Osteoblast Recruitment and Bone Formation Enhanced by Cell Matrix–associated Heparin-binding Growth-associated Molecule (HB-GAM)

Recombinant HB-GAM was produced with the aid of baculovirus in Spodopteria frugiperda cells and purified to apparent homogeneity from the culture medium as described previously (Raulo et al., 1992). N-syndecan was isolated as previously described (Raulo et al., 1994). Alcian blue-silver staining that detects both proteins and proteoglycans was used in combination with SDS-PAGE to detect fractions that contain N-syndecan.

Affinity-purified antibodies against recombinant HB-GAM were produced in rabbit as previously described (Raulo et al., 1992) and affinity-purified as previously described (Rauvala, 1989). These antibodies have been characterized by Western blotting (Raulo et al., 1992), and their specificity against HB-GAM has been verified in an immunohistochemical context (Peng et al., 1995). Affinity-purified antibodies against a synthetic peptide corresponding to the NH₂-terminal extracellular moiety of rat N-syndecan (Carey et al., 1992) were produced and verified by Western blotting and immunohistochemistry (Raulo et al., 1994; Nolo et al., 1995).

All the animals used in the study were perfusion-fixed with 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1 M phosphate buffer (PB), except for embryos that were immersion-fixed with the same fixative. Five to six serial cryostat sections (40 μm thick) were cut for each animal. Three to four sections were handled for light microscopy, while the remaining two to three sections were reserved for electron microscopy (EM) after immunostaining. Replacement of primary antibodies with nonspecific rabbit IgG served as a negative control. After avidin-biotin-peroxidase complex reaction (Vector Laboratories, Burlingame, CA), peroxidase label was developed in 0.04% 3,3-diamine-benzidine tetrahydrochloride (DAB) with nickel enhancement (0.3% nickel ammonium sulfate). Thus, positive staining produces blue color on light microscopy and aggregated intense electron-density on EM. Detailed procedures are described elsewhere (Imai et al., 1997).

The immunostained sections reserved for EM were postfixed with 1% osmium tetroxide for 1 h and dehydrated in ethanol series. The sections were flat-embedded in Epon medium and coverslipped on glass slides coated with silicon. Light microscopic observation was made at this point to select areas to be examined electron-microscopically. Ultrathin sections (60 nm) were stained with uranyl acetate and lead citrate and were examined by a transmission electron microscope (Acceleration voltage: 60 KeV; model Joel 1200; Joel, Tokyo, Japan).

A 1.8-kb fragment of N-syndecan cDNA, corresponding to base pairs 67– 1867 and containing the full coding region of the mRNA (Carey et al., 1992), was used for preparation of N-syndecan probe. A 1.25-kb cDNA containing the whole coding sequence of HB-GAM (Merenmies and Rauvala, 1990) was used to prepare HB-GAM probe. Digoxigenin-labeled single-stranded sense and anti-sense RNA probes were generated as previously described (Szabat and Rauvala, 1996). 10 serial cryostat sections (7 μm thick), which topographically correspond to the immunostained sections, were cut and mounted on glasses. In situ hybridization was performed using a modified protocol of Wilkinson (1992) and is described elsewhere (Nolo et al., 1995).

Rat osteoblast-type cells, UMR-106 (ATCC CRL-1661), and human osteoblast-type cells, Saos-2 (ATCC HTB-85), U-2 OS (ATCC HTB-96), and KHOS/NP (ATCC CRL-1544), were recently purchased from the American Type Culture Collection (ATCC; Rockville, MD). Cells were freshly prepared for the assays in media recommended by the ATCC (UMR-106: DME, Saos-2, and U-2 OS: McCoy's 5A medium, KHOS/NP: minimal essential Eagle's medium) supplemented with 10% FCS, 100 U/ml penicillin G, and 100 μg/ml streptomycin. For comparison, 3T3 fibroblast-type cells and N18 neuroblastoma cells were used as control cells. Cells were harvested by trypsin and EDTA treatment and were recovered in suspension with 10% FCS-containing media for 1.5 h before using in assay.

To characterize the used osteoblast-type cells in terms of N-syndecan expression, Northern analysis for rat N-syndecan RNA was performed. Total RNA was prepared from UMR-106, N18, and 3T3 cells by lysis with guanidinium isothiocyanate, phenol extraction, and ethanol precipitation using RNeasy™ Total RNA Kits (QIAGEN GmbH, Hilden, Germany). RNA prepared from rat perinatal brain served as control (Nolo et al., 1995). The RNA was quantified spectrophotometrically, and 10 μg of total RNA was prepared for each lane. The RNA was electrophoresed in 1.2% agarose–7% formaldehyde gel and transferred to nylon filter. A 200-bp fragment of rat N-syndecan cDNA (Nolo et al., 1995) was labeled with ³²P Ready-to-go DNA labeling kit (Amersham Pharmacia Biotek, Uppsala, Sweden). After hybridization and washing, the filter was subject to autoradiography.

To characterize the HB-GAM–binding components from UMR-106, the cells were extracted with 50 mM octyl glucoside in ice-cold PBS containing 1 mM PMSF. HB-GAM affinity column was prepared by coupling 5 mg of the recombinant HB-GAM to a 1-ml NHS-activated HI-TRAP column (Amersham Pharmacia Biotek). The extracts were centrifuged at 40,000 g for 0.5 h, and the supernatants were passed through the column that had been equilibrated with 50 mM octyl glucoside. The column was then eluted with a 0.15–2.0 M linear gradient of NaCl. One-column volume fractions were collected, run on 4–15% SDS-PAGE, and stained with alcian blue–silver nitrate. The apparent molecular mass of the proteoglycan core protein was estimated after nitrous acid depolymerization. HB-GAM–binding fractions from perinatal rat brain served as positive control and were prepared as previously described (Raulo et al., 1994).

To identify that the HB-GAM–binding fraction from UMR-106 is N-syndecan, the samples were transferred on Immobilon P (Millipore, Molsheim, France) in 10 mM CAPS, pH 9.0, for 2 h at 70 V. Affinity-purified polyclonal antibodies against the NH₂-terminal synthetic peptide of N-syndecan were applied, and alkaline phosphatase–conjugated secondary antibodies were used for the detection.

Haptotactic migratory response to HB-GAM was studied for all the used osteoblast-type cells (i.e., rat osteoblast-type cells, UMR-106; and human osteoblast-type cells, Saos-2, U-2 OS, and KHOS/NP). The assay was carried out in Transwell chambers (Corning Costar, Cambridge, MA). The chamber is divided by a polycarbonate microporous filter (15 μm thick, pore size 8 μm), and the purified recombinant HB-GAM was coated to the lower surface of the filter at doubling concentrations from 0.025 to 50.0 μg/ml overnight at 4°C. The filter was then thoroughly washed with water and dried in laminar flow hood. The chamber was filled with the FCS-free medium containing 1 mg/ml BSA 30 min before adding cells. Cells were placed in the upper chamber at ∼1.0 × 10⁵ cells/cm² and cultured for 4 h. After fixing with methanol for 15 min, staining with 0.05% toluidine blue, and removal of the cells from the upper surface of the filter, cells that had migrated and attached to the lower surface were examined under a light microscope. Five fields per filter were counted for cell number (1 field = 1/160 of entire surface of filter). Filters coated with 20 μg/ml of nonspecific rat IgG or 10 mg/ml of BSA served as controls.

To test that the cell recruitment was induced by binding of cell surface N-syndecan and substrate-bound HB-GAM, the binding was competed by adding soluble forms of N-syndecan or HB-GAM to the medium. Doubling concentrations of soluble HB-GAM (from 5 to 20 μg/ml) or soluble N-syndecan (from 1 to 4 μg/ml) were added to the cell suspension 30 min before plating the cells. UMR-106 cells were plated on the filter, the lower side of which was coated with 10 μg/ml HB-GAM. The number of cells that migrated was counted as described above.

A possible involvement of tyrosine phosphorylation in HB-GAM– induced osteoblast recruitment was tested using inhibitors of protein tyrosine kinases. Benzoquinoid ansamycin (herbimycin A), a widely used general inhibitor of protein tyrosine kinases (Migita et al., 1994), and pyrazolopyrimidine 1 (PP1), a recently defined specific inhibitor of src-type tyrosine kinases (Hanke et al., 1996), were added to the cell suspension 10 min before plating cells to inhibit N-syndecan–mediated intracellular phosphorylation. Then, cells were subject to the transfilter assay using 10 μg/ml of substrate-bound HB-GAM.

Since cells may attach nonspecifically to polycationic proteins, the following cell recruitment assay was designed to show that the three-dimensionally intact structure of HB-GAM is pertinent to the osteoblast recruitment. The purified recombinant HB-GAM was coated to chamber slides (Lab-tek: Nunc, Inc., Naperville, IL) at different concentrations at 0, 0.1, 0.5, 1, 5, 10, 25, and 50 μg/ml overnight at 4°C. The chambers were washed with water and dried in a laminar flow hood, and electron microscopy grids were placed over the coated HB-GAM. The chambers were then UV-irradiated (model UV Stratalinker TM 2400, Stratagene, La Jolla, CA; wave length 254 nm) for 1 h. After removal of the grids, the chamber was thoroughly washed with water and filled with the FCS-free media containing 1 mg/ml BSA 30 min before plating the cells. The UMR-106 cells were plated at 1.25 × 10⁵ cells/cm² and cultured in the FCS-free medium. Chamber slides coated with 20 μg/ml of nonspecific rat IgG or 10 mg/ml of BSA served as control substrates. N18 and 3T3 cells, which are shown to be N-syndecan negative in the present study, served as negative control cells.

The study included 25 male Wistar-Lewis rats, aged 7 wk (body weight: 230–250 g). Postarthritic, periosteal ossification has been characterized in an adjuvant-induced arthritic model (Imai et al., 1997). In brief, 15 rats received intradermal injection of 0.05 ml suspension of Freund's complete adjuvant (cell wall fragments of heat-killed mycobacterium butylicum; 3 mg/g body weight; Difco Laboratories, Detroit, MI) in paraffin oil into the tail base. The remaining 10 animals, serving as controls, received the paraffin oil alone. Signs of inflammation in the hindpaw appeared by 7 d after inoculation. Apparent acute inflammation subsides, and the periosteum becomes proliferated during the second week. The proliferated periosteum mineralizes during the third week (Imai et al., 1997). The animals were killed 7, 10, 14, 21, and 28 d after inoculation of the adjuvant or vehicle (three animals/arthritic group and two animals/control group). Immunohistochemistry and immuno-EM for HB-GAM and N-syndecan were performed on the ankle samples.

HB-GAM cDNA (Merenmies and Rauvala, 1990) was cloned into a BamHI restriction site in pHBApr-1-neo vector (Gunning et al., 1987). pHBApr-1-neo contains the promoter and the first intron from human β-actin gene and SV-40–polyadenylation signal. The expression cassette was removed from the plasmid by digestion with EcoRI and NdeI, separated by agarose gel electrophoresis, and purified. Fertilized eggs from FVB/NIH and FlxNMRI mice were used for microinjection. Founder animals were screened by Southern blotting from tail biopsies. Founder animals were bred with FVB/NIH mice.

Tissue samples from transgenic and nontransgenic littermates were dissected, minced with scissors, and weighed. 1 ml sample buffer (60 mM Tris, pH 6.8, 8% glycerol, 1.8% SDS, 4.5% β-mercaptoethanol) was added per 100 mg tissue, and the tissue was solubilized by boiling for 10 min. Samples were centrifuged and 10 or 20 μl of the supernatant (corresponding to 1 or 2 mg of tissue weight) was loaded on SDS-PAGE. Proteins were transferred to nitrocellulose membrane. HB-GAM was detected by affinity-purified antibodies against the NH₂-terminal synthetic peptide of HB-GAM (Rauvala, 1989).

Gross appearance of the skeletons was evaluated after maceration in potassium hydroxide and removal of the soft tissue. Immunohistochemical localization of HB-GAM was studied using longitudinal and cross-sectional sections of femora and humeri of 1-yr-old animals (two transgenic and two nontransgenic animals). Histomorphometric analyses, following the recommended procedures and nomenclature (Parfitt, 1987), were conducted on seven transgenic and nine nontransgenic animals using computer-assisted image analyzer (model Image-Pro Plus; Media Cybernetics, Los Angeles, CA).

HB-GAM was intensely expressed in the matrix of the developing cartilage (Fig. 1,a), which acts as a template for endochondral bone formation (Marks and Hermey, 1996). Little if any expression of N-syndecan was detected before beginning of ossification (Fig. 1,b). HB-GAM persisted in the matrix of the mineralized cartilage (Fig. 1,c), onto which osteoblasts deposit osteoid (Marks and Hermey, 1996; Scammell and Roach, 1996). N-syndecan–expressing cells were localized in the zone of the mineralized cartilage (Fig. 1,d). The localization of the two molecules thus overlapped in the zone of the mineralized cartilage (Fig. 1,c, arrowheads, and Fig. 1 d, arrows). The overlapping expression of the two molecules was maintained while the frontier of the ossification advanced toward the ends of the cartilage template, i.e., N-syndecan–expressing cells appeared to trace the mineralized cartilage, where HB-GAM is abundantly expressed. On the other hand, N-syndecan was not expressed in the anatomical regions where formation of bone tissue had been already completed. These light microscopic localizations of N-syndecan– expressing cells suggest that N-syndecan is expressed by the osteoblast lineage cells that are migrating toward the target matrix. As for the cellular source of HB-GAM, immunolocalization of HB-GAM and in situ hybridization of its mRNA (see below) suggest that chondrocytes supplies it to the cartilage matrix.

Localization of the mRNAs, in turn, did not overlap. HB-GAM mRNA–positive cells were localized within the unmineralized cartilage matrix, whereas N-syndecan mRNA– positive cells were localized within the mineralized cartilage and mineralized bone (Fig. 1, e and f). N-syndecan immuno-EM of the mineralized cartilage (Fig. 1,f, rectangle g) showed numerous N-syndecan–expressing cells (Fig. 1,g, ch-b) that were distinct from the hypertrophied chondrocytes (Fig. 1,g, ch-a). Immuno-EM of the mineralized bone (Fig. 1,f, rectangle h) showed that N-syndecan was expressed by a group of cells, the ultrastructural phenotypes of which were not of blood vessels, osteoclasts, or macrophages but were suggestive of an active motility with the spindle-shaped cell somas and well-developed cell processes (Fig. 1 h). These cells were often seen in the channels between the trabeculae of the mineralized bone. Based on their ultrastructural features, N-syndecan may be expressed by the osteoblast lineage cells that are migrating before adhesion to their target matrix.

The postnatal growth of bone involves numerous anatomical sites of new bone formation. The metaphyseal part of the cartilage template remains as the growth plate that is responsible for the postnatal elongation of long bone. HB-GAM was abundantly expressed in the growth plate (Fig. 2,a), which remains throughout life in rodents. The physeal part of the cartilage gives rise to two types of cartilage, i.e., the outer portion of the cartilage becomes articular cartilage, and the inner portion of the cartilage provides a substrate for osteoblast recruitment by forming a secondary ossification center. HB-GAM was intensely expressed in the cartilage matrix that forms the hollow for the secondary ossification center, with the chondrocytes being presumptive cellular source of HB-GAM. In contrast, HB-GAM was not expressed in the outer portion that becomes the articular cartilage in future (Fig. 2 a).

The woven bone that is formed to create the initial shape of bone during prenatal life is continuously resorbed by osteoclasts and macrophages and finally disappears during early postnatal life. In turn, the space created by resorbing the woven bone is replaced by lamellar bone, which continues to grow further until adulthood. HB-GAM was expressed selectively by the osteocytes of the lamellar bone, whereas it was hardly expressed by those of the woven bone (Fig. 2, b–d). HB-GAM was detected in the matrix surrounding the osteocytes (Fig. 2,e, double arrows) and in the canaliculi of the osteocytes (Fig. 2 e, arrow). Most interestingly, HB-GAM was distributed on the surface of lamellar bone, where the growth of bone takes place by adding new bone. Thus, the osteocytes of the lamellar bone may prepare the bone surface as an osteoconductive surface by distributing HB-GAM through the network of bone canaliculi. After skeletal development, however, expression of both HB-GAM and N-syndecan is downregulated (see below for details of the periosteal expression). Taken together, HB-GAM is expressed in the various anatomical sites where osteoblasts are recruited and attached for the postnatal bone formation.

To test the hypothetical osteoblast recruitment induced by HB-GAM, we first performed a haptotactic assay using several osteoblast-type cells. All of the studied osteoblast-type cells rapidly migrated to HB-GAM in a dose-dependent manner, whereas the control cells, N18 and 3T3 cells, did not respond to HB-GAM (Fig. 3 A). Thus, the migration-enhancing effect of HB-GAM appears to be a general finding in the case of osteoblast-type cells (osteogenic sarcoma-derived cell lines).

Since N-syndecan appears to be expressed by migratory cells of the osteoblast lineage in vivo (Fig. 1, g and h), we further investigated the possible functional role of N-syndecan using rat osteoblastic cells (UMR-106 cells) and the control cells (N18 and 3T3 cells). Northern analysis showed a high expression level of N-syndecan mRNA in UMR-106. The expression level was similar to or even higher than that of the perinatal brain (Fig. 3,B). Affinity chromatography showed that UMR-106 cells express a major HB-GAM–binding protein with a molecular mass similar to that found in the perinatal brain, and this protein was identified as N-syndecan by Western blotting (Fig. 3,C). All members of the syndecan family have a conserved protease-susceptible site at the extracellular juxtaposition to the transmembrane domain (Carey, 1997), and treatment of the UMR-106 with trypsin (0.01% wt/vol, for 5 min) clearly abolished N-syndecan immunostaining (Fig. 3,D). Taken together, the rat osteoblastic cells, UMR-106, express N-syndecan as their major HB-GAM–binding protein on the cell surface. In contrast, the control cells that did not respond to HB-GAM were found to be N-syndecan negative (Fig. 3 B). Thus, the responsiveness to HB-GAM in the transfilter assay appears to depend on the presence of N-syndecan.

The recruitment of UMR-106 cells was competed by adding a soluble form of HB-GAM or N-syndecan into the cell suspension (Fig. 3,E, a), suggesting that the cell recruitment was mediated by binding of the cell surface N-syndecan and the substrate-bound HB-GAM. The cell recruitment was also inhibited by adding the inhibitors of tyrosine kinases into the medium (Fig. 3 E, b), suggesting that the cytoskeletal reorganization pertinent to the HB-GAM–induced cell recruitment is mediated by tyrosine phosphorylation of intracellular components.

Since cells may attach nonspecifically to polycationic proteins, another cell recruitment assay was designed. Irradiation with UV light has a protein-denaturing effect, but it is not expected to destroy the polycationic nature of the protein. UV-irradiation has been previously used to create patterned substrates from laminin (Hammerback et al., 1985) and from HB-GAM (Rauvala et al., 1994). UMR-106 cells were clearly recruited to the intact HB-GAM from the UV-irradiated HB-GAM at the concentration of 50 μg/ml (Fig. 4, b–d), suggesting that the intact molecular structure of HB-GAM is a prerequisite for the recruitment of the UMR-106 cells. Cell migration was inactive at lower concentration of HB-GAM, and the cells did not migrate at all when the substrate was not previously coated with HB-GAM (data not shown). The N-syndecan–negative control cells, N18 neuroblastoma cells and 3T3 fibroblasts, were not recruited to the intact HB-GAM (data not shown).

Normal adult rat bone hardly expresses HB-GAM (Fig. 5) or N-syndecan (Fig. 5,c, also see the rectangle in Fig. 5,a for orientation). To study possible participation of both molecules in regenerative ossification of adult bone, we used a postarthritic ossification model. Intracutaneous inoculation of Freund's adjuvant induces a systemic autoimmune-type arthritis ∼10 d after inoculation (day 10). After the arthritic damage, the periosteum goes through a direct intramembranous ossification (Fig. 5, j and m, rectangles) (Imai et al., 1997).

HB-GAM and N-syndecan were upregulated in a site- and time-specific manner during the process. Ultrastructurally, HB-GAM was first seen around the osteocytes on day 7, and the osteocytes appeared to distribute HB-GAM into the osteocytic lacunae (Fig. 5,d). Light microscopically, numerous HB-GAM–expressing osteocytes were noted within the damaged bone (Fig. 5,e). At this stage, however, N-syndecan expression was not detected (Fig. 5 f).

On day 10, an increased number of HB-GAM–expressing osteocytes was noted in the damaged bone (Fig. 5,h). In addition to osteocytic lacunae, HB-GAM was noted on the numerous osteocytic processes adjacent to the damaged bone surface (Fig. 5,g). The surface of the damaged bone became immunoreactive to HB-GAM in a similar manner to that found during the postnatal growth of bone (Fig. 5,h, compare to Fig. 2,e). Osteocytes are extensively connected with bone surface via their cell processes that form a network within the bone (Nijweide et al., 1996); thus the osteocytes may have distributed HB-GAM to the bone surface. Because of accumulation of induced osteoblasts, the periosteum thickens during the arthritic damage (Fig. 5,j) and ossifies thereafter (Fig. 5,m) (Imai et al., 1997). The thickened periosteum became immunoreactive to N-syndecan for the first time on day 10 (Fig. 5 i).

On day 14, HB-GAM was richly expressed in the thickened periosteum (Fig. 5,k), and the expression pattern of N-syndecan was similar to that of HB-GAM (Fig. 5,l). HB-GAM and N-syndecan were downregulated at the stage of mineralization on day 21 (Fig. 5, n and o) and completely disappeared on day 28 (data not shown). During the adult periosteal ossification, HB-GAM was thus initially upregulated by the osteocytes within the damaged bone and was distributed to the bone surface to which newly induced osteoblasts are recruited. In turn, N-syndecan was expressed in the thickened periosteum, presumably by the osteoblast/osteoblast precursors that are recruited to participate in the thickening of the periosteum.

To identify the cell types that express N-syndecan during the adult periosteal ossification, a detailed electron-microscopical study was performed. Fully differentiated osteoblasts covering the surface of healthy adult bone (Fig. 5,c, rectangle) did not express N-syndecan (Fig. 6,a). Interestingly, the osteoblasts were not found on the surface of the damaged bone 7 d after adjuvant inoculation (day 7; Fig. 5,f, rectangle). Instead, numerous spindle-shaped cells and their cell processes, which may represent fibroblastic osteoprogenitor, were noted along the bone surface. However, these cells were not immunoreactive to N-syndecan (Fig. 6 b).

On day 10, numerous spindle-shaped cells with long slender cell processes were noted in the vicinity of the damaged bone (Fig. 5,i, rectangle). These spindle-shaped cells expressed N-syndecan on their cell surface (Fig. 6,c). Since osteoblast later expresses N-syndecan (see below), these N-syndecan–expressing spindle-shaped cells may represent preosteoblast, a presumptive motile precursor of osteoblast. On day 14, the thickened periosteum (Fig. 5,l, rectangle) contained numerous osteoblasts with rich rough ER and Golgi apparatus. Osteoblasts expressed N-syndecan on their cell surface (Fig. 6 d).

At mineralization stage on day 21, the thickened periosteum contained numerous osteocyte-like cells being surrounded by the osteoid matrix that they produced themselves (Fig. 5,o, rectangle). These osteocyte-like cells had not yet developed typical gap junctions with neighboring cells. Some of the osteocyte-like cells retained immunoreactivity to N-syndecan (Fig. 6,e). On day 28, the ossified matrix contains only osteocytes (i.e., the terminal differentiation stage of the osteoblast lineage), and these cells did not express N-syndecan (Fig. 6 f). These results suggest that osteoblast lineage cells express N-syndecan when the cells are motile.

12 transgenic founders were produced using the HB-GAM cDNA under a β-actin promoter (Fig. 7 A). Southern blotting and PCR were used to verify integration of the construct to genomic DNA of the founders and the first generation offspring. PCR analysis indicated that the transgene was steadily transmitted to the next generations in a Mendelian fashion (data not shown).

A preliminary study on the HB-GAM transgene expression in different tissues (brain, heart, femoral muscle, liver, and kidney) was carried out by Western blotting. Pups of three lines were found to express the transgene in heart and skeletal muscle, whereas little if any overexpression was detected in liver, kidney, or brain (Fig. 7,B, a). These three mice lines displaying a mesenchymal expression of the transgene were selected and propagated for further analysis. The littermates genotyped as transgene negative did not express HB-GAM in any of the mesenchymal tissues analyzed shortly after birth (Fig. 7 B, a).

The highest protein concentration due to the HB-GAM transgene was seen in the periosteum of adult mice with a somewhat lower expression in the heart (Fig. 7,B, b). No expression could be detected in the periosteum of the transgene-negative individuals from the same litter. Estimated from the band intensities of Western blotting (Fig. 7 B, b), ∼50 μg/g (wet tissue weight) of HB-GAM was found in the periosteum of the transgene-positive adults. This amount is comparable to the highest expression levels found in developing brain (Rauvala, 1989).

The transgenic mice were normal at birth. However, a distinct macroscopic abnormality of bone became evident in the transgenic mice by 12 mo of age. Bones from the transgene-positive mice displayed an ivory-like solid surface (Fig. 7,C). In contrast, the bones from the transgene-negative mice displayed a brownish granular appearance with some marrow structures seen through the cortical bone (Fig. 7,C). The shape of the bones was almost normal by macroscopic observation, but a detailed measurement of the bone sizes suggested an enlarged diameter at the metaphysis (Table I). Similar changes were observed among all individuals from the three different transgene-positive lines (n = 3). Other tissues did not exhibit any macroscopic changes. Adult individuals propagated from the transgene-positive lines and their nontransgenic littermates were used for further analyses of the bone phenotype.

In the transgene-positive mice, HB-GAM was abundantly expressed by the osteocytes dispersed in the cortical bone (Fig. 8,A, c). HB-GAM was localized along the surface of the cortical bone (Fig. 8,A, a) in a similar manner to that found during the adult periosteal ossification (see Fig. 5,h). It thus appears that HB-GAM is localized because of binding sites on the bone surface, which may stabilize the secreted protein and present it in a surface-bound form to the motile osteoblasts/osteoblast precursors. In contrast, the transgene-negative mice were devoid of HB-GAM expression (Fig. 8,A, d) with the exception of the epiphyseal growth plate (Fig. 8 A, b).

Measurement of bone sizes revealed that the major significant difference was the increased thickness of the cortical bone (Table I). This was in agreement with the macroscopic appearance of the transgenic mice bone (Fig. 7,C). Thickness of the cortical bone was increased up to 147.4% for the humerus and 124.2% for the femur by 1 yr of age. Histomorphometric analysis showed that the cortical bone volume (Fig. 8,B, a) and the cancellous bone volume (Fig. 8,B, b) were significantly increased in the transgene-positive mice. Bone marrow area, another parameter that is a good indicator of osteoclast function (Fig. 8 B, c) (Parfitt, 1987), was not altered. Osteoblast surface (ObS/BS,%) of the transgenic mice, an indicator of the number of osteoblasts covering the bone surface (Parfitt, 1987), was 17.2 ± 4.4 as compared with 13.4 ± 5.9 in the nontransgeic mice.

Endochondral bone formation is a complex series of sequential biological events involving template formation by the cartilage, mineralization of the cartilage, resorption of the mineralized cartilage, and finally replacement by bone (Marks and Hermey, 1996). Strictly speaking, the mineralized cartilage is partially resorbed by osteoclasts and macrophages, which create channels for recruitment of osteoblast precursors but leave some struts of the cartilage intact. These cartilage struts act as an osteoconductive substrate, permitting the osteoblast precursors to attach and initiate deposition of osteoid (Marks and Hermey, 1996; Scammell and Roach, 1996).

Our study on the developing rat bone demonstrates that HB-GAM is strongly expressed in the cartilage template (Fig. 1,a). HB-GAM persists in the mineralized cartilage (Fig. 1,c, arrowheads), onto which the osteoblast precursors are recruited (Marks and Hermey, 1996; Scammell and Roach, 1996). This finding is in agreement with the very recent report addressing the expression of HB-GAM during chicken embryogenesis (Dreyfus et al., 1998). In turn, N-syndecan is expressed by a group of cells dispersed between the struts of the mineralized cartilage (Fig. 1, d, f, and g). Ultrastructural phenotypes of the N-syndecan– expressing cells are suggestive of an active cell motility (Fig. 1 h). These observations lead us to hypotheses that (a) N-syndecan may be expressed by the motile precursors of osteoblasts; (b) HB-GAM may be expressed in the matrices of their destination; and (c) HB-GAM/N-syndecan interaction may participate in osteoblast recruitment and attachment to their target substrates in addition to the chemotactic signals provided by the diffusible factors (Bonewald, 1996).

Localization of HB-GAM can also be seen in the cartilage matrices of the epiphyseal growth plate (Fig. 8,A, a and b, arrowheads) and the secondary ossification center (Fig. 2,a), on the bone surface of the growing lamellar bone (Fig. 2 e), and in the head epithelia of the developing skull (Mitsiadis et al., 1995). All of these anatomical regions are known to be the sites where osteoblast precursors are recruited for deposition of osteoid. Osteoblasts/ osteoblast precursors are thought to be brought into the vicinity of the target matrices, presumably via the chemotactic effects of the diffusible factors (Bonewald, 1996). However, little is understood about how these cells are thereafter informed where to deposit the osteoid. Accurate guidance of axonal outgrowth, which is a particular form of cell motility, has been shown for several matrix-associated guiding molecules (Baier and Bonhoeffer, 1992). It is thus possible that the osteoblast precursors, after being brought into the vicinity of their target matrices, use a contact-dependent signal to find their destination should they express a functional receptor for the signal. N-syndecan expressed by the osteoblasts/osteoblast precursors may act as a receptor for HB-GAM to guide the migration of these cells to their target substrate.

To test the hypothetical osteoblast recruitment mediated by HB-GAM, we used several osteoblast-type cells (i.e., osteogenic sarcoma cell lines), including rat osteoblast-type cells, UMR-106, and human osteoblast-type cells, Saos-2, U-2 OS, and KHOS/NP. These cells are shown to migrate rapidly to HB-GAM in a dose-dependent manner, whereas nonosteoblastic cells, N18 and 3T3 cells, do not respond to HB-GAM (Fig. 3 A).

To further investigate how the osteoblast-type cells respond to HB-GAM, whereas the nonosteoblast-type cells do not, we studied UMR-106 cells in more detail. Based upon the profiles of osteoblastic characteristics and morphological features that are similar to osteoblast lineage cells in vivo, UMR-106 cells have been considered as preosteoblasts, the immediate precursor of the fully differentiated osteoblast (Rodan and Noda, 1991). UMR-106 cells are known to express N-syndecan as abundantly as, or even more abundantly than, the developing brain (Fig. 3. B and D), which has been the only known source of N-syndecan (Carey et al., 1992). N-syndecan is also isolated from crude extracts of UMR-106 cells by HB-GAM affinity column (Fig. 3 B).

The control cells (N18 and 3T3 cells), which did not respond to HB-GAM, were shown to be N-syndecan negative (Fig. 3,A, c). The migration of the osteoblast-type cells in the transfilter assays cannot be explained by random migration of the cells because more than 90% of the plated cells migrate to the lower surface when HB-GAM is coated at 50 μg/ml. In addition, the recruitment of the cells does not depend on the nonspecific adhesiveness provided by the polycationic property of HB-GAM, but depends on the intact molecular structure of HB-GAM (Fig. 4). Furthermore, the recruitment of UMR-106 cells is likely to depend on the interaction of the cell surface N-syndecan and substrate-bound HB-GAM since the migration can be competed by addition of either soluble N-syndecan or soluble HB-GAM (Fig. 3 E, a). Further studies are required to show whether the receptor-type protein-tyrosine phosphatase-θ/β plays a role in HB-GAM– induced osteoblast migration in addition to N-syndecan, as it is suggested for the HB-GAM–induced migratory response in neurons (Maeda et al., 1996).

Cell surface proteoglycans have been regarded as mere storage sites for growth factors or as coreceptors in growth factor signaling but have not been generally considered functional receptors that can directly activate cytosolic signaling pathways. The cytosolic domain of N-syndecan binds a protein complex containing src-family tyrosine kinases (pp60-src and fyn) and the src-substrate cortactin (T. Kinnunen et al., 1998). Binding of HB-GAM to N-syndecan leads to phosphorylation and activation of src followed by cortactin phosphorylation (T. Kinnunen et al., 1998). Phosphorylation of cortactin is known to modulate its F-actin cross-linking activity (Hung et al., 1997) and probably contributes to the cell motility (T. Kinnunen et al., 1998). Since the specific inhibitor of src-family kinases, PP1 (Hanke et al., 1996), clearly impairs the recruitment of UMR-106 cells (Fig. 3 E, b), the cytoskeletal reorganization via src-dependent phosphorylation is likely to play a role in the HB-GAM–induced osteoblast recruitment. Furthermore, methods that have been applied to study N-syndecan–mediated signaling in neurons (T. Kinnunen et al., 1998) have been very recently used to study the intracellular interactions of N-syndecan in the UMR-106 cells. These studies have shown that the cytosolic moiety of N-syndecan interacts with a tyrosine kinase–active protein complex containing cortactin in the UMR-106 cells (Kinnunen, T., S. Imai, and H. Rauvala, unpublished results). These results agree with the inhibition of osteoblast migration by PP1. Thus, also from the viewpoint of the intracellular signaling mechanism, it has been suggested that the HB-GAM–induced osteoblast migration is similar to the HB-GAM–induced migratory response in neurons.

The experimental design using the Transwell^® chamber may require an explanation for the corresponding in vivo condition. The osteoblast-type cells must have extended their cell processes to interact with the HB-GAM molecule immobilized on the lower surface (Fig. 3,A, a). In fact, N-syndecan–expressing cells in vivo are characterized by the long slender cell processes that may explore their surrounding substrates (Fig. 6,c). The lengths of the processes well exceed 15 μm, which is the depth of the pore in the filter (Fig. 3 D, a). After the induced osteoblast precursors are brought into the vicinity of the target matrices, these cells may respond to HB-GAM with their long cell processes that express N-syndecan.

While much is known about the kinetics of bone turnover, little is understood about the coordinated regulation of bone deposition and resorption at molecular level (Canalis et al., 1991). The “nerve cell” role of osteocytes has been long debated for its anatomical and cell biological characteristics (Nijweide et al., 1996). Osteocytes are dispersed throughout the mineralized bone and are connected with their neighboring osteocytes via long cell processes that form a nerve-like network within the bone. Their processes connect with each other via gap junctions (Doty, 1981), thereby allowing cell-to-cell coupling (Jeansonne et al., 1979). The matrix around their processes is permeable to macromolecules (Dillaman et al., 1991), thereby allowing rapid passage of the yet unidentified signal molecules through the canaliculi (Nijweide et al., 1996). Elevation of metabolic (Pead et al., 1988; Skerry et al., 1989) and endocrine activities (Lean et al., 1995) after mechanical or electrical stimulation has predicted a role of osteocytes as a local regulator of bone formation. It is particularly interesting that the osteocytes upregulate HB-GAM production during an accelerated periosteal ossification (Fig. 5,e) and localize it on the bone surface during recruitment of the osteoblasts (Fig. 5, h and k). By localizing HB-GAM via bone canaliculi, the osteocytes may prepare the bone surface as a substrate for the appositional ossification. The prepared bone surface may act as an osteoconductive substrate in a similar sense to that of the mineralized cartilage (see Fig. 1 c).

Detailed ultrastructural characterization suggests that N-syndecan is expressed by several groups of bone forming cells, i.e., by osteoblast-like cells (Fig. 6,d) and later by osteocyte-like cells (Fig. 6,e). This is in a clear contrast to the finding that the differentiated osteoblasts covering the healthy bone surface do not express N-syndecan (Fig. 6,a). It must be remembered that these osteoid-producing osteoblasts are not motile. In turn, expression of N-syndecan seems to depend on the motility of the osteoblast lineage cells. Its expression is the most intense in the spindle-shaped cells with the long slender cell processes (Fig. 6,c), weaker in the osteoblast-like cells (Fig. 6,d), and faint in the osteocyte-like cells (Fig. 6,e). N-syndecan completely disappears from the mature osteocytes, the immobile terminal differentiation stage of the osteoblast lineage (Fig. 6 f).

The hypothesis that HB-GAM regulates bone formation is tested by producing the transgenic mice that maintain an elevated level of HB-GAM on their bone surface. First of all, it should be noted that these mice display no developmental anomaly at birth. This is probably due to the high expression level of endogenous HB-GAM during embryonic development. The high expression level is likely to override the effect of the transgene. The high expression level is maintained also during the postnatal growth of bone, where HB-GAM is localized on bone surface with the osteocytes being its cellular source (Fig. 2,e). After skeletal maturity, however, the expression of HB-GAM is normally downregulated (Fig. 8,A, b and d; also see Fig. 5 b).

The transgenic mice, in turn, maintain an elevated level of HB-GAM even at the age of 12 mo (Fig. 7,B, b), and the localization of HB-GAM on the bone surface is similar to that of the postnatal bone growth (Fig. 8,A, a and c, see also Fig. 2, d and e) and that of the postarthritic periosteal ossification (Fig. 5 h). The transgenic mice develop a distinct bone phenotype characterized by an increased volume of the cortical and cancellous bone. This finding is compatible with the finding that local HB-GAM antagonizes bone mineral loss induced by estrogen deprivation (Matsuda et al., 1997).

The mechanism of the increased bone formation is not likely to be due to the decreased osteoclastic function since the bone marrow space, a good indicator of osteoclastic function (Parfitt, 1987), is not altered. Addition of HB-GAM to osteoblast-type cells has been shown to result in a 1.6–2.0-fold increase in the alkaline phosphatase activity and a 1.2–1.8-fold increase in the DNA content (Zhou et al., 1992). Since magnitudes of these effects are rather modest as compared with those of other growth and differentiation factors, the osteoblast-stimulating action of HB-GAM has received little attention. However, the present study demonstrates a prominent osteoblast-recruiting action of HB-GAM, and the N-syndecan expression by the osteoblasts/osteoblast precursors also supports this. HB-GAM transgene may thus lead to the increased bone volume by not only recruiting the osteoblasts to the bone surface but also by maintaining the osteoblastic activity of the recruited osteoblasts.

The hypothetical osteoblast recruitment mediated by HB-GAM explains well the yet unclear mechanism of osteoblast recruitment and attachment to the different target substrates. It is particularly noteworthy that different cell types act as the source of HB-GAM, e.g., chondrocytes during prenatal endochondral ossification, osteocytes of the lamellar bone during postnatal bone growth, and osteocytes of the injured bone during postarthritic ossification.

Induction of osteoblasts has been attributed to a variety of growth and differentiation factors (Rosen and Thies, 1992; Kingsley, 1994; Bonewald, 1996), but the induced osteoblasts must be precisely located to form an appropriate shape and mass of future bone. In addition, bone is a self-organizational organ that alters its shape and mass according to the physical and chemical environments around bone itself.

HB-GAM is upregulated on the surface of damaged bone, where recruited osteoblasts deposit new osteoid for repair process (Fig. 5). The growth and differentiation factors responsible for the osteoblast induction are probably induced in the environment of the damaged bone. These factors may also act on the damaged bone so that the damaged bone surface may be marked with HB-GAM (Fig. 9). Although HB-GAM gene expression has been shown to be induced by platelet-derived growth factor in vitro (Li et al., 1992), other factors involved in bone repair may also act for the HB-GAM induction in situ. This model suits well the yet unclear mechanism of the self-organizational control of bone formation. The osteoblasts further differentiate to osteocytes being surrounded by the osteoid matrix that they produce, and they then may become a new source of HB-GAM for the next round of osteoblast recruitment (Fig. 9). A future application of HB-GAM may contribute to the clinically desired bone formation, which has not been fully attained by application of other growth and differentiation factors.

This work was supported by grants from the Technology Development Centre, Finland, the Sigrid Jusélius Foundation, the Academy of Finland, Nakatomi Foundation, Uehara Memorial Foundation, and Research fellowships of the Japan Society for the Promotion of Science. Shinji Imai is a JSPS Research Fellow.

HB-GAM

heparin-binding growth-associated molecule

PP1

pyrazolopyrimidine 1