Identification of Dual a 4 b 1 Integrin Binding Sites within a 38 Amino Acid Domain in the N-terminal Thrombin Fragment of Human Osteopontin*

Previous work from our laboratory demonstrates that the a 4 b 1 integrin is an adhesion receptor for OPN and that a 4 b 1 binding site(s) are present in the N-terminal thrombin fragment of osteopontin (OPN) (Bayless, G. A., M., and G. E. J. Cell Sci. 111, 1165–1174). The work presented here identifies two a 4 b 1 binding sites within a recombinantly produced N-terminal thrombin fragment of human OPN. Initial experiments, using wild-type OPN containing an RGD sequence or an OPN-RGE mutant, showed identical a 4 b 1 -dependent cell adhesive activity. A strategy to localize a 4 b 1 binding sites within the thrombin fragment of osteopontin involved performing a series of truncation analyses. Removal of the last 39 amino acids (130– 168) completely eliminated adhesion, indicating all binding activity was present within that portion of the molecule. Combined mutation and deletion analyses of this region revealed the involvement of dual a 4 b 1 bind- ing sites. Synthetic peptides for both regions in OPN, ELVTDFPTDLPAT (131–143) and SVVYGLR (162–168), were found to block a 4 b 1 -dependent adhesion. The first peptide when coupled to Sepharose bound the a 4 b 1 integrin directly whereas a mutated ELVTEFPTELPAT peptide showed a dramatically reduced ability to bind. These data collectively demonstrate that dual a 4 b 1 inte- grin

Osteopontin (OPN) 1 is an extracellular matrix protein originally isolated from bone (28), and much evidence has accumulated as to its role in bone physiology (29). OPN is also secreted by many epithelial surfaces (30), and one early study supported a role for OPN in host-response to bacterial infection (31). Accumulating evidence also indicates OPN secretion is involved in inflammation and tumor progression. OPN has previously been found to be up-regulated in a variety of inflammatory, cardiovascular, and infectious diseases (32)(33)(34)(35)(36)(37) and is a major secreted product of macrophages in inflammatory settings (32)(33)(34)(35)(36). It is also associated with tumors (38 -41), particularly at the tumor-host interface (42). Recent data from knockout mice show that OPN facilitates wound healing (43), aids in host defense against viral (44) and bacterial infections (44,45), and is involved in granuloma formation (44). Other studies have suggested OPN may facilitate tumor cell metastasis (46) and decrease complement-mediated tumor cell destruction (47). Based on these data, the presence of OPN in the wound environment likely plays an important role in regulating disease progression in inflammatory and other conditions. The molecular domains in OPN that mediate its effects in these phenomena remain to be defined.
The parallels between expression of the ␣ 4 ␤ 1 leukocyte integrin and expression of OPN in wounds prompted a previous study by our laboratory to define osteopontin as a ligand for the ␣ 4 ␤ 1 integrin (14). Here, we define two binding sites for the ␣ 4 ␤ 1 integrin in the recombinant N-terminal thrombin fragment of human OPN using deletion and mutation analyses.

Preparation of Recombinant
Osteopontin; Cloning Strategy-A fulllength cDNA clone of the human osteopontin gene was obtained from the American Type Culture Collection (ATCC, Manassas, VA) (48), and the sequence is shown in Fig. 1A. Sequences encoding the wild-type N-terminal thrombin fragment of osteopontin (rOPN-(17-168)) were amplified by polymerase chain reaction using the primers 5Ј-TAGGA-TCCATACCAGTTAAACAGGCTGATTCTGGAAG-3Ј and 5Ј-GTAAGC-TTTTACCTCAGTCCATAAACCACACTATCACCTCGGCCA-3Ј(Genosys, The Woodlands, TX). Sequences encoding a mutated N-terminal fragment ([Glu 161 ]rOPN- ) in which the single RGD sequence at residues 159 -161 was changed to RGE were obtained by substituting 5Ј-TAAAGCTTTTACCTCAGTCCATAAACCACACTTTCACCTCGGC-CA-3Ј for the downstream primer used to generate the wild-type fragment. Restriction digests of the polymerase chain reaction products and the pQE 30 vector (Qiagen) were carried out overnight with BamHI and HindIII (Life Technologies, Inc.). Digested vector and insert were puri-* This work was supported by National Institutes of Health Grant HL59971 (to G.E.D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed:   where the sequence RGSHHHHHHS replaces MRIAVICFCLLGITCA at the N terminus of wild-type osteopontin. All positive clones were confirmed by sequence analysis at Lone Star Labs (Houston, TX). All subsequent constructs studied contained the RGE mutation at amino acid 161.

Production and Characterization of Recombinant Wild-type and Mutated N-terminal Thrombin Fragments of Osteopontin-Escherichia coli
strain RY2840 [MC4100 lacI Q1 lac ϩ slyD Km r ] (49) was transformed with plasmids encoding the His 6 -tagged OPN derivatives. 2 ml of overnight cultures were inoculated into 200 ml of Luria-Bertani media (Life Technologies, Inc.) containing 50 g/ml ampicillin (Sigma). Cultures were grown to an A 600 of 1.0 (ϳ2.5 h) before induction with 0.5 mM isopropylthiol-␤-D-galactoside (IPTG-Life Technologies, Inc.). Cultures were allowed to incubate for 3.5-4 h at 37°C in a shaking incubator before being placed on ice for 15 min. Bacteria were pelleted, supernatants removed, and pellets frozen at Ϫ80°C. Pellets were thawed at 25°C for 10 min, resuspended on ice in 20 ml Hepes buffered saline, pH 8.1 containing 25 mM Hepes, 150 mM NaCl and 1 mM 4-(2-aminoethyl)benzene sulfonylfluoride, HCl (CalBiochem). Bacteria were lysed and debris pelleted (20,000 ϫ g at 4°C for 20 min) before adding supernatants to 2 ml of TALON metal ion affinity column (CLONTECH) equilibrated with Hepes buffer. Columns were incubated for a minimum of 20 min at 4°C before washing with 20-column volumes of Hepes buffer. His-tagged proteins eluted with 0.2 M imidazole (Sigma) in Hepes buffer, and fractions were dialyzed (M r cutoff 7,500) against 8 liters of phosphate-buffered saline. The purity of recombinant proteins was assessed by SDS-PAGE and Western blot analysis. Protein concentration was estimated according to the method of Pace et al. (50). Yields were ϳ6 mg per 200 ml of culture.
Cell Adhesion Assays-Cell adhesion assays were performed to determine the ability of OPN to promote leukocyte adhesion. Polystyrene microwells (Corning-Costar, Cambridge, MA) were coated with 50 l of bovine OPN purified as previously described (51) or recombinant fragments of OPN at a concentration of 20 g/ml in TBS overnight at 4°C. After blocking with 100 l of 10 mg/ml BSA (Sigma, St. Louis, MO) in TBS, wells were rinsed with PSA (Life Technologies, Inc., Grand Island, NY). HL-60 promyelocytic leukemia cells and Ramos cells (ATCC) were grown in RPMI 1640 (Life Technologies, Inc.) and 10% fetal calf serum. Human umbilical vein endothelial cells were grown in M199 (Life Technologies, Inc.) supplemented with heparin (Sigma), bovine brain extract (52), and 20% fetal calf serum (Life Technologies, Inc.). Leukocytes were rinsed and resuspended in PSA at a density of 100,000 cells/well and endothelial cells at 35,000 cells per well. Media for adhesion in all leukocyte experiments contained a final concentration of 100 g/ml BSA with physiological doses of CaCl 2 (2 mM) and MgCl 2 (1 mM). HL-60 cells were activated with the ␤ 1 -activating antibody, 8A2 (53) at a concentration of 1 g/ml and a phorbol ester, 12-0-tetradecanoyl phorbol 13-acetate at a concentration of 50 ng/ml. Endothelial cells were allowed to attach in the presence of 100 g/ml BSA with 1.5 mM CaCl 2 and 1.5 mM MgCl 2 . After plating, cells were allowed to adhere for 30 -60 min at which time they were rinsed and fixed with formalin. Plates were stained with 0.1% Amido Black for 5 min and rinsed and solubilized with 2 N NaOH to obtain an absorbance reading at 595 nm, which corresponds directly to the number of cells stained in each well (54).
Peptide Synthesis and Adhesion Blockade-To confirm the findings of the truncation studies, the SVVYGLR peptide (corresponding to C-terminal amino acids 162-168) was synthesized (Sigma-Genosys). Also generated were the wild-type peptide ELVTDFPTDLPATK and aspartate mutant ELVTEFPTELPATK, representing amino acids 131-143. The molecular weight of each peptide was confirmed by massspectral analysis (Sigma-Genosys). The synthetic peptides SVVYGLR, ELVTDFPTDLPATK, and ELVTEFPTELPATK were preincubated with cells at 250, 500, and 500 g/ml, respectively under activating conditions in the presence of divalent cations for 15 min. Following the incubation period, cells were seeded, and the assay was performed as described above.
Direct Integrin Binding using Affinity Chromatography-To illustrate the integrin-binding capacity of osteopontin, the synthetic peptides ELVTDFPTDLPATK and ELVTEFPTELPATK were coupled to cyanogen-bromide 4B (Sigma) at 5 mg/ml according to the manufacturer's instructions. Ramos cells (ATCC) were surface biotinylated as described (55), and a 50-l pellet of cells was extracted with 1 ml of TBS containing 3% octylglucoside (ICN, Irvine, CA) in 1.5 mM Mg 2ϩ , 1.5 mM Mn 2ϩ , and 10 Ϫ3 M phenylmethane sulfonic acid. The HL-60 cell extracts were agitated at 5-10 min intervals with Sepharose columns (0.5 ml) over a 2-h period at 0°C. The columns were washed with 5 ml of TBS containing 3% octylglucoside, 1.5 mM Mg 2ϩ , and 1.5 mM Mn 2ϩ . This was followed with a 15-ml wash in TBS containing 1% octylglucoside, 1.5 mM Mg 2ϩ , and 1.5 mM Mn 2ϩ . Integrins were eluted with 2 ml of TBS with 1% octylglucoside ϩ 10 mM EDTA (0.25-ml fractions). 40 l of each fraction were loaded and run under nonreducing conditions on a 7% acrylamide gel and transferred to polyvinylidene difluoride membrane (Millipore). The membrane was blocked overnight at 4°C with 5% milk in 0.1% Tween 20 saline containing 2.5 mM Tris-HCl, pH 7.5. Blots were washed and streptavidin alkaline phosphatase (Sigma) was added (1: 1000) to 1% BSA in Tween 20 saline and incubated for 30 min followed by a 30-min wash in Tween 20 saline. The alkaline phosphatase activity was developed using alkaline phosphatase development kit (Bio-Rad) and stopped with water.
Integrin Immunoprecipitation-Integrins that bound to the OPN-Sepharose column were identified using immunoprecipitation. Sepharose beads conjugated with goat anti-mouse IgG (Sigma) were rinsed and suspended 1:1 with 0.5% Triton X-100 in TBS. In 1.5-ml microcentrifuge tubes, 200 l of the bead mixture was added to 5 g of monoclonal antibodies against several human integrin subunits including ␣ 4 (HP2/1, Immunotech) (56), ␤ 1 (mAb13, Becton-Dickinson) (57), and ␣ 5 (IIA1, PharMingen) (58). These mixtures were then combined with 280 l of pooled EDTA eluate from OPN-Sepharose and 700 l of 0.5% Triton X-100 in TBS. This mixture was rotated continuously at 4°C overnight after which time tubes were centrifuged and rinsed six times with 1 ml of 0.5% Triton X-100 in TBS. 75 l of 2ϫ sample buffer was added to the beads, and this mixture was boiled for 5 min. 30-l samples were run on 7% SDS-PAGE under nonreducing conditions, and blots were developed as described above.

RESULTS
To rule out the involvement of the RGD site in ␣ 4 ␤ 1 -dependent adhesion to OPN, the wild-type, RGD-containing N-terminal thrombin fragment (rOPN-(17-168)) and an RGE mutant ([Glu 161 ]rOPN-(17-168)), where Asp 161 was mutated to Glu 161 , were produced. Each clone is described as rOPN followed by the amino acids coded for in the construct (e.g. 17-168) and finally the mutation incorporated into the clone (e.g. Glu 161 for Asp 161 mutated to Glu 161 ). Clones were sequenced to confirm successful mutation, and recombinant proteins (Ͼ95% purity) were analyzed using SDS-PAGE (not shown). Western blotting experiments using a monoclonal antibody directed to the N-terminal histidine tag revealed a pattern exactly matching staining results (not shown). Proteins were tested for their ability to promote ␣ v ␤ 3 -dependent attachment ( Fig. 2A). Wild-type rOPN-(17-168) promoted endothelial cell attachment dose dependently, whereas no attachment occurred to the RGE mutant. This confirmed successful functional mutation of the RGD site in OPN. The ability of both recombinant OPN constructs to promote ␣ 4 ␤ 1 -dependent adhesion was compared using the HL-60 promyelocytic cell line in the presence of physiologic divalent cations (2 mM Ca 2ϩ , 1 mM Mg 2ϩ ). As shown in Fig. 2B, no differences were observed in the ability of either construct to promote HL-60 cell attachment. Additionally, both rOPN-(17-168) and [Glu 161 ]rOPN-(17-168) were comparable with bovine OPN with respect to their ability to promote ␣ 4 ␤ 1 -dependent cell attachment (Fig. 2C). Adhesion to both native OPN and recombinant constructs was completely inhibited by the ␣ 4 ␤ 1specific LDV peptide. The control peptide, LEV, had lesser effects compared with control (no peptide). Minimal adhesion was observed to the BSA substrate. Collectively, these data indicate that the RGD site in the N terminus of OPN is not involved in ␣ 4 ␤ 1 -dependent adhesion to recombinant OPN, as rOPN-(17-168) and [Glu 161 ]rOPN- . Consequently, subsequent constructs described contain the RGE mutation (Glu 161 ) to rule out any influence from the RGD site in OPN, although the presence of this mutation is not reflected to simplify nomenclature.
A more detailed analysis of binding activity was conducted with the constructs shown in Fig. 1B. Dose-response curves illustrating the ability of constructs to promote HL-60 cell  (rOPN-(17-168)). Thus, deletion of the last 4 amino acids on the C terminus of the thrombin fragment of OPN partially eliminated the ability of ␣ 4 ␤ 1 -dependent cell attachment to occur. The remainder of binding activity was completely removed with further truncation of the molecule by ending at amino acid residue 129. These data strongly support the concept that two binding sites exist within residues 130 -168. To examine in more detail whether the Asp 135 and Asp 139 residues located within the upstream binding region identified were important in ␣ 4 ␤ 1 -dependent cell attachment, additional constructs incorporating mutations were generated (Fig. 4) rOPN-(17-142). Interestingly, neither of the mutations completely abolished adhesion, indicating that the conservative substitution of glutamate for aspartate did not remove all activity. Collectively, these data indicate that there are dual ␣ 4 ␤ 1 binding sites in the N-terminal thrombin fragment of OPN.
To confirm both sequences in osteopontin were capable of binding the ␣ 4 ␤ 1 integrin, the synthetic ELVTDFPTDLPATK and SVVYGLR peptides, corresponding to amino acids 131-143 (with a C-terminal lysine residue added) and 162-168, respectively, were tested for their ability to block ␣ 4 ␤ 1 -dependent attachment (Fig. 5). Also, the synthetic peptide ELVTEFPTEL-PATK was created containing two conservative Asp to Glu mutations to further examine whether the aspartate residues were involved in binding to ␣ 4 ␤ 1 . In the presence of both the ELVTDFPTDLPATK and SVVYGLR peptides, ␣ 4 ␤ 1 -dependent adhesion was inhibited significantly compared with control (p Ͻ 0.001). The ELVTEFPTELPATK peptide consistently had smaller effects compared with control, similar to that observed with the LEV peptide (see Fig. 2C).
As further evidence for direct interaction of amino acids 131-143 in OPN with the ␣ 4 ␤ 1 integrin, affinity chromatography experiments were performed using surface-labeled Ramos cell extracts (Fig. 6A). Wild-type ELVTDFPTDLPATK-Sepharose, aspartate mutant ELVTEFPTELPATK-Sepharose and blank-Sepharose beads were incubated with labeled extracts, and EDTA elutions were collected and analyzed (E1-E4). Results show strong binding of the ␣ 4 ␤ 1 integrin to the wild-type peptide, whereas minimal binding occurred to the aspartate mutant. No binding of the ␣ 4 ␤ 1 integrin was observed using blank-Sepharose. Immunoprecipitation of fractions from both wild-type and mutated peptide columns revealed the presence of ␣ 4 and ␤ 1 integrin subunits (Fig. 6B), although much greater binding occurred to the wild-type peptide. Mutation of aspartate residues resulted in a reduced ability of ␣ 4 ␤ 1 to bind but did not completely eliminate binding activity. In both experiments, control integrin antibodies failed to immunoprecipitate integrins. Similar results were observed using surface-labeled HL-60 cell extracts (data not shown).

DISCUSSION
Using sequential truncation analysis of the N-terminal thrombin fragment of OPN, we observed that dual ␣ 4 ␤ 1 integrin binding sites exist in a 38-amino acid C-terminal domain. Synthetic peptides encompassing either of these regions interfered with the ability of ␣ 4 ␤ 1 -dependent attachment to occur, whereas a control peptide had lesser effects. Also, using affinity chromatography we were able to demonstrate direct binding of the ␣ 4 ␤ 1 integrin to the wild-type synthetic peptide coupled to sepharose, whereas a mutated peptide bound considerably less well. These results show that dual binding sites exist for the ␣ 4 ␤ 1 integrin in the N-terminal thrombin fragment of OPN.
Integrin Binding Sites in Osteopontin-Numerous members of the integrin family have been reported to interact with OPN including ␣ v ␤ 3 (51, 60 -62), ␣ v ␤ 1 and ␣ v ␤ 5 (63,64), ␣ 4 ␤ 1 (14), (66) and ␣ 9 ␤ 1 (67). The RGD site (68) has been reported to interact with the ␣ v ␤ 3 , ␣ v ␤ 1 , ␣ v ␤ 5 and ␣ 5 ␤ 1 integrins (51, 60 -62). The ␣ 9 ␤ 1 integrin has been shown to bind the SVVYGLR amino acid sequence (59), which comprises the last 7 C-terminal amino acids in the thrombin fragment of OPN. These results are interesting based on sequence homology in that the ␣ 9 integrin subunit is most closely related to the ␣ 4 integrin subunit (69). Both the ␣ 4 ␤ 1 and ␣ 9 ␤ 1 integrins have the ability to interact with VCAM-1 (11,70) and OPN (14,59,67), as well as sharing other common ligands (71). Here, we found that deletion of the four C-terminal amino acids of the N-terminal thrombin fragment of OPN reduced but did not eliminate ␣ 4 ␤ 1 -dependent adhesion. Our result contrasts with those observed previously for ␣ 9 ␤ 1 interaction with OPN where deletion of the YGLR sequence completely eliminated ␣ 9 ␤ 1 -dependent cell attachment, and the synthetic SVVYGLR peptide completely blocked ␣ 9 ␤ 1 -dependent cell attachment (59). Previous work by Barry et al. (72) demonstrated that the synthetic SVVYGLR peptide also interfered with ␣ 4 ␤ 1 -dependent cell adhesion, reporting that the SVVYGLR site in OPN is a binding site for the ␣ 4 ␤ 1 integrin. In the same study, using recombinantly produced peptide sequences of the OPN molecule coupled to GST, it was postulated that an additional binding site may exist for ␣ 4 ␤ 1 , yet this was not demonstrated convincingly since the FPTDLPA synthetic peptide used in the study failed to block adhesion (72). The work presented here confirms that the SVVYGLR site in the N-terminal thrombin fragment of OPN is a binding site for the ␣ 4 ␤ 1 integrin. Truncation of the last 4 amino acids, YGLR, resulted in ϳ50% decrease in adhesion, and the synthetic SVVYGLR peptide significantly blocked ␣ 4 ␤ 1 cell attachment. We also define a second binding site from amino acids 131-143, ELVTDFPTDLPAT, as demonstrated by truncation analyses, peptide blockade, and direct integrin binding. Interestingly, this peptide contains a tandem sequence consisting of two copies of a consensus Asp-hydrophobic residue-Pro that is similar to the ␣ 4 ␤ 1 binding site LDVP seen in CS-1 FN (10). In addition, the related ␣ 4 ␤ 1 binding sequence hydrophobic residue-Asp-X-Pro is seen in other ␣ 4 ␤ 1 ligands such as QIDSPL in VCAM-1 and IDAPS in FN (15)(16)(17)(18)(19)(20). The one common feature of these sequences is the presence of a proline residue two-amino acids downstream of an aspartate residue.
Additionally, we present evidence for the direct involvement of aspartate residues in the affinity of ␣ 4 ␤ 1 for ELVTDFPTDL-PAT based on evidence that a mutant synthetic peptide, containing glutamate substituted for aspartate residues, was less effective at blocking cell adhesion. It also minimally bound the FIG. 5. Peptide blocking data confirming the presence of dual ␣ 4 ␤1 integrin binding sites in OPN. Activated HL-60 cells were preincubated for 15 min at 37°C with ELVTDFPTDLPATK (500 g/ ml), ELVTEFPTELPATK (500 g/ml), and SVVYGLR (250 g/ml) peptides, added at equimolar concentrations. Control indicates the absence of peptide. Experiments were performed as in Fig. 3. Values represent averaged absorbance readings compared with control from four separate experiments performed in triplicate wells Ϯ S.D. *, p Ͻ 0.001 compared with control using Student's t test.
FIG. 6. Affinity chromatography data illustrating binding of the ␣ 4 ␤ 1 integrin to synthetic OPN peptides. A, ELVTDFPTDLPATK-, blank-, and ELVTEFPTELPATK-Sepharose columns were incubated with surface-biotinylated Ramos cell extracts as described under "Experimental Procedures." Half-ml EDTA elution fractions were collected, analyzed using Western blots, and developed for alkaline phosphatase activity. Elution pattern for all columns is shown (E1-E4). Upper arrows denote the ␣ subunit, whereas lower arrows denote the ␤ subunit. B, immunoprecipitations were performed with fractions from both ELVTDFPTDLPATK-and ELVTEF-PTELPATK-Sepharose columns. Monoclonal antibodies directed to the integrin subunits tested for are indicated above each figure. These are A (no antibody), ␣ 4 -subunit (HP2/1), ␤ 1 -subunit (mAb13) and ␣ 5 (IIA1). Upper arrows denote the ␣ 4 -subunit, whereas lower arrows denote the ␤ 1 -subunit. ␣ 4 ␤ 1 integrin in affinity chromatography experiments. Conservative substitution of glutamate for aspartate residues did not remove 100% activity in either peptide blocking studies or direct integrin binding experiments. These results are consistent with previous data from our laboratory where the control LEV peptide also had slight effects (Ref. 14 and Fig. 2C). These results correlate with previous evidence that the ␣ 4 ␤ 1 integrin recognizes a wide variety of motifs in FN, VCAM-1, OPN and denatured proteins (10,(15)(16)(17)(18)(19)(20)(21). As was previously suggested, this integrin shows a broader ligand binding specificity than most other members of the integrin family (13).
Evidence That the Thrombin Fragment of OPN Is a Matricryptin-During tissue injury, considerable alterations occur in the ECM because of enzymatic breakdown, multimerization, adsorption, mechanical forces, or denaturation to expose matricryptic sites (73). Matricryptic sites are defined as biologically active sites that are not revealed in the mature secreted form of the ECM molecules, but become exposed after structural or conformational alterations, and matricryptins represent biologically active fragments of ECM that contain exposed matricryptic sites (73). Ample evidence exists for the presence of both OPN and thrombin in injury sites (38,74), therefore increasing the likelihood that thrombin cleavage of OPN occurs in these settings (38,74). Several studies support the contention that exposure of matricryptic sites occurs within OPN following thrombin cleavage (Refs. 38, 62, 65, 67, 75; Fig. 7). Senger et al. (62) have found the thrombin fragment of OPN is more potent at promoting RGD-dependent attachment than the native protein. The immobilized thrombin fragment of OPN also stimulated greater haptotactic migration of tumor cells compared with the intact molecule (75). The ␣ 9 ␤ 1 (67) and ␣ 5 ␤ 1 (65) integrins were unable to bind OPN without thrombin cleavage. These data support the concept that matricryptic sites liberated in OPN following thrombin cleavage may be important in regulating inflammatory cell interactions with OPN in the wound environment.
An interesting feature of the above findings is that all three known integrin binding sites are located within a very limited 38-amino acid region of the thrombin fragment of OPN. The ELVTDFPDLPAT (shown here), RGD (51, 60 -62, 67) and SV-VYGLR sites (shown here and in Refs. 59,72) are all localized to a 38-amino acid region just proximal to the thrombin cleavage site (Fig. 7). These sequences are highly conserved, particularly in large species of mammals, whereas they are less conserved or absent in rodents, particularly concerning the upstream binding site. The thrombin fragment appears to have altered biological activity compared with intact OPN and contains matricryptic sites (62, 65, 67, 73, 75). Localization of these integrin binding sites directly adjacent to the thrombin cleavage site strongly implicate the physiological importance of this region of osteopontin in inflammatory and wound repair responses.