C-terminal Modification of Osteopontin Inhibits Interaction with the αVβ3-Integrin*

Osteopontin (OPN) is a multifunctional phosphorylated protein containing the integrin binding sequence Arg-Gly-Asp through which it interacts with several integrin receptors, such as the αVβ3-integrin. OPN exists in many different isoforms differing in phosphorylation status that are likely to interact differently with integrins. The C-terminal region of OPN is particularly well conserved among mammalian species, which suggests an important functional role of this region. In this study, we show that modification of the extreme C terminus of OPN plays an important regulatory role for the interaction with the αVβ3-integrin. It is demonstrated that highly phosphorylated OPN has a much reduced capability to promote cell adhesion via the αVβ3-integrin compared with lesser phosphorylated forms. The cell attachment promoted by highly phosphorylated OPN could be greatly increased by both dephosphorylation and proteolytic removal of the C terminus. Using recombinantly expressed OPN containing a tag in the N or C terminus, it is shown that a modification in the C-terminal part significantly reduces the adhesion of cells to OPN via the αVβ3-integrin, whereas modification of the N terminus does not influence the binding. The inhibited binding of the αVβ3-integrin to OPN could be restored by proteolytic removal of the C terminus by thrombin and plasmin. These data illustrate a novel mechanism regulating the interaction of OPN and the αVβ3-integrin by modification of the highly conserved C-terminal region of the protein.

Osteopontin (OPN) is a multifunctional phosphorylated protein containing the integrin binding sequence Arg-Gly-Asp through which it interacts with several integrin receptors, such as the ␣ V ␤ 3 -integrin. OPN exists in many different isoforms differing in phosphorylation status that are likely to interact differently with integrins. The C-terminal region of OPN is particularly well conserved among mammalian species, which suggests an important functional role of this region. In this study, we show that modification of the extreme C terminus of OPN plays an important regulatory role for the interaction with the ␣ V ␤ 3integrin. It is demonstrated that highly phosphorylated OPN has a much reduced capability to promote cell adhesion via the ␣ V ␤ 3 -integrin compared with lesser phosphorylated forms. The cell attachment promoted by highly phosphorylated OPN could be greatly increased by both dephosphorylation and proteolytic removal of the C terminus. Using recombinantly expressed OPN containing a tag in the N or C terminus, it is shown that a modification in the C-terminal part significantly reduces the adhesion of cells to OPN via the ␣ V ␤ 3 -integrin, whereas modification of the N terminus does not influence the binding. The inhibited binding of the ␣ V ␤ 3 -integrin to OPN could be restored by proteolytic removal of the C terminus by thrombin and plasmin. These data illustrate a novel mechanism regulating the interaction of OPN and the ␣ V ␤ 3 -integrin by modification of the highly conserved C-terminal region of the protein.
Osteopontin (OPN) 2 is a highly acidic phosphorylated glycoprotein containing an integrin binding Arg-Gly-Asp (RGD) sequence. OPN is a multifunctional protein, and it has been implicated in bioactivities including tumorigenesis, immune responses, bone remodeling, wound healing, and inhibition of ectopic calcification (1)(2)(3). A wide variety of cell types express OPN, and the protein is found both as an immobilized extracellular matrix molecule in e.g. bone and as a secreted protein in body fluids, such as milk, urine, and blood (4 -7). Nuclear magnetic resonance studies have shown that OPN has an open flex-ible conformation largely devoid of secondary structure (8).
OPN is extensively altered through posttranslational modifications, such as phosphorylation, glycosylation, sulfation, and proteolytic processing which significantly influence the function of the protein (5,13,17). The proteases thrombin, matrix metalloprotease-3 and -9, plasmin, and cathepsin D cleave OPN close to the RGD sequence, which in all cases generates N-terminal fragments containing the integrin binding RGD sequence (4,18,19). These N-terminal fragments have shown greater capability to mediate RGD-dependent cell attachment than the full-length protein presumably due to a more exposed integrin binding sequence (5,13,20).
OPN is very heterogeneously phosphorylated; and although a similar number of potential phosphorylation sites have been identified in OPN from different sources, the degree of phosphorylation varies a lot depending on the origin of the protein (1,13). The most phosphorylated form is found in milk where OPN has been shown to contain ϳ25-30 phosphate groups depending on the species (21,22). In contrast, OPN from urine and bone is only decorated by ϳ8 and ϳ10 phosphorylations, respectively (6,23). Further emphasizing the cell type-specific phosphorylation of OPN, a comparison of OPN produced by two different murine cell types showed that ras-transformed fibroblasts expressed a variant containing only four phosphate groups whereas osteoblast-derived OPN contained an average of 21 phosphorylations (24).
The widely expressed ␣ V ␤ 3 -integrin is a well characterized receptor for OPN in processes such as cell adhesion, migration, and bone resorption (18,19,25,26). However, the regulation of this interaction has not been clearly described. In one study, cell adhesion to highly phosphorylated OPN via the ␣ V ␤ 3 -integrin is significantly increased after proteolytic cleavage of OPN (19).
In another study, no difference in adhesion via this receptor between full-length and matrix metalloprotease and thrombincleaved bacterial produced OPN was observed (12), indicating that exposure of the RGD sequence after cleavage is not necessary for binding to the ␣ V ␤ 3 -integrin. These seemingly conflicting results could likely be ascribed to the difference in phosphorylation of the OPNs used in the studies. The significance of the phosphorylations in cell binding was substantiated in a direct comparison of two well characterized full-length murine OPNs. Here, a less phosphorylated isoform containing ϳ4 phosphates supported greater adhesion of the ␣ V ␤ 3 -integrinexpressing melanoma cell line MDA-MB-435 than a more phosphorylated OPN form containing ϳ21 phosphates (24). This suggests that a high degree of phosphorylation can reduce the capability of the ␣ V ␤ 3 -integrin to interact with OPN. The phosphorylation sites in OPN are located in clusters of three to five sites with stretches of unphosphorylated residues in between (21,22). In low phosphorylated OPN variants, e.g. urinary and bone OPN, only few of the potential sites are actually phosphorylated (6,23,24), whereas most of the phosphorylation sites are occupied in highly phosphorylated forms, like milk OPN (21,22). This leaves open the possibility that phosphorylation of specific clusters in OPN can influence the binding to the ␣ V ␤ 3 -integrin.
The extreme C-terminal region of OPN is highly conserved among mammalian species (see Fig. 1) and contains four serines (in human OPN) which constitute potential phosphorylation sites. The high degree of amino acid conservation could indicate an important functional role of this part of OPN as is the case for other highly conserved elements like the integrin binding sequences and sites of posttranslational modification. In support of this, it has recently been shown that monoclonal antibodies recognizing the extreme C terminus inhibited MDA-MB-435 cell adhesion to unphosphorylated recombinant OPN (27). Furthermore, a synthetic unmodified peptide corresponding to the C-terminal 18 amino acids was able to bind these cells, whereas neither a phosphorylated nor a scrambled variant had this effect (13). This suggests that phosphorylation of the extreme C-terminal region of OPN influences its cell binding capabilities.
In the present study we show that phosphorylation and other modification of the highly conserved C-terminal region of OPN inhibits its interaction with the ␣ V ␤ 3 -integrin. Adhesion mediated by C-terminally modified OPN was increased significantly after thrombin and plasmin cleavage, whereas cell attachment promoted by OPN not modified at the C terminus was only slightly increased by proteolytic removal of the C terminus. This indicates that the increase in adhesive properties is a result of removal of the modified C terminus rather than improved exposure of the RGD sequence. Collectively, these results point toward a novel mechanism by which the interaction between OPN and the ␣ V ␤ 3 -integrin can be regulated by modification of the extreme C-terminal region of the protein.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-The MDA-MB-435 human melanoma cell line (a kind gift from Dr. David T. Denhardt, Rutgers University, New Brunswick, NJ), human embryonic kidney 293T cells, Mo␣ V cells (from Mark H. Ginsberg), and the embryonic monkey kidney cell line MA104 were all maintained in Dulbecco's modified Eagle's medium (DMEM) with Glu-taMAX (Invitrogen) supplemented with 10% fetal bovine serum, and 1% antibiotics (penicillin/streptomycin). Mo␣ V is a subclone of M21 melanoma cells and has previously been shown to express high levels of the ␣ V ␤ 3 -integrin (28).
Plasmid Construction-For generation of C-terminally tagged full-length OPN (rOPN-C), human cDNA encoding residues 1-314 of OPN (the first 16 residues is the signal peptide which is not included in Figs. 1 and 2) was amplified by PCR using the forward primer 5Ј-GCTTGGATCCCTACCATGA-GAATTGCAGTG-3Ј, which contained a BamHI restriction site (underlined) and the first 5 amino acids of the signal peptide. The reverse primer was 5Ј-CTCGACTCAAGCTT-ATTGACCTCAGAAGATGCAC, which contained a HindIII restriction site (underlined) and the last 6 C-terminal residues of OPN. For generation of a C-terminally tagged fragment consisting of residues 1-228 (rOPN-228C), human cDNA was amplified by PCR using the same forward primer as above and the reverse primer 5Ј-GATCACGGAAGCTTTCTGGACT-GCTTGTGGCTGTG-3Ј, which contained a HindIII restriction site (underlined) and the residues HSHKQSR 228 of OPN. The amplified fragments were cloned into the BamHI/HindIII sites of pcDNA3.1/Myc-His(-)A (Invitrogen). These constructs encoded OPN residues 1-298 (full-length) and the fragment consisting of residues 1-228, both followed immediately by residues KLGP, the myc epitope (EQKLISEEDL), residues NSAVD and H 6 , increasing the molecular mass by 2.9 kDa.
For generation of N-terminally tagged full-length OPN, the residues 1-298 of OPN (without the signal peptide) were amplified by PCR using the forward primer 5Ј-GCTAGCA-GAGGATCGCATCACCATCACCATCACGGGGGGATACC-AGTTAAACAGGCTGATTCTGG-3Ј, which include an NheI restriction site (underlined), a H 6 tag (italic), and the first 8 residues of the secreted OPN sequence (I 1 PVKQADS). The reverse primer was 5Ј-GACCTCGAGCTACTAATTGACCT-CAGAAGATGCACT-3Ј, which include a restriction site for XhoI (underlined), two stop codons (italic), and the last 7 C-terminal residues of OPN. The amplified fragment was cloned into the NheI/XhoI sites of pEXPR-IBA42 (IBA BioTAGnology). The pEXPR-IBA42 vector contains the BM40 signal sequence resulting in secretion of the construct. This construct encoded residues ASRGS, H 6 , and the residues GG, increasing the mass with 1.4 kDa, followed by the full-length OPN sequence (residues 1-298).
Plasmids were propagated in Escherichia coli DH5␣ cells, and all constructs were verified by sequence analysis. Plasmid DNA for transfection was prepared using a NucleoSpin Plasmid QuickPure kit (Macherey-Nagel).
Protein Expression and Purification-Human embryonic kidney 293T cells were plated on 10-cm tissue culture dishes and were transfected 18 h later by calcium phosphate co-precipitation (29) using 7.5 g of plasmid DNA. After a further 48 h, the supernatants were harvested and replaced by serumfree medium (293 SFM II; Invitrogen). Serum-free medium was harvested and changed three times at 48-h intervals.
His-tagged recombinant OPN was purified from serum-free medium (ϳ200 ml) using a metal-chelate affinity column (1 ml; Qiagen) charged with nickel ions. Bound protein was eluted with 250 mM imidazole in phosphate-buffered saline (PBS). The fractions containing OPN were identified by Western blotting and desalted on a PD10 column in 25 mM ammonium bicarbonate.
Full-length and N-terminal OPN were purified from human milk and urine as described (6,19). All purified proteins were analyzed by N-terminal sequencing, MALDI-TOF MS (matrixassisted laser desorption ionization time-of-flight mass spectrometry), and SDS-PAGE with Coomassie Brilliant Blue visualization. The amount of purified OPN was determined with a BCA protein assay kit (Pierce). OPN was digested with thrombin (30 milliunits/g OPN) (Sigma) and plasmin (0.2 milliunit/g OPN) (Roche Applied Science) in 0.1 M ammonium bicarbonate at 37°C for 1 h.
Dephosphorylation of OPN-For dephosphorylation, the different OPNs were incubated with bovine alkaline phosphatase (ALP) (20 milliunits/g OPN) (Sigma) in 10 mM ammonium bicarbonate (pH 8.5) overnight at 37°C. Native and dephosphorylated OPNs were analyzed by MS in positive linear ion mode using a Voyager DE-PRO MALDI-TOF mass spectrometer (Applied Biosystems) before and after ALP treatment. The phosphate content on mOPN was also analyzed with a phosphoprotein phosphate estimation assay kit (Pierce). Finally, to ensure that ALP-treated mOPN was free of phosphates, it was digested with trypsin (Worthington Biochemical) using an enzyme to substrate ratio of 1:30 (w/w) in 20 mM ammonium bicarbonate at 37°C for 6 h. Tryptic peptides were separated on a RPC C 2 /C 18 PC 2.1/10 column (narrow-bore reversedphase chromatography; GE Healthcare) connected to a GE Healthcare SMART system. Separation was carried out in 0.1% trifluoroacetic acid (buffer A) and eluted with a gradient of 60% acetonitrile in 0.1% trifluoroacetic acid (buffer B) developed over 54 min (0 -9 min, 0% B; 9 -49 min, 0 -50% B; 49 -54 min, 50 -100% B) at a flow rate of 0.15 ml/min. The peptides were detected in the effluent by measuring the absorbance at 214 nm and analyzed for phosphorylations by MS.
For adhesion assays, 100 g of milk and urinary OPN were incubated with ALP as described above. Dephosphorylated OPN was separated from ALP and buffer components by reversed-phase HPLC as described for separation of tryptic mOPN peptides.
Carbohydrate Analyses-For removal of simple O-linked glycans, OPN was incubated with O-glycanase (0.1 milliunit/g OPN) and sialidase (0.35 milliunit/g OPN) (both from Prozyme) in 50 mM sodium phosphate (pH 7.0) overnight at 37°C. The deglycosylation was analyzed by MS before and after treatment with the enzymes.
Full-length milk OPN was additionally hydrolyzed in 2 M trifluoroacetic acid for 4 h at 100°C under oxygen-free conditions. ␣-L-Rhamnose was added as internal standard. For quantification of the more stable amino sugars, N-acetylgalactosamine (GalNAc) and N-acetylglucosamine (GlcNAc), an additional OPN sample was hydrolyzed in 4 M HCl for 6 h at 110°C. For sialic acid analysis, the protein was hydrolyzed in 0.1 M HCl for 60 min at 80°C. Monosaccharides were separated by high pH anion-exchange chromatography using a CarboPac PA1 column (4 ϫ 250 mm) (Dionex) and monitored by pulsed electrochemical detection. The neutral monosaccharides were eluted isocratically with 16 mM NaOH, whereas sialic acid was eluted with 100 mM NaOH and 150 mM sodium acetate. Standard monosaccharide mixtures of ␣-L-fucose, D-galactosamine, D-glucosamine, D-galactose, D-glucose (all 0.1 mM), D-mannose (0.5 mM), ␣-L-rhamnose (0.2 mM), and N-acetylneuraminic acid (0.02 mM) (all from Sigma) were used. Linearity was observed between the injected amounts of monosaccharide standards and the peak areas, and hence the amounts of the individual monosaccharides in the samples were deduced from standard curves.
Cell Adhesion Assays-Flat-bottom 96-well tissue culturetreated polystyrene microtiter plates (Corning) were coated overnight at 4°C with OPN diluted in PBS to the appropriate concentrations and then blocked with 1% bovine serum albumin (BSA). The cells were trypsinized, then washed twice and resuspended in DMEM containing 1 mg/ml BSA. For blocking of integrin function, cells were preincubated for 30 min at 37°C with a neutralizing antibody against ␣ V ␤ 3 (clone LM609; Chemicon) or an IgG1 isotype control mAb (Sigma) (both at 5 g/ml), or with 100 M GRGDNP or GRADSP peptides (Enzo Life Science). Subsequently, cells (5 ϫ 10 4 ) were added to coated wells and incubated for 1.5 h at 37°C in a humidified atmosphere with 5% CO 2 . Nonadhered cells were removed by washing twice with 75 l of Percoll (Sigma) (73% Percoll, 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 0.5% toluidine blue and solubilized in 50 l of 0.5% Triton X-100 before reading at 630 nm using a Microplate Autoreader EL 311 (BioTek).
Analysis of OPN Coating-The 96-well plates used in the adhesion assays were coated overnight at 4°C with OPN diluted in PBS in the same concentration range as used for the adhesion assays. Subsequently, the wells were blocked with 1% BSA in PBS for 1 h at room temperature. All washes and dilutions after this were performed in PBS-Tween (10 mM disodium phosphate anhydrate (pH 7.2), 150 mM NaCl, 0.05% (v/v) Tween 20). The coated OPNs were detected by addition of 0.5 g of the monoclonal mouse anti-human OPN antibodies 2A1, 1H3, and 3D9 (27) (a kind gift from Dr. David T. Denhardt) followed by incubation with horseradish peroxidase-conjugated polyclonal goat anti-mouse IgG (Dako) (diluted 1:1000). Both antibodies were incubated for 1 h at 37°C. Color development was obtained with TMB-one substrate (Kem-En-Tec), and the reaction was stopped by addition of 0.2 M H 2 SO 4 . Color intensity was measured at 450 nm using an ELISA reader (BioTek).
Statistical Analysis-Statistical analysis of data was done by Student's t test. Difference was considered to be statistically significant at p Ͻ 0.05.

Purification, Expression, and Characterization of Human
OPNs-The human melanoma cell line MDA-MB-435 has previously shown a much greater adhesion to a less phosphorylated murine OPN variant (4 phosphates on average) from ras-transformed fibroblasts than to a more phosphorylated mouse OPN form secreted by osteoblasts (21 phosphates on average) (24). To analyze the role of OPN modification in cell adhesion further, several different human OPN forms were either purified from natural sources or expressed recombinantly in mammalian cells. Full-length OPN (mOPN) and a naturally occurring N-terminal fragment (nmOPN) (Ile 1 -Lys 154 ; Fig. 1) (19), were purified from milk and the urinary OPN isoform (uOPN) was purified from urine. Full-length recombinant human OPN with either an N-or a C-terminal tag (rOPN-N and rOPN-C) was produced in HEK293T cells together with a C-terminally tagged fragment consisting of amino acids 1-228 (rOPN-228C; see Fig. 1). Fig. 2 shows a schematic overview of the different OPN variants, and the evaluation of their purity by SDS-PAGE. The N-terminal fragment from milk migrates and stains as a broad indistinct band in SDS-PAGE (Fig. 2, lane 3). This pattern has also been observed in a previous study, although, the purified nmOPN was shown to be a well defined homogenous molecule when analyzed by MS and RP-HPLC (19). Similarly, below the dominating band representing rOPN-N (Fig. 2, lane 5) some faint bands are present. These represents fragments of OPN; however, RP-HPLC, MS, and Western blotting analyses (data not shown) showed that these fragments of rOPN-N only constituted very minor proportions of the protein in the sample.
The degree of phosphorylation of the different OPNs was estimated by MS before and after treatment with bovine ALP ( Table 1). The two native full-length OPNs differed in the level of phosphorylation, as mOPN contained ϳ26 phosphates whereas uOPN contained ϳ8 phosphates. The N-terminal fragment, nmOPN, showed an excess mass corresponding to ϳ13 phosphates. None of the recombinant proteins was phosphorylated in any significant degree. To confirm that ALP treatment resulted in complete dephosphorylation, mOPN was also analyzed by use of a phosphoprotein phosphate estimation kit. This revealed a total of 25 phosphate groups in mOPN corresponding well with the number of phosphorylations estimated by MS. Finally, MS analysis of tryptic peptides from ALP-treated mOPN showed that none of these contained phosphate groups (data not shown).
The excess mass between the dephosphorylated proteins and the calculated mass of the amino acid sequence indicated that all the OPNs were O-glycosylated. The surplus of 3-4 kDa for uOPN and the recombinant proteins suggested that the glycans on these proteins are composed of similar monosaccharides, whereas the excess mass of 8 -9 kDa for mOPN and nmOPN indicated that human milk OPN is decorated by larger glycan structures. The glycans on uOPN, rOPN-N, rOPN-C, and rOPN-228C were of the disialylated core-1 O-glycan sialic acid- galactose-[sialic acid]-GalNAc type, as the enzymes sialidase and O-glycanase were able to remove carbohydrates from these OPNs (data not shown). In contrast, the milk OPNs were unaffected by treatment with these enzymes. Instead, the carbohydrate composition of mOPN was determined by mild acid hydrolysis followed by high pH anion-exchange chromatography separation of the resulting monosaccharides. The hydrolysates showed the presence of the five monosaccharides Gal-NAc, GlcNAc, galactose, sialic acid, and fucose (Table 2). Thus, human milk OPN is most likely decorated by glycans composed of the GalNAc-galactose-[GlcNAc] core-2 structure extended by fucosylated, and to lesser degree sialylated, poly-N-acetyllactosamine units.
Adhesion of MDA-MB-435 Cells to OPN-The adhesion of the MDA-MB-435 cells to high and low phosphorylated OPN was investigated by comparing their binding to mOPN, nmOPN, and uOPN (Fig. 3A). The number of cells increased dose-dependently on all three OPNs at low coating concentrations (0 -50 nM), and the binding was saturated at high OPN concentrations (100 -200 nM). At binding saturation, similar numbers of MDA-MD-435 cells were adhering to uOPN and nmOPN, whereas the adhesion promoted by highly phosphorylated mOPN was approximately three times lower. The different adhesion pattern could be the result of altered cell surface receptor recognition of the different OPNs. However, in all cases the binding of the cells to OPN was RGD-dependent and via the ␣ V ␤ 3 -integrin, as RGD peptides and a blocking antibody against ␣ V ␤ 3 reduced the adhesion to a level similar to the negative BSA control, whereas RAD control peptides or an isotype IgG had no effect on the adhesion (Fig. 3B).
The orientation and exposure of functional sites on the various fragments and differently phosphorylated OPNs are important parameters when using assays involving binding

TABLE 1 Mass determination of native and recombinant OPNs
The observed molecular masses of the different OPNs were determined by MALDI-TOF MS in linear mode before and after dephosphorylation with ALP. The number of phosphates is calculated by this formula: (observed mass before ALP treatment Ϫ mass after ALP treatment)/80 Da, as the mass of one phosphate group is 80 Da. The mass difference is that between the observed mass after dephosphorylation and calculated average mass.

Protein
Theoretical   (Fig. 1) were tested for recognition of the different proteins. 2A1 and 1H3 recognized all OPNs equally well at a coating concentration of 100 nM, whereas 3D9 did not detect native mOPN or rOPN-228C (Fig. 3C). The recognition pattern of 3D9 was expected as this antibody does not bind the extreme C terminus of heavily phosphorylated OPN (27) and as rOPN-228C lacks the C terminus. Similar results were obtained with coating concentration of 1 nM, 10 nM, and 25 nM (data not shown). This indicates that the different OPNs bind to the surface in a comparable manner with similar orientation and exposure of the C terminus. Therefore, the different adhesion patterns could not be ascribed to differently binding of the OPNs to the plastic plates.
A high degree of phosphorylation of full-length OPN considerably reduced the ability of the protein to interact with the ␣ V ␤ 3 -integrin. The phosphates responsible for this inhibition seemed to be located in the C terminus, as nmOPN lacking the highly phosphorylated C terminus exhibited adhesion at the same level as low phosphorylated uOPN. The monoclonal antibody 3D9 recognized uOPN but not mOPN (Fig. 3C), illustrating that the extreme C terminus is unphosphorylated or modestly phosphorylated in uOPN and highly phosphorylated in mOPN. Following ALP treatment 3D9 recognized mOPN, indicating an unmodified C terminus. If phosphorylation of the C terminus inhibited the interaction between mOPN and the ␣ V ␤ 3 -integrin, then phosphatase treatment should increase its adhesive properties. Indeed, dephosphorylated mOPN promoted significantly higher adhesion of the MDA-MB-435 cells than its phosphorylated counterpart (Fig. 3D). On the contrary, no difference was observed in the binding of the cells to uOPN before and after dephosphorylation.
To examine further the influence of modification of the extreme C terminus in cell adhesion, the three recombinant OPNs (rOPN-N, rOPN-C, and rOPN-228C) with either an Nor C-terminal tag were compared in adhesion assays. A doseresponse curve showed that maximal MDA-MB-435 binding to all of the recombinant proteins was obtained at a coating concentration of 100 nM (data not shown). At this coating concentration, rOPN-C displayed approximately one third of the bind-  FEBRUARY 3, 2012 • VOLUME 287 • NUMBER 6

JOURNAL OF BIOLOGICAL CHEMISTRY 3793
ing promoted by rOPN-N or the fragment rOPN-228C (Fig.  4A). As was the case with the native OPNs, the adhesion to the recombinant OPNs was RGD-dependent via the ␣ V ␤ 3 -integrin. These results show that modification of the extreme C terminus of OPN at Asn 298 , by a myc-His tag in this case, significantly impairs the interaction between OPN and the ␣ V ␤ 3 -integrin. The modification in form of the myc-His tag in itself is not the reason for the inhibition, as rOPN-228C did not show mitigated binding to the cells. Likewise, an N-terminal His tag in rOPN-N was also showed to be irrelevant for the adhesion properties of the protein.
If modification of the C terminus can inhibit the adhesion mediated by OPN, it could be expected that separation of the Nand C-terminal parts of OPN by proteolytic cleavage would result in an increase in cell adhesion. Proteolytic cleavage of OPN by thrombin generates an N-terminal fragment (Ile 1 -Arg 152 ) containing the RGD sequence and a C-terminal fragment (Ser 152 -Asn 298 ). Plasmin predominantly generates the N-terminal fragment Ile 1 -Lys 154 , and in contrast to thrombin, plasmin completely degrades the C terminus (19). The adhesion mediated by mOPN and rOPN-C increased ϳ5-fold after thrombin and plasmin cleavage. On the contrary, cleavage of uOPN, rOPN-N, and rOPN-228C increased the adhesion to a much lesser degree (Fig. 4B). The cell binding to plasmin-and thrombin-cleaved OPN was also neutralized by both RGD-containing peptides and a blocking antibody against the ␣ V ␤ 3 -integrin (data not shown).
Adhesion of Mo␣ V and MA104 Cells to OPN-The effect of C-terminal modification on the interaction between OPN and the ␣ V ␤ 3 -integrin was further examined using Mo␣ V cells. The Mo␣ V cells are a substrain of human melanoma cells and have previously been shown to express high levels of the ␣ V ␤ 3 -integrin (28). As shown for MDA-MB-435 cells, adhesion of these cells to OPN were also completely RGD-dependent and via the ␣ V ␤ 3 -integrin (Fig. 5A). Furthermore, the Mo␣ V cells bound to uOPN and rOPN-N with similar efficiency, which was approximately two times greater than the binding to mOPN and rOPN-C (Fig. 5B). Next, MA104 embryonic monkey kidney cells were investigated to see whether the tendency also applied to cells that were not of melanoma origin. Adhesion of the kidney cells was inhibited ϳ75% by the ␣ V ␤ 3 -integrin-blocking antibody, and addition of RGD-containing peptides reduced the binding to a level comparable with the negative BSA control (Fig. 5C). This showed that the ␣ V ␤ 3 -integrin is the principal OPN receptor on the MA104 cells. The kidney cells exhibited approximately ϳ25-40% higher adhesion to uOPN and rOPN-N compared with mOPN and rOPN-C, respectively (Fig. 5D). Thus, the interaction between OPN and other types of ␣ V ␤ 3 -integrin-expressing cells was also mitigated by high phosphorylation of the C terminus or the presence of a tag at the C-terminal Asn 298 .

DISCUSSION
In the present study modification of the C terminus in terms of both phosphorylation and a covalently attached tag to Asn 298 was demonstrated to significantly reduce the adhesion of different cell lines to human OPN via the ␣ V ␤ 3 -integrin. This indicates a potential new mechanism for regulation of the interaction between OPN and integrins. The alignment shown in Fig. 1 illustrates that the extreme C-terminal region of OPN is highly conserved among mammalian species. Several functional important motifs like phosphorylation sites, glycosylation sites, the RGD sequence, and the cryptic integrin binding motif SVVYGLR are highly conserved. The conservation of the C terminus suggests that this region may also be important for the function of OPN.
OPN is O-glycosylated at 3-5 conserved threonines in an unphosphorylated region to the N-terminal side of the RGD sequence (6,(21)(22)(23)(24). All the OPN isoforms used in the present study were also observed to be O-glycosylated. The carbohy- drates on the recombinant OPNs and uOPN consisted of a disialylated GalNAc-galactose core. This glycan structure has previously been described in uOPN and in other isoforms of the protein from bone and fibroblasts (6,23,24). The excess mass after dephosphorylation and the carbohydrate composition analysis of human milk OPN showed that this isoform is glycosylated differently from the other OPNs. The presence of large fucosylated N-acetyllactosamine units on mOPN shown here has also been demonstrated in another study (30). The difference in glycan structure between the OPNs is nevertheless not likely to cause the observed functional differences because milk OPN lacking the C terminus (nmOPN and proteolytically cleaved mOPN) but still containing the large O-glycan structures promoted strong cell attachment; furthermore, rOPN-C decorated by the small glycans exhibited only minor cell adhesion.
The significant reduction of ␣ V ␤ 3 -integrin-mediated cell adhesion to C-terminally modified OPN was consistent using two different melanoma cell lines. The tendency also applied to other cell types as seen for the MA104 kidney cells. The difference between the OPNs was not as pronounced for the kidney cells as for the melanoma cells. This could be explained by the presence of several OPN-binding integrins on the MA104 cells, as complete inhibition of adhesion was not obtained using anti-␣ V ␤ 3 -integrin antibodies. In both melanoma cell types, complete inhibition of cell adhesion was obtained when the ␣ V ␤ 3integrin was blocked.
The adhesion mediated by highly phosphorylated OPN via the ␣ V ␤ 3 -integrin on the MDA-MB-435 melanoma cells was negligible and increased after dephosphorylation; however, previous studies have shown that partial enzymatic dephosphorylation of OPN from bovine milk and bone results in decreased ␣ V ␤ 3 -integrin-dependent adhesion of osteoclasts (25,31). Similarly, in vitro phosphorylation of recombinant (nonphosphorylated) rat OPN increased the number of osteoclasts adhering to OPN, although only 1-2 phosphates were incorporated by the kinase (32). These studies all used in vitro enzymatic treatment to either add or remove phosphates without characterization of the resulting products. In addition, comparison of results across different experimental set-ups using different OPN species varying in amino acid sequence can also be difficult. In the present study thoroughly characterized human OPN forms were utilized, thereby making a direct comparison in functionality more straightforward. Consistent with the present study, a direct comparison of two well characterized murine OPNs showed that a less phosphorylated isoform containing ϳ4 phosphates supported much higher adhesion of the MDA-MB-435 cells, also used in this study, than a more phosphorylated form containing ϳ21 phosphates (24). Monoclonal antibodies directed against the extreme C-terminal region of OPN were able to inhibit OPN-mediated cell adhesion (27). These findings are in agreement with the present study and support our conclusion that modification of the C terminus of OPN can influence the RGD-mediated interaction with the ␣ V ␤ 3 -integrin.
The mechanism by which phosphorylation of the C terminus can regulate the integrin-mediated cell binding of OPN is not clear. It has been demonstrated that OPN can mediate insideout signaling through the CD44 receptor resulting in increased RGD-dependent integrin adhesion (33). An unidentified region in the C-terminal half of OPN has been implicated as the CD44 binding site (34); hence, the extreme C terminus of OPN could potentially contain the CD44 binding site. If this were the case, phosphorylation of the C terminus could affect CD44 binding and thereby indirectly regulate RGD-dependent integrin activation. However, both OPN containing (uOPN and rOPN-N) as well as lacking (nmOPN and rOPN-228C) the C terminus promoted strong cell attachment, which indicates that involvement of CD44 inside-out signaling is not responsible for cell binding. OPN is considered to be an intrinsically disordered protein; however, the protein has been shown to exhibit a long range intramolecular interaction between the N-and C-terminal regions (10), and recently it was demonstrated that OPN consists of several segments, e.g. the integrin binding region, that display local secondary structure (11). In addition, ␤-sheet conformations of the region immediately after the RGD sequence and of the C-terminal part of OPN have been predicted to bring these two regions in proximity of each other (13). Phosphorylation of the C terminus could potentially have an influence on the suggested structural elements in the vicinity of the RGD sequence and thereby influence the integrin binding properties of the OPN.
The adhesion mediated by mOPN and rOPN-C increased 4 -5-fold after thrombin and plasmin cleavage. This illustrates that the inhibition of OPN functionality by C-terminal modification can be eliminated by removal of the C terminus after cleavage close to the RGD sequence. The increased cell binding after proteolytic cleavage cannot only be ascribed to exposure of the RGD sequence because only small increases in adhesion after thrombin and plasmin cleavage of uOPN, rOPN-N, and rOPN-228C were observed. Furthermore, involvement of integrins recognizing the SVVYGLR motif exposed after cleavage of the OPNs is also not likely because all adhesion to both cleaved and uncleaved OPN was neutralized by RGD-containing peptides and antibodies against the ␣ V ␤ 3 -integrin.
Adhesion of various cell types via the ␣ V ␤ 3 -integrin to different full-length OPN forms has in some cases been demonstrated to be very negligible and increases considerably after protease cleavage (18 -20), whereas other reports have observed either strong adhesion to full-length OPN or no dif-ference between cleaved and uncleaved OPN (12,26,27). The novel mechanism observed in this study for melanoma and kidney cells, in which phosphorylation of the C terminus can inhibit the interaction between OPN and the ␣ V ␤ 3 -integrin, might offer a possible explanation for these seemingly conflicting observations. However, whether the described mechanism also is true for adhesion to osteoclasts which are the classical cells that bind OPN via the ␣ V ␤ 3 -integrin remains to be investigated. Different structural requirements for binding of several integrins to OPN have been demonstrated (12,35). Further studies are needed to clarify whether modification of the C terminus affects the interaction between OPN and other integrins than ␣ V ␤ 3 .
In summary, we have shown that modification of the extreme C terminus can regulate OPN interaction with the widely expressed ␣ V ␤ 3 -integrin. The presented data indicate that enhanced adhesion mediated by OPN via this integrin after proteolytic cleavage probably is the result of removal of an inhibitory part of the protein, e.g. the C terminus rather than better exposure of the RGD sequence. This adds a new mechanism to the in vivo control of this multifunctional protein.