Cell Type-specific Post-translational Modifications of Mouse Osteopontin Are Associated with Different Adhesive Properties*

Osteopontin (OPN) is a highly modified integrin-binding protein found in all body fluids. Expression of OPN is strongly correlated with poor prognosis in many different human cancers, suggesting an important but poorly understood role for this protein in tumorigenesis and metastasis. The protein exists in a number of different isoforms differing in the degree of post-translational modifications that are likely to exhibit different functional properties. This study examines for the first time the post-translational modifications of OPN from transformed cells and the effects of these modifications on cell biology. We have characterized the complete phosphorylation and glycosylation patterns of OPN expressed by murine ras-transformed fibroblasts (FbOPN) and differentiating osteoblasts (ObOPN) by a combination of mass spectrometric analyses and Edman degradation. Mass spectrometric analysis showed masses of 34.9 and 35.9 kDa for FbOPN and ObOPN, respectively. Enzymatic dephosphorylation, sequence, and mass analyses demonstrated that FbOPN contains approximately four phosphate groups distributed over 16 potential phosphorylation sites, whereas ObOPN contains ∼21 phosphate groups distributed over 27 sites. Five residues are O-glycosylated in both isoforms. These residues are fully modified in FbOPN, whereas one site is partially glycosylated in ObOPN. Although both forms of OPN mediated robust integrin-mediated adhesion of mouse ras-transformed fibroblasts, the less phosphorylated FbOPN mediated binding of MDA-MD-435 human tumor cells almost 6-fold more than the heavy phosphorylated ObOPN. These results strongly support the hypothesis that the degree of phosphorylation of OPN produced by different cell types can regulate its function.

phosphorylation (19). The interactions between OPN and some receptors are dependent on the PTM state of the protein. For instance, OPN promotes RGD-dependent osteoclast attachment provided that the protein is phosphorylated. Phosphorylation of recombinant OPN increases osteoclast attachment in vitro (20), whereas partial dephosphorylation of bovine OPN by tartrate-resistant acid phosphatase abolishes the ability to mediate osteoclast attachment (21). Similarly, OPN can stimulate bone resorption only if it is properly phosphorylated (22). OPN has cytokine-like properties and can influence the production of cytokines by interaction with surface receptors on immune cells. Phosphorylation of OPN is required for integrin binding and subsequent induction of interleukin-12 expression in macrophages (7) and for mediating their activation and spreading (23). The protein has also been shown to promote trophoblastic cell migration in a process dependent on the level of phosphorylation of the protein (24). In addition, the importance of OPN phosphorylation in mineralization became evident from studies of dephosphorylated OPN that had lost the ability to inhibit hydroxylapatite formation (5,25). Further emphasizing the importance of the phosphorylations, it was recently shown that highly phosphorylated milk OPN containing ϳ28 phosphorylations promotes hydroxylapatite formation and growth, whereas bone OPN modified by only 13 phosphates inhibits formation (26). Similar results have been obtained in other studies showing that phosphorylation is crucial for OPN inhibition of calcium oxalate crystallization in urine (27) and calcification of vascular smooth muscle cells (28).
OPN is encoded by a single copy gene but exists in a number of different isoforms that differ mainly in the extent of PTM. Normal rat kidney cells secrete both phosphorylated and nonphosphorylated variants of OPN that also vary in glycosylation patterns and in their ability to associate with fibronectin (29). Likewise, differentiating rat osteoblasts produce two forms of OPN as seen on SDS-polyacrylamide gels; a 55-kDa form that contains little phosphorylation and a highly phosphorylated 44-kDa form (30). The extent of phosphorylation of OPN in osteoblast and epidermal cells is responsive to hormonal influences such as 1,25-dihydroxyvitamin D 3 (31,32), and oncogene-transfected Rat-1 cells switch from the synthesis of sialylated to non-sialylated OPN upon phorbol ester treatment (33).
The only OPN isoforms that have been thoroughly characterized with regard to the PTM pattern are those from bovine and human milk and rat bone. Bovine milk OPN contains 27 phosphoseryl residues and one phosphothreonine residue (34). Up to 34 phosphoserines and two phosphothreonines were identified in the human counterpart (35). Studies of OPN purified from rat bone revealed up to 29 potential phosphorylation sites; however, in this isoform, the average level of phosphorylation was estimated to be only ϳ10 -11 phosphates/OPN molecule (36). Phosphorylation sites in OPN are located predominantly at serines in the recognition motifs of the Golgi kinase/ mammary gland casein kinase ((S/T)X(E/S(P)/D)) (37,38) and/or casein kinase II (SXX(E/S(P)/D)) (37,39). In addition to phosphorylation, the characterized OPN isoforms all contain glycosylations. Bovine and human milk OPNs contain three and five O-glycosylated residues, respectively, whereas four gly-cans have been identified in the sequence of rat bone OPN (34 -36).
Here, we have purified and extensively characterized two isoforms of OPN expressed by 275-3-2 murine ras-transformed fibroblasts (FbOPN) and MC3T3-E1 murine differentiating immortalized osteoblasts (ObOPN). This is the first comparative characterization of OPN isoforms expressed by cells from the same species, and it is the first study to characterize the PTMs of OPN produced by transformed cells. Major differences in the degree of phosphorylation of these proteins were found, and these correlated with differences in biological activity, confirming that the function of OPN produced by different cell types is distinct.

EXPERIMENTAL PROCEDURES
Materials-Sequencing grade modified trypsin was from Promega (Madison, WI); thermolysin, Percoll, bovine serum albumin (BSA), bovine alkaline phosphatase, and fibronectin were from Sigma; and endoproteinase Asp-N was from Roche Diagnostics (Penzberg, Germany). The RPC (narrow-bore reverse-phase chromatography) C 2 /C 18 PC 2.1/10 and the Superdex 75 PC 3.2/30 columns were from Amersham Biosciences AB (Uppsala, Sweden). Reagents used for sequencing were purchased from Applied Biosystems (Warrington, UK). 2,5-Dihydroxybenzoic acid was from LaserBio Labs (Sophia-Antipolis Cedex, France). The peptide N-glycosidase F kit was obtained from New England Biolabs (Beverly, MA). The GRGDNP and GRADSP peptides were from BIOMOL International LP (Plymouth Meeting, PA). All other chemicals used were of analytical grade.
Cell Lines and Culture-The highly mineralizing murine osteoblast MC3T3-E1 subclone 4 cell line (40) was a kind gift from Dr. R. Franceschi (University of Michigan) and was cultured in ␣-minimal essential medium (Invitrogen) supplemented with 10% fetal bovine serum (HyClone, Logan, UT), 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM glutamine (all from Invitrogen). For differentiation, cells were grown until confluent and then switched to medium containing 100 g/ml ascorbic acid and 10 mM ␤-glycerophosphate (both from Sigma) for an additional 10 -12 days before generating conditioned medium for ObOPN purification. The 275-3-2 murine ras-transformed fibroblast cell line (41) and the MDA-MB-435 human breast cancer cell line (American Type Culture collection) were maintained in Dulbecco's modified Eagle's medium (Mediatech, Inc., Herndon, VA) supplemented with 10% fetal bovine serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM glutamine. The 275-3-2 cells were used to make conditioned medium when ϳ80% confluent for purification of FbOPN.
Purification of OPN-OPN was purified from media conditioned by MC3T3-E1 murine differentiating immortalized osteoblasts (ObOPN) and 275-3-2 murine ras-transformed fibroblasts (FbOPN). Cell lines were grown as described above, and the conditioned media were generated by incubating the cells in serum-free medium overnight. OPN was affinity-purified using monoclonal antibody 2A1 (42) coupled to protein G-agarose beads (Pierce). Approximately 50 ml of conditioned medium was incubated with 1 ml of antibody-coupled beads overnight at 4°C with end-over-end rotation. The beads were gently pelleted, washed with phosphate-buffered saline, and packed into 2-ml disposable columns (Pierce). OPN was eluted with 100 mM glycine and 500 mM NaCl (pH 2.5) and immediately neutralized. Fractions were analyzed by SDS-PAGE, and proteins were visualized by non-ammoniacal silver staining or by Western blotting with antibody 2A1. Positive fractions were pooled, desalted on PD-10 columns (GE Healthcare), and lyophilized.
Analysis of OPN-Native and dephosphorylated FbOPN and ObOPN were analyzed by mass spectrometry (MS) using a Voyager DE-PRO MALDI-TOF mass spectrometer (Applied Biosystems). Samples for MS analyses were prepared by mixing the sample with a saturated solution of 2,5-dihydroxybenzoic acid at a 1:1 ratio directly on the MS target probe. All spectra were obtained in positive linear ion mode using a nitrogen laser at 337 nm and an acceleration voltage of 20 kV. Typically, 50 -100 laser shots were added per spectrum and calibrated with external standards. The masses were assigned using the halfheight method. For dephosphorylation, 2.5 g of each OPN isoform was incubated with 0.07 units of bovine alkaline phosphatase in 10 mM ammonium bicarbonate (pH 8.5) overnight at 37°C.
Generation and Separation of Peptides-OPN was digested with trypsin using an enzyme/substrate ratio of 1:30 (w/w) in 0.1 M ammonium bicarbonate at 37°C for 6 h. Tryptic peptides were separated by reverse-phase pressure liquid chromatography (RP-HPLC) on a RPC C 2 /C 18 PC 2.1/10 column connected to a GE Healthcare SMART system. Separation was carried out in 0.1% trifluoroacetic acid (solvent A) and eluted with a gradient of 60% acetonitrile in 0.1% trifluoroacetic acid (solvent B) developed over 54 min (0 -9 min, 0% solvent B; 9 -49 min, 0 -50% solvent B; and 49 -54 min, 50 -100% solvent B) at a flow rate of 0.15 ml/min. The peptides were detected in the effluent by measuring the absorbance at 214 nm. A large fragment not susceptible to trypsin cleavage (Gln 35 /Gln 54 -Arg 128 ) was separated from the tryptic peptides by gel filtration on a Superdex 75 PC 3.2/30 column. The column was equilibrated with 0.1 M ammonium bicarbonate and operated at a flow of 0.05 ml/min. The fractions containing the fragment in question were pooled, lyophilized, and further digested with thermolysin in 0.1 M pyridine acetate and 5 mM CaCl 2 (pH 6.5) at 56°C for 6 h. The resulting peptides were separated by RP-HPLC as described for the tryptic digest. Fractions Fb-Th1 (where Th is thermolysin) and Ob-Th1 (see Fig. 3, C and D) from the thermolytic digests were further digested with endoproteinase Asp-N (2 g) in 50 mM sodium phosphate buffer (pH 8.0) at 37°C for 18 h.
Characterization of Peptides-Peptides were characterized by MS and amino acid sequence analyses. All mass spectra were obtained in both positive reflector ion and positive linear ion mode as described above. The theoretical peptide masses and sequence coverage were calculated using the GPMAW program (Lighthouse Data, Odense, Denmark). Amino acid sequence analyses were performed on an Applied Biosystems Procise HT protein sequencer with online identification of phenylthiohydantoin (PTH)-derivatives. Glycosylated serine/threonine residues were identified by the lack of a PTH-derivative in the cycles in which these amino acids were modified.
Enzymatic Deglycosylation-N-Linked deglycosylation was performed using the peptide N-glycosidase F deglycosylation kit as recommended by the manufacturer. OPN (15 g) was  incubated with 500 units of peptide N-glycosidase F at 37°C for 20 h. The reaction products from enzymatic deglycosylation were analyzed by SDS-PAGE.
Cell Adhesion Assays-Flat-bottom 96-well tissue culture treated polystyrene microtiter plates (Corning Corp., Corning, NY) were coated with 100 l of OPNs (10 g/ml) or fibronectin (2.5 g/ml) in phosphate-buffered saline overnight at 4°C and then blocked with 1% BSA. MDA-MB-435 and 275-3-2 rastransformed fibroblast cells were trypsinized, washed twice, and resuspended in Dulbecco's modified Eagle's medium containing 1 mg/ml BSA. Cells were preincubated with 100 M GRGDNP or GRADSP control peptide at 37°C for 30 min. Cells (5 ϫ 10 4 ) were then added to the coated wells and adhered for 1 h (275-3-2 cells) or 18 h (MDA-MB-435 cells) as described (43) at 37°C in a humidified atmosphere with 5% CO 2 . Nonadhered cells were removed by washing twice with 75 l of 73% Percoll and 0.9% NaCl, and adherent cells were fixed with 50 l of fixative (10% glutaraldehyde in Percoll). Fixed cells were stained with 100 l of 0.1% crystal violet and solubilized in 50 l of 0.5% Triton X-100 before reading at 570 nm in an MRX Revelation absorbance reader (Thermo Labsystems, Franklin MA).
Statistical Analysis-Statistical analysis of data was done by Student's t test. Differences were considered to be statistically significant at p Ͻ 0.05.

RESULTS
Molecular Masses-OPN was purified from the media conditioned by the mouse MC3T3-E1 osteoblasts and 275-3-2 rastransformed fibroblasts as described, and the purity was ana-lyzed by SDS-PAGE and silver staining (Fig. 1). There was a clear difference in the migration during SDS-PAGE of the two OPN isoforms, suggesting that they are differently modified by PTMs. More exact estimation of the average masses of the OPN isoforms was determined by linear MALDI-TOF-MS, showing molecular masses of ϳ34.9 and 35.9 kDa for FbOPN and ObOPN, respectively (Fig. 2). The mass spectra show broad peaks, which suggest that both OPNs are heterogeneously modified. To estimate the total number of phosphate groups present, FbOPN and ObOPN were treated with bovine alkaline phosphatase. The molecular mass of dephosphorylated FbOPN was 34.6 kDa, corresponding to a loss of approximately four phosphate groups ( Fig. 2A). Phosphatase treatment of ObOPN reduced the mass to 34.2 kDa, corresponding to a loss of ϳ21 phosphorylation groups (Fig. 2B). Parallel control experiments with bovine OPN (data not shown) showed that the dephosphorylation reaction was complete. Subtraction of the observed average mass of the dephosphorylated OPNs from the theoretical mass of the mouse OPN polypeptide (30,746 Da) leaves ϳ3.9 and 3.5 kDa for glycosylation sites on FbOPN and ObOPN, respectively, suggesting that FbOPN contains slightly more carbohydrates than does ObOPN.
Phosphorylation Sites-FbOPN and ObOPN were digested with trypsin, and the resulting peptides were separated by RP-HPLC (Fig. 3, A and B). A large acidic fragment of OPN (Gln 35 / Gln 54 -Arg 128 ) was not susceptible to trypsin cleavage and had to be isolated from the tryptic digests of FbOPN and ObOPN by gel filtration (data not shown). The purified fragments were further digested with thermolysin, and the resulting peptides were separated by RP-HPLC (Fig. 3, C and D).
All fractions from the RP-HPLC separations of tryptic and thermolytic peptides were analyzed by MALDI-TOF-MS ( Table 1). Detection of phosphorylated peptides and peptide mapping of the OPN sequence were performed as described previously (35). In total, 16 and 27 phosphorylation sites were identified in the sequences of FbOPN and ObOPN, respectively. Several of the phosphopeptides contained more serines/ threonines than the observed number of phosphorylation sites. In these situations, the phosphate groups were assigned to residues fitting the target sequence of the Golgi kinase/mammary gland casein kinase ((S/T)X(E/S(P)/D)) based on the localization of phosphorylated residues in other OPN isoforms (34 -36). Data from the peptide phosphorylation analysis of the two proteins are summarized in Table 1, and the resulting map of modifications is shown in Fig. 4. Characterization of the proteins revealed that most peptides exist in several states of phosphorylation. The only phosphorylated residues identified as fully phosphorylated were Ser 8 , Ser 10 , and Ser 11 in ObOPN (Table 1). As shown in Table 1, it is evident that ObOPN contains significantly more phosphorylations than does the fibroblast analog. Furthermore, based upon how often each residue was encountered as phosphorylated, it is noticeable that the most abundant species of a given peptide from FbOPN was unmodified or singly phosphorylated, whereas the corresponding peptide in the osteoblastic counterpart was predominantly phosphorylated at all potential sites. This corresponds well with the data from the dephosphorylation studies (Fig. 2) and further emphasizes that the potential sites of phosphorylation  in ObOPN are phosphorylated to a much higher degree than in FbOPN. RP-HPLC (Fig. 3) showed that there is a significant structural difference between the two OPNs. This is especially true for the separation of thermolytic peptides (Fig. 3, C and D), indicating that the proteins have different PTMs and perhaps also different proteolytic cleavage. For instance, the peptides starting at Ser 59 , Asn 60 , Ser 61 , and Ser 64 were detected only in the FbOPN digest ( Table 1), suggesting that thermolysin cannot cleave ObOPN in this region. This could be explained by the heterogeneous phosphorylation of these serines, and intensive phosphorylation at Ser 59 , Ser 61 , and Ser 64 in ObOPN could potentially hinder thermolytic cleavage in this region.
O-Glycosylation Sites-All O-glycosylated residues in FbOPN were detected in fraction Fb-Th1 (Fig. 3C). N-terminal sequencing of this peak revealed the presence of three peptides starting at Val 88 , Ser 99 , and Ile 118 , respectively. Linear MALDI-TOF-MS showed that all of the peptides contain species with mass increments of ϳ365 and/or 291 Da, corresponding to different amounts of N-acetylhexosamine (HexNAc), hexose (Hex), and sialic acid (Sia) residues (Table 2) Table 2). The lack of PTH-  Thr at Thr 121 during Edman sequencing indicates that this amino acid is modified in the peptide. Furthermore, sequencing of peptide Ser 99 -Arg 128 showed no PTH-derivative in the cycles corresponding to Thr 107 , Ser 109 , and Thr 110 , which, together with the MS data in Table 2, indicates O-glycosylation at these residues.
To confirm the O-glycosylation sites and their structures, the glycopeptides in fraction Fb-Th1 were further digested with endoproteinase Asp-N. This resulted in peptides cleaved N-terminal to Asp 103 and Asp 113 , which were subsequently separated by RP-HPLC (data not shown). Edman sequencing of the peptides showed no identifiable PTH-derivatives in the cycles corresponding to Thr 107 , Ser 109 , Thr 110 , Thr 116 , and Thr 121 ( Table  2). No signals of peptide 103 DETVTASTQA 112 were observed by MS. However, masses corresponding to Asp 103 -Arg 128 were detected in linear MALDI-TOF-MS mode, with m/z values consisting of a series of peaks separated by ϳ365 or 291 Da, showing that the peptide contains multiple glycosylation sites composed of up to 7 HexNAc-Hex and 10 Sia units, respectively. This peptide contains all five amino acids in question (Thr 107 , Ser 109 , Thr 110 , Thr 116 , and Thr 121 ), and the masses strongly indicate that they are all glycosylated. Furthermore, the monoisotopic masses of peptide 113 DTFTPIVPTVD-VPNGR 128 showed that this peptide exists in different glycosylation variants. The most abundant species had an excess mass of 1895.67 Da, corresponding to two HexNAc-Hex and four Sia units, indicating that Thr 116 and Thr 121 are modified by the Sia-Hex-(Sia)-HexNAc structure.
Fraction Ob-Th1 (Fig. 3D) contained peptides Val 88 -Arg 128 and Ile 118 -Arg 128 ( Table 2). It was determined that Thr 121 was modified by O-glycans consisting of Sia-Hex-(Sia)-HexNAc or Sia-Hex-(Sia-Hex-HexNAc)-HexNAc moieties and that peptide Val 88 -Arg 128 contains multiple HexNAc, Hex, and Sia units ( Table 2). Furthermore, fraction Ob-Th2 (Fig. 3D) was observed only in the RP-HPLC analysis of thermolytic ObOPN peptides and was not seen in the corresponding chromatogram for FbOPN. This fraction contained Val 106 -Arg 128 as determined by MS and sequencing. Linear MS of Ob-Th2 showed a series of m/z values at 6181.98, 5891.57, 5600.66, and 5309.40 Da (Fig. 5B). These masses correspond to peptide 106 VTASTQADTFTPIVPTVDVPNGR 128 , containing four HexNAc-Hex moieties with varying amounts of Sia. Amino acid sequencing suggested modifications of Ser 109 , Thr 110 , Thr 116 , and Thr 121 , which, in combination with the MS data, indicates O-glycosylation sites at these residues.
As with FbOPN, the glycopeptides from Ob-Th1 (Val 88 -Arg 128 and Ile 118 -Arg 128 ) were further digested with endoproteinase Asp-N. A peptide starting at Asp 103 was detected by N-terminal sequencing with no identifiable PTH-derivative in the cycles corresponding to Thr 107 , Ser 109 , and Thr 110 . Sequencing and MS analysis of peptide Asp 113 -Arg 128 confirmed that Thr 116 and Thr 121 are glycosylated as the corresponding peptide in FbOPN (Table 2).
N-Glycosylation-The mouse OPN sequence contains a single asparagine (Asn 62 ) in a putative N-glycosylation motif. MS analysis of peptides containing this residue did not show glycosylation in either of the two proteins. This observation was supported by N-terminal sequencing of the large tryptic peptide isolated by gel filtration, which clearly showed PTH-Asn in the cycle corresponding to Asn 62 . In addition, incubation of either of the proteins with peptide N-glycosidase F did not result in altered migration upon SDS-PAGE (data not shown).
In summary, these data show that Thr 107 , Ser 109 , Thr 110 , Thr 116 , and Thr 121 are glycosylated in FbOPN. The absence of any trace of PTH-derivatives in these positions indicates that they are fully glycosylated. The corresponding residues were also found to be glycosylated in ObOPN, but with heterogeneity at Thr 107 , as this residue was observed both with and without glycans. The heterogeneity at Thr 107 in ObOPN could account for the excess mass observed in dephosphorylated FbOPN compared with dephosphorylated ObOPN (Fig. 2). The masses of the glycopeptides in both proteins show that each glycosylated amino acid can be modified by various glycan structures, although in most cases, they con-  Table 2).
Adhesion Assays-To assess the biological consequences of these different patterns of phosphorylation, we tested the ability of FbOPN and ObOPN to mediate cell adhesion. The 275-3-2 cells (mouse transformed embryonic fibroblasts) were able to adhere to both FbOPN and ObOPN: adhesion of these cells to FbOPN was slightly less (30% lower) than to ObOPN (Fig. 6A). These cells adhered to both proteins at Ͼ50% of the maximal adhesion (to fibronectin). A strikingly different response was seen when the human breast cancer cell line MDA-MB-435 was tested. Adherence of these cells to FbOPN was robust, Ͼ50% of the adhesion to fibronectin. On the other hand, ObOPN supported minimal cell adhesion, Ͻ10% of that to fibronectin and 6-fold lower than that to FbOPN (Fig. 6B). In every case, the addition of the RGD peptide reduced adhesion to OPN to control (BSA) levels. An RAD peptide was ineffective in blocking adhesion. These results suggest that both cell lines adhere to OPN through integrins that bind the RGD sequence motif, thus excluding CD44 and the ␣ 4 ␤ 1 and ␣ 9 ␤ 1 integrins. Together, these results support the hypothesis that integrin-mediated binding of some cell types to OPN is modulated by the degree of phosphorylation of the protein.

DISCUSSION
We have presented here the first direct comparison of OPNs expressed by different cell types originating from the same species and the first comprehensive analysis of PTMs of OPN associated with transformed cells. We have identified 27 and 16 sites of phosphorylation in ObOPN and FbOPN, respectively. However, the difference in phosphorylation among the proteins is even more pronounced because ObOPN contains, on average, ϳ21 phosphate groups compared with an average of only four phosphate groups in FbOPN. The presence of only 4 mol of phosphate/OPN molecule makes FbOPN the least phosphorylated OPN isoform characterized yet. Many experiments have demonstrated the tumorigenic and metastatic abilities of ras-transformed fibroblasts, and ras family proto-oncogenes are one of the most widely mutated in tumors. The FbOPN characterized in this study produced by ras-transformed fibroblasts can be hypothesized to be similar to tumorproduced OPN and to cause similar functional differences.
The phenomenon of heterogeneous phosphorylation observed in FbOPN and ObOPN has also been reported in human milk, rat bone, and chicken osteoblast OPNs (35,36,45). An alignment of the two murine proteins with the previously characterized milk and rat bone OPN isoforms, including the sites of PTM, is shown in Fig. 7.
The two mouse OPNs contain five O-glycosylated residues (Thr 107 , Ser 109 , Thr 110 , Thr 116 , and Thr 121 ). The unique serine glycosylation has not previously been shown in OPN, further emphasizing the individual character of the PTMs among OPN isoforms from different cells and species (Fig. 7). The five glycosylation sites in FbOPN are fully substituted, whereas Thr 107 in ObOPN is only partially glycosylated.
The amino acid sequence of murine OPN contains a single potential N-glycosylation site (Asn 62 ). Analysis of thermolytic peptides covering this residue (Table 1), as well as treatment with peptide N-glycosidase F, indicated that neither FbOPN nor ObOPN is N-glycosylated. The absence of N-glycosylation sites is consistent with data on milk and rat bone OPNs, which also lack N-glycosylation sites (34 -36). OPN is a member of the SIBLING (small integrin-binding ligand N-linked glycoprotein) family of phosphoglycoproteins (19). However, the presence of N-linked glycans in OPN is debatable, and data unambiguously showing this modification in OPN remain to be presented.
The differences in PTMs of FbOPN and ObOPN translated into functionally distinct proteins exhibiting dissimilar adhesion of 275-3-2 and MDA-MB-435 cells (Fig. 6). To investigate which receptors mediate the binding between either of the two cell lines and OPN, we examined the effect of an RGD peptide in the adhesion assay. As shown in Fig. 6A, the ability of the 275-3-2 cell line to adhere to either FbOPN or ObOPN could be reduced to BSA control levels by the addition of this peptide, but not an RAD peptide-containing control. Similarly, the adhesion supported by FbOPN to the MDA-MB-435 cells was completely inhibited upon treatment with the RGD peptide. The adhesion of the MDA-MB-435 cells to ObOPN was not significantly above the negative BSA control in any case (Fig. 6B). The ability of the RGD peptide but not the RAD peptide control to block all observed adhesion suggests that OPN binding is integrin-mediated in both the 275-3-2 and MDA-MB-435 cells. These data are supported by previous experiments showing that MDA-MB-435 cells are unable to adhere to RGD mutant recombinant (non-phosphorylated) OPN (43).
These data suggest that OPN-mediated cell adhesion is a complex event that is influenced by the PTMs of OPN. For example, all binding of the MDA-MB-435 cells to FbOPN is via integrins, and these integrins cannot recognize ObOPN, perhaps due to conformational changes. Cellular interactions of OPN are complicated by the multiple integrins that can act as receptors for OPN. Involvement of the ␣ 4 ␤ 1 and ␣ 9 ␤ 1 integrins in this study can be excluded because they bind only the N-terminal part of the protein resulting from thrombin cleavage (14,15), but several other RGD peptide-binding integrins could be mediating the binding to OPN we observed. Further work is required to understand which integrins are involved.
In summary, OPN produced by ras-transformed fibroblasts (FbOPN) contains, on average, approximately four phosphate groups, whereas OPN produced by immortalized osteoblasts (ObOPN) contains ϳ21 phosphate groups. The phosphate groups are distributed on 16 and 27 sites in FbOPN and ObOPN, respectively. Both proteins are O-glycosylated in a region N-terminal to the integrin-binding RGD sequence. FbOPN contains five fully substituted O-glycosylation sites, one of which is only partially substituted in the ObOPN counterpart. Both OPNs contain a single O-glycosylated serine, a novel type of modification in OPN. The fact that different cell types produce diversely modified OPNs with different cell binding properties indicates that PTMs are very important in the regulation of OPN function.