Release of intact and fragmented osteocalcin molecules from bone matrix during bone resorption in vitro.

Osteocalcin detected from serum samples is considered a specific marker of osteoblast activity and bone formation rate. However, osteocalcin embedded in bone matrix must also be released during bone resorption. To understand the contribution of each type of bone cell in circulating osteocalcin levels, we used immunoassays detecting different molecular forms of osteocalcin to monitor bone resorption in vitro. Osteoclasts were obtained from rat long bones and cultured on bovine bone slices using osteocalcin-depleted fetal bovine serum. In addition, human osteoclasts differentiated from peripheral blood mononuclear cells were used. Both rat and human osteoclasts released osteocalcin from bovine bone into medium. The amount of osteocalcin increased in the presence of parathyroid hormone, a stimulator of resorption, and decreased in the presence of bafilomycin A1, an inhibitor of resorption. The amount of osteocalcin in the medium correlated with a well characterized marker of bone resorption, the C-terminal telopeptide of type I collagen (r > 0.9, p < 0.0001). The heterogeneity of released osteocalcin was determined using reverse phase high performance liquid chromatography, and several molecular forms of osteocalcin, including intact molecule, were identified in the culture medium. In conclusion, osteocalcin is released from the bone matrix during bone resorption as intact molecules and fragments. In addition to the conventional use as a marker of bone formation, osteocalcin can be used as a marker of bone resorption in vitro. Furthermore, bone matrix-derived osteocalcin may contribute to circulating osteocalcin levels, suggesting that serum osteocalcin should be considered as a marker of bone turnover rather than bone formation.

Osteocalcin messenger RNA has also been detected in tissues other than bone, but it appears to be processed properly only in the bone microenvironment (4,5). The structure of osteocalcin is characterized by three glutamic acid residues, which undergo a vitamin K-dependent carboxylation. The ␥-carboxyglutamic acid residues (Gla) provide osteocalcin with the ability to bind bone hydroxyapatite with a high affinity (6,7). Osteocalcin is the second most abundant protein in the bone matrix, and it is highly conserved among all vertebrate species (8). The biological function of osteocalcin is probably related to the regulation of bone turnover and/or mineralization (9,10).
The expression of osteocalcin is a marker of late osteoblast differentiation and is induced only after the expression of other osteoblastic markers such as alkaline phosphatase and type I collagen (11,12). Newly synthesized osteocalcin is mostly (60 -90%) adsorbed to the bone hydroxyapatite via the Gla residues, but a part of it leaks into the circulation where it can be detected (13,14). Although osteoblasts synthesize only intact osteocalcin (15), osteocalcin may further undergo intracellular processing or be degraded after secretion, leading to the generation of smaller fragments. Only intact molecules are able to bind to the bone hydroxyapatite, and osteocalcin fragments lose their binding ability probably because of an altered conformation and subsequent loss of affinity for bone mineral (6,16). Circulating osteocalcin has been widely used in clinical investigations as a marker of bone formation (17), whereas protein expression has served as an index of osteoblastic phenotype and bone formation in vitro (11).
Earlier studies have suggested that circulating osteocalcin originates exclusively from biosynthesis in osteoblasts and not from the breakdown of bone matrix (14, 18 -20). However, later studies on patients with different metabolic bone diseases have suggested that not all of osteocalcin fragments are derived from the metabolism of osteocalcin in the circulation or peripheral organs but also from osteocalcin embedded in bone (21)(22)(23). Thus, part of osteocalcin found in the blood may also originate from the resorption process, when osteocalcin embedded in the bone matrix is released during bone degradation (Fig. 1). During bone resorption, osteoclasts secrete protons into the space between the bone surface and cells using vacuolar type H ϩ ATPase, and the acidification results in the dissolution of inorganic mineral. Organic bone matrix degradation is mediated by proteolytic enzymes, primarily cathepsin K, and the released material is endocytosed for further degradation in transcytotic vesicles in the resorbing osteoclast. Eventually, the degraded material is excreted into the extracellular space via a functional secretory domain (24,25). Because osteocalcin is rather susceptible to proteolysis in vitro (16,26), acids and proteases may also attack osteocalcin during bone degradation. Although the secretion of osteocalcin from osteoblasts has been widely studied, the release of osteocalcin molecules from the bone matrix and their potential contribution to the circulating osteocalcin pool has been discussed (27)(28)(29) but not clearly documented. If such a contribution exists, the circulating osteocalcin should rather be considered an indicator of bone turnover and not merely a marker of bone formation.
The purpose of the present study was to investigate whether osteocalcin detectable by immunoassays is released from bone during osteoclastic bone resorption, in addition to osteocalcin synthesized during bone formation. Furthermore, our aim was to evaluate different immunoassays for their capability of detecting bone matrix-derived osteocalcin in osteoclast cultures and to gain insight into the molecular forms of osteocalcin released during bone resorption in vitro.

EXPERIMENTAL PROCEDURES
Reagents-␣-Modified minimum essential medium, fetal bovine serum (FBS), 1 M HEPES solution, and antibiotics (penicillin and streptomycin) were purchased from Invitrogen. Macrophage colony-stimulating factor was purchased from R & D Systems, and the receptor activator of nuclear factor B ligand (RANKL) and tumor necrosis factor ␣ were from Peprotech. Dexamethasone, parathyroid hormone (PTH), bafilomycin A1 (BafA1, an inhibitor for vacuolar type H ϩ ATPase), trans-epoxysuccinyl-L-leucylamido- [4-guanidino]butane (E64, an inhibitor for cysteine proteases), Hoechst 33258, and leukocyte acid phosphatase kit 387-A were purchased from Sigma. Calcium was determined with Calcium Roche/Hitachi reagents from Roche Applied Science. Streptavidin-coated microtitration plates, Assay® buffer, Delfia® wash solution, Delfia® enhancement solution, and Victor 2 Multilabel Counter were from PerkinElmer Life Sciences, and synthetic human osteocalcin (residues 1-49 with Gla at positions 17, 21, and 24) was purchased from Advanced Chemtech (Louisville, KY). The monoclonal antibodies (MAb) for osteocalcin have been described in detail previously (30). Briefly, MAb 3G8 requires the full-length molecule for recognition, MAb 8H12 binds to residues in region 7-19 and MAbs 2H9 and 3H8 have an epitope on the residues spanning positions 20 -43. In addition, MAb 3H8 favors the Gla-containing forms of osteocalcin. The MAbs were raised either against bovine osteocalcin (3G8 and 3H8) or the fusion protein of glutathione S-transferase and human osteocalcin (8H12 and 2H9). The Alexa 647-labeled goat anti-mouse antibody, TRITC-labeled phalloidin, and succimidyl ester of carboxyfluorescein were purchased from Molecular Probes (Eugene, OR).
Osteocalcin Immunoassays-MAb 3G8 or 8H12 were used as biotinylated capture antibodies and MAb 2H9 or 3H8 as europium-labeled tracer antibodies resulting in three different two-site combinations: 3G8/2H9 (I-OC, for intact OC), 8H12/2H9 (M-OC, for the majority of OC), and 8H12/3H8 (T-OC, for total OC). The antibodies were biotinylated with 50-fold molar excesses of biotin-isothiocyanate and labeled with 200-fold molar excesses of europium(III) chelate as described previously (30). Synthetic human osteocalcin 1-49 was used as a calibrator. Samples or calibrators (10 l of each) were added to the wells of streptavidin-coated plates. A mixture containing 100 ng of bio-MAb and 100 ng of Eu-MAb in 50 l of Assay® buffer containing 5 mmol/liter EDTA was added to each well. After 2 h of shaking at room temperature (22°C), the plates were washed six times with Delfia® wash solution, and 200 l of Delfia® enhancement solution was added to each well. After 30 min of shaking, time-resolved fluorescence was measured using the Victor Multilabel Counter. The calibration curve covered a range from 0.4 to 59 ng/ml. Analytical detection limits were defined as the concentration corresponding to the mean value ϩ 3 S.D. of 12 determinations of the zero calibrator and were set for the I-OC, M-OC, and T-OC assays as 0.02, 0.06, and 0.50 ng/ml, respectively. The withinassay and between-assay coefficient of variations (CVs) were determined using a control sample prepared from FBS and were found to be less than 10% (n ϭ 12).
Osteoclast Cultures-A mixed rodent bone cell population was cultured on bovine bone slices as described in detail previously (31) and originally introduced by Boyde et al. (32) and Chambers et al. (33). Briefly, osteoclasts were mechanically isolated from the long bones of 1-day-old Sprague-Dawley rats and allowed to attach to devitalized slices of bovine cortical bone (thickness, ϳ0.15 mm) for 30 min, after which the nonadherent cells were washed away. The osteoclasts were cultured on 24-well plates in ␣-modified minimum essential medium (1 ml/well) supplemented with 10% osteocalcin-depleted FBS, 20 mM HEPES, and 100 units/ml penicillin, 100 g/ml streptomycin for 3-5 days at ϩ37°C and 5% CO 2 . Controls consisting either of bone slices alone or mixed bone cell population plated on glass coverslips were included in each experiment. PTH (10 nM), BafA1 (3 nM), and E64 (50 M) were added at the beginning of the culture when indicated, and medium samples (30 -50 l/well) were collected daily and stored at Ϫ20°C until analyzed. Human osteoclasts were induced to differentiate from peripheral blood mononuclear cells as published elsewhere (34). Briefly, mononuclear cells were isolated from human peripheral blood using the Ficoll-Paque TM technique (Amersham Biotech). The cells were washed four times with phosphate-buffered saline (PBS), and 1,000,000 cells/bone slice were allowed to adhere for 2 h. The nonadherent cells were washed away, and the monocytes adhered on the bone were cultured in ␣-modified minimum essential medium supplemented with 10% regular FBS, 20 mM HEPES, antibiotics, 10 ng/ml of macrophage colony-stimulating factor, 20 ng/ml of RANKL, 10 ng/ml of tumor necrosis factor ␣, and 10 Ϫ8 M dexamethasone for 12 days. Half of the medium was replaced with fresh medium containing 2-fold concentrations of cytokines every 4 days. Additionally, to study the release of the inorganic matrix in the absence of osteoclasts, some bovine bone slices were exposed to 0.6 M HCl at ϩ4°C for 24 h.
Osteocalcin-depleted Fetal Bovine Serum-The FBS used in the rat osteoclast cultures was depleted of bovine osteocalcin prior to use. Equal amounts of MAbs 8H12, 2H9, and 3H8 (1 mg of MAb mixture/1 ml of matrix) were coupled to a gel matrix (Affi-Gel 10; Bio-Rad) according to the manufacturer's instructions using sterile reagents. FBS (14 ml) was mixed with the coupled matrix (1 ml) in an end-over-end rotator for 1 h at ϩ4°C and centrifuged for 10 min at 1000 rpm. The supernatant, i.e. osteocalcin-depleted FBS, was collected and stored at Ϫ20°C. The matrix was washed two times with PBS and osteocalcineluted with 0.5 M glycine-HCl, pH 2.5, in an end-over-end rotator for 15 min at ϩ4°C. The matrix was then washed twice with PBS prior to the preparation of the next batch.
Evaluation of Osteoclast Cultures-The osteocalcin immunoassays I-OC, M-OC, and T-OC described above and a competitive osteocalcin enzyme-linked immunosorbent assay (Rat-MID osteocalcin assay; Nordic Biosciences) were used to measure osteocalcin concentration in the medium. The amount of degraded bone matrix was assayed by measuring the C-terminal cross-linked telopeptide of type I collagen (CTX) from the medium (CrossLaps for Culture; Nordic Biosciences), and the activity of tartrate-resistant acid phosphatase isoenzyme 5b (TRACP5b) in the medium was assessed as described previously (35). The osteoclasts were fixed with 3% paraformaldehyde and cells were stained for TRACP enzyme activity with a leukocyte acid phosphatase kit. The nuclei were visualized with Hoechst staining, and the TRACPpositive multinucleated cells (at least 3 nuclei) were counted as osteoclasts. In addition to the medium, the osteocalcin and calcium levels were also determined from the supernatants collected from the HCltreated bone slices.
Immunostainings-The organic bone matrix components were visualized with fluorescein-labeled bone. The bone slices were incubated in a bicarbonate solution (pH 8.3) containing a succimidyl ester of carboxyfluorescein for 2 h with gentle stirring and then washed with PBS before usage. Rat osteoclasts were cultured on labeled bone slices for 48 h, fixed with 3% paraformaldehyde, permeabilized with 0.1% Triton X-100 for 5 min on ice, and washed once with 2% bovine serum albumin FIG. 1. Possible sources of osteocalcin in circulation. The newly synthesized osteocalcin (biosynthetic OC) is partly incorporated into the matrix and partly released into the circulation. Osteocalcin embedded in bone matrix is released during bone resorption (resorptive OC), but it is not precisely known how extensively osteocalcin is degraded during the resorption process. Osteocytes express osteocalcin, but their contribution to the circulating levels of osteocalcin is unknown.
in PBS (BSA-PBS). The anti-osteocalcin MAb 3H8 (1000 ng/bone slice) diluted in 0.5% BSA-PBS was added and incubated for 45 min at room temperature. After washing with BSA-PBS, the cells were incubated with the Alexa 647-conjugated anti-mouse antibody and TRITC-phalloidin in BSA-PBS for 45 min at room temperature. After washing, the cells were evaluated with a Leica TCS-SP confocal laser scanning microscope equipped with an Argon-Krypton laser (Leica Microsystems). Fluorescein-labeled samples were visualized using a 495-530-nm filter, TRITC-labeled samples were visualized using a 580 -630-nm filter, and Alexa 647-labeled samples were visualized using a 660 -740-nm filter. Confocal images of triple staining were acquired by a sequential scanning method, i.e. all channels were scanned separately to avoid overlapping.
Fractionation of Osteocalcin from Cell Culture Media and Bovine Serum-The culture medium (4 ml) collected from the rat osteoclasts cultured for 5 days with 10 nM PTH and the supernatant obtained from the bone slices incubated with HCl at ϩ4°C for 24 h were used for the analysis. Further, osteocalcin was isolated from fetal bovine serum using the Affi-Gel 10 matrix coupled to the MAbs 8H12, 2H9, and 3H8 as described above. The samples (the culture medium or the supernatant as such and the osteocalcin isolated from the serum) were extracted in solid phase extraction cartridges (Sep-Pak Plus C18; Milli-pore) using 40% acetonitrile for elution. The extracted material was fractionated on a Vydac C4 reverse phase high performance liquid chromatography (HPLC) column (2.1 ϫ 150 mm) equipped with a Vydac C4 guard column (both from The Sep/a/ra/tions Group). The solvent gradient used was 2-30% B (0-23 min) (A ϭ 0.1% trifluoroacetic acid/ water and B ϭ 0.08% trifluoroacetic acid/acetonitrile), 30 -60% B (23-45 min), 60 -80% B (45-65 min), 80% B (65-70 min), and 80 -2% B (70 -75 min) with a flow rate of 150 l/minute. Fractions (50 l) were collected during elution, diluted in Assay® buffer and analyzed for osteocalcin with the I-OC, M-OC, and T-OC assays. Fractions collected from the bovine serum fractionation were also analyzed with matrixassisted laser desorption ionization time of flight mass spectrometry (MALDI-MS) and N-terminal sequencing as described previously (36).
Statistical Analysis-Comparison between groups was performed with the nonparametric Wilcoxon's test using the Statistical Analysis System Enterprise Guide 2 program (SAS Institute). Bonferroni adjustment was used in multiple comparisons, and a p value of less than 0.05 was considered statistically significant. Because of normal distribution, one-way analysis of variance was used in the comparison of cultures performed with normal versus osteocalcin-depleted FBS, and Pearson correlation coefficients were used in correlation studies. All of the results are presented as the means Ϯ S.E. can be found partially in the same compartments inside resorbing cells. Rat osteoclasts were stained for osteocalcin after a 2-day culture on fluorescein-labeled bovine bone slices and z sections (A and B) and x-y sections from the nuclear level (C and D) were obtained with a confocal laser scanning microscope. Actin staining was used to visualize cell boundaries (indicated with a dotted line) and to identify osteoclasts characterized by actin rings.

RESULTS
Osteocalcin (T-OC assay) was detected in the culture medium of rat osteoclasts after 2-3 days of culture on bovine bone, and the amount of osteocalcin increased in a time-dependent manner ( Fig. 2A). The concentration of osteocalcin was higher in the osteoclast cultures than in the corresponding controls, and the difference was statistically significant after 3 days of culture (p ϭ 0.0002 for the bone only control and p ϭ 0.0007 for the cells only control, n ϭ 36) and even more pronounced at the end of the culture period (p Ͻ 0.0001 for the bone only and p ϭ 0.0014 for the cells only, n ϭ 30). The release of osteocalcin into the medium was significantly increased when PTH, a known stimulator of bone resorption, was added to the medium, and osteocalcin was almost undetectable when BafA1, a potent inhibitor of the vacuolar type H ϩ ATPase and bone resorption, was included in the culture medium (Fig. 2B). Osteocalcin levels in the BafA1-treated cultures were similar to those of the controls at each time point (p values Ͼ 0.05). In addition to the T-OC assay, osteocalcin was also detected with the I-OC and M-OC assays, and the levels of all detectable forms of the protein were significantly reduced in the presence of BafA1 (I-OC, p ϭ 0.012; M-OC, p Ͻ 0.0001; and T-OC, p Ͻ 0.0001; n ϭ 17; Fig. 3). Furthermore, an increase in osteocalcin in response to PTH was significant for all of the three assays (p values less than 0.0001; data not shown). Stimulation with PTH also re-sulted in the detection of osteocalcin when a competitive osteocalcin enzyme-linked immunosorbent assay was used (212 Ϯ 21 ng/ml, n ϭ 5). However, in the unstimulated cultures, the osteocalcin levels detected with this assay (22.8 Ϯ 9.1 ng/ml, n ϭ 5) did not differ from the reported detection limit (21.1 ng/ml). Osteocalcin was detected inside the resorbing osteoclasts by staining with a monoclonal antibody against a midmolecular epitope (MAb 3H8, epitope in the fragment 20 -43) (Fig. 4, B and D). The bone matrix endocytosed from fluorescein-labeled matrix was also clearly detectable inside the cells (Fig. 4, A and C) and vesicles containing labeled bone partially co-localized with the vesicles positive for osteocalcin.
The osteocalcin detected in the culture medium of rat osteoclasts had a statistically significant positive correlation to the bone resorption rate as measured by CTX. The correlation coefficient for T-OC and CTX was 0.949 (p Ͻ 0.0001, n ϭ 11) at the end of the culture period (day 5) in the cultures performed with osteocalcin-depleted FBS and not treated with stimulators or inhibitors (Fig. 5A). Osteocalcin detected in human osteoclast cultures also had a statistically positive correlation to bone resorption as evaluated by the CTX assay (Fig. 5B). The correlation coefficient at the end of the culture (day 12) was highest for T-OC (r ϭ 0.934, p Ͻ 0.0001, n ϭ 48) but also highly significant for the other osteocalcin assays (I-OC, r ϭ 0.916, p Ͻ 0.0001; M-OC, r ϭ 0.923, p Ͻ 0.0001) and osteocalcin enzymelinked immunosorbent assay (p ϭ 0.902, p Ͻ 0.0001).
The treatment of osteoclast cultures with two inhibitors of bone resorption, BafA1 and E64, resulted in distinct responses in putative bone degradation markers (Fig. 6). The amount of all detectable forms of osteocalcin was significantly reduced in the presence of BafA1 compared with the untreated cultures. In particular, the amount of M-OC and T-OC were decreased to ϳ10% (p Ͻ 0.0001, n ϭ 17) and also I-OC levels reduced to about 40% (p ϭ 0.019, n ϭ 17). In the presence of E64, the amount of osteocalcin was also reduced, but the inhibition was less pronounced. The M-OC and T-OC levels decreased to about trast, the treatment with BafA1 or E64 resulted in a pronounced reduction in the CTX levels to less than 5% (p Ͻ 0.0001 and p ϭ 0.0018, respectively), and the inhibitory effects of both compounds on CTX levels were indistinguishable from each other.
The fractionation of osteocalcin from the osteoclast culture medium resulted in one predominant peak that was detectable by all three assays I-OC, M-OC, and T-OC. Also two minor peaks eluting earlier in fractionation were identified, and neither of them contained intact osteocalcin (I-OC). (Fig. 7A). In contrast, only one single peak was observed when osteocalcin released by the acid treatment of bone was fractionated. This peak was detectable with all three assays, including I-OC. (Fig.  7B). The elution profile of bovine serum osteocalcin was more heterogeneous and consisted of at least four main peaks eluting approximately at 43, 45, 47, and 49 min. The peaks at 43 and 45 min were predominantly positive for T-OC, and the peaks at 47 and 49 min were detectable by all three assays I-OC, M-OC, and T-OC (Fig. 7C). The MALDI-MS analysis of these four peaks revealed several prominent ions with molecular masses between 2713 and 5720 Da. The predominant ions identified with MALDI, the N-terminal sequencing results, and the corresponding bovine osteocalcin fragments are summarized in Table I.
Osteocalcin was efficiently removed from fetal bovine serum by immunoaffinity depletion and collected osteocalcin-depleted FBS contained less than 1% of the original osteocalcin (T-OC in FBS 293 ng/ml and in osteocalcin-depleted FBS 1.1 ng/ml, mean of five separate batches). Bone resorption by rat osteoclasts was not disturbed when regular FBS in the culture medium was replaced by osteocalcin-depleted FBS. After 3 days, TRACP-positive multinucleated cells were detected on bone slices (Fig. 8, A and B), and there was no difference in the number of osteoclasts between these two cultures (p ϭ 0.67, n ϭ 8) (Fig. 8C). Also, the concentrations of TRACP5b and CTX in the medium were similar (p ϭ 0.89, n ϭ 4 and p ϭ 0.92, n ϭ 11, respectively).

DISCUSSION
Since the discovery of osteocalcin in the late 1970s (13,37,38), it has been used as a specific marker of osteoblast activity both in vivo (17,39,40) and in vitro (1,41) because of the restricted expression in the mature cells of osteoblastic lineage (15,42,43). Osteocalcin content in adult bone has been reported to range from 0.28 (human) to 2-2.5 (cow) mg/g of dry bone, and it represents up to 20% of noncollagenous bone matrix proteins (8,44). During bone remodeling, osteocalcin embedded in the matrix is exposed to osteoclastic bone resorption and to the proteolytic microenvironment needed for bone degradation. In vitro, osteocalcin is susceptible to proteolysis and several proteases, e.g. plasmin, trypsin, and cathepsins can cleave human osteocalcin into smaller fragments (16,26). However, because the molecular weight of osteocalcin is small (ϳ6 Da) compared with many other bone matrix proteins, it may not require extensive cleavage for detachment from the matrix in vivo.
We studied the detachment of osteocalcin from the bone matrix during bone resorption in vitro. Osteocalcin was clearly released into the medium during osteoclastic resorption both in the rat and human in vitro models we used. The concentration of osteocalcin increased in the presence of PTH, a known stimulator of bone resorption, and was almost undetectable in the presence of bafilomycin A1, an inhibitor of bone resorption. A minor amount of osteocalcin was detected in medium incubated with bone slices alone, but the level was not increased after the first 24 h. Also, the rat bone cells cultured on glass coverslips did not produce detectable levels of biosynthetic osteocalcin. Osteocalcin was detected with assays measuring either intact osteocalcin exclusively (I-OC) or those also detecting fragments in addition to intact molecules (M-OC and T-OC), and these results were also reproducible by a commercially available enzyme-linked immunosorbent assay. The concentration of osteocalcin in the culture medium demonstrated a strong (r Ͼ 0.9) and highly significant (p Ͻ 0.0001) correlation to the concentration of CTX, which is frequently used to quantify resorption in bone culture supernatants. The correlation between osteocalcin and CTX was observed both in the rat and human osteoclast cultures. Taken together, these results suggest that the osteocalcin detected by the immunoassays is derived from degraded matrix and can thus be used as an index of bone resorption in vitro. The correlation to CTX was nearly similar with all osteocalcin assays, further suggesting that several different molecular forms of osteocalcin can be used in monitoring cultures. We have previously reported that the same immunoassays are suitable in monitoring osteocalcin production in rat primary osteoblasts (45), indicating that the very same osteocalcin immunoassays can be applied to monitor both de novo synthesized osteocalcin in osteoblast cultures, as well as the bone matrix-derived osteocalcin in osteoclast cultures.
In addition to the immunoassays, osteocalcin was demonstrated inside bone-resorbing osteoclasts by immunostaining, and intracellular osteocalcin was predominantly located in the same vesicles as the endocytosed bone matrix. Although the labeling of bone with fluorescein has probably produced small amounts of labeled osteocalcin, which can be responsible for some of the co-localization, this does not invalidate the conclusion that osteocalcin can be detected inside osteoclasts in the very same compartments as the proteins originally present in the bone matrix. In addition to intracellular vesicles, osteocalcin immunostaining was also observed at the bottom of the resorption lacunae but not on intact bone surfaces, suggesting that mineral dissolution was obligatory to expose the binding epitopes. The immunoassay results and microscopic data together support a concept for the release of immunodetectable osteocalcin during bone resorption. In agreement with our results, Kurihara et al. (46) have provided evidence for osteocalcin immunoreactivity in the supernatant of human bone particle-derived osteoclasts cultured on human bone slices  (51,52), all calculations of theoretical molecular masses were performed using the sequence of bovine osteocalcin (average molecular masses) with glutamic acid at positions 17, 21, and 24 instead of ␥-carboxyglutamic acid. b ND, not determined.
with a sandwich assay for the N-terminal epitope of human osteocalcin.
The correlation between CTX and intact osteocalcin indicated that part of the osteocalcin was able to escape complete degradation in osteoclasts and was released from resorption as intact, unfragmented molecules. This was further supported by the HPLC fractionation of osteocalcin from the osteoclast culture medium, because the largest peak in the elution profile was positive for the I-OC assay and thus consisted of intact osteocalcin molecules. Two smaller peaks in the elution profile could not be identified due to the small amount of protein, but on the basis of earlier elution time they most likely consist of osteocalcin fragments. Peaks were eluted approximately at the same time (45 and 47 min) as two fragments of serum osteocalcin. Thus, the peaks in the osteoclast culture medium might contain osteocalcin fragments similar to those detected in se- rum. Because the elution profiles of both osteocalcin isolated from the culture medium and serum had similarities, osteocalcin released during bone resorption may represent a part of osteocalcin found in circulation and could be able to contribute to the circulating pool of both osteocalcin fragments and intact molecules. Osteocalcin is, however, further metabolized in the circulation and peripheral organs such as the kidneys, liver, and lungs (14,47,48), adding more complexity to serum osteocalcin and making it difficult to draw conclusions on the basis of in vitro studies only.
The circulating osteocalcin appeared to be a complex mixture of several molecular forms. One predominant form was identified as intact osteocalcin 1-49 on the basis of both immunoreactivity (I-OC) and size (molecular mass, 5719 Da) and another one as osteocalcin fragment 8 -33 according to mass spectrometric data (molecular mass, 2897 Da). The latter form was recognized by the T-OC assay only. Interestingly, Gundberg et al. (49) have demonstrated that the degradation of osteocalcin by cathepsin K results predominantly in osteocalcin fragments spanning residues 8 -33 and 8 -35. Other forms identified in the bovine serum were osteocalcin fragments 8 -31, 8 -37, and 8 -49. Because cathepsins cleave osteocalcin specifically between residues Gly 7 and Ala 8 , the circulating osteocalcin fragments starting from Ala 8 could putatively all result from the degradation by cathepsins (26,49). Serum may, of course, contain additional shorter osteocalcin fragments, which remain to be characterized. It has been previously shown by Garnero et al. (50), that circulating human osteocalcin consists of intact osteocalcin (one-third), osteocalcin fragments 1-43 (one-third), and smaller, unidentified fragments (one-third). We were unable to demonstrate the presence of fragment 1-43 in bovine serum, but this can be attributed in part to the different methodology, because Garnero et al. utilized immunoassays instead of the purification and mass spectrometric characterization adopted in this study.
The results obtained with the inhibitors of resorption suggest that different molecular forms of osteocalcin might be released during different stages of osteoclastic bone resorption. The inhibition of acidification in the resorption lacunae by BafA1 resulted in a complete inhibition in the release of CTX as well as of all of the molecular forms of osteocalcin. When the proteolytic activity of cathepsins was disturbed with E64, an almost complete inhibition in the release of CTX was evident, indicating that the osteoclasts were unable to degrade the organic bone matrix. However, the inhibition of cathepsins did not have such a profound effect on the release of osteocalcin, suggesting that part of osteocalcin is liberated from the bone matrix via a cathepsin-independent pathway. Moreover, the release of intact osteocalcin was almost unaffected by E64, which suggests that the detachment of intact molecules does not require the activity of cathepsins. The osteocalcin released during the dissolution of mineral consisted almost exclusively of intact molecules (HPLC), thus supporting the concept that the acidification step, and not cathepsin-mediated degradation, is the main source of intact osteocalcin during bone resorption. Thus, when the mineral is dissolved and its component ions released from hydroxyapatite during acidification, osteocalcin bound to calcium via the Gla residues might be released as intact molecules independent of the proteolytic pathways. After demineralization, proteolytic enzymes cleave collagen and other matrix proteins into smaller fragments. At this step, the osteocalcin that has not been mobilized during matrix dissolution will be susceptible to degradation by proteases, especially cathepsins, resulting in the generation of osteocalcin fragments, probably starting predominantly from the residue Ala 8 , with subsequent heterogeneity observed in the resorptive osteocalcin.
It is well documented that osteoblasts produce osteocalcin, which is used as a marker of bone formation. We have demonstrated that osteocalcin is released during bone resorption as well as being produced by osteoblasts during bone formation. A part of osteocalcin escapes the proteolytic degradation during bone resorption and is released both as intact molecules and fragments. The detection of osteocalcin in the culture medium provides a useful tool for monitoring the bone resorption rate in osteoclast cultures. These results also suggest that a fraction of intact osteocalcin and osteocalcin fragments in the circulation may actually be derived from bone resorption, although it is difficult to estimate the relative proportion of serum osteocalcin that would come from each process in a clinical setting. In conclusion, serum osteocalcin should preferentially be considered a bone turnover marker instead of a pure marker of bone formation.