Extracellular bone acidic glycoprotein-75 defines condensed mesenchyme regions to be mineralized and localizes with bone sialoprotein during intramembranous bone formation.

Bone acidic glycoprotein-75 is expressed very early during in vivo models of intramembranous bone formation, highly enriched in condensing osteogenic mesenchyme after marrow ablation and the osteoprogenitor layer of tibial periosteum. Bone sialoprotein accumulates within bone acidic glycoprotein-75-enriched matrix areas at a later stage in both models. Decalcification of initial sites of mineralization consistently revealed focal immunostaining for bone acidic glycoprotein-75 underneath these sites suggesting that mineralization occurs within bone acidic glycoprotein-75-enriched matrix areas. Ultrastructural immunolocalization of bone acidic glycoprotein-75 does not support a direct association with banded collagen fibrils, but rather suggests it is a component of a separate, amorphous scaffold occupying interfibrillar spaces. Double immunogold labeling demonstrated that a sizeable proportion of bone sialoprotein particles were located within a 50-nm radius of bone acidic glycoprotein-75. These results define bone acidic glycoprotein-75 as the earliest bone-restricted, extracellular marker of osteogenic mesenchyme. Based on this early bone-restricted expression pattern and a previously documented propensity of bone acidic glycoprotein-75 to form supramolecular complexes through self-association, bone acidic glycoprotein-75 may serve a key structural role in setting boundary limits of condensing osteogenic mesenchyme.

The molecular mechanism responsible for the biomineralization of bone remains unresolved (for reviews see Refs. [1][2][3][4][5][6][7][8]. Several issues remain controversial including whether biomineralization of bone is an active or passive process (9). If an active process, it is unknown whether the same cell biological and biochemical mechanism is used for both endochondral and intramembranous pathways of new bone formation (10). In terms of the identity of an active initiation or nucleation complex, a large body of evidence supports a role for both extracellular membrane limited vesicles (2,5,6) and phosphoproteincollagen complexes (1,3,4). However, the physiological roles for each of these initiation mechanisms in embryonic and postnatal bone development and repair is yet to be determined.
Marrow ablation of the rodent tibia and the postnatal tibial periosteum represent two non-pathological in vivo models of intramembranous bone formation. After removal of the marrow, the intramedullary cavity of the tibia rapidly fills with primary bone over a period of 7-8 days (11)(12)(13)(14). This bone is then removed through osteoclastic resorption coincident with the intramedullary cavity refilling with marrow tissue. The diaphyseal periosteum comprises morphologically separate layers, each representing different stages of osteogenesis and cell differentiation (15). Its outermost and innermost cell layers reflect osteoprogenitor and osteoblastic cell stages of differentiation, respectively (16,17). Many investigators have used these models to relate the onset of expression of specific cell surface receptors such as syndecan (18) and extracellular matrix markers such as bone sialoprotein (BSP) 1 (19) to morphological landmarks (and the underlying regulatory mechanisms) of intramembranous bone development.
Bone acidic glycoprotein-75 (BAG-75) is an acidic glycoprotein of 75 kDa in apparent size containing elevated contents of Asp/Glu (29%), sialic acid, and covalently bound phosphate (44 residues/mol) (20). Immunostaining of mouse and rat tissues, immunoprecipitations with a range of biosynthetically labeled tissues, and Northern blotting with a range of tissue-derived mRNA samples, 2 all indicate that BAG-75 is restricted in its expression to actively forming intramembranous bone and dentin (10,21,22). Purified BAG-75 displays a strong propensity to self-associate in vivo and in vitro into supramolecular spherical complexes (10 -20 m diameter) composed of multiple 10 -12-nm diameter microfibrils (23,24). A similar "fine filamentous" network is present at sites of primary bone formation (25,26). Electronegative complexes composed of BAG-75 can sequester millimolar quantities of phosphate ions (24); BAG-75 also binds up to 135 atoms of Ca 2ϩ /molecule at saturation (27). Interestingly, we have shown that the initial nucleation of hydroxyapatite crystals mediated by cultured osteoblastic UMR-106 cells occurs within supramolecular biomineralization foci (BMF) containing BAG-75 and BSP (28).
Bone sialoprotein has also been proposed to function as an apatite nucleator in bone tissue. It is found in the mineralizing boundaries of bone, dentin, and calcifying cartilage tissues (29 -31) and co-localizes to small foci within osteoid (32). Osteoblastic culture models forming biologic apatite express BSP prior to and during their respective mineralization periods (33,34); and purified BSP nucleates a small amount of apatite in vitro (35). Tye et al. (36) have shown that polyglutamate sequences within BSP mediate its binding to hydroxyapatite.
The goal of this study was to determine whether initial nucleation of hydroxyapatite in two in vivo models of intramembranous bone development occurs within pre-existing extracellular structures enriched in both BAG-75 and BSP that we have termed biomineralization foci. BMF are envisioned as three-dimensional matrix volumes representing a sequestered environment within developing bone matrix that accumulates and organizes secreted materials (alkaline phosphatase containing matrix vesicles, nucleators, enzymes) required for initial nucleation of calcium phosphate crystals.

Materials
Male Sprague-Dawley rats (250 -276 g) were purchased from Sasco, Inc. (Willington, MA). All animal procedures performed were pre-approved by the University of Missouri-Kansas City institutional review committee. Monoclonal anti-BSP (WV1D1(9C5)) antibodies were produced by M. Solursh and A. Franzen and obtained from the Developmental Studies Hybridoma Bank under the auspices of the NICHD, National Institutes of Health, and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA. Anti-BAG-75 protein antibodies (number 504) were previously characterized (21).

Methods
Bilateral Marrow Ablation Surgery-All animals were kept two per standard cage and allowed free access to food (Purina number 5001 lab rodent chow) and water. Marrow ablation surgery was performed bilaterally as described previously (13) under an approved University of Missouri-Kansas City laboratory animal protocol. Prior to surgery, rats were anesthetized intraperitoneally using a mixture of 80 mg/kg ketamine HCl and 12 mg/kg xylazine. After both hind limbs were shaved and scrubbed with betadine, a 2-cm longitudinal incision was made along the medial aspect of each proximal tibia, and the periosteum was exposed. While irrigating with saline, a 1-mm hole was drilled through the cortex 2-4-mm inferior to the epiphyseal growth plate using a round burr and high speed Dremel drill. A 20-gauge intravenous catheter connected to a vacuum line was used to aspirate the marrow tissue; aspiration was alternated with flushing of the intramedullary cavity with sterile saline. After 4 -5 cycles of aspiration and flushing, the ablated marrow cavity was filled with sterile saline and the wound was sutured with wound clips. Ablated rats begin to ambulate within 3-4 h of the surgery. Rats were euthanized 5, 7, and 10 days following surgery. Tibiae were dissected free of soft tissues and muscle and the bones were split in half to expose the intramedullary cavity that was filled with developing bone at these time points. New primary bone was carefully removed in one or two large segments (about 1 cm in length), leaving the more dense cortical bone behind. Primary bone was immediately covered with OCT and then flash frozen in liquid nitrogen prior to cryomicroscopy.
Electron Microscopy-Marrow ablation tissue used for electron microscopy was obtained 7 days after surgery, from the region most distal to the drill site. This region represents the earliest osteogenic phase (13). For conventional electron microscopy, tissue was fixed for 24 h with 2% glutaraldehyde in 150 mM sodium cacodylate, pH 7.2, at 4°C, then dissected into samples of Ͻ1 mm 3 . Samples were post-fixed in 1% osmium tetroxide, dehydrated, embedded, sectioned, and stained according to published protocols (37). For ultrastructural immunocytochemistry, distal tissue samples were fixed with 2% p-formaldehyde in phosphate-buffered saline overnight at 4°C. Samples were dissected into final-sized blocks, cryoprotected, frozen in liquid nitrogen, and sectioned as described previously (38,39). Frozen sections of 80 -100-nm thickness were processed for indirect immunolabeling using secondary antibodies coupled to 6-or 12-nm colloidal gold (Jackson ImmunoResearch, West Grove, PA), under conditions designed to minimize cross-reactivity and nonspecific binding of antibodies (40). In some cases, sections were decalcified in ice-cold 5% trichloroacetic acid for 2 h prior to immunolabeling. Sections were examined using a JEOL 1200EX transmission electron microscope at an accelerating voltage of 100 kV.

Confocal Microscopy Protocol on Decalcified Frozen Sections of Marrow Ablation Tissue at Days 7 and 10: Color Segmentation Analyses-
The fluorescence signal per pixel in a confocal image is assigned an 8-bit indexed color value from 0 to 255 for red and green color bands (the blue color band in these RGB images is set to zero). A mean background fluorescence value was determined separately for red and green color bands by averaging several areas in the image that do not exhibit an obvious signal. This mean value was subtracted from all pixel-indexed color values in its respective color band to correct for background fluorescence. An automated algorithm developed for a Silicon Graphics work station by the Whitaker Biomedical Imaging Laboratory (Dept. of Biomedical Engineering, The Cleveland Clinic Foundation) checks each pixel in the 512 ϫ 512 pixel image array for its indexed color value. Color values in each pixel above the threshold set for each color band are assigned a new color (yellow) when both green and red signals co-localize in that pixel. There are two major parameters that need to be set for accurate quantitation of fluorophore co-localization. First, lowlevel fluorescence values that may represent nonspecifically bound fluorescent reporter in the tissue sections need to be excluded. This procedure restricts the co-localization analysis to pixels exhibiting high intensity fluorescence as a result of specifically bound fluorescent reporter. If there is only one color band present and the value of that color band is greater than or equal to a preset value, then it is represented as that color in the image. Otherwise, it is turned black and becomes part of the background. In these particular color segmentation analyses, the value for high level fluorescence was set at 40 or higher of 256 indexed color values, thus preserving all pixels with fluorescence levels greater than 15% of saturated levels. A second parameter to be set is the red/green (or green/red) ratio calculated for each pixel. If the intensity of the minor color band is less than a preset proportion of the intensity of the major color, then that pixel is assigned the major color and is not considered to exhibit a significant level of co-localization. If the intensity of the minor color band is above the preset threshold, then it is considered to exhibit a significant level of co-localization and the pixel color is changed to yellow. This assures that any pixels considered to contain co-localized color bands have a significantly high ratio of one fluorophore to the other. Here, in these analyses, we set that ratio to be 1:4; that is, the intensity of the minor colors must be equal to or greater than 25% of the intensity of the major color.
Color segmentation analysis emphasizes co-localization as defined by the thresholds chosen and de-emphasizes intensity differences between the two color signals. Without use of color segmentation, visual recognition greatly underestimates the extent of yellow or co-localization.
Paraffin Embedding and Colorimetric Immunostaining for Marrow Ablation Tissue Sections-Ablated tibias were fixed for 2 days in Bouin's solution and then decalcified in 4% p-formaldehyde containing 0.85% sodium chloride and 10% acetic acid (13). Tissues were dehydrated in a series of ethanol solutions and infiltrated with xylene prior to embedding in paraffin and cutting into 5-m sections. Sections were de-paraffinized through sequential immersion in xylene and a graded series of ethanol solutions according to conventional procedures. Following re-hydration in Tris-buffered saline (TBS), tissue sections were immunostained using a Vectastain ABC kit with glucose oxidase (Vector Labs, Inc.), tetranitro blue tetrazolium substrate, and anti-BAG-75 protein antibodies (number 504) according to instructions supplied by the manufacturer. Photomicrographs of stained sections were taken with an Olympus microscope.
Visualization of Mineral and Immunofluorescent Staining of Primary Bone-Cryosections of undecalcified primary bone were fixed in ice-cold 70% ethanol for 1 h. Fixed sections were then washed three times with TBS, rinsed with de-ionized water, and then stained for 10 min with 0.4 mM alizarin red S (AR-S) dye, pH 4.2, at room temperature (34,41). Sections were then rinsed five times with water followed by a TBS wash to reduce nonspecific AR-S staining and cover slipped. Photomicrographs were taken with a Nikon TE2000 microscope equipped with a DN100 digital camera controlled with Image Analysis software.
Selected sections were decalcified and then immunostained. Briefly, coverslips were gently removed by soaking in TBS and sections were incubated for 16 h at 4°C in TBS containing 50 mM EDTA. Removal of AR-S stain was confirmed by microscopy. Decalcified sections were then processed for immunofluorescent staining with anti-BAG-75 protein antibodies (number 504) as described (28). Digital images were taken as described above. Digital images were assembled into collages using Photoshop software (version 7.0).
Removal of Periosteal Layer from Tibial Diaphysis-Full thickness periosteum was microdissected from a 7-mm portion of the mid-diaphyseal region of tibia recovered from 21-day-old male Sprague-Dawley rats (Harlan Labs) as described in Ref. 42. Periosteum tissue was fixed in 2% formaldehyde in phosphate-buffered saline for 3 h at 4°C, and then infiltrated with 20% sucrose in phosphate-buffered saline overnight at 4°C. Tissues were then placed into cryomolds, cambium side downward, covered with TBS Tissue Freezing Medium (Triangle Biomedical Sciences, Durham, NC), and frozen on dry ice. Specimen blocks were stored at Ϫ80°C until ready for cryostat sectioning at 10-m thickness. Alkaline phosphatase staining was done using the Vector Red kit (Vector Labs., Burlingame, CA)

BAG-75 Expression Precedes BSP and Alkaline Phosphatase Expression within the Periosteal Layer Lining Growing Rat
Bone-Radial growth in bone cortex width is mediated by the periosteum and follows an intramembranous pathway (15). Cambium, or osseous, periosteum exhibits a developmental polarity with osteoprogenitor cells in the apical layers, and preosteoblasts and mature osteoblasts in the basal layer ( Fig.  1, left panels) (42). The osteogenic nature of this tissue is demonstrated by its positive staining for alkaline phosphatase activity ( Fig. 1, middle panels), and BSP expression in its basal cell layers (Fig. 1, right panels). Of particular interest are the confocal staining patterns of BAG-75 and BSP within this osteogenic tissue. BAG-75 displays two distinctive deposition patterns: a fibrillar or layered pattern evident in the apical cell layers (Fig. 1, right panels, yellow arrowhead), and a second more focal pattern in the basal cell layers (Fig. 1, right panels, white arrowhead). By comparison, BSP (and alkaline phosphatase) appears as focal distributions restricted to the basal cell layers. The size and shape of these focal deposits of BAG-75 and BSP in the basal cell layers range from 5 to 15 m across, and are similar to the smaller-sized BMF observed in marrow ablation tissue sections and osteoblastic cultures (28). This change in BAG-75 staining pattern in the cambium periosteum coincides with the known developmental transition of osteoprogenitor cells to osteoblasts in this tissue, and suggests that the physical state of BAG-75 may change during osteoblastic dif- ferentiation. In addition, these data suggest that BAG-75 is expressed at an earlier stage of osteogenesis than either BSP or alkaline phosphatase. We next examined BAG-75 and BSP expression in the rat marrow ablation model of intramembranous bone development to extend our findings to an in vivo system that forms hydroxyapatite crystals.
Initial Mineral Deposits in the Marrow Ablation Model Occur in Focal Areas Enriched in BAG-75-Earlier work with the ablation model documents that intramembranous bone forms in the intramedullary cavity within 7-8 days after surgery (13). Fig. 2A depicts a region at day 7 containing condensed mesenchyme in the shape of thin trabeculae that stain lightly with AR-S. The presence of initial mineral crystals is evident at focal sites of intense AR-S staining (Fig. 2, A-C and E, arrows). These focal AR-S-stained sites are roughly circular in shape with overall diameters of 50 -75 m (Fig. 2, B and E, arrows) and appear to be composed of multiple, smaller, 10 -20-m diameter, spherical complexes staining intensely for mineral (Fig. 2C). This two-dimensional organization of mineralizing matrix is similar to that of the two-sized populations of biomineralization foci observed in UMR cultures (28). While not rigorously proven, we assume that the focal sites marked by the box in Fig. 2A represent a cross-sectional view through tubular shaped trabeculae. After decalcification and immunostaining the same regions, it is apparent that patches of BAG-75 underlie all areas that earlier contained AR-S-stained mineral deposits (compare Fig. 2, B/C and E with D and F, respectively). Remarkably, the shape and position of the BAG-75 immunofluorescent signal closely paralleled that of the AR-S colorimetric stain (compare Fig. 2, B and D). However, the relative intensity of the BAG-75 signal did not correlate directly with the intensity of AR-S fluorescence (compare Fig. 2, C and D).
Structures analogous to large BMF observed in UMR cultures (28) were also detectable in primary bone after hematoxyln and eosin staining. As shown in Fig. 3, a spherical acellular central region ϳ50 -75 m in diameter and stained pink with eosin (marked by arrowheads) is mostly surrounded by osteoblastic cells staining intensely for alkaline phosphatase expression (arrows). Close inspection suggests that the central region displays a reticular substructure that is reminiscent of the internal punctate composition of large BMF from UMR cultures (28). Based on these striking morphological similarities and the specific localization of extracellular BAG-75 protein prior to nucleation of mineral, we conclude that the sites of initial mineralization in primary bone and the UMR culture model share key structural features, yet may differ in their three-dimensional organization.
Condensing Mesenchyme Deposits BAG-75 Extracellularly in a Pattern Reflecting Future Dimensions of Mineralized Trabeculae-Primary bone formation after marrow ablation follows a process whereby the blood clot, which initially fills the emptied marrow cavity, is replaced by a porous loose connective tissue that subsequently undergoes condensation, differentiates into bone, and mineralizes. This process is illustrated at 5 days after ablation surgery where immunostaining of BAG-75 in decalcified tissue sections suggests an early organizational role during condensation of the osteogenic mesenchyme (Fig.  4). Mesenchymal condensation occurs at the edge of the residual fibrin clot (arrowheads) (Fig. 4C). Importantly, immunostaining for BAG-75 reveals a strong signal co-incident with White arrowheads outline the residual blood clot that displays autofluorescence. White arrow marks new brightly stained mesenchymal condensation that formed near the edge of the residual clot. Bar, 134 m. B, higher power view of same region of day 5 ablated tibia immunostained for BAG-75. Large arrow denotes the same mesenchymal condensation identified in A. Small arrows outline the segment of preexisting trabecular bone that is labeled primarily on its periphery. Bar, 67 m. C, hematoxylin-and eosin-stained new intramembranous bone at day 5 after marrow ablation. Black arrow demarks new, porous, eosin-stained mesenchymal condensation, whereas black arrowheads outline the area containing residual blood clots. This section represents an adjacent section to that shown in A. Bar, 67 m. D, hematoxylin-and eosin-stained new intramembranous bone at day 7 after marrow ablation. Black arrow denotes eosin-stained mesenchymal condensation, whereas black arrowheads outline the area containing residual blood clots. Note the increased intensity of eosin staining and higher density of condensed mesenchymal trabeculae on day 7 as compared with that on day 5 in B. Bar, 67 m. the first morphological appearance of developing bony trabeculae (Fig. 4, A-C). In this context, growth of condensed mesenchyme is progressing from right to left, replacing the retracting clot. Extracellular BAG-75 is expressed throughout the developing trabeculae and defines sharp boundaries with the loose connective tissue that will form the marrow (Fig. 4, B and C). Positive immunostaining of segments of pre-existing bone for BAG-75 (small arrows, Fig. 4B) is restricted primarily to their outer edge. Finally, comparison of the histological appearance of intramembranous bone at day 5 versus day 7 after ablation reveals several interesting points. First, the average width of condensed mesenchymal trabeculae does not change noticeably from their formation on day 5 to their mineralization on day 7 (Fig. 4, C and D) (13). However, the density of eosin staining increases dramatically over this 2-day period (for internal reference please refer to residual ossicles of bone within each view); this likely reflects the continued accumulation of type I collagen and other components required for mineralization. These results suggest that extracellular BAG-75 may participate in condensation of osteogenic mesenchymal cells and in defining the dimensions of future mineralized bone trabeculae.
FIG. 6. Quantitative color segmentation analysis of the relative co-distribution among BAG-75, BSP, and mineral in developing bone. Marrow ablation tissue was processed using thresholds for color segmentation of 15% of peak immunostaining intensity and an immunostaining ratio of 1 to 4. A, segmentation analysis supports co-localization of BAG-75 and BSP at m resolution. Results of quantitative color segmentation analysis at two different magnifications (ϫ16 and 400) and at two different times (days 7 and 10) during primary bone formation. B, relative distribution of BAG-75 and BSP with alizarin red S-stained mineral within day 7 primary bone. Results of quantitative color segmentation analysis of alizarin red S-stained frozen sections of undecalcified day 7 primary bone at two different magnifications (ϫ16 and 400).  7. Ultrastructural immunogold labeling studies demonstrate that BAG-75 exists as part of an amphorous extracellular network distinct from type I collagen fibrils. Primary bone tissue was cryosectioned and processed directly in A and B for BAG-75 immunogold electron microscopy using 12-nm particles. In C-E, thin sections of day 7 primary bone were double immunolabeled for BAG-75 (12-nm colloidal gold particles) and BSP (6-nm colloidal gold particles) before or after a short decalcification step (see "Methods"). Note that C and

BAG-75 Defines Matrix Regions to be Mineralized in Vivo 25460
On Days 7 and 10 after Ablation, BSP and BAG-75 Exhibit a Tight Spatial Co-localization during Primary Bone Formation-To determine whether BAG-75 is co-expressed with BSP, a suggested nucleator of hydroxyapatite crystal formation in bone (35,36), we carried out double label confocal immunofluorescence studies on decalcified cryosections of days 7 and 10 primary intramedullary bone. Decalcified tissue sections were used to avoid the potential effect of mineral on antibody access and to obtain a more accurate appraisal of actual co-localization of these two matrix proteins. These time points during the marrow ablation osteogenic process represent the maximum osteogenic response time (day 7) and the beginning of active bone tissue resorption (day 10). Fig. 5 (top panel) depicts the confocal detection of each matrix protein distribution separately at low magnification (ϫ16), whereas Fig. 5 (middle  panel) shows the color overlay and segmentation analyses for the ϫ16 images. The overlay view depicts an 8-bit depth combined color spectrum in an RGB color format, whereas color segmentation objectively depicts where green and red indexed color bands co-exist in each pixel within the image, and the computer assigns a yellow color where they co-localize (see "Methods" for details). According to our criteria for co-localization, the intensity of the minor color band must be equal to or greater than 25% of the intensity of the major color. Several features of these analyses are noteworthy. First, at day 7, the extent of co-localization of BAG-75 and BSP is almost complete. This is particularly well illustrated in side-by-side comparisons of segmentation views with conventional overlay images (Fig.  5, middle panel). In contrast on day 10, where the rate of formation is predicted to be less, BSP expression is substantially reduced and areas containing only BAG-75 are prominent (Fig. 5, top and middle panels). However, in areas where it is expressed, BSP is consistently accompanied by an elevated content of BAG-75, higher than in BSP non-expressing regions (Fig. 5, top panel). Whereas these specimens appear more condensed than in Figs. 2A and 4D, each displays a trabecular organizational pattern (Fig. 5, top panel). What seems to differ is the amount of loose connective tissue separating individual trabecular segments that may reflect local variability within the ablated intramedullary cavity. As shown in Fig. 5 (bottom  panel), substantial co-localization of BAG-75 and BSP in day 7 and 10 bone is still evident at 5-m resolution.
A quantitative morphometric analysis of BSP and BAG-75 co-localization based on color segmentation (Fig. 5) is graphically depicted in Fig. 6A. Results plotted are the average of two separate experiments. On day 7, BSP and BAG-75 co-localized with each other in 81% of the intramedullary tissue area at low magnification (ϫ16), and a substantial level of co-localization continued at 10 days after marrow ablation (52%). These trends persisted at higher magnification (ϫ400) yielding 78 and 64% co-localization within the intramedullary cavity on 7 and 10 days after ablation, respectively. At day 10, 34 -45% of the BAG-75 signal is distinct, whereas BSP alone is rarely found at either time point (Fig. 6A). Thus, at m resolution, a high level of co-localization exists between BSP and BAG-75 in marrow ablation tissue in vivo during primary bone formation.
Both BSP and BAG-75 Exhibit Substantial Spatial Co-localization with the Mineral Phase during Primary Bone Formation in Vivo-Because results depicted in Fig. 2 show that mineral was deposited exclusively at sites enriched in extracellular BAG-75, it was of interest to quantitate the extent of co-localization of BAG-75 or BSP with the mineral phase using undecalcified tissue sections. Similar to that for BAG-75 and BSP co-localization, a morphometric approach based on color segmentation results was used and results are plotted in Fig.  6B. Both matrix proteins exhibited a high degree of co-localization with the mineral phase, detected by alizarin red S, at either low or high magnification fields. In both cases, 28 -63% of the pixels in these images co-localized with the mineral phase (Fig. 6B). Importantly, there was 3-4 times more detectable protein immunostaining associated with the mineral phase than that identified separately. Unlike the co-localization analysis shown in Fig. 6A, these values likely underestimate the degree of BAG-75 or BSP co-localization because of the mineral phase blocking antibody access to the protein epitopes (Fig. 6B).

Ultrastructural Localization of BAG-75 within Primary Bone Does Not Support a Direct Association with Banded Collagen
Fibrils-To investigate the ultrastructural distribution and location of BAG-75, undecalcified sections of primary bone were subjected to an immunogold labeling approach with the same anti-BAG-75 antiserum used in the above mentioned confocal studies. As shown in Fig. 7, A and B, BAG-75 immunogold particles were neither randomly distributed within the extracellular matrix nor aligned in a periodic manner with banded fibrils. Rather, in one view, clusters of BAG-75 immunogold particles (5-10 per group) appeared to be concentrated near discrete regions of increased electron density (large arrows) within the unmineralized collagenous matrix adjacent to mineralized deposits (small white arrows, Fig. 7A). In a second view, which may be related to the first through a 90 degree rotation, BAG-75 was associated with an amorphous layer of electron density (large arrows) that appeared to overlay and obscure underlying banded collagen fibrils (small arrows, Fig.  7B). Areas of moderate labeling density often exhibited sharp boundaries with adjacent areas displaying background levels (bottom half of Fig. 7B). Non-immune serum-treated control sections were found to be almost completely devoid of clustered immunogold particles (see Fig. 7E).
Double Immunogold Labeling after Decalcification Reveals Ultrastructural Co-distribution of BAG-75 and BSP-Frozen thin sections of day 7 primary bone were double immunolabeled for BAG-75 (12 nm particles) and BSP (6 nm particles) both before (Fig. 7C) and after mild decalcification with EDTA (Fig. 7D). Non-immune serum-treated control sections displayed negligible immunostaining (Fig. 7E). Under the conditions used, the density of anti-BAG-75 particles appeared to be higher than that for anti-BSP particles, although this could be because of technical issues, e.g. antibody affinity differences or antibody access to epitopes. Whereas the same section was not analyzed both before and after decalcification, severalfold D represent semi-serial sections from the same tissue sample. Results shown are representative of replicate analyses. Scale bars in A-E, 200 nm. A, BAG-75 immunogold particles are clustered in groups of 4 -10 at focal sites of moderate electron density (black arrows) within the unmineralized collagen containing matrix. Adjacent denser regions containing mineral crystals are marked by small white arrows. Inset depicts the selected area after 2-fold enlargement. B, this section depicts two regions, a lower region enriched in a loose matrix containing banded collagen fibrils (small black arrows) and an upper region containing an additional, amorphous layer that is enriched in BAG-75 immunogold particles. The amorphous BAG-75 network (area denoted by large black arrows) does not appear to be aligned or oriented relative to the collagen matrix; lighter areas of electron density are largely devoid of BAG-75 immunogold particles. Inset depicts the selected area after 1.5-fold enlargement. C, undecalcified section. Arrows denote positions of selected 6-nm BSP particles. D, decalcified section. Arrows mark positions of selected 6-nm BSP particles. Inset shows a magnified view of smaller BSP immunogold particles (arrowheads) located within 50 nm of larger BAG-75 particles. Scale bar for inset: 80 nm. E, negative control. Undecalcified section was treated the same as experimental sections except that non-immune rabbit serum was substituted for primary antiserum.
higher immunogold particle labeling densities were consistently observed in decalcified specimens when semi-serial sections were compared (Fig. 7, C and D). BAG-75 appears to be located predominantly in regions of higher electron density. Within the aforementioned regions, 12-nm BAG-75 particles seem to be preferentially clustered in groups of 2 to 6 ( Fig. 7C). Some of the BSP immunogold signal was seen to co-distribute within 50 nm of the larger anti-BAG-75 particles (inset, Fig.  7D), whereas other 6-nm particles were present in regions of lighter electron density located up to 300 nm from the nearest BAG-75-associated particle (Fig. 7, C and D). In this way, ultrastructural immunolabeling results establish that BAG-75 and BSP are in part co-distributed with each other in primary bone within an approximate 50 nm radius. DISCUSSION The results presented in this paper support the following conclusions. 1) BAG-75 protein is expressed very early during the process of intramembranous bone formation in two in vivo models. BAG-75 protein appears coincident with morphological identification of the condensed mesenchyme stage after marrow ablation; it is deposited in a trabecular pattern that appears to define the boundaries of mineralized ossicles evident 1-2 days later. Within the growing rat periosteum, BAG-75 appears as thin layers in the fibroblastic layer that changes to a punctate staining pattern in the cambium layer. Temporally, BAG-75 is expressed prior to alkaline phosphatase and BSP within the periosteum. 2) BSP and BAG-75 proteins exhibit a close spatial co-localization with each other and individually with mineral deposits in the marrow ablation model. Given that BSP was not found in the absence of BAG-75, yet areas of developing bone and periosteum containing only the latter are readily apparent, BAG-75 appears to define a larger osteogenic matrix area than BSP. BSP accumulation occurs preferentially within BAG-75-enriched bone matrix, a coincidence that was also noted within initial sites of mineral nucleation (biomineralization foci) in osteoblastic cultures (model in Fig. 9, accompanying article, Ref. 28). 3) Ultrastructural immunolocalization of BAG-75 within the developing primary bone does not support a direct association with banded collagen fibrils, but rather indicates it is a component of a separate, amorphous scaffold or network that occupies interfibrillar spaces. 4) Double immunogold labeling demonstrated that a sizeable proportion of BSP molecules were located within a 50-nm radius of BAG-75. Taken together, these results define BAG-75 as the earliest bone-restricted, extracellular marker of osteogenic mesenchyme.
Initially discovered in 1988 (20), studies on BAG-75 have been limited by its strong propensity to form supramolecular complexes in vivo and in vitro, its restricted expression to primary bone, and its phosphoprotein nature (44 phosphates/ mole) (20,24,43). However, when such complexes are accounted for, the quantity of BAG-75 protein in primary bone is near that for osteopontin, a prominent non-collagenous protein.
Mesenchymal condensation is the first morphologically identifiable stage of embryonic cartilage and bone formation (48 -50). Syndecan and tenascin are known to function embryologically in defining the boundaries of condensing osteogenic or chondrogenic mesenchyme with loose marrow tissue (18,44). While early expression of syndecan paralleled that of tenascin, tenascin expression was reduced during diaphyseal intramembranous bone formation. It is interesting to speculate that the appearance of BAG-75 protein within the diaphyseal periosteum may be inversely correlated with that for tenascin.
Direct immunolocalization at the light and electron microscopic levels within developing bone indicates BAG-75, or structures in which it is contained, can assume either a supramolecular spherical shape or fibrillar layers. The detailed structure of these forms within bone is not yet known because immunogold particles, whereas not directly associated with collagen fibrils, labeled amorphous patches of increased electron density. It is noteworthy that purified BAG-75 can assemble in vitro into thin microfibrils ϳ12 nm in diameter and these microfibrils can coalesce to yield supramolecular spherical structures of 10 m in diameter (23,24). These structures are remarkably similar to the focal BAG-75 sites found in basal cell layers of the cambium periosteum. Because these structures can restrict the movement of phosphate ions in vitro, we have proposed that assemblies of BAG-75 could act as anion barriers because of their high electronegative fixed charge density, which approximates that of bone proteoglycans (24).
Debate continues regarding the biochemical mechanism of vertebrate biomineralization. Schinke, Karsenty and coworkers (9) hypothesize that calcification reactions in vivo are passive processes occurring readily in vascular, connective, and bone tissues in the absence of osteoid, and must therefore be under constant suppression by inhibitors. This conclusion is based largely upon the spectacular phenotype of the matrix GLA protein knockout mouse (51), which dies because of extensive vascular calcification. Other views of active biomineralization have also been proposed. In one, Glimcher (1) has proposed that initial biomineralization is mediated by phosphoprotein nucleators associated with the "hole region" of type I collagen fibrils. In another, phospholipid membrane-delimited vesicles released into the extracellular matrix by osteoblasts are believed to be the sites of initial mineral nucleation in bone (2,6,11,52,53). Matrix vesicles are small (50 -300 nm in diameter) and contain alkaline phosphatase and pyrophosphatase (54) activities required for mineral nucleation. Calcified, spherically shaped structures with a lower size limit of 300 -600 nm in diameter, termed crystal ghost aggregates, have also been observed in bone tissue (55)(56)(57) and in primary osteoblastic cultures (58 -60). BSP has been shown to co-localize with these structures in osteoid (32) and cultures (58,60,61). Based on the presence of polyacidic domains and a capacity to nucleate hydroxyapatite in vitro, BSP has been proposed to act as a nucleator in bone and cartilage (35,36). Interestingly, although BSP is associated with the first needle-like crystals produced in developing bone or culture models, BSP has not been reported to be a component of matrix vesicles (31, 59 -61). In this way, BSP appears to define a biochemically distinct population of particles, e.g. crystal ghost aggregates, which are larger than matrix vesicles and that can also become mineralized (Refs. 32, see also Midura et al. (28), accompanying article).
The relationship of BAG-75 and BSP to the subsequent nucleation of mineral crystals during intramembranous bone development provides a new higher-order context for biomineralization. We envision that a BAG-75-containing matrix scaffold could serve several heretofore unrecognized functions. First, extracellular BAG-75 could serve to define the boundary limits and volume of condensed mesenchyme to be subsequently differentiated into bone. In an analogous way, pre-existing BAG-75 containing BMF structures represent the sites of future mineral nucleation in osteoblastic cultures (model in Fig.  9; Midura et al. (28)). It is interesting to note that the dimensions of new BAG-75 positive trabeculae do not seem to change appreciably from their initial formation (condensation) to their subsequent mineralization. However, the staining density of trabeculae increases substantially reflecting the accumulation of collagen, BSP, and perhaps matrix vesicles (28) during the osteoblast differentiation stage. In this way, an underlying BAG-75 network may attract and distribute BSP and vesicle populations within trabeculae. The fact that a substantial portion of BSP immunogold particles were found within 50 nm of BAG-75 particles is consistent with this rationale. Temporal analyses in mineralizing osteoblastic cultures have already shown that BSP and matrix vesicles accumulate within BAG-75 containing BMF only after addition of a phosphate source (model in Fig. 9, Midura et al. (28)). BAG-75 may resemble the matrilin family of proteins that readily form oligomeric filamentous networks in connective tissues through their von Willebrand factor type A-like modules (62). In addition, matrilin-2 displays strong protein-protein interactions with other non-collagenous proteins such as fibronectin, nidogen-1, laminin-1, and fibrillin-1 (63), which are believed to be important in assembly of complex extracellular structures. Finally, a BAG-75-containing network could restrict access of mineralization inhibitors, e.g. matrix GLA protein, osteocalcin, osteopontin, and fetuin, to sites of initial mineral nucleation. Future studies will investigate potential protein-protein interactions of monomeric and oligomeric BAG-75 complexes with BSP and biomineralization vesicles.