Confocal Laser Raman Microspectroscopy of Biomineralization Foci in UMR 106 Osteoblastic Cultures Reveals Temporally Synchronized Protein Changes Preceding and Accompanying Mineral Crystal Deposition*

Mineralization in UMR 106-01 osteoblastic cultures occurs within extracellular biomineralization foci (BMF) within 12 h after addition of β-glycerol phosphate to cells at 64 h after plating. BMF are identified by their enrichment with an 85-kDa glycoprotein reactive with Maackia amurensis lectin. Laser Raman microspectroscopic scans were made on individual BMF at times preceding (64–76 h) and following the appearance of mineral crystals (76–88 h). The range of variation between spectra for different BMF in the same culture was rather small. In contrast, significant differences were observed for spectral bands at 957–960, 1004, and 1660 cm-1 when normalized BMF spectra at different times were compared. Protein-dependent spectral bands at 1004 and 1660 cm-1 increased and then decreased preceding the detection of hydroxyapatite crystals via the phosphate stretching peak at 959–960 cm-1. When sodium phosphate was substituted for β-glycerol phosphate, mineralization occurred 3–6 h earlier. Irrespective of phosphate source, the Raman full peak width at half-maximum ratio for 88 h cultures was similar to that for 10-day-old marrow ablation primary bone. However, if mineralization was blocked with serine protease inhibitor 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 64–88-h BMF spectra remained largely invariant. In summary, Raman spectral data demonstrate for the first time that formation of hydroxyapatite crystals within individual BMF is a multistep process. Second, changes in protein-derived signals at 1004 and 1660 cm-1 reflect events within BMFs that precede or accompany mineral crystal production because they are blocked by mineralization inhibitor 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride. Finally, the low extent of spectral variability detected among different BMF at the same time point indicates that mineralization of individual BMF within a culture is synchronized.

gated, facilitating further growth and expansion of the initial mineral phase into the larger, territorial collagenous matrix. The latter research focuses on the in vivo functionality of the mineralized bone product where hydroxyapatite crystal formation is envisioned to occur in a manner that facilitates subsequent vascular access to the crystals and placement of crystals within the organic matrix so as to optimize mechanical support for organs, joints, muscles, and tendons.
Bone osteoid is enriched in phosphoproteins, acidic glycoproteins, and proteoglycans, some of which like BSP or its fragments perform in vitro as nucleators of hydroxyapatite crystals (14,15). We have shown that phosphoglycoprotein BAG-75 delineates future extracellular sites of mineralization in vivo within primary bone and in vitro in osteoblastic cultures termed biomineralization foci (9,10). Specifically, BMF are 10 -25 m in diameter, spherical extracellular complexes containing several sizes of vesicles that are sites of the first mineral crystals in several osteoblastic models. Following plating, UMR 106-01 cells proliferate over the first 60 -64 h and attain a competency to initiate mineralization in BMF, if supplemented with a phosphate source (16). BSP has also been localized to discrete mineralizing nodules or crystal ghost aggregates in rat bone, which are analogous to BMF (17)(18)(19). BSP incorporation into BMF peaks at ϳ8 h following phosphate addition, several hours prior to the initial appearance of mineral crystals (18). Based on these findings, we proposed that BMF complexes function in an active mineralization process initiated and controlled by osteoblastic/osteocytic cells.
To characterize the structure and composition of BMF complexes, we have used laser capture microscopy to isolate mineralized BMF complexes. Initial studies revealed that, as predicted, mineralized BMF complexes were preferentially enriched in BAG-75 and BSP. In addition, however, purified BMF also contained 45-50-kDa fragments of BAG-75 and BSP (10). Importantly, cleavage of BAG-75 and BSP could be specifically blocked by treatment of osteoblastic cultures with 4-(2aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF), a serine protease inhibitor. Nucleation of mineral crystals within BMF complexes was also inhibited in AEBSF-treated cultures. We hypothesize that mineral crystals in these models of primary bone formation are the product of a biochemical pathway controlling the proteolytic activation and assembly of nucleating complexes within BMF.
Confocal laser Raman spectroscopy is a nondestructive and sensitive method capable of identifying individual spectral bands arising from both the inorganic calcium phosphate crystals and the protein/lipid constituents of BMF (20). Several recent studies illustrate the usefulness of this method when applied to studies of bone and bone cell cultures. First, Raman spectroscopy was able to quantitate the mineral phosphate content of bone ossicles at 1 or 3 weeks. The phosphate signal (v 1 at 927-979 cm Ϫ1 ) (21,22) and that for hydroxyproline (862-899 cm Ϫ1 ) were integrated, and a ratio for each ectopic ossicle was then calculated (21). Second, Jell et al. (23) used real time Raman scans to detect changes in primary osteoblastic cultures treated with bioglass conditioned media over a 2-week period. Third, two-dimensional mapping of collagen-derived and mineral-derived signals in cortical bone from genetically modified strains of mice indicated correlations of physical structure and composition with the elastic modulus mechanical properties (24). Fourth, comparative Raman analyses were able to detect substantive changes in the osteocyte lacunae surrounding the cells in glucocorticoid-treated bone. The observed reduction in mineral to matrix ratio correlated well with a similar reduction in elastic modulus localized to these sites (25). Carden et al. (26) used Raman spectroscopy to identify changes in bone caused by mechanical deformation. These initial analyses suggest that deformation brought about breakage of collagen cross-links. Finally, the method was used to characterize the mineral deposits induced by oyster shell nacre-derived biomolecules when added to MC3T3-E1 cultures (27).
The goal of this study was to use confocal laser Raman microspectroscopy to characterize the chemical nature of the inorganic and organic components comprising individual biomineralization foci as a function of time following addition of BGP or sodium phosphate to UMR 106-01 cultures. Our findings indicate for the first time that protein-derived spectral signals from BMF undergo reproducible temporal changes that precede and accompany the event of mineral crystal formation occurring within these complexes. In contrast, when mineralization is blocked with protease inhibitor AEBSF, the BMF Raman spectral profile remains unchanged over the same time period. Our findings demonstrate that mineralization within a majority of BMF in a culture follows a similar multistep pathway displaying a remarkable temporal synchronization.

Materials
Male Sprague-Dawley rats (250 -275 g) were purchased from Sasco Inc. (Willington, MA). All reagents used were either biotechnology grade or reagent grade. UMR 106-01 cells were cloned by Midura et al. (28). Texas red-conjugated neutravidin and biotinylated MAA lectin were purchased from Invitrogen and E-Y Laboratories, Inc., respectively.

Methods
Marrow Ablation Surgery-Marrow ablation surgery was performed bilaterally as described previously (29,30). Rats were maintained two per standard cage, and all rats had unrestricted access to food (Purina 5001 lab rodent chow, 1% calcium) and water. Briefly, after exposure of the proximal region of the tibia, a 1.5-2.0-mm diameter hole was drilled through one wall of the cortex at a site 3-4 mm distal to the subchondral plate, and the marrow was removed through several rounds of suction and flushing with sterile saline. Animals were euthanized on days 7 and 10 following surgery by exsanguination via heart puncture under anesthetic. Tibias from each rat were dissected free of soft tissues and processed as follows. One tibia was split longitudinally in half and the medullary cavity exposed. The contents of the cavity were carefully separated from the cortical wall and removed, usually in one piece, and placed in a preweighed vial, and a dry weight was obtained prior to acid digestion. Primary bone was removed from the second tibial intramedullary cavity and frozen in OCT media, and 7-mthick frozen sections were cut for Raman scans. Alternatively, some ablated tibiae were processed for histology as noted below.
Tibiae were fixed/decalcified for 2 days in Bouin's solution and then for 6 days in 4% formaldehyde containing 0.85% sodium chloride and 10% acetic acid. The latter reagent was changed every 2nd day over this period. Tissues were then dehydrated in a series of ethanol solutions and infiltrated with xylene prior to embedding in paraffin and cutting 5-m sections. Sections were de-paraffinized through sequential immersion in xylene and a graded series of ethanol solutions according to conventional procedures; sections were then stained with Alizarin red S dye and/or MAA lectin.
Fluorescent Staining of Bone Sections and UMR 106 Cultures with MAA Lectin-Frozen sections of new intramedullary bone from the marrow ablation model and glass slides with ethanolfixed osteoblastic cell layers were hydrated with Tris-buffered saline (pH 7.5) containing 0.05% Triton X-100 and stained as usual with Alizarin red S dye. Following destaining, sections were viewed and photographed using a Nikon TE-2000 fluorescent microscope. The positions of stained trabeculae were noted relative to fixed landmarks. Sections were then decalcified in cold 0.05 M EDTA (pH 7.5) for 45 min and then exchanged back into TBS (TBS: Tris-buffered saline (pH 7.5) containing 0.05% Triton X-100) prior to treating with 200 g/ml biotinylated MAA lectin for 90 min at room temperature. Controls did not receive lectin. Slides were washed four times in a coplin jar with TBS and then incubated with 40 g/ml Texas red-conjugated neutravidin for 60 min in the dark. Slides were then washed four times with TBS. Slides were drained, coverslipped, and imaged with a fluorescent microscope. Using landmarks, prior fields were located and photographed again under both bright field and fluorescent illumination.
Routine Culture of UMR Cells-UMR 106-01 BSP cells were passaged and cultured at 37°C and 5% carbon dioxide as described previously (10).
Growth and Mineralization of UMR Cells on Glass Slides for Raman Spectroscopy-Cells were seeded at a density of 1.0 ϫ 10 5 cells/cm 2 in Growth Medium (Growth Medium: Eagle's minimal essential medium supplemented with Earle's salts, 1% nonessential amino acids, 10 mM HEPES (pH 7.2), and 10% fetal bovine serum). After 24 h, the medium was exchanged with Growth Medium containing 0.5% BSA (catalog number A-1933, Sigma) in place of fetal bovine serum. Sixty four hours after plating, the culture medium was exchanged with Mineralization Media (Mineralization Medium: Growth Medium containing either 0.1% BSA and 6.5 mM ␤-glycerol phosphate or 5.0 mM sodium phosphate). (Addition of a supplemental phosphate source to UMR 106-01 cultures prior to 64 h had no effect on the formation of mineral crystals. 3 ) Cultures were then incubated for up to an additional 24 h at which time the culture slide was removed, rinsed very briefly with 0.05 M ammonium bicarbonate (pH 7.5), drained, and air-dried quickly in a laminar flow hood. Acellular controls were prepared by coating glass microscope slides with bovine serum albumin. Slides were incubated with media containing 0.1% bovine serum albumin for 64 h at which time 6.5 mM BGP was added, and the plates were incubated for an additional 24 h. Slides were then rinsed briefly with ammonium bicarbonate (pH 7.5), air-dried, and analyzed.
In some culture experiments, serine protease inhibitor AEBSF (EMD Biosciences, Inc.) was added at 64 h after plating at a final concentration of 0.1 mM in Mineralization Media. We have shown previously (10) that this concentration is the minimum dosage able to effectively block mineralization of BMF within UMR cultures. Cultures were then incubated with AEBSF for 0 -24 h at which time the culture slide was removed and processed for Raman spectroscopy or Alizarin red staining.
Isolation of Mineralized BMF from UMR Cultures by Laser Capture Microscopy-UMR cells were grown as described above on Fisher Plus microscope slides (Fisher), fixed, and stained with Alizarin red S dye. Immediately prior to laser capture, slides were dehydrated through a graded series of ethanol washes and xylene rinses. Dried slides were stored at Ϫ20°C in a sealed box with desiccant until used. Mineralized BMF were collected onto standard caps using an Arcturus Pixel IIe microscope. Collection films were pooled and stored in 70% ethanol at Ϫ20°C until ϳ6200 BMF were collected. LCM-captured BMF were then mixed in 70% ethanol to dislodge the purplestained particles that were then microcentrifuged to remove the ethanol. BMF pellets were extracted twice sequentially over a 2-day period at 4°C with 1.1 ml of 0.1 M Tris acetate buffer (pH 7.8) containing with 0.5% octyl glucoside, 0.05% SDS, 0.05 M EDTA, and 0.02% sodium azide. Extracts were then dialyzed first against 0.01 M Tris acetate buffer (pH 7.8) containing 8 M urea, 0.05% SDS, 0.1% octyl glucoside, 0.05 M EDTA, and second against 0.01 M Tris acetate buffer (pH 7.8) containing 8 M urea, 0.05% SDS, and 0.1% octyl glucoside. Controls consisted of glass slides containing the total cell layer fractions from ϩBGP or ϪBGP cultures; control slides were extracted using a similar protocol. The resultant dialyzed extracts were used for comparative immunoblotting studies where identical protein amounts were loaded per gel lane.
SDS-PAGE and Immunoblotting-Protein samples were electrophoresed under reducing conditions on 4 -20% linear gradient gels (ISC BioExpress) according to Laemmli (31) and electroblotted onto polyvinylidene difluoride membranes (Millipore Corp.) for 2 h at 100 V. The transfer buffer was composed of 10 mM CAPS buffer (pH 11.0) containing 10% methanol. Blots were processed essentially as described previously (32). Immunoblotting with digoxygenin-MAA lectin followed a procedure described earlier (32) except that secondary horseradish peroxidase-conjugated anti-digoxygenin antibody was used for chemiluminescent detection. Films were digitized using a flat bed scanner. To detect proteins directly, some gels were stained with Sypro Ruby stain.
Raman Spectral Analysis of UMR Cultures and Marrow Ablation Tissue Sections-All Raman spectra were recorded on a Jasco NRS-2000 Raman microspectrometer using 514.5 nm excitation from an argon ion laser (Spectra-Physics, model 2020, Mountain View, CA). The laser beam was focused down to a diameter of ϳ1 m through a ϫ100 objective, generating a power of ϳ6 milliwatts at the sample. Backscattered (180°) Raman photons were collected with the same objective, passing through a Notch filter (Kaiser Optical Systems, Inc.) to cut off undesired Rayleigh scattering, and detected by a liquid nitrogen-cooled CCD detector (Princeton Instrument LN/CCD-1100 PF). Data were obtained from 10 or more individual sampling sites (either BMF, UMR 106-01 cell layers, or bone tissues from day 10 marrow ablation model), and the results were then averaged prior to graphical display. Statistical treatment of the data and data processing were carried out either with program Origin 6.0 or with SigmaStat for Windows version 3.11. The Raman instrument is equipped with a confocal microscope, which allows taking micrographs in situ to identify the exact positions where the spectra were recorded. The instrument was calibrated using a silicon standard.
Energy-dispersive X-ray Spectrometry-Following Raman analyses, UMR-106 cell cultures grown on glass slides were removed by scraping with a razor blade, and the resultant powder mounted for spectroscopic elemental analysis. All observations were carried out using a IMIX-PC analyzer and Prism detector (Bruker AXS, Ewing, NJ) and field emission scanning electron microscope FEG XL30 (FEI Co., Hillsboro, OR). Spectra were acquired for 30 s at an accelerating voltage of 15 kV and spot size of 5. Quantification was performed utilizing a synthetic ceramic hydroxyapatite standard and ZAF factors for matrix correction (33).

MAA Lectin Defines Sites of Initial Mineral Nucleation in
Culture and in Day 10 MA Primary Bone-Biomineralization foci are extracellular supramolecular protein-and vesicle-containing complexes that are the sites of mineral crystal nucleation within UMR 106 osteoblastic cultures. As shown in Fig. 1, A, D, and G, BMF are morphologically distinguishable under bright field illumination as spherical 10 -25-m diameter extracellular structures localized to the apical surfaces of adherent UMR 106-01 osteoblastic cells. We have previously shown that Maackia amurensis lectin binds specifically to BSP and FIGURE 1. MAA lectin identifies sites where initial mineral crystal formation will occur in UMR 106-01 osteoblastic cultures and in developing primary bone. A and B, bright field and MAA lectin-stained images of the same field from UMR culture at 64 h after plating (prior to addition of BGP). Arrows denote pre-BMF complexes. C, 7-day-old primary bone from marrow ablation model stained with Alizarin red S dye. Arrow demarks mineralized trabeculae. D and E, bright field and MAA lectin-stained images of the same field from UMR culture at 76 h after plating. Arrows denote BMF complexes. F, alizarin red S stained 7-day-old primary bone section after decalcification (same field as in C). G, un-mineralized BMF in 88-h UMR culture without BGP (Alizarin red-stained). Arrows denote pre-BMF. H, mineralized BMF at 88 h after Alizarin red staining. Arrows demark mineralized BMF complexes. I, 7-day-old primary bone from marrow ablation model was decalcified as shown in F and then stained with fluorescent MAA lectin. Arrow demarks demineralized trabeculae. J, scanning electron microscopic overhead view of 82-h culture with white arrow demarking one BMF within a cluster. K, backscatter electron microscopic view of field shown in J shows the small particulate nature of the mineral deposits with BMF at 82 h. White arrow identifies BMF containing smaller dense spherical mineral deposits. BAG-75 in Western blots of cell layer extracts of mineralized UMR 106 osteoblastic cells (10). MAA lectin recognizes oligosaccharide chains containing sialic acid conjugates (␣-sialyl-[233]-lactose), and bone matrix glycoproteins BSP and BAG-75 are biomarkers for biomineralization foci. In particular, BAG-75-stained BMF are detectable prior to mineralization of UMR 106 cultures, whereas binding of BSP to pre-BMF peaks at 8 h following addition of BGP to competent cultures (8,18). Based on this rationale, we show here for the first time that precursor BMF structures can also be stained with MAA lectin coupled to Texas red dye (Fig. 1, A and B (arrows)). Importantly, UMR 106 cells associated with these BMF complexes do not stain as prominently with MAA lectin (Fig. 1B, arrows). As predicted from previous work (8,10), formation of Alizarin red S-stained mineral crystals occurs exclusively within BMF structures (compare Fig. 1, E versus H). BMF maintain a distinctive biochemical composition over the 24-h mineralization time period as evidenced by their specific staining with MAA lectin (Fig. 1, B and E, arrows). Note the strict one to one correspondence between the BMF complexes visible under bright field illumination and those identified by fluorescent lectin staining. Finally, backscatter electron images of 82-h cultures made with the scanning electron microscope show that mineralized BMF contain small, particulate deposits of less than 0.5-m in diameter ( Fig. 1, J and K).
In an analogous way, day 10 MA primary bone from the marrow ablation fracture healing model appears to mimic the process in osteoblastic cultures. Briefly, as shown in Fig. 1C, bony trabeculae are readily detected using Alizarin red S dye. Importantly, when this mineral (and stain) is removed by decalcification with EDTA, residual MAA lectin reactivity is found to reside solely in the BAG-75-enriched trabeculae (9,34), not the intervening soft mesenchyme (Fig. 1I). A control bright field image demonstrates that decalcification was complete as evidenced by the absence of Alizarin red stain (Fig. 1F). Taken together, our results demonstrate that un-mineralized pre-BMF have a distinctive biochemical composition and morphological appearance and can be readily identified in UMR cultures prior to and after deposition of mineral crystals. These focal sites of mineral nucleation and propagation display a close biochemical correspondence to newly mineralized trabeculae in day 10 MA primary bone in that both are enriched in phosphoglycoproteins reactive with MAA lectin. Fig. 1 demonstrate that nucleation and propagation of mineral crystals in UMR osteoblastic cultures occurs within pre-existing extracellular supramolecular complexes reactive with MAA lectin termed biomineralization foci. To further validate this correspondence, we isolated mineralized Alizarin red-stained BMF by laser capture microscopy (10). Extracts of purified BMF were then subjected to SDS-PAGE followed by protein staining or by immunoblotting with digoxygenin-conjugated MAA lectin. For comparison, the total cell layer fractions of mineralized and non-mineralized UMR cultures, containing both BMF and cells, were also extracted similarly and are denoted as ϩCL and ϪCL, respectively. Fig. 2 depicts the comparative results of general protein staining and MAA-lectin labeling for BMF and total control cell layer extracts. It is apparent that the laser microscope captured BMF preparation displays a distinctive protein composition evident after Sypro Ruby staining. Interestingly, a single MAA-reactive band of about 85 kDa is dramatically enriched in the BMF sample as compared with the total cell layer fractions from mineralized and non-mineralized cultures (Fig. 2). Although not quantitative, this comparison indicates that the MAA-reactive glycoprotein is preferentially enriched in mineralized BMF and represents an independent confirmation for the co-localization of MAA lectin reactivity with the initial sites of mineral deposition. This outcome is very similar to the enrichment also observed with phosphorylated glycoprotein BAG-75 in isolated BMF (8,10). Given the similar appearance and size of the BAG-75 and BSP bands to that of the MAA lectin-reactive band, it is very likely that they are all identical. Because fluorescent MAA lectin staining not only identifies sites of initial mineral crystal formation, but also accurately predicts these locations before mineral deposition ( Fig. 1), the mechanism for MAA lectin-reactive glycoprotein localization cannot be due solely to an affinity for hydroxyapatite.

Confocal Laser Raman Microscopic Scans of Biomineralization Foci Reveal Temporal Changes in Protein-and Mineral-
derived Signals-Individual biomineralization foci within monolayer UMR osteoblastic cultures were subjected to confocal laser Raman spectroscopic analysis to evaluate the kinetics of mineral crystal deposition and the chemical signatures localized to these sites (Fig. 3). Specifically, cultures were grown on glass slides for 0 -24 h after addition of BGP. At specified times, cultures were removed from dishes, rinsed quickly with ammonium bicarbonate (pH 7.8), and air-dried immediately. Serumdepleted conditions were used to avoid the presence of serum proteins. As shown in Fig. 3A (arrows), BMF are readily identifiable under bright field illumination both before (64, 70, 73, Alizarin red-stained BMF were isolated from mineralized 88-h UMR cultures by laser capture microscopy as described previously (10). BMF were then decalcified, and the proteins were solubilized prior to SDS-PAGE under reducing conditions. The same amount of protein (6.5 g) was applied to each lane. Left panel, image after electroblotting onto polyvinylidene difluoride membrane and luminescent detection with digoxygenin-conjugated MAA lectin staining (see under "Methods"). Right panel, gel image after staining with Sypro Ruby dye (reprinted here by permission from the American Society for Biochemistry and Molecular Biology). Molecular weight estimates refer to blue pre-stained standards co-electrophoresed on the same gel. Results depicted are representative of two separate BMF preparations. Buffer, extraction buffer alone; BMF, represents proteins extracted from laser microscope purified BMF; ؉CL, total cell layer extract of BGP-treated cultures at 88 h after plating; ϪCL, total cell layer extract of cultures not treated with BGP at 88 h after plating. and 76 h) and after mineralization (82 and 88 h). In this model, mineral crystals form exclusively within BMF complexes and appear as focal white deposits in the 82-and 88-h cultures. The ϳ1-m diameter laser beam used to interrogate the samples is also visible within the un-mineralized BMF (Fig. 3A, light blue dot). The normalized average of 10 separate spectral scans of individual BMF at different times of mineralization are depicted in Fig. 3B. Interestingly, individual spectra from the 10 different BMF sampled/time point were found to exhibit close overall structural similarity as evidenced by the low noise levels in averaged spectral line tracings (Fig. 3B). Importantly, variation among BMF replicates on the same culture slide was noticeably less than that between BMF from different time points during the mineralization period. As a result, the data indicate that the process of BMF mineralization in UMR 106-01 cultures is temporally synchronized.
Spectral scans of BMF were found to sort into roughly two different spectral profiles that correlated with time of mineralization, e.g. 64 -76 and 82-88 h. Unexpectedly, comparison of individual scans from 64 to 88 h shows dramatic changes in protein-derived peaks (1000 -1800 cm Ϫ1 ) during the period preceding, during, and after initial mineral crystal formation within BMF (Fig. 3B). It is known from previous work (18) with this model that BSP localizes to BMF during the mineralization period reaching a maximum around 72-74 h. Also Huffman et al. (10) showed that BAG-75 and BSP are cleaved during the process of mineralization within BMF, another change in composition. Therefore, the composition of BMF complexes and its Raman spectrum would be expected to change with time. Formation of hydroxyapatite mineral crystals within BMF complexes is readily detected after 76 h based on the assignment of the 957-962 cm Ϫ1 band to the v 1 symmetric stretch signal for PO 4 Ϫ3 (35). Such changes are more evident when first normalized relative to the signal at 1450 cm Ϫ1 (N-H bend; methyl deformation; CH 2 -wag) (Fig. 3C), which was relatively invariant over the experimental time course. When plotted in this way, it appears that the 1004 and 1660 cm Ϫ1 Raman signals first maximize at about 70 h and then decrease to a minimum at 82 and 88 h (p ϭ 0.05 compared with 64-h non-mineralized control). Interestingly, the mineral-derived phosphate signal near 960 cm Ϫ1 exhibits a complementary inverse relationship with that for the 1004 band (assigned to phenylalanine) and 1660 cm Ϫ1 peaks (assigned to amide I, COCON stretch) (Fig. 3C). Finally, the band centered at 1340 cm Ϫ1 also decreases slightly over the mineralization period (p Ͻ 0.05 at 82 and 88 h compared with 64 h non-mineralized control); this signal has been assigned to protein ␣-helices where its intensity is sensitive to molecular orientation (36). Although the identities of protein substituents contributing to the 1004, 1340, and 1660 cm Ϫ1 peaks are not yet known, the temporal correlations of these transitions imply these groups may play a direct role in mineral crystal nucleation and propagation within BMF. However, not all Raman spectral bands, e.g. 1243 cm Ϫ1 (assigned to amide III, CONOH stretch band) or 1307 cm Ϫ1 , change substantially during the mineralization period (Fig. 3C).
In view of the unusual nature of our findings, we attempted to rule out an artifactual explanation with several additional controls. First, direct Raman spectral analyses of UMR 106 cells that underlie or are located adjacent to BMF demonstrated an absence of hydroxyapatite spectrally because no peak was evident at 957-960 cm Ϫ1 (Fig. 4B). Furthermore, confocal microscopy failed to detect hydroxyapatite crystals (white deposits) within osteoblastic cells of the cell layer (Fig. 4A). Importantly, the shape and pattern of protein and lipid signals in the range of 1100 -1700 cm Ϫ1 in 64 -88-h cultures is invariant, in contrast to that for BMF spectral scans (compare Fig. 4, B and C, with Fig. 3, B and C). Like the scans depicted in Fig. 3, each spectral tracing is the average of 10 different cellular sites, and the excellent agreement between different sites sampled is reflected by the low level of noise or fluctuation exhibited by the tracing line (Fig. 4B). However, in contrast to results with BMF, the normalized signals at 1004 and 1656 cm Ϫ1 did not change significantly over the 24-h mineralization period (p Ͼ 0.50 compared with 64-h non-mineralization control) (Fig. 4C). A major peak at 1100 cm Ϫ1 is evident in the scans of UMR 106 cells but is not present in BMF (compare with spectra in Figs. 3B and 4B). Interestingly this signal was also found to be a prominent characteristic of control glass slides (without cells) that had been incubated with BSA-containing media. Taken together, these control studies confirm that mineralization in UMR 106 cultures is an extracellular event because phosphate signals derived from mineral crystals only occur within BMF, not intracellularly in osteoblastic cells (compare Figs. 3 and 4). Also, temporal changes observed in 1004, 1340, and 1660 cm Ϫ1 bands in BMF spectral scans did not occur in the underlying cell layer.
As a second control, blank glass slides were incubated in BSA-containing culture media alone for 88 h and then rinsed and air-dried. Importantly, these slides gave rise to Raman spectra devoid of all of the major peaks identified in mineralizing cultures. For example, the 1004 cm Ϫ1 peak observed to vary with time of mineralization in BMF spectra was absent from control glass slides (not shown). These results demonstrate that the protein-and hydroxyapatite-dependent Raman signals Averaged signals at 1004 and 1656 cm Ϫ1 depicted in B were normalized relative to an invariant signal at 1450 cm Ϫ1 and were plotted as a function of time of mineralization. Data were analyzed by a one-way ANOVA using multiple comparisons versus the 64-h control group with application of the Holm-Sidak method. There was not a statistically significant difference (p Ͼ 0.500). Error bars represent means Ϯ S.E.
require the presence of osteoblastic cells and are not present with glass slides exposed to BSA-containing media alone.
Third, results in Fig. 3 raise the question whether Raman spectral scans for mineralized BMF are dependent upon the position sampled within these complexes. To address this, Raman spectra were taken at five different positions within a single representative BMF structure chosen randomly for the separate mineralization time points. Resultant spectral data from the 64-, 73-, and 88-h time points are plotted in Fig. 5. Comparison of Raman spectral profiles in Fig. 3, which represent the normalized average of 10 individual BMF from each culture time point, with those for different positions within a single BMF reveal that the same basic scan profile and spectral changes, particularly those at 957-960, 1004, and 1660 cm Ϫ1 , are observed qualitatively in the two cases. Temporal changes occurring in the 957-960 and 1660 cm Ϫ1 bands were statistically significant at 88 h (p ϭ 0.05 compared with 64-h nonmineralized control). Thus it does not make a substantial difference what place within the BMF was scanned; the same basic spectrum was obtained.
Temporal Changes in Raman Spectral Profiles Are Blocked by Serine Protease Inhibitor AEBSF-AEBSF-treated cultures represent an excellent spectroscopic control because all steps and procedures are the same in the two experiments except for the addition of inhibitor at 64 h. BMF complexes are visible at 64 h ( Fig. 1) and can be scanned throughout the culture period even in the presence of a minimum effective inhibitory dosage of AEBSF. As shown in Fig. 6, AEBSF-treated cultures display largely invariant spectra at all time points sampled from 64 to 88 h. Each Raman spectral profile represents the average of 10 separate scans of individual BMF. As illustrated both by the spectral tracings and line plots of peak heights (Fig. 6, A and B), all AEBSF-treated BMF display spectra similar to that for 64-h BMF in mineralizing cultures (Fig. 3, B and C). Importantly, not only is the hydroxyapatite peak at 957-960 cm Ϫ1 largely absent from AEBSF-treated 76-and 88-h cultures, but tem-poral protein-dependent changes preceding mineralization at 1004 and 1660 cm Ϫ1 are also missing. These results suggest that shifts in Raman protein signals between 64 and 76 h are dependent, either directly or indirectly, upon a functional serine protease. Close inspection suggests that the mineral content of inhibitor-treated BMF does still increase slightly, albeit only 10 -15% of that observed without inhibitor (compare Fig. 6B and Fig. 3C) (p Ͻ 0.05 at 76 and 88 h compared with 64-h non-mineralized control). Also, a band centered at 1340 cm Ϫ1 assigned to protein ␣-helices was found to increase over the mineralization period in AEBSF-treated BMF (p ϭ 0.05 at 88 h compared with 64-h non-mineralized control). Interestingly, this orientation-sensitive signal decreased when mineralization was allowed to proceed (compare Fig. 6B and Fig. 3C).

Substitution of Sodium Phosphate for ␤-Glycerol Phosphate Leads to Faster Production of Mineral Crystals within BMF
Complexes-Because mineralization in UMR cultures can be blocked by levamisole (16), it is assumed that alkaline phospha-  tase is required to first cleave ␤-glycerol phosphate, releasing inorganic phosphate ions, before mineralization can proceed within BMF. In view of the temporal synchronization observed with ␤-glycerol phosphate-supplemented UMR 106-01 cultures (Fig. 3, A-C), we asked whether the mineralization process would be altered if sodium phosphate was substituted for the organic ␤-glycerol phosphate. Because preliminary studies have shown that breakdown of BGP is ϳ80% complete within mineralizing UMR 106 cultures at 88 h (data not shown), 5.0 mM sodium phosphate was added to cultures instead of 6.5 mM BGP. Equivalent amounts of mineral crystals were deposited within BMF when either 5 mM sodium phosphate or 6.5 mM ␤-glycerol phosphate was used (Fig.  7A). Similarly, mineralization occurring with either supplemental phosphate source was effectively blocked by serine protease inhibitor AEBSF (Fig. 7A).
Following a similar protocol as used previously, Raman spectral scans were made on 10 different BMF complexes at each different time point over the 24-h mineralization period. Data were normalized relative to the invariant 1450 cm Ϫ1 signal and plotted versus culture time in Fig. 7B. Several features of the data are noteworthy. First, the time at which the rapid phase of mineral crystal production begins advanced from 76 to 73 h with sodium phosphate. Second, the 960/1450 cm Ϫ1 ratio was found to be two to three times larger than that seen with BGPtreated cultures (compare Fig. 3C and Fig. 7B). Third, spectral signal ratios at 1660 and at 1004 cm Ϫ1 displayed qualitatively similar temporal profiles with either supplemental phosphate source (compare Fig. 3C and Fig. 7B). The 1004 cm Ϫ1 signal displayed the most reproducibility, exhibiting a significant decrease between 70 and 82 h with either phosphate source. However, it is noteworthy that the timing of this drop in 1004/ 1450 cm Ϫ1 signal occurred 3-6 h earlier in sodium phosphatesupplemented cultures versus ␤-glycerol phosphate cultures (compare Fig. 3C versus Fig. 7B). In this way, the decrease in 1004 cm Ϫ1 signal consistently preceded the dramatic rise in BMF mineral content despite a temporal shift in the onset of mineralization. Fourth, the mineralization process occurring within different BMF complexes in sodium phosphate supplemented cultures is also temporally synchronized.
After Raman spectral analysis, cell layers were carefully scraped from glass culture slides, and the resultant powder was subjected to energy-dispersive x-ray analysis for calcium and phosphorus (  (8 -10). To test this hypothesis, primary bone was isolated from the intramedullary cavity of marrowablated rats (29). Day 10 MA primary bone contains new mineralized bony trabeculae separated by un-mineralized areas resembling soft marrow tissue. Resultant Raman spectral scans of the mineralized and un-mineralized areas of day 10 MA pri- FIGURE 7. Comparison of mineralization occurring within BMF with supplemental sodium phosphate versus ␤-glycerol phosphate. A, mineralization with 5 mM sodium phosphate is equivalent to that with 6.5 mM BGP, and both processes are blocked by serine protease inhibitor AEBSF. UMR106-01 cultures were grown according to the mineralization protocol (see under "Methods") except that in some cultures 5 mM sodium phosphate was substituted for 6.5 mM ␤-glycerol phosphate. AEBSF (0.1 mM) was also added to some cultures at 64 h. Control cultures without supplemental phosphate or BGP yielded background levels of Alizarin red S staining and are indicated by the dashed line. Cultures were stopped at 88 h, and mineralization was quantitated using an Alizarin red dye binding assay (16). Data were analyzed by a one-way ANOVA using multiple comparisons. Asterisks for AEBSF-treated cultures represent values significantly different from their respective controls (p Ͻ 0.00002); the amount of mineral present in sodium phosphate versus BGF-supplemented cultures was not significantly different (p ϭ 0.188). B, substitution of sodium phosphate for BGP leads to more rapid mineralization of BMF, but yields similar temporally dependent Raman spectral changes. Please note that only the 960 cm Ϫ1 data are plotted relative to the right axis, whereas all other Raman ratios are plotted relative to the axis on the left. UMR106-01 cultures were grown on glass slides as described under "Methods," and 5 mM sodium phosphate was added to the culture medium at 64 h. mary bone were compared with the Raman spectra for 88-h mineralized BMF (Fig. 8). However, the identity of the un-mineralized regions of day 10 MA primary bone scan cannot be assigned with assurance in these unstained sections, e.g. marrow, osteoid, and blood clot. Un-mineralized regions of day 10 MA primary bone were missing the 1004 cm Ϫ1 band and mineral signal at 960 cm Ϫ1 , both of which are evident in spectra from mineralized day 10 MA primary bone trabeculae and in 88-h BMF. Some minor distinctions are apparent when comparing the two mineralized samples. For example, the shape of the major peak ranging from 1050 to 1150 cm Ϫ1 is different in 88-h mineralized cultures versus day 10 MA primary bone trabeculae. Although the shape of the band ranging from 1200 to 1375 cm Ϫ1 is similar for both mineralized samples, that for day 10 MA primary bone is shifted by about 50 cm Ϫ1 toward lower wave numbers (Fig. 8). Table 2 provides a comparison of key parameters describing the hydroxyapatite mineral crystals present in day 10 MA primary bone with that for BMF in UMR cultures. Because Raman spectroscopy has proven to be a useful tool in evaluating mineral crystal structure/composition, the data support a structural similarity between hydroxyapatite crystals deposited within BMF and that formed developmentally in new primary bone. Spectral peak position identifies the classification of a crystal attribute, whereas the peak width (full width at halfmaximum, FWHM) reflects the relative crystallinity. Spectral features of minerals from different sources are listed in Table 2, including chemically synthesized hydroxyapatite, 70-, 73-, and 88-h UMR 106 cultures grown in sodium phosphate or BGP, and primary day 10 MA bone. Chemically synthesized hydroxyapatite was used here as a common standard to estimate the relative crystallinity of the different mineral sources. Relative crystallinity was based on values for the FWHM between chemical standard and the biological mineral. The most immature mineral detected in UMR cultures, that exhibiting the highest FWHM, is at 73 h in BGP-supplemented cultures where the peak position at 957 cm Ϫ1 suggests a transition state of octacalcium phosphate or, alternatively, small hydroxyapatite crystals exhibiting a high degree of imperfection and substitution. By comparison, the mineral components in both 88-h UMR 106 cultures and day 10 MA primary bone gave identical peak position and lower FWHM (Table 2). When taken together with similarities in Raman peak positions noted earlier, these latter results support use of the UMR culture model to simulate primary bone formation.

DISCUSSION
The UMR 106-01 osteoblastic cell culture model represents a model of primary bone mineralization (8,9). We show here that initial mineral crystal deposition occurs within 10 -25-m diameter extracellular BMF structures enriched in an 85-kDa glycoprotein reactive with MAA lectin. Mineral crystal formation occurs within ϳ9 -12 h following the addition of an exogenous phosphate source to mineralization competent cells. Interestingly, mineralization occurred earlier when sodium phosphate was substituted for ␤-glycerol phosphate. To better characterize the process of nucleation, confocal laser Raman microspectroscopic scans were made on individual biomineralization foci at different times preceding (64, 70, 73 h) and following the appearance of mineral crystals (76, 82, 88 h). Importantly, the range of variation between spectra for different randomly selected BMF from the same culture dish was found to be small (ϳ5-10%). In contrast, statistically significant differences were observed for BMF bands at 960 (hydroxyapa- FIGURE 8. Mineralized BMF and trabeculae from day 10 MA primary bone exhibit similar Raman spectral profiles. UMR 106 cultures were grown as described under "Methods." Primary bone was harvested on day 10 from the intramedullary cavity of marrow ablated tibiae; 7-m-thick frozen sections (n ϭ 6) of primary bone were selected for Raman spectroscopic analysis. Separate mineralized areas and un-mineralized areas were then scanned, averaged, and spectra normalized as noted previously and compared directly with that for 88-h BMF (see Fig. 3B). Spectra are offset to improve clarity and are plotted on the same scale.  tite), 1004 (phenylalanine), 1340 (␣-helix), and 1660 cm Ϫ1 (amide I) when averaged and normalized spectra at different times were compared. In particular, the protein-dependent spectral signal at 1004 cm Ϫ1 decreased over the 9 -12 h preceding the appearance of hydroxyapatite crystals detected visually and via the v 1 -phosphate stretching peak at 959 -960 cm Ϫ1 . It is noteworthy that Penel et al. (37) also consistently detected a similar 1004 cm Ϫ1 Phe signal in developing rabbit bone when viewed by intravital Raman microspectroscopy. When mineralization in UMR 106 cultures was blocked by addition of a minimum dose of serine protease inhibitor AEBSF, the aforementioned protein-dependent temporal changes were largely blocked. Instead, the average BMF spectrum remained largely unchanged over the 24-h mineralization period. Mineral crystals produced initially at 70 -73 h displayed a higher FWHM (lower relative crystallinity) compared with that at 88 h, the latter value of which was consistent with day 10 MA primary bone hydroxyapatite. In summary, confocal laser Raman spectroscopy demonstrates for the first time that nucleation and propagation of hydroxyapatite crystals within individual extracellular BMF complexes is a multistep process involving both protein and mineral crystal transitions. Second, changes in protein-derived signals at 1004 cm Ϫ1 , as well as possibly at 1340 and 1660 cm Ϫ1 , reflect events within BMFs that are associated with and precede mineral crystal formation because they are blocked by AEBSF, an inhibitor of mineralization (10). Finally, the low degree of spectral noise or variability observed among different BMF at the same time points indicates that mineralization of individual BMF is temporally synchronized within UMR 106 cultures.
We have proposed that UMR 106-01 osteoblastic cells represent a model of primary bone formation (8,9) and have used several features of this cell line to experimentally probe the mechanism of mineralization. UMR cells attain a competency to mineralize 60 h after plating (16,18). After addition of an exogenous phosphate source at 64 h after plating, the first hydroxyapatite crystals are formed about 9 -12 h later at specific sites we have termed BMF. In contrast, primary calvarial osteoblastic cultures take 12 or more days to accomplish the same outcome. Biomineralization foci are 10 -25-m diameter extracellular, vesicle-enriched complexes found in primary fetal osteoblastic cell cultures, common osteoblastic cell lines, and in primary bone in vivo (8,9). Although BMF complexes are the sites in which the first mineral crystals of hydroxyapatite are deposited, these macromolecular structures can be readily detected prior to mineralization by virtue of their distinctive spherical appearance, apical location, and enrichment in biomarker protein bone acidic glycoprotein-75. The distinctive biochemical composition of pre-or un-mineralized BMF complexes was also shown here by their preferential reactivity with MAAlectin. In view of this ability to identify BMF complexes in UMR 106-01 cultures before and after mineralization, we undertook a laser Raman confocal microscopic analysis of protein and mineral changes during the 24 h after addition of supplemental phosphate to cells.
Prior proteomic analyses on BMF also used cells grown on glass slides (10). In those studies, a minimum of 500 BMF, puri-fied by laser capture microscopy, was pooled to carry out combined mass spectral peptide mapping and Western blotting (10). In contrast, the sensitivity of Raman microspectroscopy and the small beam size (1 m) of the laser Raman probe permitted spectral scans to be made on individual BMF yielding new information on the temporal progression of biochemical changes occurring within. When BMF from 64-, 70-, 73-, 76-, 82-, and 88-h cultures were analyzed, the range of variation between replicates at the same time point was quite small. Importantly, the extent of variation among 10 replicate BMFs at each time point was less than observed between spectra at different time points. The same pattern of spectral changes was consistently and reproducibly detected in culture studies performed by two different investigators over 20 different passages of UMR 106-01 cells over a 24-month period.
Regardless of phosphate source, the data document a progressive pattern of spectral changes arising from BMF proteins over the 9 -12 h preceding deposition of hydroxyapatite crystals within these structures. In particular, the inflection point for a drop in the 1004/1450 cm Ϫ1 ratio shifted ϳ6 h earlier in sodium phosphate-supplemented cultures where mineralization begins at 76 h versus ␤-glycerol phosphate-supplemented cultures where the first mineral crystals are detected at 82 h. Furthermore, the shift in intensity of the 1004/1450 cm Ϫ1 ratio did not occur in cultures where mineralization was blocked by AEBSF or within UMR 106 cells associated with mineralizing BMF. Based on these findings, we speculate that the 1004 cm Ϫ1 spectral change detects a protein change within BMF required for mineralization. Although the identity of individual protein(s) contributing to this altered phenylalanine Raman signal is not known, BSP represents a potential candidate based on its properties. It is recruited to pre-BMF complexes just prior to mineralization (18); it is fragmented within BMF complexes (10) by an AEBSF-sensitive protease, and it has been proposed to function as a nucleator of mineralization (14,15). Additional studies will be necessary to identify the protein(s) responsible for the spectral changes occurring before and during mineral nucleation within BMF.
The term crystallinity refers to the order of a solid where a highly crystalline material displays long range order among its component atoms, and amorphous materials, i.e. glasses, exhibit short range order only. The full width of the Raman hydroxyapatite v 1 -phosphate stretching signal at half-height (FWHM) is used as a measure of the relative crystallinity or structural disorder of a solid (38), where a smaller value reflects increased crystallinity. The extent of crystallinity can vary in two ways, the order of the component atoms can become more or less perfect, and the overall size of individual crystals can change. Neither Raman nor IR spectroscopy distinguish these different possibilities or provide a readout convertible in units of crystallite size, yet the FWHM provides a useful estimate of crystalline order (Table 2). Specifically, initially detected mineral crystals at 70 -77 h displayed a significantly larger FWHM than that at 82-88 h, where the rate of mineralization appeared to plateau. Importantly, we have shown earlier by x-ray diffraction that 88-h BMF contain similarly sized hydroxyapatite crystals regardless of whether produced in sodium phosphate or BGP-supplemental UMR 106 cultures (16).
We have used the marrow ablation model to produce primary bone within the intramedullary cavities of rat tibiae (29,30). An advantage of this model is that primary bone forms throughout the intramedullary cavity in the absence of cartilage. As a result, segments of primary bone of 2-5 mm in diameter and up to a centimeter in length may be dissected free of the bony cortex for analysis. Comparison of the mineral phases present within BMF complexes with that in mineralized trabeculae of day 10 MA primary bone revealed several key similarities. Overall, spectral profiles exhibit a general similarity in band shape and intensity reflecting a similar overall composition. In contrast, Raman spectral profiles for un-mineralized areas were noticeably different; however, identification, e.g. marrow, osteoid, or blood clot, is problematic in unstained sections. Positions of the v 1 -phosphate stretch peaks in day 10 MA primary bone and that for 88-h BMF are both 959 cm Ϫ1 , which is very close to that for a hydroxyapatite standard (960 cm Ϫ1 ) and published values for the v 1 -phosphate shift of 957-959 cm Ϫ1 (35,39). Also, the v 1 -phosphate shift for both day 10 MA primary bone and 88-h BMF gave the same FWHM. Based on their FWHM, the different crystalline samples can be ordered as follows: e.g. hydroxyapatite chemical standard Ͼ88 h BMF and 10-day MA primary bone Ͼ70 h BMF (in sodium phosphate) Ͼ73 h BMF (in ␤-glycerol phosphate). The 957 cm Ϫ1 v 1 -phosphate shift position for 73-h BMF (in ␤-glycerol phosphate) is consistent with either a very small, poorly crystalline, highly substituted hydroxyapatite (40) or with a octa-calcium phosphate/amorphous calcium phosphate (41) precursor phase.
An unexpected result of this study was the degree of spectral similarity observed between different BMF complexes at each culture time point. The variation was found to be about 5-10% regardless of when the spectra were taken over the 24-h mineralization period. In the same way, BMF from AEBSF-treated cultures gave remarkably similar spectral scans regardless of when the cultures were stopped, e.g. 64, 76, or 88 h. Taken together, these results suggest that the process of mineralization within BMF is temporally synchronized. Because other osteoblastic cell culture models have not been shown to be synchronized, this property could be due to either the clonal nature of UMR 106-01 cells (28) or the robust and rapid capacity for mineralization of this cell line (42). However, our results raise intriguing questions regarding the process of mineralization within extracellular BMF. First, how do UMR 106-01 cells become synchronized, e.g. what is the signal that triggers the initial mineral crystal nucleation? Interestingly, Beck and coworkers (43)(44)(45) have proposed that extracellular inorganic phosphate ion is "sensed" by osteoblastic cells leading to the induced expression within 12-24 h of key mineralization-related genes. Expression of transcription factor ERG-1 seems to be particularly responsive to phosphate ion (45,46).
The mineralization protocol for UMR 106-01 cells involves an exchange with either 5 mM phosphate or 6.5 mM BGP-supplemented media at 64 h after plating. Plots of the 960/1450 cm Ϫ1 ratio demonstrate the first mineral crystals are formed ϳ9 to ϳ12 h later. Although the Raman 960/1450 cm Ϫ1 ratio was higher in sodium phosphate-treated versus BGP-treated BMF, quantitative calcium and phosphorous analyses indicated that the amount of mineral crystals formed in 82-88-h cultures was similar. This implies that the 1450 cm Ϫ1 signal (N-H bend; methyl deformation; CH 2 -wag) is reduced under these conditions. Because mineralization of UMR 106 cultures occurs 3-6 h earlier when cultures are treated with sodium phosphate and because ␤-glycerol phosphate must first be cleaved by alkaline phosphatase to release inorganic phosphate, we propose that exogenously supplied phosphate ion may be responsible for synchronization of the mineralization process within BMF. We speculate that mineral is formed more quickly in sodium phosphate-treated versus BGP-treated cultures because the initially higher media phosphate ion concentration in the latter case serves as a better inducer of mineralization-related genes and/or facilitates the nucleation of mineral crystals.
Because mineral crystal formation in UMR 106-01, MC3T3-E1, and primary fetal calvarial cultures occurs within extracellular BMF complexes (8), it is interesting to speculate how synchronized osteoblastic cells could initiate mineralization. In view of the ability of AEBSF to block both mineralization (10) and Raman spectral changes within BMF, we hypothesize that UMR cells may secrete or shed a serine protease that acts as a direct biochemical mediator of the process of mineralization. For example, the in vivo functionality of osteocyte-enriched dentin matrix protein 1 in phosphate regulation largely resides in a 57-kDa proteolytic fragment (47). Prior proteomics work also provides indirect support for such a mechanism in mineralizing UMR cultures. For example, fragments of both BSP and BAG-75 are localized within mineralized BMF, whereas AEBSF blocks this fragmentation and incorporation within BMF (10). BSP has been shown to act as a nucleator in in vitro mineralization assays (14,15); however, the effect of proteolytic cleavage on this property has not been tested directly. Similarly, activation of procollagen C-terminal propeptidase enhancer protein is also blocked by AEBSF (10); the activated enhancer stimulates procollagen processing and fibrillogenesis (48,49). Finally, initial studies demonstrate that extracellular expression of proprotein convertase serine protease SKI-1 (S1P) activity peaks in UMR 106 cultures at 76 h (data not shown), the same time as the first mineral crystals appear within BMF (50 (52) showed that mouse calvarial primary cultures formed mineral giving rise to Raman bands characteristic of a poorly carbonated apatite. Specifically, the cells formed a ␤-tri-calcium phosphate-like mineral as evidenced by a Raman peak at 975 cm Ϫ1 , which is not normally expected during mineral growth. Raman spectral scans of mouse coronal suture cultures were also interpreted as evidence for an octa-calcium phosphate-like transient intermediate phase (41). The octa-calcium phosphate-like phase was characterized based on a Raman band at 955 cm Ϫ1 . As shown here, the mineral present in UMR 106 cultures grown in sodium phosphate or ␤-glycerol phosphate exhibits the characteristics of an immature crystalline, carbonated hydroxyapatite mineral phase, e.g. elevated FWHM and reduced calcium/phosphate ratio.
Type I collagen and non-collagenous extracellular matrix proteins have been implicated as nucleators in bone mineralization. However, Raman microspectroscopic evidence supporting protein structural or conformational changes associated with mineralization is still limited. Bohic et al. (53) studied the osteoinductive effects of leukemia inhibitory factor and oncostatin M on cultures of rat bone marrow stromal cells. Raman spectra of the mineral phase contained a dominant 960 cm Ϫ1 apatitic phosphate peak with an FWHM of about 19 cm Ϫ1 , which is close to the value of 20 cm Ϫ1 found in rat 10 day MA bone mineral and in UMR 106 cell cultures. Interestingly, the major vibrational modes of protein amide I and III bands in the cell layer also changed in response to leukemia inhibitory factor or oncostatin M treatment. In this study of biomineralization foci, time-resolved, progressive changes in the amide I, ␣-helix, and phenylalanine signals were observed to occur before, during, and/or after mineralization in BGP-treated cultures. However, when mineralization was blocked by serine protease inhibitor AEBSF, not only was the mineral phosphate signal largely eliminated but also the temporal changes in protein amide I, and phenylalanine signals were predominantly blocked. In summary, confocal laser Raman microspectroscopy demonstrates directly for the first time that formation of hydroxyapatite crystals within individual extracellular BMF complexes is a multistep process involving both protein and mineral crystal transitions. Finally, the results suggest that mineralizations of individual BMF within UMR 106 cultures are temporally synchronized, a feature facilitating future regulatory studies of the underlying mineralization pathway.