium Accumulation-BMSCs (2 ϫ 10 5 /well) were plated into a 12-well plate and cultured until confluent. Culture medium was then replaced with osteogenic induction medium in the presence or absence of 200 ng/ml bone morphogenetic protein 2 (BMP2) (Wyeth Research, Cambridge, MA). After 6 weeks, calcium deposits were detected by staining with 2% Alizarin Red S (pH 4.2, Sigma). To quantify the stained nodules, the stain was solubilized with 0.5 ml of 5% SDS in 0.5 N HCl for 30 min at room temperature (29). Solubilized stain (0.15 ml) was transferred to wells of a 96-well plate, and absorbance was measured at 405 nm.
In Vivo Bone Formation-Approximately 3 ϫ 10 6 BMSCs were mixed with 40 mg of hydroxyapatite/tricalcium phosphate ceramic powder (Zimmer Inc., Warsaw, IN) and then transplanted subcutaneously into the dorsal surface of 10-week-old immunocompromised beige mice (NIH-bg-nu-xid, Harlan Sprague-Dawley, Indianapolis, IN) as previously described (30). These procedures were performed in accordance to specifications described in an animal protocol approved by an institutional review board (NIDCR 02-222). The transplants were harvested after 8 -9 weeks, fixed in freshly prepared 4% paraformaldehyde at 4°C for 2-3 days, decalcified with buffered 10% EDTA (pH 8.0) at 4°C for 1-2 weeks, and then processed for standard paraffin embedding. Three representative H&E-stained histological sections for each transplant were chosen for the quantitation of new bone generated. The percentage of new bone formed in transplants was measured by computer assisted histomorphometry using Scion Image (Scion Corp., Frederick, MD) or ImageJ (NIH Image, rsb.info.nih.gov/nih-image/).
Northern Blot Analysis-Confluent BMSCs were treated with 200 ng/ml BMP2 or vehicle in osteogenic induction medium for 2 weeks. The total RNA was isolated using RNA STAT-60 Reagent (TEL-TEST "B" Inc., Friendswood, TX) according to the manufacturer's recommendations. Northern blot analyses were performed as previously described (31).
Enzyme-linked Immunosorbent Assay-BMSCs (5 ϫ 10 4 /well) were plated into a 24-well plate and cultured to confluence (5% CO 2 , 37°C). The cells were cultured in ␣-minimal essential medium media with 3% bovine serum albumin overnight, treated with vehicle or 2 ng/ml TGF-␤1 in ␣-minimal essential medium media containing 0.1% bovine serum albumin for 30 min. The culture medium was collected for measurement of TGF-␤ concentration. Attached cells were washed twice with PBS and then fixed with freshly prepared 2% paraformaldehyde for 30 min at room temperature for analysis of matrix-bound TGF-␤1. Concentrations of TGF-␤1 in the culture media were measured using the DuoSet Elisa Development System (R&D Systems). The TGF-␤1 binding to cell surface and sequestered in the ECM was detected using a biotin-conjugated TGF-␤1-specific antibody (chicken, 36 g/ml, R&D Systems) after blocking nonspecific binding with Reagent Diluent (R&D Systems). Streptavidin-horseradish peroxidase (R&D Systems) was added for 20 min to bind the immobilized biotinylated anti-TGF-␤1 antibody. The substrate (tetramethylbenzidine/hydrogen peroxide, R&D Systems) reacted with horseradish peroxidase to develop color. Color development was stopped with 2 N H 2 SO 4 , and the intensity of the color was measured at A 450 nm and corrected at A 560 nm using a microplate reader. The amount of TGF-␤1 was calculated from a TGF-␤1 standard curve and was normalized by protein content. The protein content of attached cells was measured using a BCA Protein Assay Kit (Pierce).
Apoptosis of BMSCs-BMSCs (3 ϫ 10 4 /well) were plated into an 8-well chamber slide and cultured for 2-3 days (5% CO 2 , 37°C). The cells were then treated with vehicle or 1 M staurosporine (Alexis, San Diego, CA) for 4 h or 2 ng/ml TGF-␤1 for 48 h. At indicated time points, the cells were washed twice with PBS and then fixed with freshly prepared 1% paraformaldehyde for 10 min at room temperature. Cells undergoing apoptosis were detected by immunohistochemical staining with a terminal deoxynucleotidyl transferase-mediated dUTP and nick end-labeling (TUNEL) assay using the ApopTag Plus Peroxidase In Situ Apoptosis Detection Kit (Chemicon, Temecula, CA), following procedures recommended by the manufacturer.
Caspase-3 Activity of BMSCs-Confluent BMSCs were treated with vehicle, 2 ng/ml TGF-␤1, or 2 ng/ml TGF-␤1 plus 3 M caspase-3 inhibitor (benzyloxycarbonyl-DEVD-fluoromethyl ketone, BioVision, Mountain View, CA) in osteogenic induction medium for 5 or 6 h. The caspase-3 activity of BMSCs was measured using the BD ApoAlert TM Caspase Colorimetric Assay Kit (BD Biosciences) following procedures recommended by the manufacturer except that the caspase-3 activity was normalized by protein content. The protein content of cells was measured using a BCA Protein Assay Kit (Pierce).
Immunohistochemistry-Mouse embryos were removed immediately prior to birth (E19.5), dissected, fixed in cold freshly prepared Millonig's Modified Phosphate Buffered formalin (Surgipath, St. Louis, MO) at 4°C for 3 days, and then transferred into 70% ethanol. The fixed samples were processed for standard paraffin embedding and sectioning. Enzymatic immunochemistry staining of TGF-␤ was performed using the broad spectrum immunoperoxidase AEC kit (SuperPicTure TM Polymer Detection Kit, Zymed Laboratories Inc.), following procedures recommended by the manufacturer. The sections were blocked with 10% goat serum (Zymed Laboratories Inc.) and incubated with specific primary antibody recognizing TGF-␤1 (mouse, clone 9016, 1:50 in 10% goat serum, R&D Systems) or mouse IgG1 control at the same dilution (R&D Systems) overnight at 4°C. Apoptotic cells were detected by immunohistochemical staining with a TUNEL assay using the KLE-NOW-FragEL TM DNA Fragmentation Detection Kit (Oncogene, San Diego, CA), following procedures recommended by the manufacturer.
Microcomputed Tomography Analysis-Femurs were scanned and reconstructed with 8-m isotropic voxels on a CT system (eXplore MS, GE Medical Systems, London, Ontario, Canada). A bone standard (SB3, Gammex RMI, Middleton, WI) was also scanned using the same protocol and used for the calibration of bone mineral density measurements. The reconstructed three-dimensional imagines of distal femurs were analyzed using MicroView (GE Medical Systems, London, Ontario, Canada). A fixed threshold was used to separate the bone and marrow phase. The trabecular bone mineral density and bone volume per tissue volume in the distal femur were measured in an elliptical cylinder (1.3 ϫ 0.6 ϫ 1.9 mm) within the metaphysis.
Statistical Analysis-All figures are representative of at least three experiments. All data are presented as mean Ϯ S.E. Statistical analyses were done by using Student's t test.

Osteogenesis of BMSCs Is Decreased in the Absence of Bgn and Dcn in Vitro and in Vivo-
We have previously shown that bgn/dcn-deficient mice exhibit a more severe reduction in bone mass compared with wild-type (WT) mice or mice deficient in either bgn or dcn (17). Microcomputed tomography (CT) analysis showed that the trabecular bone mineral density and bone volume per tissue volume in the metaphyses of the distal femur of bgn/dcn-deficient mice were lower than WT mice at 6 weeks of age (Fig. 1). Moreover, the osteopenic phenotype was more profound with aging ( Fig. 1). To investigate the cellular mechanism of this severe osteopenia in the absence of both Bgn and Dcn, we hypothesized that the defect in BMSCs could be related to impaired bone formation in bgn/dcn-deficient mice. To test this hypothesis, we assessed the ability of osteogenic stem cells to calcify in vitro using Alizarin Red S staining that allows visualization of Ca 2ϩ accumulation. Ca 2ϩ accumulation in bgn/ dcn-deficient BMSC cultures was much lower than that in WT cell cultures ( Fig. 2A). Because Ca 2ϩ accumulation was relatively low in both WT and bgn/dcn-deficient cultures, we treated the BMSCs with BMP2, a stimulator of osteogenesis (32,33). BMP2 dramatically increased Ca 2ϩ accumulation ( Fig.   2A, right). Quantitation of Alizarin Red S staining showed that Ca 2ϩ accumulation in the cultures of BMSCs from bgn/dcndeficient mice was 51.6 Ϯ 4.4% less than that of WT cell cultures in the presence or absence of BMP2 ( Fig. 2A, right). These results indicated that the in vitro calcification ability of BMSCs was decreased in the absence of Bgn and Dcn.
To further extend our in vitro results, we assessed the ability of BMSCs to form bone in vivo using transplantation experiments. In those experiments, BMSCs from bgn/dcn-deficient or WT mice were mixed with hydroxyapatite/tricalcium phosphate and transplanted subcutaneously into the dorsal subcutis of immunocompromised mice. Although bone tissue was formed from both bgn/dcn-deficient and WT BMSCs at 8 weeks after transplantation (Fig. 2B, left), BMSCs from bgn/dcn-deficient mice generated 82 Ϯ 4% less bone than the cells from WT mice (Fig. 2B, right). Collectively, both in vitro and in vivo results suggested that the defect in osteogenesis by BMSCs contributed, at least in part, to the impaired bone formation in bgn/dcn-deficient mice. The impaired bone formation in the absence of Bgn and Dcn could be due to a defect in the differentiation of BMSCs. However, the expression levels of core binding protein (Cbfa1/Runx2), an essential regulator of osteoblast differentiation, and the osteoblast marker osteopontin, were not affected by the lack of both proteoglycans (Fig. 2C). The expression level of bone sialoprotein in bgn/dcn doubledeficient BMSCs was higher than in WT cells in both the presence or absence of BMP2 (Fig. 2C). Those results suggested that the impaired bone formation in the absence of Bgn and Dcn was not due to a decrease in differentiation capacity of BMSCs.
The Number of BMSCs Decreases in the Absence of Bgn and Dcn-Because the differentiation of BMSCs was not decreased in the absence of Bgn and Dcn, we speculated that a decrease in the number of osteogenic stem cells might be the reason for decreased bone formation. To measure this, colony forming efficiency assays were performed to determine the number of BM-SCs present in bone marrow. The results showed that there were substantially fewer colonies formed in the cultures derived from bgn/dcn-deficient bone marrow compared with the WT bone marrow (Fig. 3A, left). Quantitative evaluations showed that 66 Ϯ 5% fewer colonies were formed in bgn/dcn-deficient bone marrow cell cultures compared with WT cultures (average of seven independent experiments) (Fig. 3A, right). This data demonstrated that the number of osteogenic stem cells was significantly decreased in the absence of both Bgn and Dcn.
Apoptosis of BMSCs Increases in the Absence of Bgn and Dcn-Changes in the proliferation or survival of BMSCs could affect the number of osteoprogenitor cells and subsequently alter the number of mature osteoblasts. We, therefore, theorized that a decrease in proliferation and/or survival of BMSCs was responsible for the lower number of BMSCs in bgn/dcndeficient mice. To determine the proliferation rate of BMSCs in vitro, we used BrdUrd incorporation assay. As shown in Fig.  3B, the number of dividing cells in the bgn/dcn-deficient BM-SCs was 44.5 Ϯ 7.9% higher than that of WT cells after overnight growth (Fig. 3B, left). This experiment demonstrated that proliferation of BMSCs was increased in the absence of Bgn and Dcn.
Because the decreased number of BMSCs in the absence of Bgn and Dcn is not due to decreased proliferation of BMSCs, we next determined the effect of Bgn and Dcn on the survival of BMSCs in vitro. TUNEL staining of BMSCs showed that the number of apoptotic cells in the bgn/dcn-deficient BMSC culture was higher compared with the WT cell culture (Fig. 3C).
When BMSCs were treated with a potent apoptotic inducer, staurosporine, the number of apoptotic BMSCs from bgn/dcndeficient mice was still higher compared with WT cells (Fig.  3C). Therefore, the decrease in the number of osteogenic stem cells in the bgn/dcn-deficient mice is due to increased apoptosis rather than decreased proliferation of BMSCs.
Our in vitro findings suggested that increased apoptosis of osteogenic stem cells might lead to an insufficient number of mature osteoblasts and decreased bone formation. To confirm this in vivo, we examined apoptosis in the endosteum of the growth plates of the legs and ribs. The endosteum that lines the surface of the marrow cavity is composed of thin layers of mesenchymal cells and contains abundant osteoprogenitors at this embryonic stage. TUNEL staining showed that there were substantially more TUNEL-positive cells (arrows) in the endosteum (Fig. 4) at the bone collar region of growth plates in bgn/dcn-deficient mice (Fig. 4, c and d) compared with WT mice (Fig. 4, a and b). Increased apoptosis was also found in the bone marrow of bgn/dcn-deficient mice compared with WT mice (Fig.  4, e and f). Some of the apoptotic cells were observed on the surface of trabecular bone (asterisk, Fig. 4, g and h). Similarly, in ribs more apoptotic endosteum cells were found in bgn/dcndeficient mice as compared with WT mice (Fig. 4, i and j). These results suggested that there was increased apoptosis in osteoprogenitors and pre-osteoblasts. Thus, our in vitro and in vivo results demonstrate that the decreased number of cells in the osteoblastic lineage in bgn/dcn-deficient mice was due to increased apoptosis of osteoprogenitors.
Sequestration of TGF-␤ in the ECM Is Disrupted in the Absence of Bgn and Dcn-To understand the underlying mechanisms that cause increased proliferation and apoptosis of BM-SCs in the absence of Bgn and Dcn, we focused on the growth factors and cytokines that are regulated by Bgn and Dcn. It has been shown that both Bgn and Dcn bind to TGF-␤ (7) and regulate its effects on osteoblasts and their precursors (27). It was predicted that Bgn and Dcn would bind TGF-␤ and sequester it within the ECM. We further speculated that the distribution of TGF-␤ might be altered in the absence of Bgn and Dcn. Immunohistochemistry of sections of femur with a monoclonal antibody specific for TGF-␤1 showed that the immuno-reactivity in the regions containing hypertrophic chondrocytes, trabecular bone, and bone marrow of WT mouse femur section was remarkably greater than comparable bgn/dcn-deficient tissues (Fig. 5A). To quantify the levels of TGF-␤, we did an in vitro experiment to determine the relative amount of TGF-␤1 that binds to the cell layer (cell surface and ECM) in confluent cultures of BMSCs using a biotin-conjugated TGF-␤1-specific antibody. Consistent with immunohistochemistry of femoral sections, the amount of TGF-␤1 was much higher in the cell layer of WT BMSCs compared with bgn/dcn-deficient BMSCs (Fig. 5B). These results suggested that the amount of TGF-␤ sequestered in the ECM was decreased in the absence of Bgn and Dcn. We, therefore, theorized that TGF-␤ cannot be trapped in the ECM when both Bgn and Dcn are absent, subsequently leaving TGF-␤ free. To test this concept, we decided to measure the concentration of TGF-␤ in the media versus in the cell layers. However, the basal level of TGF-␤ in the culture media was undetectable. We took a different approach, in which we added 2 ng/ml TGF-␤1 into the confluent culture of BMSCs and then measured the TGF-␤1 that remained in the medium and bound to the cell layer. Fig. 5C (left) shows that the level of TGF-␤1 remaining in the bgn/dcn-deficient BMSC culture medium was higher than that in WT culture medium. In contrast, the level of TGF-␤1 binding to the cell layer of bgn/dcn-deficient BMSCs was lower than that binding to the WT layer (Fig. 5C, right). Overall, these results supported our hypothesis that Bgn and Dcn sequester TGF-␤ in the ECM.
BMSCs Are Hyper-responsive to TGF-␤ in the Absence of Bgn and Dcn-In vivo and in vitro analysis of the BMSC microenvironment suggested that TGF-␤ could not be sequestered in the ECM in the absence of Bgn and Dcn and therefore, the excess TGF-␤ would be free. We, therefore, hypothesized that the free TGF-␤ would bind to its receptors to activate its signal transduction pathway. To confirm our hypothesis, we examined the amount of TGF-␤1 bound to the surface of WT and bgn/dcn-deficient BMSCs in the sub-confluent cultures of BM-SCs. Confocal microscopic study showed substantially higher levels of active TGF-␤1 on the surface of bgn/dcn-deficient BMSCs compared with WT cells (Fig. 6A). Furthermore, Western blot analysis showed that the expression of TGF-␤ receptors I and II in bgn/dcn-deficient BMSCs was higher than that in WT cells (Fig. 6B), possibly due to positive regulation by TGF -␤1 (34, 35). We therefore predicted that TGF-␤ signaling would be increased in bgn/dcn-deficient BMSCs as compared with WT cells. To test this, we examined the expression of molecules that are critical to the TGF-␤ signaling transduction. TGF-␤ binds to its receptors and signals specifically through phosphorylation of members of the Smad family, Smad2/3, which then form heterodimers with Smad4. The heterodimers translocate to the nucleus and act as transcription factors regulating cell growth and survival (36,37). The expression levels of both Smad2 and Smad3 were higher in bgn/dcn-deficient BMSCs compared with WT cells in the presence or absence of TGF-␤1 (Fig. 6C). Smad2 was rapidly phosphorylated in both WT and bgn/dcn-deficient BMSCs upon treatment with 2 ng/ml TGF-␤1 (Fig. 6C). However, the expression level of p-Smad2 in bgn/dcn-deficient BMSCs was much higher than that in WT cells. These data clearly show that TGF-␤ signaling in BMSCs was increased in cells lacking Bgn and Dcn.
Increased TGF-␤ Signaling in the Absence of Bgn and Dcn Leads to Increased Apoptosis of BMSCs-Increased TGF-␤ signaling might alter the functions of TGF-␤ in regulating BMSCs. When BMSCs were treated with 2 ng/ml TGF-␤1, the cells rapidly proliferated (data not shown) and formed condensed nodules in both WT and bgn/dcn-deficient cultures. However, BMSCs in bgn/dcn-deficient cultures detached from tissue culture plates and died after 6 days of treatment with TGF-␤1. In contrast, WT BMSCs were unaffected (Fig. 7A). These results suggest that BMSCs from bgn/dcn-deficient mice responded differently to TGF-␤1 and underwent premature death.
It has been reported that overexpression of Smad3 in the presence of TGF-␤ significantly induces apoptosis (38,39). To test whether the premature death of BMSCs in the absence of Bgn and Dcn was due to increased apoptosis, we examined apoptosis of BMSCs in the presence or absence of TGF-␤1. TUNEL staining showed that there were substantially more TUNEL-positive cells in bgn/dcn-deficient BMSC cultures (Fig. 6B). TGF-␤1 up-regulated the expression of Smad-3 (Fig. 6C) and accordingly increased apoptosis of BMSCs (Fig. 7B). Caspase-3 is a key protease in the execution of apoptosis (40). To examine whether the TGF-␤1-induced death of BMSCs was mediated by caspase-3, we added a caspase-3 inhibitor to the TGF-␤1-treated BMSC cultures. As expected, the caspase-3 inhibitor did prevent TGF-␤1induced death of BMSCs (Fig. 7A).
To understand the molecular mechanisms of the accelerated apoptosis in bgn/dcn-deficient BMSCs, we examined the expression of pro-apoptotic Bcl-2 family proteins and the protease activity of caspase-3. Fig. 7 shows that the expression of proapoptotic proteins, Bax and Bad, was higher in bgn/dcn-deficient BMSCs than that in WT cells. The expression of both the active and the truncated form of Bad was slightly higher in bgn/dcn-deficient cells compared with WT cells in the presence and absence of TGF-␤1 (Fig. 7C). TGF-␤1 increased the expression of Bax in both WT and bgn/dcn-deficient cells. However, the expression level of Bax was higher in bgn/dcn-deficient BMSCs than in WT cells. The protease activity of caspase-3 was significantly higher in bgn/dcn-deficient BMSCs than that in WT cells (Fig. 7D). TGF-␤1 increased the activity of caspase-3 in both WT and bgn/dcn-deficient BMSCs (Fig. 7D). The protease activity of caspase-3 was still significantly higher in bgn/dcn-deficient BMSCs (Fig. 7D). This partially explains the early onset of TGF-␤1-induced death of bgn/dcn-deficient BMSCs. Taken together, it appears that the TGF-␤1 signaling is increased in the absence of Bgn and Dcn, leading to switch the fate of BMSCs from growth to apoptosis. DISCUSSION We have previously shown that bgn/dcn double-deficient mice have lower bone mass in trabecular and cortical bones compared with WT mice (17). We now could demonstrate that this reduction in bone mass was at least in part due to increased apoptosis of bgn/dcn-deficient BMSCs and consequently dramatic reduction in the number of BMSCs. A deeper investigation of the mechanisms showed that the lack of Bgn and Dcn prevents TGF-␤ from proper sequestration within the ECM, resulting in higher levels of free TGF-␤ available to bind to its receptors and, consequently, overactivating its signaling through the Smad pathway. Overactivated TGF-␤ signaling played a central role in promoting apoptosis in the bgn/dcndeficient BMSCs. Our results demonstrate that these two ECM proteoglycans are involved in controlling proliferation and survival of BMSCs and, subsequently, maintaining appropriate numbers of mature osteoblasts by controlling the availability of TGF-␤ to osteoprogenitors.
TGF-␤ is abundant in the ECM of bone and regulates proliferation, differentiation, and mineralization of osteoprogenitors and osteoblasts (41)(42)(43). Both large and small proteoglycans in the ECM interact with cytokines and growth factors and regulate their function (2,12,44). The ECM serves as storage for TGF-␤ and controls its activity (43). TGF-␤ is a well studied growth factor in terms of binding to proteoglycans, such as betaglycan, Bgn, Dcn, and fibromodulin (7,45). In vivo administration of Dcn prevents fibrosis caused by overexpression of TGF-␤ in human and experimental animal models (46 -48). In vitro, Dcn inhibits the binding of TGF-␤ to macrophages, and in turn reverses the TGF-␤-induced proliferation and adhesion of macrophages (49). Overexpression of Bgn or Dcn could neutralize TGF-␤-induced cell growth, collagen gel retraction, and expression of matrix protein (6,50,51). However, the TGF-␤mediated stimulation of collagen gel retraction and expression of Bgn is not affected by adding free Dcn to the culture medium (51), suggesting that Dcn controls TGF-␤ biological activity by sequestering it into the ECM. Nevertheless, those studies have not shown evidence that Bgn and Dcn sequester TGF-␤ in the ECM to regulate its activity. Here, we present for the first time direct evidence that Bgn and Dcn modulate TGF-␤ biological activity by sequestering it in the ECM. Similarly, BMP2, another member of the TGF-␤ superfamily, might also be regulated by Bgn and Dcn with the same mechanism. The expression of its signaling components, Smad1 and phosphorylated Smad1, were higher in bgn/dcn-deficient BMSCs compared with WT cells (data not shown), which could also explain the higher expression of bone sialoprotein (Fig. 2C).
We clearly demonstrate that TGF-␤ signaling in BMSCs increases in the absence of both Bgn and Dcn due to loss of the sequestration of TGF-␤ in the ECM. Dysfunction of TGF-␤ activation has been found in other pathological conditions (52,53). TGF-␤ is an important regulator of osteogenesis (43,54). Although the induction of TGF-␤ signaling in the absence of Bgn and Dcn is modest, the impact of this modest induction on tissues could be very significant, because it is constitutive. Increased TGF-␤ signaling could be the major mechanism that causes the osteopenic phenotype in the bgn/dcn-deficient mice. Consistent with our findings, a similar phenotype is observed in the transgenic mice that overexpress TGF-␤2 in osteoblasts via an osteocalcin promoter-driven transgene (55). Furthermore, inhibition of TGF-␤ signaling by expressing the dominant-negative form of TGF-␤ II receptor in transgenic mice causes a high bone mass phenotype (55,56). Therefore, alteration of the TGF-␤ signaling could affect bone mass.
TGF-␤ regulates cell growth, differentiation, and apoptosis in a variety of cells (57)(58)(59)(60). It has been reported that increased activation of TGF-␤ promotes apoptosis in different tissues (52,53,61). TGF-␤ stimulates proliferation of osteoprogenitor cells and ECM formation, but inhibits differentiation and mineralization of mature osteoblasts (41)(42)(43)62). Our studies revealed multiple roles for TGF-␤ in the regulation of BMSCs during osteogenesis. TGF-␤ stimulated proliferation of BMSCs and subsequently induced apoptosis. The switch in the fate of BMSCs from growth to programmed cell death depends on the activity, concentration, and duration of TGF-␤ treatment. The "TGF-␤ Hyperactive" BMSCs in the absence of Bgn and Dcn proliferated faster and underwent apoptosis prematurely. It has been shown in other cell culture systems that the biological effects of TGF-␤ could change under different circumstances, which can result in switching the fate of cells from one dominant activity to the other (59,63). For example, Hagedorn et al. (63) reported that the concentration of TGF-␤ controls the fate of neural progenitor clusters between neurogenesis and apoptosis. The majority of neural progenitor clusters choose neuronal fate at lower concentrations of TGF-␤, but apoptosis at higher concentrations. Not only TGF-␤, but also another member of the TGF-␤ superfamily, BMP2, favored neural differentiation at lower concentrations and apoptosis at higher concentrations (64). Apoptosis is a necessary process during development and maintenance of homeostasis in adult tissues (65). TGF-␤ controls the appropriate numbers of mature osteoblasts through its dual effects on proliferation and apoptosis of FIG. 7. The expression of pro-apoptotic proteins in BMSCs increases in the absence of Bgn and Dcn. A, phase contrast light microscopy (ϫ200) shows the post-confluent BMSCs cultures after treatment with vehicle, 2 ng/ml TGF-␤1, or 2 ng/ml TGF-␤1 plus 3 M caspase-3 inhibitor benzyloxycarbonyl-DEVD-fluoromethyl ketone (Z-DEVD-FMK)) in osteogenic induction medium for 6 days. B, apoptosis of BMSCs was measured by TUNEL staining after treatment with vehicle or 2 ng/ml TGF-␤1 for 48 h. Bar ϭ 50 m. Apoptotic cells were quantified by scoring all of the cells in three randomly chosen fields (ϫ400) on a chamber slide of each sample. Data are reported as means Ϯ S.E. (n ϭ 3). *, p Ͻ 0.03 and **, p Ͻ 0.007 compared with the percentage of TUNEL-positive WT BMSCs. C, Western blotting analysis shows the expression of pro-apoptotic proteins, truncated (T) Bad, full-length (F) Bad and Bax 48 and 72 h after BMSCs were treated with 2 ng/ml TGF-␤1. Hsp90 expression level was used as a loading control. D, the protease activity of caspase-3 was measured by the detection of the chromophore p-nitroanaline (pNA) after pNA was cleaved by caspase-3 from pNA-labeled caspase-3-specific substrate. The units of caspase-3 activity were quantified using a standard curve of free pNA. The protease activity of caspase-3 was normalized with protein content. Data are reported as means Ϯ S.E. (n ϭ 3). *, p Ͻ 0.05 and **, p Ͻ 0.0005 compared with the caspase-3 activity of WT BMSCs at the same time point. osteogenic stem cells during skeletal development and bone remodeling (66,67). Alteration of TGF-␤ activity could interrupt the balance between growth and death. The induction of premature death in osteogenic stem cells by an overactivated TGF-␤ signaling pathway would then lead to insufficiency of mature osteoblasts in bgn/dcn-deficient mice, causing a decrease in osteogenesis.
In summary, we show for the first time that ECM proteoglycans control the fate of bone marrow stromal cells. Our findings are important for several reasons. First, knowing the functions of ECM proteoglycans in regulating osteogenesis is critical for understanding how the skeleton is developed and, specifically, how the integrity of skeleton is maintained, and for understanding age-related skeletal diseases, such as osteoporosis. Second, our results reveal the importance of micro-environment in developing and maintaining the skeletal system. Third, our findings on the regulation of osteogenic stem cells by ECM proteoglycans provide a new target for developing therapeutic approaches for osteoporosis.