Identification of the Phosphorylated Sites of Metabolically 32 P-Labeled Osteopontin from Cultured Chicken Osteoblasts*

Osteopontin (OPN) is one of the major secretory phosphoproteins in both calcifying and non-calcifying tissues. Evidence has accumulated for the biological im-portance of the phosphoproteins and, in particular, the phosphate groups in bone formation, resorption, and calcification. The precise locations of the phosphate groups in the OPN molecule were determined by metabolically labeling OPN with 32 P in cultured chicken osteoblasts, followed by purification to homogeneity. N-terminal sequencing showed a single sequence of WPVSKRQHAISA, consistent with that deduced from both cDNA, and previous amino acid sequencing of the protein isolated from chicken bone. Three 32 P-labeled peptides were isolated by reverse-phase high performance liquid chromatography of thrombin-digested, 32 P-labeled OPN. The N-terminal sequencing of each of these thrombin fragments gave single sequences as fol-lows: WPVSKSRQHAIS, SHHTHRYHQDHVD, and ASKLRKAARKL, with approximate molecular masses of 5, 30, and 20 kDa. These data demonstrate that 32 P was incorporated throughout the N- to C-terminal sequence of the protein. Thrombin specifically cleaved chicken OPN at two sites: between Arg-22 and Ser-23, which generated the 5-kDa N-terminal end fragment, and an-other between Lys-138 and Ala-139, which generated the 30- and 20-kDa fragments. To further define the exact locations of the phosphorylated

It has been well established that the processes of phosphorylation and dephosphorylation of proteins catalyzed by protein kinases and phosphatases, respectively, play a major role in the initiation, regulation, and termination of a wide range of intracellular biochemical processes with significant functional consequences. Such processes may also play an important role in a wide variety of intercellular mechanisms, including general cell-cell signal transduction of extracellular agonists via specific transmembrane receptors, often causing alterations of intracellular concentrations of cAMP, calcium, inositol polyphosphates, and/or diacylglycerol. These intracellular modulators in turn induce a cascade of biochemical processes by mechanisms that also involve phosphorylation of specific rate-determining enzymes or proteins by protein kinases (1)(2)(3). Phosphoproteins containing o-phosphoserine and o-phosphothreonine are not only found in the intracellular compartments of cells, but have also been identified in extracellular matrices, particularly those of normal vertebrate mineralized tissues and in experimental and human pathologically calcified tissues (4 -16). There is uncertainty as to the biological functions of these proteins and the precise role of the covalently bound phosphate groups per se. The presence of these phosphorylated proteins in all of the normal and pathologically mineralized tissues of vertebrates, and their ultrastructural localization (17)(18)(19)(20)(21)(22)(23), have combined to suggest that one biological function of these phosphoproteins and specifically of the covalently bound phosphate groups is their critical role in the nucleation and growth of inorganic calcium-phosphate crystals (24 -26). This view is also supported by in vitro nucleation experiments, which attempted to simulate the postulated in vivo nucleation substrate of bone tissue by cross-linking the resident phosphoproteins in situ in their native positions and comparing the efficacy (induction or lag time) of this nucleation substrate with samples containing the collagen-phosphoprotein complexes after the covalent phosphate groups were enzymatically cleaved. Not only did the decalcified bone samples containing the collagen-phosphoproteins complexes markedly decrease the nucleation induction time, but this property was lost when the phosphate groups alone were removed enzymatically, leaving the dephosphorylated collagen-phosphoprotein completely intact and, hence, pinpointing the role of the phosphate groups in this in vitro crystal nucleation event (27,28). In addition to their postulated role in calcification, these phosphoproteins have been implicated in many other biological functions (13)(14)(15)(16)29).
Two major glycosylated phosphoproteins in bone and many other mineralized tissues, osteopontin (OPN) 1 and bone sialo-protein (BSP), have been extensively studied in the past decade. The potential roles of these two phosphoproteins in the calcification process have not been clearly defined. Data have been reported demonstrating that these two components can induce or inhibit calcification in several different in vitro models of calcification (30 -33). The results may be due to the experimental techniques and models utilized to assess crystal nucleation and the accumulation of increasing amounts of the solid crystalline phase of calcium phosphate by secondary nucleation and possibly by growth of the crystals. However, their presence in pathologically calcified soft tissue such as breast tumors, urinary stones, and atherosclerotic plagues (13)(14)(15)(16) supports the general hypothesis that they are involved in the deposition of a solid phase of calcium phosphate in normal and pathologically calcified tissues.
OPN is one of the most characterized of the secreted phosphoproteins. It is synthesized in a highly phosphorylated form by osteoblasts (34,35), and can be isolated from the mineralized extracellular matrix of bone tissue (12,26). In contrast, except for milk, it is essentially unphosphorylated in non-mineralized tissues (37,38). Clearly, such differences in the extent of phosphorylation of the same protein when expressed in widely different tissues suggests different functional properties of both the unphosphorylated and phosphorylated protein, and further suggests the presence of tissue-specific kinase(s) that modify this protein. The protein kinase activity found in the bone, which phosphorylates the endogenous proteins of bone, was first observed three decades ago (39). More recently, the presence of membrane-bound casein kinase type II activity was detected in 14-day-old embryonic chicken bone, which phosphorylated endogenous bone phosphoproteins and dephosphorylated casein (40). Further studies using cultured chicken osteoblasts as the source of protein kinase from bone cells led to the conclusion that the microsomal casein kinase II (CKII) type activity of chicken osteoblasts was the predominant enzyme that phosphorylated purified chicken OPN and recombinant mouse OPN (41,42). In these studies assessment of this enzyme with other protein substrates, including casein and synthetic peptide substrates, indicated that, while its activity is very similar to that of casein kinase II, it also possessed unique enzymatic activities that distinguished it from that of classical casein kinase II (41,42). This enzyme was found to be localized in the Golgi apparatus of chicken osteoblasts (43). Furthermore, in vitro studies also identified the major protein kinase responsible for the phosphorylation of both bovine OPN and BSP to be a CKII (44). Other investigators using ROS 17.28 osteosarcoma cells have also isolated a membrane-associated CKII, which phosphorylates the major dentin phosphoprotein, phosphorphoryn (45). This enzyme was also found in the endoplasmic reticulum and Golgi apparatus (46,47). The subcellular location of the CKII to the endoplasmic reticulum and Golgi apparatus is significant mechanistically, since both OPN and BSP are secretory proteins and the enzyme participating in their post-translational modification must encounter them.
Extensive studies have been carried out in this laboratory to isolate, purify, and characterize the protein kinase(s) responsible for the phosphorylation of the secreted extracellular matrix phosphoproteins of bone. Quantitative amounts (mg) of 32 P-secreted osteopontin from cultured chicken osteoblasts were isolated and purified to define both the location of the phosphorylated amino acids in the protein and the amino acid sequences adjacent to the phosphorylated residue. This paper reports on the isolation and purification of 32 P-labeled peptides of OPN and the solid-phase amino acid sequencing of seven peptides, which unequivocally identify the peptide motifs and the specific sites of the phosphorylated OPN.

Methods
Labeling and Isolation of Secreted 32 P-Labeled Osteopontin from Mineralized Cultured Chicken Osteoblasts-Chicken calvarial osteoblasts were isolated by sequential collagenase/trypsin treatment of 17-day-old embryonic calvaria. Cells were initially maintained in minimal essential media supplemented with 10% fetal bovine serum until confluence. They were then put into BGJ b media supplemented with 10% fetal bovine serum, 12.5 g/ml ascorbic acid, and 10 mM ␤-glycerol phosphate. Cultures were grown for 3 weeks in the latter medium to promote extracellular matrix mineralization. To generate 32 P-labeled OPN, mineralized cultures (15 100-mm culture dishes, 15 ϫ 10 6 cells/ dish) were initially incubated with phosphate-free Dulbecco's modified Eagle's medium for 1 h, followed by incubation of Dulbecco's modified Eagle's phosphate-free medium (5 ml/dish) containing 12.5 g/ml ascorbic acid (1 mCi of 32 P/dish) and incubated for 18 h. All protein extraction buffers and wash buffers were ice-cooled to ϳ4°C. The medium was removed, and the mineralized cell layers were washed (three times) with 10 ml of 10 mM phosphate buffer, pH 7.2. The cells were lysed with ice-cooled phosphate buffer (2 ml/dish), pH 7.2, containing 0.5% Tween 20, 0.05% deoxycholate, 50 mM NaCl, 1 mM MgCl 2 , 50 mM NaVO 3 , 10 mM NaF, and 1 mM phenylmethylsulfonyl fluoride for 10 min on ice. The lysis buffer was removed, and the mineralized cell layer was washed (three times) with 10 ml of 10 mM phosphate buffer (pH 7.2)/dish. The mineralized cell layer was then extracted 0.1 M HCl (adjusted to pH ϳ2.0 with citric acid) 10 ml/dish over ice for 3 h. The acid extracted proteins were lyophilized, and 32 P-labeled OPN was purified by reversephase HPLC on a C-4 column as described below.
Purification of Secreted 32 P-Labeled Osteopontin by Reverse-phase HPLC on a C-4 Column-The lyophilized acid extract was resuspended in ϳ0.5 ml of H 2 O, 0.1% trifluoroacetic acid (v/v), and aliquots (0.25 ml) were subjected to reverse-phase HPLC on a C-4 column (15 ϫ 0.38 cm). After injection, the column was washed for 10 min with H 2 O, 0.1% trifluoroacetic acid (v/v) followed by elution using a linear gradient from H 2 O, 0.1% trifluoroacetic acid (v/v) to 60% CH 3 CN, 0.55% trifluoroacetic acid (v/v) at 60 min at a flow rate of 0.72 ml/min. Fractions of 0.72 ml were collected, and absorbance ϭ 230 nm was recorded by a continuous on-line chart recorder/integrator. Aliquots of 0.025 ml from each fraction were counted for 32 P. The radiolabeled peaks were pooled, freezedried, and rechromatographed under same conditions as described above. An aliquot of the purified osteopontin was then N-terminally sequenced.
Thrombin Digestion of Purified 32 P-Labeled Osteopontin and Isolation of Thrombin Fragments by HPLC on a C-4 Column-Approximately 200 g of purified 32 P-labeled OPN was incubated with bovine thrombin (1 unit/4 g of protein) in 0.20 ml of 50 mM Tris buffer, pH 8, containing 2 mM CaCl 2 for 2 h at 37°C. The reaction products were treated with 50 l of H 2 O, 0.1% trifluoroacetic acid (v/v) and HPLCchromatographed on C-4 column (15 ϫ 0.38 cm) as described above. The 32 P-radiolabeled fragments were identified by counting for 32 P, separately pooled, and freeze-dried. Aliquots of each thrombin fragment were N-terminally sequenced.
Trypsin Digestion of Purified 32 P-Labeled Osteopontin and Isolation of the Labeled Peptides by Reverse-phase HPLC on a C-18 Column-The purified 32 P-labeled OPN (ϳ0.5 mg) from the HPLC C-4 column was digested with trypsin (2% w/w) in 0.2 ml of 50 mM NH 4 HCO 3 , pH ϳ8.0, for 20 h at 37°C. The reaction products were then treated with 50 l of H 2 O, 0.1% trifluoroacetic acid (v/v) and subjected to HPLC on a C-18 tein; ATZ, anilinothiazoline; PTH, phenylthiohydantoin; DTT, dithiothreitol; CKII, casein kinase II; HPLC, high performance liquid chromatography; EDAC, N-ethyl-NЈ-(dimethylaminopropyl)carbodiimide; PAGE, polyacrylamide gel electrophoresis; MES, 2-(N-morpholino)ethanesulfonic acid. column (25 ϫ 0.46 cm). After injection the column was washed for 10 min, followed by linear gradient elution from H 2 O, 0.1% trifluoroacetic acid (v/v) to 60% CH 3 CN, 0.055% trifluoroacetic acid (v/v) over 120 min, with a second gradient from 60% CH 3 CN to 80% CN 3 CN over 30 min at a flow rate of 0.5 ml/min. The absorbance at 219 nm was recorded continuously on an on-line chart recorder-integrator using a Gilson HM Holochrome detector. Fractions of 0.5 ml were collected. Aliquots of 0.025 ml from each fraction were counted for 32 P radioactivity, and the radioactive fractions were then separately pooled for each peak, freezedried, and rechromatographed. Each purified 32 P-labeled peptide was then sequenced by automated solid-phase amino acid sequencing technique described below.
N-terminal Sequencing Utilizing Biobrene-treated Glass Filters-Nterminal sequencing was carried out by Edman degradation (48) using an automated model 477A sequenator (Applied Biosystems Inc., Foster City, CA) under the same conditions as described previously (49). Glass filters coated with Biobrene were used, and the proteins or peptides were adsorbed on these filters. This approach was used for N-terminal sequence analysis in which the prime interest was to identify the N-terminal sequence, for example, of purified secreted [ 32 P]OPN and large thrombin fragments. For identification of the 32 P-labeled peptide regions and the specific sites of phosphorylation, the sequencing conditions are described below.
Automated Solid-phase N-terminal Sequencing in Identification of Sites of Phosphorylation-The sequencing method for identification of phosphorylated sites required an approach that would optimize the identification of phosphorylated residues such as Ser, Thr, or Tyr. This is because these derivatives as well as inorganic phosphate that may be released during sequencing conditions are highly hydrophilic and not extracted by the usual sequence extraction solvent (1-chlorobutane). Therefore, there is significant limitation in attaining sufficient 32 P counts for identification of the site of phosphorylation. A solid-phase sequencing technique utilizing GF/F glass filters derivatized by p-aminophenyltrimethoxysilane, followed by covalent attachment of the phosphorylated peptide through its C terminus (50), was used, which allowed the use of a sufficiently hydrophilic extraction solvent (S 3 ) without loss of the peptide, which would be the case if the "adsorptive" Biobrene method was used.
The glass GF/F Whatman filters (5.0 cm diameter) were treated with trifluoroacetic acid overnight at 37°C. These filters were then freed of trifluoroacetic acid, air-dried in a fume hood, and treated with p-aminophenyltrimethoxysilane overnight at 37°C. The filters were removed from the solution and extensively washed with MeOH and dried. In the present study, trimethoxysilane was used instead of "triethoxysilane" since the latter reagent was no longer commercially available. The derivatized filters were stored in dry, dark glass dishes until needed. The covalent attachment of the 32 P-labeled peptides was carried out on derivatized GF/F glass filters cut in circular sizes of ϳ1 cm in diameter and using water-soluble carbodiimide (EDAC) (50).
The automated N-terminal sequencing of the 32 P-labeled peptides were performed directly from these filters. Further modifications of the sequencing conditions included the following. The extraction solvent S 3 was CH 3 CN instead of 1-chlorobutane and contained 0.1% thiourea, and the acid cleavage step was performed using HCl/MeOH (1 ml of acetyl chloride ϩ 13 ml of methanol ϩ 0.1% thiourea). Thiourea was used in the sequencing solvents instead of DTT. This is more advantageous than DTT since phosphoserines under sequencing conditions can be derivatized by nucleophilic attacks by DTT. For identification of sites of phosphorylation, one third of the ATZ products of each cleavage step were converted to PTH-derivatives and analyzed by on-line HPLC (model 120A, Applied Biosystems) and two thirds were collected as ATZ-derivatives for 32 P counting. To quantify yields of sequenced peptides, initial yield (I o ) and repetitive yield (R) were calculated by linear regression analysis of the observed yield (M) at each cycle (n): log 10 (M) ϭ n log 10 (R) ϩ log 10 (I o ).
Evaluation of Efficiency of the Solid-phase N-terminal Sequencing of 32 P-Labeled Peptides by Using Synthetic Peptide Substrate, Kemptide, 32 P-Phosphorylated by cAMP-dependent Kinase-Kemptide (LR-RASAV), 20 g (26 nmol) was phosphorylated using cAMP-dependent kinase (1 g) and 150 M [ 32 P]ATP (4 mCi/mmol) in 0.1 ml of KH 2 PO 4 / NaHPO 4 buffer, pH 7.4, I (ionic strength) ϭ 0.2 mol liter Ϫ1 for 30 min at room temperature. The phosphorylated Kemptide was isolated by HPLC on a C-18 column using a linear gradient from H 2 O, 0.1% trifluoroacetic acid (v/v) to 50% CH 3 CN, 0.055% trifluoroacetic acid (v/v) in 40 min. An aliquot of the 32 P-labeled Kemptide (ϳ3 nmol; ϳ26,500 dpm) in 40 l of CH 3 CN/H 2 O (1:3) was applied onto a derivatized Whatman GF/F glass filter (ϳ1.0 cm in diameter) and dried. 40 l of 160 mM MES buffer, pH 4.0, containing (5 mg/ml) EDAC was added and incubated for 1 h at 37°C. The filter was washed with H 2 O and counted. 17,200 dpm remained on the filter, which was equivalent to ϳ65% covalent coupling. N-terminal sequencing was then carried out directly from the filter with one third of the Edman degradation products converted to PTH-derivatives and analyzed on an on-line HPLC model 120A (Applied Biosystems), and two thirds were collected as ATZ-derivatives and counted for 32 P.
SDS-PAGE and Immunoblotting-SDS-PAGE (10%) polyacrylamide gel electrophoreses were carried out as described previously (51). Samples of the isolated proteins from mineralized bone and [ 32 P]ATP-labeled proteins were prepared in 10% sample buffer (20% sucrose, 10% SDS, 0.12 M Tris (pH 6.8), 0.25% bromphenol blue, 10% 2-mercaptoethanol, and 25% glycerol) and heated at 100°C for 5 min. Autoradiography was carried out on 32 P-labeled proteins, which had been resolved by SDS-PAGE after the gels had been stained and unstained. Gels were dried on a Bio-Rad dryer (model 224) for 2-3 h and exposed to X-Omat AR film (Eastman Kodak Co., Rochester, NY) at Ϫ70°C.
Western Blotting-Following SDS-PAGE with prestained low M r protein standards (Bio-Rad), the gels were soaked in blotting buffer (48 mM Tris, 39 mM glycine, 20% methanol, pH 9.2) with Immobilon P and 6 pieces of Whatman filter paper for 30 min. The electrophoretic transfer of proteins onto Immobilon was carried out at 14 V and 280 mA for 1 h using Bio-Rad Trans-Blot SD blotter. The membranes were blocked overnight at room temperature by 10% nonfat milk in 20 mM phosphate buffer, pH 7.4, containing a 0.15 M NaCl and 0.1% Tween 20. The Immobilon P-bound proteins were subjected to specific reaction with anti-chicken OPN (19) and 170 ng/ml of the same buffer without the nonfat milk antibodies for 1 h. The membranes were then washed three times with the same buffer and incubated with horseradish peroxidaseconjugated goat anti-rabbit IgG antibodies for 1 h. After several washings, the membranes developed with 3,3Ј,5,5Ј-tetramethylbenzidine substrate. Nonspecific reactions were evaluated using primary and secondary antibodies with chicken, mouse, and rabbit serum albumin.

RESULTS
Our initial attempts at large scale purification of 32 P-labeled secreted OPN of mineralized cultured chicken osteoblasts directly from the culture media were unsuccessful. However, milligram amounts of 32 P-labeled OPN were extracted from the calcified cell layer matrix in 0.1 M HCl after 14 days of culture and subsequently purified by two successive reverse-phase HPLC chromatographies on a C-4 column (Fig. 1). The purified OPN had a single N-terminal sequence (12 residues): WPVSKRQHAISA, which was consistent with the sequence deduced from cDNA (52) and direct amino acid sequencing of chicken OPN (18). Most of the OPN (Ͼ90%) was recovered from the cell-calcified matrix layer; less than 10% was found in the culture medium. Analysis of the purified 32 P-labeled OPN on SDS-PAGE gave a molecular mass of ϳ58 kDa. The protein reacted with a polyclonal antibody of chicken osteopontin (Fig.  2, panel A). OPN, isolated and purified from mineralized chicken bone (44) and used as a substrate for the protein kinases isolated from the osteoblast membrane fractions (42), had the same characteristics as the secreted OPN from the cultured osteoblasts (Fig. 2, panel B), indicating that the cultured osteoblasts post-translationally modified the protein in a fashion similar to that of osteoblasts in vivo.
After cleavage of the 32 P-labeled OPN with thrombin, three radiolabeled fragments were isolated by reverse-phase chromatography on a C-4 column (Fig. 3), fractions 37, 42, and 45. N-terminal sequencing revealed a single sequence in each of the three peptides: F 37 , WPVSKSRQHAIS; F 42 , SHHTHRY-HQDHVD; and F 45 , ASKLRKAARKL.
SDS-PAGE of the whole thrombin digest revealed three components with molecular masses of 5, 20, and 30 kDa (Fig. 3,  inset).
The performance efficiency of the solid-phase sequencing was evaluated using 32 P-labeled synthetic peptide, Kemptide (a substrate for cAMP-dependent protein kinase). Sequence analysis showed that the procedures adopted and developed were effective in identifying both the phosphorylated amino acid sequences and the site of phosphorylation. The identification of a particular phosphorylated residue was accomplished during sequencing by collecting two thirds of the Edman degradation products of each cycle as ATZ-amino acid derivative and radioactive counting for 32 P radioactivity; one third was converted to the PTH-amino acid derivative and analyzed by on-line HPLC to identify the amino acids. 32 P-Labeled Kemptide (1.95 nmol, 17, 200 dpm), covalently attached through its C terminus to derivatized Whatman GF/F glass filter, was sequenced as LR-RASAV, with an initial yield of I ϭ 251 pmol and a repetitive yield R ϭ 82%. Counts for 32 P for each cycle of Edman degradation showed a sharp increase in 32 P released at cycle 5, which corresponded to the serine residue, indicating that the phosphate group was present in this residue. At cycle 5 residual serine and dehydroalanine (a ␤ elimination product of phosphoserine) were identified. Thus the total count in this cycle reflected 32 P i ϩ 32 P-Ser. Total radioactivity recovered (10,270 dpm) through the sequencing steps was 6846 dpm from two thirds of the Edman degradation and 3423 dpm. from the one third used for PTH-amino acids, reflecting a recovery of ϳ60%. The filter retained 4390 dpm (25% of the initial total 32 P dpm) after eight cycles.
To identify the specific sequences adjacent to the phosphorylated serine residue, 32 P-labeled OPN was digested with trypsin and the 32 P-labeled peptides isolated by reverse-phase HPLC on a C-18 column (Fig. 4). Seven major radiolabeled peptides were isolated and N-terminally sequenced. Typically, 200 -300 pmol of the 32 P-labeled peptides from secreted OPN were recovered and covalently attached to derivatized GF/F glass filters, representing 65-85% of the total as determined from the 32 P counts after washing the filters. The total 32 P radioactivity on the filter ranged from 2000 dpm to 6000 dpm. 32 P released during sequencing and counting of two thirds of each cycle gave a background of 30 -40 dpm, with the cycle containing the phosphorylated residues releasing 100 -300 dpm depending on the initial 32 P level in a given peptide. Typical data are shown in Fig. 5, which demonstrates that the repetitive yields of sequencing were approximately 85%, with initial yields ranging between 50 and 300 pmol, depending on the amount of a given peptide. It is worth noting that during sequencing, some of the Ser(P) led to the formation of some dehydroalanine by ␤-elimination and thus the 32 P counts of a specific cycle reflected a combination of 32 P i and 32 P-Ser. This method permitted simultaneous identification of the amino acid sequence of the phosphorylated peptide and the phosphorylated residue. While unequivocal identification of each peptide sequence and the site of phosphorylation was possible, the size of one peptide (Fig. 4, fraction 6) with an amino acid sequence starting at Tyr-29 and extending to Lys-78 did not allow the unequivocal identification of the site of phosphorylation. The sequencing of this peptide up to 20 amino acid did not release any 32 P radioactivity, indicating that the phosphate label was not on the Thr-26, Ser-36, Ser-38, or Thr-45. The 32 P-labeled peptide containing residues Tyr-29 to Lys-78 was then further digested with endopeptidase Asp-N, from which a  179 -191, 194 -205, and 214 -219 contain recognition sequences for both CKII(s) as well as cGMPdependent kinases (XSR(K)X). In addition, the phosphorylated serine in peptide 194 -205 had a sequence recognition for Ca 2ϩ /Calmodulin-dependent protein kinase (RXXS). Two peptides, 179 -191 (ELAQHQSVENDSR) and 239 -248 (HSIENNEVTR), were also phosphorylated. The time lapse between HPLC purification of these peptides and their solid--phase sequence analyses resulted in a sufficient decay of 32 P so that the release of 32 P during sequencing was not high enough to define which of the two serines were phosphorylated. Based on the data obtained from the other peptides, it appeared that the serine residues in the "SXE" sites were the only residues phosphorylated. However, phosphorylation of the second serines "S(T)R" in these peptides could not be excluded. DISCUSSION Our initial survey of the culture medium and the cell-calcified matrix layer during the isolation and purification of quantitative (milligram) amounts of 32 P-labeled secreted OPN from cultured chicken osteoblasts indicated that Ն90% of the secreted OPN was in the calcified matrix of the cell layer. Our quantitative analysis of OPN in the medium and mineral layer during purification indicated that 93% of the secreted 32 Plabeled OPN was in the mineral and 7% in the medium. These results were based on calculations from the HPLC-purified material in terms of the 32 P counts and protein content. Furthermore, from such data the calculated ratio of 32 P label/OPN protein for medium and mineralized layer were comparable, suggesting that the phosphorylated forms of OPN in these two compartments were of very similar nature.
The unique cleavage of OPN by thrombin (19,(53)(54)(55)(56)(57), which cleaved chicken OPN at two specific sites (one between Arg-22, and Ser-23 and the other between Lys-138 and Ala-139) (Fig.  6), was used to establish that phosphorylated sites were distributed throughout the length of the OPN molecule, which is synthesized and secreted by cultured chicken osteoblasts. In vitro studies of the extent of phosphorylation and of the protein kinases that predominantly phosphorylate bovine bone OPN FIG. 4. Isolation and purification of tryptic peptides of purified 32 P-labeled OPN. 32 P-Labeled OPN (ϳ0.5 mg) was digested with trypsin (2% w/w) in 0.2 ml of 50 mM NH 4 HCO 3 , pH 8.0, for 20 h at 37°C. The peptides were separated by reverse-phase HPLC on a Dynamax C-18 column (25 ϫ 0.46 cm) using a linear gradient from H 2 O, 0.1% trifluoroacetic acid (v/v) to 60% CH 3 CN, 0.055% trifluoroacetic acid (v/v) in 120 min, at a flow rate of 0.5 ml/min. Absorbance at 219 nm was recorded continuously (solid line). Fractions of 0.5 ml were collected and aliquots (0.05 ml) counted for 32 P radioactivity (dashed line). Inset, rechromatography of fraction F 6 using a linear gradient from H 2 O, 0.1% trifluoroacetic acid (v/v) to 60% CH 3 CN, 0.055% trifluoroacetic acid (v/v) in 60 min. All of the other radiolabeled fractions were also rechromatographed prior to sequence analyses (data not shown). Fig. 4. Initially the total 32 P on the filter was 3200 dpm; at the end of 16 sequencing cycles the 32 P remaining on the filter was 1005 dpm (30% of the initial 32 P). Initial yield, I o ϭ 50 pmol; repetitive yield, R ϭ 85%. The total radioactivity recovered from the sequencing and counting of the two thirds of Edman degradation products (862 dpm) ϩ one third analyzed as PTHamino acid (431 dpm) ϭ 1293 dpm (40%). B, peptide F 6 from Fig. 4 had an initial total 32 P count on the filter of 4,915 dpm; after sequencing, 1820 dpm (37%) remained on the filter, recovered from sequencing and counting of two thirds of Edman degradation products (1685 dpm) and one third analyzed as PTH-amino acid (843 dpm) ϭ 2,528 dpm (52%). Initial yield, I ϭ 316 pmol; repetitive yield, R ϭ 92%. Ⅺ, log 10 (pmol of amino acid) at each cycle; E, log 10 (pmol ("dehydroalanine")) generated from Ser(P); q, log 10 (pmol (dehydroalanine ϩ Ser)).

FIG. 5. Solid phase N-terminal sequence analysis of purified 32 P-labeled tryptic peptides of metabolically labeled OPN from cultured chicken osteoblasts. A, peptide F 3 from
showed that bovine bone OPN was highly phosphorylated (ϳ9 mol of phosphate/mol of OPN) by CKII, and that these phosphates were distributed on both the N-terminal and C-terminal halves of the molecule (44), consistent with the results obtained from metabolically labeled chicken OPN of cultured osteoblasts.
The present results indicate that three peptides (8 -18, 48 -65, and 149 -162) concentrated in the N-terminal half of the molecule are phosphorylated only by a CKII, while the remaining four peptides concentrated in the C-terminal half of the molecule appear to be phosphorylated by a combination of CKII and cGMP-dependent kinase (peptide 179 -191). Peptide 194 -205 can be phosphorylated on Ser-196, either by an CKII or a Ca 2ϩ /calmodulin-dependent kinase, and Ser-204 of this peptide can be phosphorylated by cGMP-dependent protein kinase. The results indicate that the CKII is the predominant kinase that post-translationally modifies chicken bone OPN (6 serines and 1 threonine residue; ϳ70% of the total phosphorylated residues), while the remaining ϳ30% of the total phosphorylated residues (3 serines) are phosphorylated by other kinases. The determination of the total number of moles of phosphate/mol of OPN in the metabolically 32 P-labeled OPN synthesized by cultured chicken osteoblast was not possible using 32 P count and the content of the protein. This is because the specific activity of 32 P during pulse-chase experiments cannot be accurately established. This is the result of the synthesis of ATP, which occurs from both unlabeled P i and 32 P i within the cells during the pulse-chase, leading to dilution of the added 32 P i , the extent of which is unknown. However, a good estimation of the total of the content of 32 P for each specific tryptic peptide and the amino acid sequences was possible from the data obtained. For example, peptides F 1 , F 2 , F 3 , F 4 , and F 7 all contained a single phosphorylated residue but contained different concentrations of 32 P. It can be assumed that the peptide F 3 , which contained the largest concentrations of 32 P, represented 1 mol of phosphate/mol of peptide (i.e. 100% phosphorylation). Using this as a base line, the relative distribution of phosphates on each peptide and phosphorylated residue was calculated (Table I). It is interesting to note that metabolic phosphorylation led to a wide range of degrees of phosphorylation on different peptides (30 -100%). Additionally, peptides that contained multiple juxtaposed phosphorylated serines, e.g. peptides F 5 and F 6 , showed different extents of phosphorylation of the adjacent serine residues. In F 5 , the three serines were about equally phosphorylated (30% of each, i.e. a total 0.88 mol of phosphate/ mol of peptide), whereas the first serine of peptide F 6 was phosphorylated approximately 2-fold greater than the second serine (67% and 33%, respectively, with a combined total of 1.0 mol of phosphate/mol of peptide). The total number of moles of phosphate/mol of metabolically phosphorylated OPN calculated from the values in Table I (including an unidentified peptide between F 5 and F 6 ; see Fig. 4) was 6.1. This occurred despite the fact that there were a total of seven different phosphorylated peptides identified, containing up to 10 phosphorylated residues, as well as an eighth unidentified phosphorylated peptide. These findings indicate that not all the poten-FIG. 6. Topographically phosphorylated peptide regions and specific residues of secreted OPN from chicken osteoblasts relative to the primary amino acid sequence and to that of bovine milk OPN (58). The complete primary amino acid sequences were from cDNA for chicken bone OPN (52) and bovine milk OPN (59). The sequences enclosed in the shaded boxes represent the phosphorylated peptide regions that occurred in both and the residues with phosphate indicated by the letter P. The solid arrows represent the thrombin cleavage sites for chicken OPN identified from the isolated thrombin fragments by reverse-phase HPLC on a Vydac C-4 column and N-terminally sequenced (present work) and that of bovine bone OPN (44). The cell binding sequence (GRGDS) is boxed with a bold line, and the bold circled P on bovine milk OPN represents the phosphorylated sites that are absent in chicken OPN. tial phosphorylation sites are phosphorylated on every OPN molecule in a given population. On average, most of the peptides contain approximately 0.76 mol of phosphate/mol peptide, with an average overall extent of phosphorylation per peptide per site of ϳ53%. The biological significance of the variable extent phosphorylation of the potential phosphorylation sites and its variation with age and rate of synthesis and the possible rate of dephosphorylation are not known at this time, but present intriguing avenues for future research.
In vitro phosphorylation of purified OPN obtained from mineralized chicken bone and recombinant mouse OPN by four pure protein kinases, namely cAMP-dependent kinase, protein kinase C, cGMP-dependent kinase, and CKII, showed that there was essentially no phosphorylation of OPN by cAMP-dependent kinase, minor phosphorylation by protein kinase C and cGMP-dependent kinase (each introducing Յ1 mol of phosphate/mol of OPN), whereas CKII significantly phosphorylated OPN (ϳ9 mol of phosphate/mol of OPN) (41,42). These data are completely consistent with the conclusions obtained in the present study on the extent and sites of phosphorylation in 32 P-labeled OPN synthesized by chicken osteoblasts. Similar in vitro phosphorylation data were obtained for bovine bone OPN (44). It was of interest that in vitro, CKII phosphorylated all the potential sites (9 mol of phosphate/mol) of chicken OPN in contrast to the phosphorylation of OPN in osteoblast cell culture, in which case only partial phosphorylation of the potential sites for phosphorylation was observed (6.1 mol of phosphate/ mol of [ 32 P]OPN (61%)). A possible explanation of the more complete phosphorylation that occurs in vitro is that the overall reaction conditions are optimized in vitro. For example, the relative concentrations of the reactants such as the enzyme, ATP, and OPN would be very different when compared with those in the secretory pathway of the protein in cell culture and/or the length of the reaction time.
Although there are some conflicting reports with respect to whether phosphorylation of matrix phosphoproteins occur extracellularly or intracellularly (40,46), recent work carried out in our laboratory with OPN and BSP and a series of purified protein kinases isolated from osteoblasts, as well as the present study, strongly support the conclusions that phosphorylation of OPN and BSP occurs "intracellularly." Recently, 27 phosphorylated serines and one phosphorylated threonine were identified in bovine milk OPN (58). The predominant sites for phosphorylation contained recognition sequences for casein kinase II. Fig. 6 compares the phosphorylated regions of chicken bone OPN obtained in the present study with those of bovine milk OPN. Although most of the phosphorylated regions in both proteins are highly conserved, bovine milk OPN was more phosphorylated than chicken bone OPN. This is probably related to the fact that chicken OPN is only ϳ60% homologous with the primary structure of bovine milk OPN. Of the 28 phosphorylated sites in bovine milk OPN, representing 60% of the potential sites for phosphorylation, 12 of the serines with recognition regions SXE/SXSSEE are either absent in chicken bone OPN or the serine residue is substituted by amino acids such as I, P, or A. However, the phosphorylated sites in the chicken bone OPN were similar to those found in bovine milk OPN. Therefore, it is apparent that the lower level of phosphorylation of chicken OPN is due to the absence of the specific amino acid sequence, SXE/SXSSEE, or to the substitution of serine residues by other amino acids. In view of the fact that bovine milk OPN has a very high homology with OPN from other mammalian species, it may be expected that the extent of phosphorylation in mammalian bone OPN may be similar to milk OPN. Nevertheless, although the number of potential sites for phosphorylation in mammalian OPN may be generally the same in each species (regardless of tissue origin), the extent of phosphorylation can vary due to differences in the types and numbers of protein kinases in the various tissues or organs. As a result of different levels of phosphorylation, interactions of the OPN from various tissues with other proteins and conformational changes collectively may reflect different biological functions and open new avenues of research.
As often is the case in complex biological systems, the local environment and type of matrix with which OPN is interacting may play an important role whereby a given state of phosphorylation is optimal for induction or inhibition of calcification. One may therefore intuitively postulate that OPN at a particular state of phosphorylation in different matrix environments can behave and function very differently. The variable extent of phosphorylation of OPN from different tissues or organs and the variable degree of phosphorylation in a given population of OPN molecules, all signify the possibility that an important and a wide range of biological functions of this molecule may be exerted frequently by the capacity of a given cell population to secrete OPN at a particular state of phosphorylation. Additionally, the biological functions of OPN may be further modulated in the extracellular matrix by the varying degrees of dephosphorylation that can take place by the actions of the phosphatases.