Phosphorylation of purified bovine bone sialoprotein and osteopontin by protein kinases.

The large number of covalently bound phosphates on the extracellular phosphoproteins osteopontin (OPN) and bone sialoprotein (BSP) have been implicated in biological functions such as mineral deposition and osteoclast binding. In the present study the state of phosphorylation of BSP and OPN was evaluated by in vitro 32P labeling using a series of protein kinases and quantification. Both the purified bovine BSP and OPN were radiolabeled by [32P]ATP and factor-independent protein kinase. Quantification of 32P radioactivity incorporated on dephosphorylated BSP and OPN provided 6.6 and 8.9 mol of phosphate incorporated/mol, respectively. Native OPN incorporated 1.07 and BSP 2.46 mol of phosphate/mol by factor-independent protein kinase. These data led to calculations that OPN and BSP, respectively, contain 7.83 and 4.14 mol of phosphate/mol in their natural state. Thrombin digests of 32P-labeled BSP showed radioactivity to be associated with fragment of ∼molecular mass values 30 kDa (N-terminal half), with no observable radioactivity associated with the 40-kDa fragment (C-terminal half). Similar experiments with 32P-labeled OPN provided two radiolabeled thrombin fragments, with molecular mass 30 kDa (N-terminal half) and 20 kDa (C-terminal half), both were radioactive. The major phosphorylation was associated with the N-terminal half containing 7.0 mol of phosphate, and 1.9 mol of phosphate were associated with the C-terminal half. Additional experiments of in vitro phosphorylation of OPN and BSP by several other known protein kinases were carried out. cAMP-dependent protein kinase showed no phosphorylation of OPN or BSP, while protein kinase C and cGMP-dependent protein kinase led to minor phosphorylation, each of the latter introduced about 1 mol of phosphate/mol of OPN and BSP molecule.

The large number of covalently bound phosphates on the extracellular phosphoproteins osteopontin (OPN) and bone sialoprotein (BSP) have been implicated in biological functions such as mineral deposition and osteoclast binding. In the present study the state of phosphorylation of BSP and OPN was evaluated by in vitro 32 P labeling using a series of protein kinases and quantification. Both the purified bovine BSP and OPN were radiolabeled by [ 32 P]ATP and factor-independent protein kinase. Quantification of 32 P radioactivity incorporated on dephosphorylated BSP and OPN provided 6.6 and 8.9 mol of phosphate incorporated/mol, respectively. Native OPN incorporated 1.07 and BSP 2.46 mol of phosphate/mol by factor-independent protein kinase. These data led to calculations that OPN and BSP, respectively, contain 7.83 and 4.14 mol of phosphate/mol in their natural state. Thrombin digests of 32 P-labeled BSP showed radioactivity to be associated with fragment of ϳmolecular mass values 30 kDa (N-terminal half), with no observable radioactivity associated with the 40-kDa fragment (C-terminal half). Similar experiments with 32 P-labeled OPN provided two radiolabeled thrombin fragments, with molecular mass 30 kDa (N-terminal half) and 20 kDa (C-terminal half), both were radioactive. The major phosphorylation was associated with the N-terminal half containing 7.0 mol of phosphate, and 1.9 mol of phosphate were associated with the C-terminal half. Additional experiments of in vitro phosphorylation of OPN and BSP by several other known protein kinases were carried out. cAMP-dependent protein kinase showed no phosphorylation of OPN or BSP, while protein kinase C and cGMP-dependent protein kinase led to minor phosphorylation, each of the latter introduced about 1 mol of phosphate/mol of OPN and BSP molecule.
In addition to a major collagen matrix, bone contains several other non-collagenous proteins. Of the non-collagenous proteins, glycosylated phosphoproteins have been the subject of intense study in the past decade or so. Two such proteins, osteopontin (OPN) 1 and bone sialoprotein (BSP) are probably the best known (1)(2)(3)(4)(5)(6)(7). The importance of these phosphoproteins in initiation, regulation, and stability of hydroxyapatite crys-tals have received increasing attention (8 -11). BSP has been purified from several species including chicken, bovine, rat, human, and rabbit bones (12)(13)(14)(15)(16), and the cDNA primary amino acid sequence deduced for BSP from bovine, porcine, human, and rat (5,14,17). Both OPN and BSP possess an "RGD" (Arg-Gly-Asp) sequence region found to be involved in cell binding (6,18,19). The susceptibility of OPN to thrombin cleavage has been observed for OPN from human (20), porcine (14), mouse (21), rat (22)(23)(24), and chicken (25) leading to fragments of molecular size ranging between 23 and 45 kDa. However, the physiological importance of thrombin cleavage of OPN is not yet clearly understood. Suggestions have been made that since the thrombin cleavage sites are "close" to the RGD cell binding region, it may be involved in altering the cell attachment properties of OPN (14,20,22,26,27). BSP also possesses the RGD cell binding sequence and has been shown to have cell attachment properties (6,18,19). However, thrombin action on BSP from chicken bone (13), a rat osteosarcoma cell line (28), and porcine bone (14) led to no observable cleavages, and BSP from bovine bone is found to be specifically cleaved by thrombin 140 amino acids from the RGD cell-binding sequence, as reported in the present work.
The number of acidic amino acids and covalently bound phosphates in both OPN and BSP seems to enable them to bind simultaneously to hydroxyapatite and cells (2,7). This process may become particularly significant when the cells in question are osteoclasts (involved in bone resorption) or metastatic tumors. At present, there is little known about the nature of the phosphorylated regions and how these may be coupled with the above function, or the protein kinases responsible for the phosphorylation of these proteins. The sites of phosphorylation of purified bovine milk OPN have been determined by an indirect method using S-ethanethiol derivatization followed by sequencing, which led to determinations of 27 phosphoserine and 1 phosphothreonine residues (29). In our laboratory, a more direct method was used to metabolically 32 P label secreted OPN from cultured chicken osteoblasts, and sequence analysis led to identification of phosphorylated peptide regions with up to a total of 8 phosphorylated residues (7 P-Ser and 1 P-Thr) (30,31). This work on metabolically 32 P-labeled secreted OPN from cultured chicken osteoblasts revealed that the phosphorylation regions and sites on chicken OPN were predominantly with sequences SSEE and SXEE (i.e. recognition aa sequences for FIPK). However, there were also several other phosphorylated regions with recognition amino acid sequences for other protein kinases. In a separate study in this laboratory using cultured chicken osteoblast cytosolic and microsomal enzyme preparations, we have shown that the cytosolic preparations contained several protein kinases (FIPK, cGMP-and cAMP-dependent kinases, protein kinase C, and Ca 2ϩ /calmodulin-dependent kinase), whereas the microsomal preparation contained predominantly FIPK. Use of these osteoblast enzyme preparations in phosphorylation of purified chicken OPN and recombinant mouse OPN in combination with specific inhibitors (heparin in particular) and "qualitative analysis" led to the conclusion that these proteins were predominantly phosphorylated by FIPK (31,32). Despite significant advances that have been made in the study of OPN and BSP, there is still uncertainty and controversy with respect to the state of phosphorylation and protein kinases that are involved. In the present work, both purified bovine bone OPN and BSP were subjected to in vitro phosphorylation by several known protein kinases. We report on the quantification of phosphates (moles of phosphate/mole of OPN and BSP) that different protein kinases introduce, and use native and dephosphorylated forms of OPN and BSP to determine the extent of naturally occurring phosphorylation. Additionally, thrombin cleavage of 32 P-labeled OPN and BSP was used to evaluate the proportions of the phosphates on different domains.

MATERIALS AND METHODS
Isolation and Purification of Bovine BSP and OPN-Cortical bone of the mid-portion of the femora of 5-6-week-old calves was prepared and extracted in dilute HCl, and initially fractionated by DE52 ion exchange chromatography as described previously (3,4), with the exception that 7 M urea was included in the buffers. The fractions containing the majority of the phosphoproteins were dialyzed free of salt, dried, redissolved in a solution containing 7 M urea, 0.4 M NaCl, 0.05 M KH 2 PO 4 / KOH buffer, pH 6.8, and aliquots of 0.25 ml molecularly filtered on a TSK-3000 HPLC column (60 ϫ 0.75 cm) at a flow rate of 1 ml/min using the same buffer solution. The major peak was pooled, concentrated, and adjusted to 0.3% trifluoroacetic acid in H 2 O. Samples were then chromatographed by reverse-phase HPLC using a Brownlee C-4 column (22 ϫ 0.46 cm) with a linear gradient: 10% acetonitrile and 0.3% trifluoroacetic acid in H 2 O to 60% acetonitrile and 0.3% trifluoroacetic acid, at a flow rate of 1 ml/min.
The bovine bone OPN was purified by a similar procedure, with the exception that the DE52 ion exchange chromatography was carried out at pH 4 as described previously (2). Further purification of OPN was achieved by reverse phase HPLC on a Vydac C-4 column (15 ϫ 0.38 cm).
Thrombin Digestion-Aliquots of 200 g of purified BSP were digested separately with highly purified human and bovine thrombin (5000 and 2000 units/mg, respectively, Sigma). 1 unit of thrombin was used per 8 g of OPN or BSP protein in 0.1 M NH 4 HCO 3 , pH 8.0, or in 0.05 M Tris-HCl, pH 8.0, buffer containing 10 mM CaCl 2 for 2 h at 37°C. A control sample of BSP or OPN was also incubated in the same buffer without the addition of thrombin. An aliquot of the thrombin digest (5 g of protein) and of the control sample was subjected to SDS-PAGE on 10% acrylamide gels (5 ϫ 7 cm) at 10 mA for 3 h, fixed, stained, and destained ( Fig. 1). The major portion of the thrombin-digested sample of BSP (195 g of protein) was chromatographed by reverse-phase HPLC on a Brownlee C-4 column (22 ϫ 0.46 cm) using a linear gradient from 0.3% trifluoroacetic acid in H 2 O to 80% acetonitrile and 0.3% trifluoroacetic acid with a flow rate of 1 ml/min (Fig. 2). Fractions 1-4 were separately pooled, freeze-dried, and an aliquot subjected to SDS-PAGE ( Fig. 1). N-terminal sequences were obtained from fractions 1, 2, and 3. Fraction 4 is predominantly thrombin which co-elutes with the small amounts of uncleaved BSP.
The bovine OPN was digested similarly and the OPN fragments were separated by SDS-PAGE (Fig. 3) followed by blotting onto Immobilon P and N-terminal sequencing directly from strips of Immobilon P.
Amino Acid Sequencing-The N-terminal amino acid sequences of purified BSP and OPN, and the peptides generated after digestion with thrombin were carried out by Edman degradation (33) using an Applied Biosystems model 477A automated sequenator essentially as described in Ref. 34.
Pico Tag HPLC Analysis for Phosphorylated Residues and Total Amino Acids-100 g of purified bovine BSP or OPN was mildly acid hydrolyzed in 0.5 ml of 4 M HCl ϩ 1% phenol for 4 h at 110°C in a Pyrex-sealed tube. The hydrolysate was dried in vacuo, resuspended in 20 l of a mixture of 200 proof ethanol:water:triethylamine (2:2:1) and dried in vacuo. The sample was then derivatized to phenylthiocarbamyl (PTC) derivatives for 30 min at room temperature with a solution of 200 proof ethanol:water:triethylamine:phenylisothiacyanate (7:1:1) as described in the Waters Associates Pico Tag manual. After redrying in vacuo, the PTC-derivatives were dissolved in 20 l of Waters sample diluent buffer and analyzed using the Waters HPLC system and Pico Tag column (0.39 ϫ 15 inch) maintained at 38°C. The PTC-derivatives were eluted by a set of gradients, from 94% buffer A to 51% buffer B in 10 min (flow rate 1.0 ml/min), from 51% buffer B to 100% buffer B in 0.2 min, maintained at 100% buffer B for 1.3 min at 1.0 ml/min flow rate, followed by 100% buffer B for 0.8 min at a flow rate of 1.5 ml/min.
The PTC-phosphoamino acids were identified by analyzing initially the standard mixture of PTC-derivatives of phosphoserine, phosphothreonine, and phosphotyrosine of 125 pmol of each alone, and in a mixture of standard PTC-derivatives (65 pmol each), followed by analysis of the PTC-derivatives of the bovine BSP or OPN acid hydrolysate. Comparison of the elution times of the standard PTC-phosphoamino acids with that of the PTC-derivatives of BSP amino acids enabled determination of the absence or presence of particular phosphoamino acids.
For total amino acid analysis 50 g of BSP or OPN was hydrolyzed in 6 N HCl at 110°C for 24 h. The derivatization to PTC-derivatives was as described above. The Pico Tag analysis in quantifying the total amino acids was accomplished using standard PTC-derivatives.
Dephosphorylation of Bovine OPN and BSP by Acid Phosphatase-OPN (90 g) and BSP (100 g) were each suspended in 0.2 ml of NaAc buffer, pH 5.0, I (ionic strength) ϭ 0.1 mol/liter and 10 units of acid phosphatase (potato, Sigma) was included and incubated for 3 h at 37°C. Each sample (OPN and BSP) was then isolated rapidly by reverse-phase HPLC on a Vydac C-4 column (15 cm ϫ 0.38 cm). This step separates both the cleaved phosphates and the acid phosphatase from the dephosphorylated OPN and BSP. The extent of dephosphorylation was determined by analyzing for phosphoamino acids after partial HCl hydrolysis as described above.
In a separate experiment, 20 g of bovine BSP and 15 g of OPN were first dephosphorylated with 5 units of acid phosphatase at pH 5.0 for 3 h at 37°C twice, followed by isolation of each dephosphorylated protein using reverse-phase HPLC on a Vydac C-4 column (Nerst Co.). Both BSP and OPN were then [ 32 P]phosphorylated using FIPK and [ 32 P]ATP as described below and passed through a HPLC Vydac C-4 column. The radiolabeled OPN and BSP were then incubated with 5 units of acid phosphatase in NaAc buffer, pH 5.0, for 3 h at 37°C twice. The reaction mixture was then rechromatographed using a HPLC Vydac C-4 column. 1-ml fractions were collected and aliquots counted for 32 P radioactivity. The degree of dephosphorylation was calculated from the loss of 32 P after acid phosphatase treatment.
Phosphorylation of Bovine OPN and BSP by Factor-independent Protein Kinase-The dephosphorylated OPN (40 g) and BSP (39 g) were phosphorylated by [ 32 P]ATP (specific activity 100 mCi/mmol) and factor-independent protein kinase (FIPK) (50 ng, Upstate Biotechnology, Inc.) in 0.2 ml of KH 2 PO 4 /Na 2 HPO 4 (0.1 M) buffer, pH 7.4, containing 5 mM MgCl 2 and 1 mM EGTA for 1 h at room temperature (22°C). In a separate experiment, the time dependent phosphorylation indicated that within 30 min, the phosphorylation of OPN and BSP by FIPK is virtually complete. After incubation of 1 h, the phosphorylated OPN and BSP were rapidly isolated by reverse-phase HPLC on a Vydac C-4 column (15 ϫ 0.38 cm) by linear gradient from 100% H 2 O ϩ 0.1% trifluoroacetic acid to 60% CH 3 CN ϩ 0.055% trifluoroacetic acid in 60 min at a flow rate of 0.5 ml/min. The absorbance at 230 nm was recorded continually and fractions of 0.5 ml were collected. Aliquots of 0.05 ml from each fraction were counted for 32 P radioactivity in 5 ml of biodegradable scintillant. Fractions corresponding to OPN and BSP absorption and 32 P-labeled ( Fig. 4) were pooled and freeze dried. For example, in a similar experiment bovine OPN (6.42 ϫ 10 Ϫ10 mol, 21.3 g determined by Lowry's protein assay as well as total amino acid analysis) and BSP (1.36 ϫ 10 Ϫ9 mol, 45 g) were phosphorylated by FIPK in the presence of [ 32 P]ATP (specific activity 11 mCi/mmol). The labeled proteins were isolated by reverse-phase HPLC on a Vydac C-4 column with total 32 P radioactivity 1.38 ϫ 10 5 dpm (6.26 ϫ 10 Ϫ8 Ci) and 2.16 ϫ 10 5 dpm (9.82 ϫ 10 Ϫ8 Ci) on OPN and BSP, respectively. Using specific activity of [ 32 P]ATP ϭ 11 mCi/mmol led to 5.69 ϫ 10 Ϫ9 mol and 8.93 ϫ 10 Ϫ9 mol of 32 P incorporated in OPN and BSP, respectively.
Phosphorylation of Bovine OPN and BSP by cGMP-dependent Protein Kinase and Protein Kinase C-Phosphorylation of bovine OPN and BSP by cGMP-dependent kinase and protein kinase C were carried out essentially as above with the exception that (a) for cGMP-dependent kinase (Promega), the reaction mixture contained no EGTA, but ϩ 1 mM CaCl 2 ϩ 2.0 M cGMP ϩ 50 ng of cGMP-dependent protein kinase (from Promega), or (b) for protein kinase C ϩ 1 mM Ca 2ϩ ϩ 0.1 mM Lphosphatidyl-L-serine ϩ 50 ng of protein kinase C (Promega Co.). In a separate experiment, dephosphorylated OPN and BSP were first thrombin digested followed by 32 P labeling by the above protein kinases.

Reverse-phase HPLC of Thrombin-digested 32 P-Labeled OPN and BSP and Quantification of Moles Phosphate/Mol of Protein on the
Fragments-Aliquots of the original 32 P-labeled OPN and BSP and thrombin digests were subjected to SDS-PAGE followed by autoradiography (Fig. 5). The 32 P-labeled OPN and BSP and thrombin fragments were separated by reverse-phase HPLC on a Vydac C-4 column and aliquots were counted for 32 P radioactivity (Fig. 6). Furthermore, two different thrombin digests of both OPN and BSP 32 P-labeled by FIPK were subjected to reverse-phase HPLC separation on a Vydac C-4 column (Fig. 6). One thrombin digest used 1 unit of thrombin/8 g of OPN or BSP of which aliquots were run on SDS-PAGE (Fig. 5), the other thrombin digest was using 5 units of thrombin/8 g of OPN and BSP (i.e. 5 times more thrombin). For this latter case, SDS-PAGE was also run (Fig. 6). The reverse-phase HPLC separation profiles of the two different thrombin digestions are compared (Fig. 6) and the 32 P radiolabeled thrombin fragments of both OPN and BSP were N-terminal sequenced, as well as 32 P radioactivity and total protein amount of each fragment were determined (from Pico Tag total amino acids) for calculations of moles of phosphate/mol of thrombin fragment.
Quantification of Number of Moles of Phosphate/Mol of OPN or BSP Introduced by FIPK-32 P-Labeled samples of OPN and BSP (free of excess [ 32 P]ATP, and phosphorylation reaction buffers from "Phosphorylation of Bovine OPN and BSP by Factor-independent Protein Kinase" and "Phosphorylation of Bovine OPN and BSP by cGMP-dependent Protein Kinase and Protein Kinase C" as described above) were each resuspended separately in 0.8 ml of H 2 O and the following aliquots were removed from each for analysis: (a) 5 l for 32 P count, (b) 0.15 ml for Lowry's protein assay, (c) 0.15 ml for total amino acid analysis and phosphoamino acid analysis, (d) 0.15 ml for thrombin digestion followed by SDS-PAGE and autoradiography (Fig. 5).
Determination of the Accurate Specific Activity of [ 32 P]ATP-Purchased [ 32 P]ATP (specific activity 3000 mCi/mmol, DuPont NEN) was diluted approximately 100-fold by cold ATP. Due to rapid hydrolysis of purchased [ 32 P]ATP and that of cold ATP (the latter being used to dilute the former to appropriate specific activity) and since the errors introduced by the apparent knowledge of the concentrations could lead to enormous errors in the later calculations of the moles of phosphate/mol of OPN and BSP using the 32 P counts; the absolute specific activity of the working solutions of [ 32 P]ATP was calculated by complete phosphorylation of Kemptide (specific substrate for cAMP-dependent kinase, Sigma) by cAMP-dependent kinase. For example, 15 g of Kemptide (LRRASVA) was incubated with 4 g of cAMP-dependent kinase (catalytic subunit, Sigma) in the presence of 5 mM CaCl 2 , 5 mM MgCl 2 , 150 M [ 32 P]ATP working stock for 30 min at room temperature in a total reaction volume of 0.15 ml. The reaction was terminated by addition of 0.1 ml of 100% H 2 O ϩ 0.1% trifluoroacetic acid and subjected to HPLC on a Dynamax C-18 column (Rainin Co.). The HPLC conditions were: a linear gradient from 100% H 2 O to 50% CH 3 CN in 40 min, a flow rate 0.5 ml/min, absorbance ϭ 219 nm was recorded continually and fractions of 0.5 ml/fraction collected. Kemptide eluting at 35% CH 3 CN was counted for 32 P radioactivity by taking aliquots of 0.1 ml from 0.5-ml fractions. The total radioactivity incorporated was 4.7 ϫ 10 5 dpm equivalent to 2.15 ϫ 10 Ϫ7 Ci. The total Kemptide injected was 15 g, equivalent to 1.95 ϫ 10 Ϫ8 mol. Hence specific activity of the stock working [ 32 P]ATP was 11.05 Ci/mol (or 11.05 mCi/mmol).

Bone Sialoprotein
Purified bovine BSP was isolated from the major phosphoprotein containing fractions separated by DE52 ion exchange chromatography followed sequentially by molecular filtration and reverse-phase HPLC. SDS-PAGE of the major BSP peak obtained by reverse-phase HPLC chromatography revealed either a single band or two of very closely adjacent bands of 66 kDa molecular mass (Fig. 1), with a single N-terminal amino acid sequence (LSMKNLNRRAK) identical to the sequence of bovine BSP derived from the cDNA sequence (6). This approach provided ϳ 2.0 mg of BSP per 100 g of bovine bone powder. Both O-phosphoserine and O-phosphothreonine have been identified in the non-collagenous phosphoproteins (3,35,36). Significant amounts of both phosphoamino acids were also identified in purified OPN (12,13,37). However, the only previous analyses of phosphoamino acids in BSP have been discussed in the report by Heinegard (38), where significant amounts of O-phosphoserine (27 residues/1000 total residues) were observed, with only very small amounts of O-phosphothreonine (0.5 residues/1000 total residues). Analysis in the present report for phosphoamino acids in purified bovine BSP led to similar results, with O-phosphoserine (19 residues/1000 total residues) and O-phosphothreonine (0.3 residues/1000 total residues). The above values reflect nanomoles of amino acids recovered and were not corrected for loss due to partial acid hydrolysis.
Evaluation of the State of Phosphorylation of BSP by Protein Kinases-To further our understanding of the state of phsophorylation and the protein kinases that are involved in this process, both purified bovine bone OPN and BSP were dephosphorylated by acid phosphatase followed by in vitro phosphorylation using pure protein kinases. By utilizing the 32 P radiolabel and phosphoamino acid analysis, it was established that tartrate-resistant acid phosphatase under the experimental conditions used removed ϳ65-75% of the phosphates on OPN and BSP. Repetition of this process twice led to ϳ90% dephosphorylation. The numerical values in Table II are not corrected to the residual 10% natural phosphates remaining on OPN and BSP.
The dephosphorylated BSP was 32 P-phosphorylated by FIPK incorporating 6.6 mol of phosphate/mol of BSP, Fig. 4 and Table  II. Further studies with protein kinases: (a) cAMP-dependent, (b) protein kinase C, and (c) cGMP-dependent led to no phosphorylation by cAMP-dependent kinase and minor phosphorylation (ϳ1.0 mol of phosphate/mol of BSP) by each of protein kinase C and cGMP-dependent kinase. Similar study carried out using FIPK and the above enzymes using native (nondephosphorylated) BSP led to significantly reduced amounts of 32 P incorporation compared with the dephosphorylated form. FIPK introduced 2.46 mol of phosphate/mol of native BSP, indicating that BSP contained 62% naturally phosphorylated sites. Therefore, native purified BSP contained 4.14 mol of phosphate/mol that were naturally occurring and phosphorylated by FIPK type enzyme in vivo (Table II). The overall results indicate that the major protein kinase that phophorylates BSP is FIPK.
Thrombin Cleavage of Native and 32 P-Labeled BSP-Previous reports from several laboratories have utilized the cleavage of OPN by thrombin (13,14,20,22,25) to distinguish between OPN and BSP since it has been uniformly reported that BSP is not cleaved by thrombin. When purified bovine BSP was incubated with thrombin under conditions usually used for thrombin cleavage of OPN (14, 20, 22-24), we found that bovine BSP also undergoes such specific fragmentation by thrombin. Both human and bovine thrombin fragmented bovine BSP and in the same way (Fig. 1, lanes 3 and 3Ј, respectively). The thrombingenerated polypeptides were separated and isolated by reversephase HPLC on C-4 column (Fig. 2, fractions 1-4). Thrombin generated two fragments of very similar molecular mass (40 kDa) eluted from reverse-phase HPLC at 16 and 20% acetonitrile concentrations, respectively (Fig. 2, fractions 1 and 2), and a third smaller molecular mass (30 kDa) fragment eluted around 32% acetonitrile (Fig. 2, fraction 3). The original residual 66-kDa BSP and thrombin co-eluted at 40% acetonitrile (Fig. 2, fraction 4). The N-terminal amino acid sequencing of thrombin fractions 1 and 2 led to a single sequence KAGAT-AGKKA in each case. This sequence corresponds to the sequence starting with Lys-123, indicating cleavage to be between Arg-122 and Lys-123. The third lower molecular mass fragment ( Fig. 2 fraction 3) provided a sequence of LSMKNL-NRRAK, corresponding to the N-terminal of the original 66-kDa BSP. The presence of two C-terminal end thrombin fragments with the same N-terminals and very similar molecular masses (40 kDa), but eluting at sufficiently enough different acetonitrile concentrations suggests that the naturally occurring BSP may exist in at least two different molecular forms similar to findings reported for OPN (22). 32 P-Labeled BSP by FIPK was subjected to thombin cleavage followed by HPLC analysis and SDS-PAGE (Figs. 5 and 6, SDS-PAGE gel insets). In Fig. 5, lane 3, the small amount (Ͻ5%) of the total radioactivity present in that lane associated with high molecular mass region represent aggregates of BSP, which is not observed in lane 4 of thrombin cleaved BSP, and not observed for OPN or its thrombin cleaved samples. Although BSP is known to not stain easily by Coomassie Blue, with the amounts of BSP (10 g) used which gave very intense staining by "Stains All," we observe some staining by Coomassie Blue at levels 10 -15% that of Stains All (Fig. 1). This observation is probably partly due to the large amount of pure BSP used and possibly some desulfated/deglycosylated states being present in the purified BSP sample. Analysis of 32 Plabeled bovine BSP thrombin fragments by HPLC and SDS-PAGE ( Fig. 6B and Table II) showed that of the total 6.6 mol of phosphate incorporated by FIPK (or 4.14 mol of naturally occurring phosphates) on this protein, no significant phosphorylation was observed on the 40-kDa (C-terminal half) fragment. The predominant phosphorylation was associated with the ϳ30-kDa (N-terminal half) thrombin fragment and minor phosphorylation (25%) found on a lower molecular mass fragment (22 kDa). This latter 22-kDa fragment was found to increase when higher amounts (5 ϫ) of thrombin were used. For instance, compare HPLC profiles of 32 P counts, solid line and dashed lines, and autoradiographs ( Fig. 5 and Fig. 6B) for two different thrombin concentrations. The proportions on the 30-and 22-kDa N-terminal fragments of BSP were ϳ75 and ϳ25%, respectively, when 1 unit of thrombin/8 g of BSP was used, and this proportion changed to ϳ55 and ϳ45% when 5 units/8 g of BSP (5 ϫ more) thrombin was used. The Nterminal sequence analysis of this fragment showed that it has the N-terminal amino acid sequence of the original BSP. Therefore, it appears that the easily generated N-terminal half residues Leu-1 to Arg-122 (30 kDa) can further fragment at a much lower rate to generate the 22-kDa fragment which still has the N-terminal sequence of BSP. Similar treatment of OPN, that is, 5 ϫ more thrombin, however, led to no further fragmentation of either the 30-or 20-kDa fragments (Fig. 6A).
Studies of thrombin action on BSP from chicken bone (13), a rat osteosarcoma cell line (28), and porcine bone (14) apparently lead to no observable cleavage. We have repeated the experiment using purified chicken BSP and thrombin under the same experimental conditions used in this study which lead to specific cleavage of bovine BSP and OPN, and found that, consistent with the previous report, chicken BSP is not cleaved by thrombin. The present report is the first to date to demonstrate thrombin cleavage of BSP and identify the site of cleavage. Furthermore, while the cleavage site of the bovine BSP contains the recognition sequence (underlined), Arg-Lys-Ala-Gly, for thrombin specificity, the site of cleavage (i.e. between Arg and Lys) has not been observed previously for other proteins susceptible to thrombin action, e.g. OPN or fibrinogen, where cleavage takes place between Arg/Lys and residues Ala, Gly, and Ser. Thus, it appears that the thrombin cleavage of bovine BSP is unique and may be referred to as atypical.
The physiological importance of thrombin cleavage of OPN found in bone has been emphasized as a modulator of functionality of this protein through possibly changing its cell-binding property (12,22,26,38). However, both OPN and BSP possess the RGD cell-binding sequence found to be involved in cell attachment through specific receptors on the cells. It appears that modulation by thrombin cleavage does not seem to have the same applicability for BSP as that reported for OPN in the literature. BSP from several species are found not to be susceptible to thrombin cleavage and bovine BSP is cleaved 140 amino acids removed from the RGD cell-binding sequence. In the case of rat BSP, it has been suggested that the non-RGD peptide region containing the sequence KKAGDA (residues 123-128) may be involved in cell attachment (28), in particular osteoclasts (39,40). This sequence is also found in human BSP and is analogous to the non-RGD sequence KQAGDV in fibrinogen known to be involved in cell attachment (41). It is interesting to note that this sequence in bovine BSP is RKAGAT (residues 122-127) and the thrombin cleavage occurs between Arg and Lys ( Table I) suggest that sequences other than RGD are involved in modulating osteoclast cell behavior. Recently, the importance of phosphate groups during cell attachment has been highlighted by the changes in osteoclast binding toward "phosphorylated" and "dephosphorylated" OPN and BSP (46).

Osteopontin
The bovine bone OPN (0.5 mg/100 g of bone powder) was purified similarly to BSP. The N-terminal sequence analysis provided LPVKPTSSGSKE, and on SDS-PAGE 64 kDa (Fig.  3).
Evaluation of the State of Phosphorylation of OPN by Protein Kinases-FIPK phosphorylated in vitro bovine OPN with incorporation of 8.9 mol of phosphate/mol of OPN, Fig. 4 and Table II. Similar studies were carried out using OPN and protein kinases: (a) cAMP-dependent protein kinase, (b) protein kinase C, and (c) cGMP-dependent protein kinase. cAMPdependent protein kinase showed no phosphorylation, and minor phosphorylation (ϳ1.0 mol of phosphate/mol) was observed by each of cGMP-dependent protein kinase and protein kinase C. Evaluation of the degree of phosphorylation by FIPK and other enzymes of native (non-dephosphorylated) OPN led to incorporation of 32 P at a significantly reduced level compared with dephosphorylated forms (Table II). For example, FIPK introduced only 1.07 mol of phosphate/mol. These values indicate that bovine OPN contained 88% naturally phosphorylated sites that were recognition sites for FIPK. Hence, native purified OPN contained 7.83 mol of phosphate/mol that were naturally occurring and phosphorylated by FIPK-type enzyme in vivo. The overall results, summarized in Table II, clearly identify FIPK as the predominant protein kinase that phosphorylates OPN with minor phosphorylation carried out by cGMP-dependent protein kinase and protein kinase C.
Thrombin Cleavage of Native and 32 P-Labeled OPN-Thrombin digestion of OPN led to two major fragments, one with molecular mass of 20 kDa and the other 30 kDa (Fig. 3). N-terminal sequencing of these cleavage products led to sequences starting with Ser-164 (SKKFRRFMVQ) for the 20-kDa fragment and Leu-1 (LPVKPTSSGSKE) for the 30-kDa fragment. Thus thrombin specifically cleaved bovine OPN between Arg-163 and Ser-164. This cleavage site, about 10 amino acids from the RGD cell-binding region, is typical of thrombin cleavage of OPN from other species (Table I).
Aliquots of OPN 32 P-labeled by FIPK were subjected to thrombin cleavage followed by SDS-PAGE and autoradiography (Figs. 5 and 6). For quantification of the moles of phosphate introduced/mol of OPN and its thrombin fragments, the total 32 P radioactivity associated with OPN protein (Fig. 4) or thrombin fragments (Fig. 6) and total amounts of protein present were determined after reverse-phase HPLC. The HPLC separation of 32 P-labeled OPN and its thrombin fragments provided recoveries in the range of 75-85%. A loss of 15-25% material, however, does not affect quantitation since both the 32 P radioactivity and protein contents of the samples were determined after HPLC. The proportions of 32 P on the thrombin fragments were also determined from the autoradiographs by densitometric scanning. It is worth noting that this latter method only provides relative radioactivity on the total bands present, and in the absence of absolute quantification that was obtained from HPLC separated fractions this would not be of considerable use. However, in conjunction with the absolute quantification from the HPLC method, it has provided 32 P label on Nand C-terminal fragments that were consistent with what was obtained from the HPLC method. For instance, the HPLC method gave 32 P associated with N-terminal 78% and with C-terminal 22% for bovine OPN, while two different experiments of SDS-PAGE and scanning of autoradiographs gave values for N-terminal 74 and 84% (average 79%) and for Cterminal 26 and 16% (average of 21%). The HPLC separated 32 P-labeled thrombin fragments were also N-terminal sequenced to reveal their identity. The major phosphorylated fragment of bovine OPN was found to be the 30 kDa which contained 79 Ϯ 5% of the total phosphates and the 20-kDa fragment that contained 21 Ϯ 5% of the total phosphates on OPN (Fig. 6A and Table II). This indicates that the N-terminal

TABLE I Relationship between thrombin cleavage site and RGD cell binding sequence of OPN and postulated non-RGD cell binding sequence for BSP from different species and comparison to that of bovine BSP and OPN
For OPN cell binding sequence RGD (solid box) relative to the thrombin cleavage site between residues RS indicated by solid arrow, throughout several species. For BSP the postulated "non-RGD" cell binding sequence analogous region to that of fibrinogen (solid box), with thrombin cleavage thus far only found for bovine BSP beween RK indicated by solid arrow. half of OPN (30 kDa) starting with Leu-1 and extending to Arg-163, contained about 6.9 mol of the total phosphates incorporated by FIPK (or 6.19 mol naturally occurring phosphates). The C-terminal half (20 kDa) starting with Ser-104 and ending with the C-terminal contained 1.9 mol of the total phosphates incorporated (or 1.64 mol naturally occurring phosphates) on OPN. Thrombin cleavage showed that of the total phosphates introduced (ϳ1.1) by each of cGMP-dependent kinase and protein kinase C on OPN, ϳ75% (0.8 mol) of the total phosphate was associated with the N-terminal half and remaining ϳ0.3 mol was on the C-terminal half. Implications of the in Vitro Phosphorylation of Bovine OPN and BSP-The total number of phosphates introduced in vitro by FIPK on OPN and BSP is clearly different, about 25% less phosphate on BSP compared to OPN. The naturally occurring phosphates calculated using the difference in the extent of phosphorylation on native and dephosphorylated forms by FIPK also emphasized these differences, where OPN was found to contain ϳ7.8 and BSP 4.1 mol of phosphate/mol. This indicates that BSP has ϳ47% less naturally occurring phosphates compared with OPN. Whether these natural variations are indications of important physiological processes such as, the two proteins are phosphorylated on the potential phosphorylation sites to different extent prior to secretion or that BSP specifically undergoes partial dephosphorylation while it is resident in the extracellular matrix is not easy to discern. The present in vitro phosphorylation experiments were carried out on samples of OPN and BSP using the uncleaved molecules and prior thrombin cleaved samples followed by phosphorylation. It was found that as far as the phosphorylation was concerned regarding several kinases used, there was no significant difference in the extent of phosphorylation on the two halves of OPN and BSP. Thus a single cleavage of OPN or BSP did not alter significantly the ability of the kinases to recognize the specific phosphorylation regions. This may have been expected since small synthetic peptide substrates of 10 -15 amino acid residues size with recognition sequences toward different kinases are phosphorylated with relative ease, e.g. Kemptide, syntide 2, etc. It is noteworthy that the possibility of some of the observed phosphates introduced in vitro reactions by FIPK were the result of contaminant kinases phosphorylating Tyr residues may be excluded, since our analysis of the 32 Plabeled proteins for phosphoamino acids led to identification of no P-Tyr, but predominantly P-Ser. Furthermore, the P-Ser peak elution time of Pico Tag analysis was correlated with the 32 P count released from HPLC during analysis. Similarly, the possibility of sugars on OPN and BSP being phosphorylated can be excluded since we observe no other additional 32 P-peak in the Pico Tag HPLC analysis profile than P-Ser/P-Thr. It may be perceived that the presence of significant amounts of glycosylation on both OPN and BSP, and since these glycosylations occur on Ser/Thr sites, the possible deglycosylation naturally occurring in the matrix or by contamination of acid phosphatase (by glycosidases) may expose the Ser/Thr sites normally occupied by glycosyl moi- FIG. 4. Reverse-phase HPLC profiles of 32 P-labeled OPN and BSP. Dephosphorylated BSP (39 g) and OPN (40 g) were phosphorylated by FIPK (50 ng) using [ 32 P]ATP (specific activity 100 mCi/mmol) at room temperature for 1 h followed by separation on a Vydac C-4 HPLC column. Elution was achieved by initial wash with 100% H 2 O ϩ 0.1% trifluoroacetic acid for 20 min followed by linear gradient from 100% H 2 O ϩ 0.1% trifluoroacetic acid to 60% CH 3 CN ϩ 0.055% trifluoroacetic acid in 60 min. A: BSP, total 32 P-incorporated 1.77 ϫ 10 6 dpm; B: OPN, total 32 P-incorporated 2.45 ϫ 10 6 dpm. Solid lines are absorbance ϭ 230 nm and dotted lines 32 P count. Void (fractions [15][16][17][18][19][20] contains the excess free 32 P from [ 32 P]ATP.

TABLE II
In vitro phosphorylation of bovine OPN and BSP by protein kinases, quantification of phosphate introduced and proportional distribution of the phosphates on the N-terminal and C-terminal halves of these proteins after specific thrombin cleavage Quantification of moles of phosphate incorporated by protein kinases per mol of OPN and BSP before and after dephosphorylaion. Proportion of the total phosphates incorporated on two halves of OPN and BSP was determined after thrombin cleavage and quantifying 32 P label after isolation of each thrombin fragment from HPLC using Vydac C-4 column. The naturally occurring phosphates per mol of OPN and BSP were calculated by utilizing the difference between amount of phosphate introduced in vitro for native and dephosphorylated forms: (a) phosphate incorporated in vitro and (b) calculated naturally occurring.  (29). Also, the recognition amino acid sequences around Oglycosylation sites thus far known in the literature (29,47), bear no amino acid recognition sequences to those of FIPK. Hence, it is not expected that the two post-translational modifications to compete for the same sites, such that degrees of deglycosylation may cause increase in the moles of phosphate introduced/mol of OPN and BSP. Furthermore, similar quantitative analysis using FIPK and recombinant mouse OPN (which contains no post-translational modifications) has provided ϳ9 mol of phosphate/mol (32), consistent with what has been observed for the purified bovine bone OPN. With both proteins, the predominant phosphorylation sites are on the N-terminal half with minor or no phosphorylation on the C-terminal half. The recent report by Ek-Rylander et al. (46) in which the phosphates on OPN and BSP are found to be important in osteoclast binding, when taken together with the present findings, indicate that osteoclast binding is strongly assisted by the heavily phosphorylated N-terminal halves of both OPN and BSP, which are the non-RGD domains. Therefore, one would postulate that the overall cell attachment properties of these proteins require more than a single functional domain or protein sequence. The most plausible conclusion is that overall cell attachment/modulation involves participation of (a) the RGD sequence region, (b) phosphorylated regions, and (c) possibly non-RGD amino acid sequence "KKAGDA," in a coupled fashion. The precise functional consequence of coupling or synergistic effect of these different moieties is as yet not known. However, involvement of the RGD sequence in cell attachment is well established, and thus far there is considerable evidence that the covalently bound phosphates on OPN and BSP are important in mineral deposition (48,49) and osteoclast cell attachment (46). Since these proteins are inti-mately involved in biological events in the extracellular matrix, then the state of phosphorylation can affect the functional properties of OPN and BSP with perturbations of the matrix biology. Although the KQAGD region was found not to affect osteoclast attachment to OPN, this region appears to have more of a signaling effect toward change of osteoclast behavior, i.e. osteoclast cell "rounding" has been reported in response to short peptides KQAGD and AGDV (40).
It has been well established that the recognition sites of FIPK for phosphorylation are predominantly Ser/Thr-X-Glu-Glu-Glu or Ser/Thr-XX-Glu-Glu (50 -52). Recent findings of the naturally occurring sites of phosphorylation of bovine milk OPN (29)  without HPLC. Lane a, phosphorylated by FIPK; Lane b, phosphorylated by cGMP-dependent kinase; and Lane c, phosphorylated by protein kinase C. B: BSP, 32 P-labeled by FIPK and digested by human thrombin followed by reverse-phase HPLC on a Vydac C-4 as A. Solid line, 32 P count of thrombin fragments when 1 unit of thrombin was used per 8 g of BSP (total BSP used 5 g) and dotted line represents counts when 5 units of thrombin were used per 8 g of BSP (total BSP used 6 g). Inset: autoradiography following SDS-PAGE (10% acrylamide slap gel, 17 ϫ 20 cm) of 32 P-labeled BSP thrombin digest (5 units of thrombin per 8 g of BSP) without HPLC. Lane a, phosphorylated by FIPK; Lane b, phosphorylated by CGMP-dependent protein kinase; and Lane c, phosphorylated by protein kinase C. of peptide regions: residues 5-14 PTSSGSSEEK and 42-50 NSVSSEET. Work in this laboratory on the identification of the sites of metabolically 32 P-labeled secreted chicken OPN (30), showed phosphorylation sites on the N-terminal half of the molecule at regions with peptide sequences: residues 8 -18 QHAISASSEEK and 55-62 THYSSEEN. With both OPN from bovine milk or chicken osteoblasts, the N-terminal heavy phosphorylations are distinctly carried out by FIPK, consistent with in vitro phosphorylation by FIPK. Inspection of bovine bone OPN sequence revealed two major potential phosphorylation peptide regions, analogous to those found by sequence analysis above: one corresponds to residues 5-14 PTSSGSSEEK and the other with residues 42-50 QNSVSSEET. Similar analysis of the N-terminal half from the thrombin cleavage site of bovine BSP revealed potential phosphorylated sites by FIPK to be peptides with residues 12-18 LDSEEN, 42-64 FAVQSSSDS-SEENGNGSSEEE and 69 -74 TSNEEG. The potential phosphorylated sites by FIPK on the C-terminal half of bovine OPN are peptides with residues 159 -176 DATEEDFTSHIESEEM and 237-244 DHKSEEDK. Unlike OPN, the C-terminal half of BSP does not appear to be phosphorylated.
Identification of the FIPK as the major enzyme involved in the phosphorylation of OPN and BSP through in vitro studies as presented here has pertinence to the identity of the types of enzyme(s) phosphorylating these proteins in vivo. Our work using cultured chicken osteoblasts as the source of protein kinases led to the conclusion that FIPK of osteoblasts, found in the microsomal fraction, was the predominant enzyme that phosphorylated purified chicken OPN and recombinant mouse OPN (31,32). This enzyme was found to be localized in the Golgi apparatus of the chicken osteoblasts (53). In these studies, FIPK was also found in the cytosolic cell compartment, and at this time there is no further information whether the microsomal FIPK is the same or an isoform of the cytosolic enzyme. Previously, the presence of membrane-associated FIPK activity in 14-day embryonic chicken bone that can phosphorylate endogeneous chicken bone phosphoprotein was reported (54). Also, using ROS 17.28 osteosarcoma cells led to isolation of membrane-associated FIPK which phosphorylated dentin-derived phosphophoryns (55). This latter work was extended to localization of the enzyme to the endoplasmic reticulum (56). The subcellular location of the FIPK to endoplasmic reticulum or Golgi is of significance mechanistically since both OPN and BSP are secretory proteins and the enzyme participating in their post-translational modification must encounter them. Although there are some conflicting reports with respect to whether phosphorylation of extracellular matrix phosphoproteins are intracellular or extracellular (55,56), work carried out in this laboratory thus far indicated that the phosphorylation of OPN and BSP occurs "intracellularly." Classical casein kinase II (FIPK) has been shown to be localized predominantly in the cytoplasm and nucleus where it phosphorylates a wide range of protein substrates with important functional consequences (50 -52). In addition, casein kinase II has also been found in a membrane-bound form (57), and mammary gland casein kinase II which phosphorylates the secretory protein (casein) was localized in the Golgi. A significant analogy exists between these latter well established studies and what has been thus far found for the secretory phosphoproteins (OPN and BSP) and protein kinases involved in the bone tissue.