Adhesive and Migratory Effects of Phosphophoryn Are Modulated by Flanking Peptides of the Integrin Binding Motif

Phosphophoryn (PP) is generated from the proteolytic cleavage of dentin sialophosphoprotein (DSPP). Gene duplications in the ancestor dentin matrix protein-1 (DMP-1) genomic sequence created the DSPP gene in toothed animals. PP and DMP-1 are phosphorylated extracellular matrix proteins that belong to the family of small integrin-binding ligand N-linked glycoproteins (SIBLINGs). Many SIBLING members have been shown to evoke various cell responses through the integrin-binding Arg-Gly-Asp (RGD) domain; however, the RGD-dependent function of PP is not yet fully understood. We demonstrated that recombinant PP did not exhibit any obvious cell adhesion ability, whereas the simultaneously purified recombinant DMP-1 did. A cell adhesion inhibitory analysis was performed by pre-incubating human osteosarcoma MG63 cells with various PP peptides before seeding onto vitronectin. The results obtained revealed that the incorporation of more than one amino acid on both sides of the PP-RGD domain was unable to inhibit the adhesion of MG63 cells onto vitronectin. Furthermore, the inhibitory activity of a peptide containing the PP-RGD domain with an open carboxyl-terminal side (H-463SDESDTNSESANESGSRGDA482-OH) was more potent than that of a peptide containing the RGD domain with an open amino-terminal side (H-478SRGDASYTSDESSDDDNDSDSH499-OH). This phenomenon was supported by the potent cell adhesion and migration abilities of the recombinant truncated PP, which terminated with Ala482. Furthermore, various point mutations in Ala482 and/or Ser483 converted recombinant PP into cell-adhesive proteins. Therefore, we concluded that the Ala482-Ser483 flanking sequence, which was detected in primates and mice, was the key peptide bond that allowed the PP-RGD domain to be sequestered. The differential abilities of PP and DMP-1 to act on integrin imply that DSPP was duplicated from DMP-1 to serve as a crucial extracellular protein for tooth development rather than as an integrin-mediated signaling molecule.

Phosphophoryn (PP) (alternatively referred to as dentin phosphoprotein or dentin phosphophoryn), which is the carboxylterminal cleaved product of DSPP, contains the RGD domain and repeat sequences of Ser-Ser-Asp (SSD) (there are over 200 tandem copies in humans and approximately 100 copies in mice) (Fig. 1B). Most SSD repeats are phosphorylated and may be some of the most acidic proteins in the human body [5,6]. The Dspp gene is known to be primarily expressed in odontoblasts and, to a lesser extent, in osteoblasts [7,8]. Dspp is also expressed in other tissues such as the salivary glands, lungs, and kidneys [9][10][11]. Functional analyses in genetically altered mouse models mainly elucidated the function of DSPP as an inducer of mineralization in the extracellular matrix [12][13][14]. An in vitro overexpression study revealed that PP induced mineral nodule formation, even in NIH3T3 fibroblast cells [15].
DMP-1 was found to be the most similar to DSPP among the SIBLING members, and these share many similarities in both their gene and protein structures and play important roles in the development of hard tissue (Fig. 1C) [12,[16][17][18][19][20][21][22]. A previous study indicated that DSPP was created due to gene duplications in the ancestor DMP-1 genomic sequence of toothed animals [23]. DSPP and DMP-1 are both cleaved into two protein chains; the N-terminal regions are proteoglycans that contain chondroitin sulfate chains, and the C-terminal regions are highly phosphorylated. As shown in Figure 1C, PP and carboxyl-terminal DMP-1 (C-DMP-1) both contain the integrin binding site RGD, which is colored red, while PP also includes long SSD repeats, which are colored green. DMP-1 was previously shown to aid adhesion to various cells through integrin receptors [24]. Bone morphogenetic protein 1 (BMP-1) and its alternatively spliced isoform, tolloid (TLD) are known to cleave full-length DMP-1 and DSPP proteins into two proteins [25][26][27][28].
Yamakoshi et al. recently proposed that DSPP should be classified into intrinsically disordered proteins (IDPs) due to its high net charge and low hydrophobicity [29,30]. IDPs generally do not adopt a defined three-dimensional structure, but, nevertheless, possess important functions in vivo and in vitro [31]. Since IDPs are known to be highly susceptible to bacterial proteases due to their flexible unfolded structures [32], the mammalian expression system is considered to be better for generating recombinant PP proteins. However, Marschall et al. reported that only small amounts of PP-related proteins were secreted from transfected mammalian cells due to their extremely acidic nature and SSD repeats [25]; therefore, the purification of recombinant PP proteins by a mammalian expression system was considered to be difficult. In the present study, we successfully generated recombinant PP using a mammalian expression system and evaluated its integrinmediated adhesive effects by simultaneously analyzing the effects of recombinant C-DMP-1 and the well-known integrin ligand vitronectin. Wells coated with recombinant PP did not facilitate cell adhesion, whereas recombinant C-DMP-1 and vitronectin did. Further analyses utilizing various recombinant proteins and peptides containing PP-RGD indicated that the Ala-Ser site flanking the RGD domain was a key peptide bond that allowed the PP-RGD domain to be sequestered.

Results
Generation of a rabbit anti-PP antibody and recombinant PP (rPP) protein We first generated an affinity-purified rabbit anti-PP polyclonal antibody to detect rPP. The carboxyl-terminal amino acid sequences of PP (DSEGSDSNHSTSDD) were selected as the antigen peptide based on low sequence similarities. The rabbit anti-PP antibody was generated by serial vaccinations using the antigen peptide as described. To examine the antigen recognition capacity of the rabbit anti-PP antibody, the affinity-purified rabbit anti-PP antibody, whole antisera (unpurified anti-PP antisera), and column flow-through solution were titrated using an enzymelinked immunosorbent assay (ELISA) with serial dilutions ranging from 1:1,000 to 1:64,000. Relative to the column flow-through solution, the affinity-purified rabbit anti-PP antibody as well as whole antisera showed a positive dilution from 1:64,000, which indicated that the rabbit anti-PP antibody retained its antigen recognition ability following affinity purification ( Fig. 2A). To examine the specificity of this rabbit anti-PP antibody, a dentin extract was dot-blotted onto a nitrocellulose membrane and PP was detected with the rabbit anti-PP antibody. The rabbit anti-PP antibody only reacted with the dentin extract of wild type mice and not with that of DSPP-null mice. The rabbit anti-PP antibody reacted with neither bovine serum albumin (BSA) nor Dulbecco's Phosphate-Buffered Saline (DPBS) (Fig. 2B).
As shown in Figure 2C, rPP appeared as a single band that migrated to 60 kDa on SDS-PAGE, as observed using Stains-All staining, and the molecular size of this single band was concomitant with the protein band identified by the anti-PP antibody and anti-6xHis antibody (data not shown). rC-DMP-1 proteins appeared as a single band that migrated to 50 kDa on SDS-PAGE, as observed using Stains-All and Coomassie brilliant staining, and the molecular size of this single band was consistent with the protein band identified by the anti-6xHis antibody (data not shown). These results confirmed that the purities of rPP and rC-DMP-1 were suitable for further experimental use.

Differential cell adhesive abilities of rPP, rC-DMP-1, and vitronectin
Utilizing MG63 and MC3T3-E1 cells (unless otherwise noted, MC3T3-E1 cells refer to subclone 4 in this study), the cell adhesive ability of rPP was simultaneously examined with rC-DMP-1 and vitronectin. These cells were unable to attach to wells coated with 20 or 100 nM rPP, but did attach to wells coated with 20 or 100 nM rC-DMP-1 and vitronectin. The adhesive potency of these cells was more profound on vitronectin than on rC-DMP-1 ( Fig. 3A and B). We then evaluated the effects of MnCl 2 , CaCl 2 , and MgCl 2 on cell adhesion to these proteins. As shown in Figure 3C and D, none of the divalent cations enhanced the cell adhesion of MG63 and MC3T3-E1 cells to rPP. In contrast, the addition of MnCl 2 strongly potentiated cell adhesion to rC-DMP-1.
Six other cells were also unable to attach to rPP Since rPP was incapable of aiding cell adhesion to human osteosarcoma MG63 cells or mouse pre-osteoblastic MC3T3-E1 cells, we examined various cell adhesions on rPP in parallel with rC-DMP-1 and vitronectin with the addition of MnCl 2 . As shown in Figure 4, three human dental pulp (hDPC) cells, the human osteosarcoma cell line, SaoS2, parental heterogeneous MC3T3-E1 cells, and mouse myoblast cell line, C2C12 were clearly unable to attach to wells coated with 100 nM rPP, but did attach to wells coated with 100 nM rC-DMP-1 and vitronectin, which was similar to the results obtained for MG63 and MC3T3-E1 cells.
The removal of SSD repeats in PP had no apparent influence on its adhesive potency The most characteristic feature of PP is the SSD repeat in its central portion. Therefore, we speculated that SSD repeats may somehow negatively affect cell adhesion to mask the RGDdependent adhesive potency of rPP. To examine this hypothesis, we generated rPP-DSSD in which the SSD repeats were removed (the deduced amino acid sequences of rPP and rPP-DSSD were shown in Fig. 5A). As shown in Figure 5B and C, MG63 and MC3T3-E1 cells were unable to attach to PP-DSSD, which was consistent with the results obtained for rPP. These results indicated that the lack of binding potency by rPP could not be attributed to the SSD repeats.
rC-DMP-1 and vitronectin, but not rPP directly associated with integrin avb3 and avb5 Since rC-DMP-1 and vitronectin were able to support MG63 cell adhesion, we pre-incubated these cells with neutralizing antibodies against human integrin b1, avb3, and avb5 before seeding onto rC-DMP-1 and vitronectin (Fig. 6A). The number of cells that attached to rC-DMP-1 was significantly lower following the preincubation with antibodies against integrin avb3 and avb5 than that with control IgG. Moreover, the co-addition of neutralizing antibodies against avb3 and avb5 more potently reduced the number of attached cells. The number of cells that attached to vitronectin was profoundly inhibited by the preincubation with the neutralizing antibody against integrin avb5 and was slightly inhibited by the preincubation with the neutralizing antibody against integrin avb3. The preincubation with the neutralizing antibody against integrin b1 was unable to inhibit MG63 cell adhesion to either rC-DMP-1 or vitronectin. Therefore, we hypothesized that the inability of rPP to facilitate MG63 cell adhesion may be attributed to the inability of rPP to associate with integrin avb3 and avb5. We examined the binding abilities of rPP, rPP-DSSD, rPP-RGE, which was the RGD-inactivated mutant of rPP, rC-DMP-1, and vitronectin to integrin avb3 and avb5. As shown in Figure 6B, integrin avb3 and avb5 were able to associate with rC-DMP-1 and vitronectin in dose-dependent manners, with the binding capability of vitronectin being higher than that of rC-DMP-1. However, integrin avb3 or avb5 did not bind to rPP, rPP-DSSD, or rPP-RGE. This result demonstrated that the adhesive inability of rPP was presumably due to its inability to bind to the integrin receptors.

Identification of flanking amino acid sequences to allow the RGD domain to become inactive
We hypothesized that a specific flanking amino acid sequence must exist in the vicinity of the RGD domain to sequester the function of PP-RGD. Therefore, we generated various kinds of mouse PP peptides, as shown in Table 1: H-SESANESGSRG-DASYTSDESS-OH (closed RGD), H-SANESGSRGDA-SYTSDE-OH (closed 7-RGD-7), H-NESGSRGDASYTS-OH (closed 5-RGD-5), H-SGSRGDASY-OH (closed 3-RGD-3), and H-SRGDA-OH (1-RGD-1). These peptides contained the RGD domain at their centers and 9, 7, 5, 3, and 1 juxta-amino acids were attached at both the amino-and carboxyl-terminal sides. We preincubated MG63 cells with these peptides before seeding onto vitronectin and then examined whether these peptides were able to inhibit the adhesion of MG63 cells to vitronectin. As shown in Figure 7A, MG63 cell adhesion to vitronectin was significantly lower when these cells were preincubated with 1-RGD-1 than with closed RGD, closed 7-RGD-7, closed 5-RGD, and closed 3-RGD-3. As shown in Figure 7B, we then generated a peptide with 2 amino acids at both sides of RGD (closed 2-RGD-2) ( Table 1) and found that the preincubation of MG63 cells with this peptide was unable to inhibit binding to vitronectin, which was also observed with closed 3-RGD-3. This result was supported by the same inability of the preincubation with peptides having two and three (closed 2-RGD-3) (Table 1), and three and two (closed 3-RGD-2) ( Table 1) amino acids at the amino-and carboxyl-terminals of RGD, respectively, to inhibit binding to vitronectin. We demonstrated that the incorporation of more than one amino acid sequestered the ability of the RGD domain. However, it remains unknown whether the incorporation of amino acids at both the amino-and carboxyl-terminal sides or at either one of these sides was sufficient to allow the PP-RGD domain to become inactive. Thus, we newly synthesized various kinds of PP peptides as shown in Table 1 H-DSSDSSDSSDSSDSSNSS-OH (SSD repeats) had the representative SSD repeats. H-PSGNGVEEDEDTGSGDGE-OH (DSP) was the well-conserved glycosaminoglycan attachment site in the dentin sialoprotein (DSP), which is the amino-terminal cleaved product of DSPP. As described above, we preincubated MG63 cells with these peptides and evaluated their RGD domain activities by analyzing their inhibitory effects on the adhesion of MG63 cells to vitronectin. The adhesion of MG63 cells was significantly inhibited by the preincubation with C-opened RGD, but not by that with its RGD-inactivated mutant, C-opened RGE. Ninety-six-well plates were precoated with 20 or 100 nM rPP, rC-DMP-1, and vitronectin, seeded with MG63 (A) and MC3T3-E1 (B) cells in serum-free medium, and incubated for 1 hr. MG63 (C) and MC3T3-E1 (D) cells were seeded onto 100 nM rPP, rC-DMP-1, and vitronectin with either MnCl 2 , CaCl 2 , or MgCl 2 (1 mM) in serum-free medium, and incubated for 1 hr. After washing non-adherent cells, the attached cells were stained with 0.2% crystal violet and dissolved in 1% SDS solution. Absorbance was measured at 570 nm. Each value represents the mean of triplicate determinations; bars mean 6SD. Statistical analysis was performed by a one-way ANOVA, followed by Dunnett's test. ***p,0.001 indicates significantly lower and + p,0.05 and +++ p,0.001 indicate significantly higher than rC-DMP-1-coated wells at the same concentration (A and B) or rC-DMP-1-coated wells with the same divalent cation (C and D). doi:10.1371/journal.pone.0112490.g003 The preincubation with N-opened RGD slightly inhibited cell adhesion; however, its effect was not as marked as that of Copened RGD. As expected, closed RGE and N-opened RGE did not alter the number of attached cells. SSD repeats and the DSP peptide did not affect the adhesion of MG63 cells to vitronectin.
We then performed a competitive integrin binding assay. As shown in Figure 7D, among closed RGD, closed RGE, C-opened RGD, C-opened RGE, N-opened RGD, N-opened RGE, SSD repeats, and the DSP peptides, the competitive addition of Copened RGD or N-opened RGD (20 nM) significantly reduced biotinylated GRGDS (5 nM) binding to integrin avb3. To determine the binding potencies of these peptides, we used serially-diluted cyclic-(GRGDSP) to compete with biotinylated GRGDS. Dose-dependent inhibition by cyclic-(GRGDSP) was regressed to a sigmoid curve (R = 0.99) and the inhibitory values of these peptides were determined by converting their absorbances to the inhibitory value of cyclic-(GRGDSP). The activities of 20 nM C-opened RGD and N-opened RGD were equal to 5.27 (26.5%) and 3.30 (16.5%) nM of cyclic-(GRGDSP), respectively. The activity of the same amount of cyclic-(GRGDSP) (20 nM) was assigned a value of 100%. Since neither closed RGD, closed RGE, C-opened RGE, N-opened RGE, SSD repeats, nor the DSP peptide exhibited inhibitory effects, their converted values were not determined.
Cell adhesion and migration to rPP with an exposed carboxyl-terminal side of the RGD domain Since the carboxyl-terminal side of RGD was considered to be more essential than the amino-terminal side in sequestration of the ability of PP-RGD, we generated a hypothetical truncated rPP fragment (rPP-(Ala terminal) protein) that terminated with Ala 482 next to the RGD domain and compared the binding abilities of rPP-DSSD and rPP-(Ala terminal) to MG63 cells by flow cytometry. We mixed rPP-DSSD and rPP-(Ala terminal) with the Alexa Fluor 488 tetrafluorophenyl ester at the same dye: protein molar ratio (30: 1) for conjugation and their equivalent labeling degrees were validated by measuring their fluorescence intensities (FI) in 96-well plates (Fig. 8A). As shown in Figure 8B, MG63 cells incubated with Alexa-rPP-(Ala terminal) had an apparently higher FI than those incubated with Alexa-rPP-DSSD and the control cells. Cells incubated with Alexa-rPP-DSSD showed a slightly higher FI than that of the control cells. We then analyzed the cell-adhesive ability of rPP-(Ala terminal) and found that the abilities of MG63 and MC3T3-E1 cells to attach to rPP-(Ala terminal)-coated wells were similar to their abilities to attach to rC-DMP-1-coated wells ( Fig. 8C column 2 and 4). On the other hand, MG63 and MC3T3-E1 cells were unable to attach to rPP-RGE-(Ala terminal)-coated wells (column 3), which was the RGDinactivated mutant of rPP-(Ala terminal), or to rPP-coated wells (column 1). Moreover, MG63 cells spread and formed actin stress fibers on rPP-(Ala terminal)-coated wells as well as rC-DMP1- . Six other mesenchymal cell lines were unable to attach to rPP. Three hDPC (human dental pulp cells) cells from different donors, Saos2 (human osteosarcoma cell line), parental MC3T3-E1 (heterogeneous cell population), and C2C12 (mouse myoblast cell line) were seeded onto rPP, rC-DMP-1, and vitronectin (100 nM) with MnCl 2 (1 mM). The number of attached cells was evaluated as described before. Each value represents the mean of triplicate determinations; bars mean 6SD. Statistical analysis was performed by a one-way ANOVA, followed by Dunnett's test. ***p, 0.001 indicates significantly lower and ++ p,0.01 and +++ p,0.001 indicate significantly higher than rC-DMP-1-coated wells. doi:10.1371/journal.pone.0112490.g004 coated wells (Fig. 8D, boxes 2 and 4). However, cell adhesion and organization were delayed and cells remained round on rPP (box 1), rPP-RGE-(Ala terminal) (box 3), and non-coated wells (box 5). The organization of the actin stress fibers of MC3T3-E1 cells was similar on these substrates (data not shown).
Cell matrix interactions between RGD domains and integrin receptors commonly mediate cell migration as well as cell adhesion. Therefore, we examined the effects of rPP, rPP-(Ala terminal), rPP-DSSD, and rC-DMP1 proteins on haptotaxis cell migration by the modified Boyden chamber assay using the Transwell system. We coated the bottom side of the membrane with rPP, rPP-(Ala terminal), rPP-DSSD, or rC-DMP1 and counted the number of MG63 cells that migrated from the upper side to the bottom side of the membrane. We found that rPP-(Ala terminal) and rC-DMP1 were able to induce haptotaxis migration by MG63 cells, whereas rPP and rPP-DSSD were not (Fig. 8E).
The Ala-Ser bond was the key flanking peptide bond that allowed primate and mouse PP-RGD domains to become inactive Since the release of Ser 483 from the vicinity of RGD disclosed PP-RGD ability, we speculated that the Ala-Ser peptide bond may be the key to sequestering the ability of the RGD domain to act on integrin. Therefore, we newly generated rPP-DSSD-RGDTSYT in which Ala 482 was replaced with Thr, rPP-DSSD-RGDDSYT in which Ala 482 was replaced with Asp, rPP-DSSD-RGDACYT in which Ser 483 was replaced with Cys, rPP-DSSD-RGDAIYT in which Ser 483 was replaced with Ile, rPP-DSSD-RGDAVYT in which Ser 483 was replaced with Val, and rPP-DSSD-RGDNPYT in which Ala 482 -Ser 483 bond was replaced with an Asn-Pro bond to mimic the human and mouse DMP-1-RGD domains. We then examined their cell-adhesive potencies by quantifying the number of attached MG63 cells on wells coated with normal rPP-DSSD and these mutated rPP-DSSD proteins. As shown in Figure 9, MG63 cells were able to attach to the plates coated with 1 mM of these mutated rPP-DSSD proteins, with rPP-DSSD-RGDDSYT and rPP-DSSD-RGDAVYT showing higher binding capacities. In contrast, MG63 cells were unable to attach to plates coated with 1 mM of normal rPP-DSSD. The binding ability of MG63 cells was higher on plates coated with mutated rPP-DSSD proteins than with rPP-DSSD at 250 nM; however, the cell-adhesive effects of mutated rPP-DSSD proteins were not significant at this concentration.

Discussion
In the present study, we demonstrated that neither rPP nor rPP-DSSD exhibited any adhesive or migratory ability, whereas the simultaneously purified rC-DMP-1 did. Further analyses utilizing various peptides containing the PP-RGD domains revealed that the Ala 482 -Ser 483 flanking sequence next to the RGD domain was the key peptide bond that allowed the PP-RGD domain to become inactive.
Vitronectin is an extensively examined adhesive glycoprotein that has been purified from the plasma and ligands of integrin avb3 and avb5 [33][34][35][36][37][38][39]. We showed that that adhesion of MG63 cells to rC-DMP-1 and vitronectin was mediated by integrin avb3 and avb5 (Fig. 6). Thus, we investigated the adhesive potency of rPP using the known adhesive proteins, rC-DMP-1 and vitronectin as positive controls. We mainly used the human osteosarcoma cell line MG63, which has been shown to express several integrin receptors of SIBLING members [24]. As shown in Figure 3A and . The inability of rPP, rPP-DSSD, and rPP-RGE to bind to integrin avb3 or avb5. (A) MG63 cells were preincubated with the neutralizing antibodies against integrin avb3 (LM609), avb5 (P1F6), and b1 (4B7) and control IgG (10 mg/ml) for 15 min, and were then seeded onto rC-DMP-1 and vitronectin (100 nM) in serum-free medium. Statistical analysis was performed by a one-way ANOVA, followed by Dunnett's test. **p, 0.01 and ***p,0.001 indicate significantly lower than the cells preincubated with control IgG. (B) The binding of integrin avb3 and avb5 to rPP, rPP-DSSD, rPP-RGE, rC-DMP-1, and vitronectin was analyzed using solid phase binding assays. Various amounts (0,400 ng) of integrin avb3 and avb5 were added to 96-well plates precoated with either 100 nM of PP, PP-DSSD, PP-RGE, rC-DMP-1, or vitronectin. Bound integrin avb3 and avb5 were then detected with the anti-integrin av antibody, an appropriate HRP-conjugated secondary antibody, and TMB. Each value represents the mean of triplicate determinations; bars mean 6SD. doi:10.1371/journal.pone.0112490.g006 B, rPP did not exhibit any adhesive potency. Since divalent cations such as MnCl 2 , CaCl 2 , and MgCl 2 have been shown to potentiate RGD-integrin-mediated signaling [40,41], we seeded MG63 and MC3T3-E1 cells with divalent cations; however, no divalent cation clearly induced cell adhesion to rPP (Fig. 3C and D). A recent study examined the adhesion of C3H/10T1/2 cells to a relatively low amount (750 ng/ml) of rPP [42] (No description regarding the expression system was provided in this study. Bacterial bovine rPP was used in a related study [43]). The cell adhesion assay performed after 4 or 24 hours of incubation may have led to their conclusion. In our study, we performed all cell adhesion assays within a short time frame (1 hr) without serum in order to exclude the possibility that any pro-adhesive proteins secreted from seeded cells unexpectedly aided adhesion to the coated proteins. We also coated wells with various concentrations (1.7,1333 nM i. e. 0.1,80 mg/ml) of intact rPP to determine whether MG63 and MC3T3-E1 cells were able to attach to the wells, and found that no concentration of rPP aided cell adhesion (data not shown). We also confirmed that C3H/10T1/2 cells were unable to attach to wells precoated with rPP (100 nM) within a short time frame (1 hr) without serum. Therefore, we speculated that C3H/10T1/2 cells may secrete a certain amount of proadhesive proteins to aid their adhesion to coated proteins in the long term.
The inability of rPP to facilitate cell adhesion was observed when not only MG63 and MC3T3-E1 cells, but also six different mesenchymal cell lines were seeded onto rPP (Fig. 4). Therefore, we concluded that the adhesive inability of rPP was universally observed rather than being osteoblastic cell-specific. As shown in Figure 5, rPP-DSSD did not have any positive effects on cell adhesion, which was similar to rPP. This result was supported by the SSD representative peptide being unable to inhibit cell adhesion (Fig. 7C). Thus, we concluded that the lack of binding potency was not due to the SSD repeats of rPP, but could be attributed to their inability to bind to integrin receptors (Fig. 6B).
RGD-containing proteins do not always possess the ability to act on integrin because the RGD domain may not be localized on their surfaces [36]. We preheated rPP at 70uC for 20 min before coating onto 96-well plates; however, MG63 cells were still unable to adhere to these wells (data not shown). This result indicated that the inaccessibility of PP-RGD was not due to the threedimensional structure of rPP. The accessibility of integrin to the RGD domains is also known to be influenced by the flanking amino acid sequences of the RGD domain [36]. Therefore, we hypothesized that the flanking amino acid sequences may be responsible for allowing the PP-RGD domain to be sequestered. An adhesion assay utilizing several peptides narrowed and identified the exact amino acids of PP in order to allow the PP-RGD domain to become inactive (Fig. 7). The inhibitory ability of 20 nM of the C-opened RGD peptide (having 5.27 nM of the binding activity of cyclic-(GRGDSP) to integrin avb3) was more potent than that of 20 nM of the N-opened RGD peptide (having 3.30 nM of the binding activity of cyclic-(GRGDSP) to integrin avb3) (Fig. 7D). Therefore, we generated the hypothetical recombinant amino-terminal products of PP, which terminated at Ala 482 next to the RGD domain, and demonstrated that Alexa-rPP-(Ala terminal) was able to bind to MG63 cells using flow cytometry. MG63 cells incubated with Alexa-rPP-DSSD also had a higher FI than control cells. A recent study reported that the highly acidic carboxyl-terminal of rat PP could be endocytosized into different mesenchymal cells even at 4uC for 15 min [44]. Since rPP-DSSD includes part of the corresponding amino acid sequence, the slight increase in FI in MG63 cells incubated with Alexa-rPP-DSSD could be attributed to this unique endocytic mechanism. We then demonstrated the RGD-dependent effects of rPP-(Ala terminal) on cell adhesion and migration (Fig. 8). These results suggested that the Ala-Ser bond next to the RGD domain was indispensable for sequestering the activity of PP-RGD.
Previous studies showed that the exogenous addition of the porcine PP-RGD peptide, which had Thr next to the RGD  domain, induced the cellular migration of human dental pulp cells [45]. Therefore, we determined whether point mutations in Ala and/or Ser altered the ability of the PP-RGD domain (Fig. 9). We examined 6 kinds of mutations and found that all converted rPP-DSSD into a cell-adhesive protein. We compared the cell-adhesive abilities of rPP-DSSD (having the RGDASYT sequence), rPP-DSSD-RGDTSYT, to mimic the porcine PP, and rPP-DSSD-RGDDSYT, to mimic the elephant PP, and found that rPP-DSSD-RGDDSYT exhibited the most potent cell-adhesive ability followed by rPP-DSSD-RGDTSYT. Since the hydropathy index  This result indicated that the existence of a non-polar hydrophobic amino acid next to Ala 482 appeared to be more effective for activating the PP-RGD domain. rPP-DSSD-RGDNPYT, which had Asn and Pro next to the RGD domain to mimic the human and mouse DMP1-RGD domains, was also able to aid cell adhesion; however, MG63 cells were unable to attach to wells coated with lower than 250 nM of rPP-DSSD-GDNPYT. Since rC-DMP-1 exhibited cell-adhesive potency when wells were coated at concentrations of 20 and 100 nM (Fig. 3), the celladhesive potency of rPP-DSSD-GDNPYT was not as strong as that of rC-DMP-1. These results indicated that other flanking amino acids also influenced sequestration of the cell-adhesive ability of the PP-RGD domain. However, most importantly, a mutation in the 1 st and/or 2 nd amino acids next to the PP-RGD domain was sufficient to confer cell-adhesive potency to PP. An evolutionary analysis of the DSPP genomic sequence indicated that ancestor DSPP was created by a DMP-1 duplication [23]. In contrast to C-DMP-1, neither primate nor mouse PP was able to act on the integrin receptor through their RGD domains; therefore, we assumed that PP lost its RGD activity during the evolutionary process after the generation of ancestor DSPP. The PP-RGD domain is not ubiquitously conserved between toothed animals (e. g. rat PP lacks the RGD domain.). Thus, the PP-RGD domain may not play a significant role in tooth development.
A recent in vitro study suggested that the inhibition of endogenous DSPP expression repressed the carcinogenesis features of human oral squamous cells [46], and the expression of DSPP was recently shown to be upregulated in several cancer tissues [4,47,48]. Aberrant conditions such as the release of intracellular proteases from cancer cells or a local low pH environment may degrade PP to generate the PP fragments included the RGD domain with an open amino-or carboxylterminal side. The integrin accessibility of one of the SIBLINGs, osteopontin, was shown to be enhanced by multiple juxtapositions in RGD amino acid cleavage by plasmin and cathepsin D [49]. A search of the MEROPS (http://merops.sanger.ac.uk/index. shtml), an information resource for peptidases, revealed that more than 80 kinds of peptidase, including cathepsins such as cathepsin D, B, and S, cleaved their substrates at the Ala-Ser bond. Therefore, we incubated 5 mg of rPP with purified cathepsin D, B, and S (sigma) (substrate: enzyme = 75:1) for 3 hrs at 37uC in 10 mM sodium acetate buffer at an appropriate pH (pH 3.3 for cathepsin D, 6.0 for cathepsin B, and 6.5 for cathepsin S), and determined whether rPP was cleaved by these enzymes; however, no visible cleaved products were observed by Stains-all staining (data not shown). The functional importance of PP-RGD domains has not yet been evaluated in vivo. DMP-1 is known to contribute to the invasion of colon cancer cells in a RGD-independent manner by bridging MMP-9 to CD44 [50]. Therefore, PP may possess such RGD-independent effects as features for carcinogenesis. Future studies to explore degraded PP fragments and the proteases responsible are warranted in order to determine the post-cleavage functions of PP in carcinogenesis as well as hard tissue development.

Ethics statement
Animal experiments to collect protein samples from DSPP-null and wild type mice were performed in compliance with the Hiroshima University guidelines on the care and use of laboratory animals. All experimental procedures were approved by the Committee of Research Facilities for Laboratory Animal Science of Hiroshima University (Permit Number: A10-81-2). The generation of three human dental pulp cell lines was performed in compliance with the Hiroshima University ethical guidelines for epidemiological research. All experimental procedures were approved by the Committee of Research Ethics of Hiroshima University (Permit Number: D88-3). Figure 9. Replacement of the 1 st and/or 2 nd carboxyl-terminal amino acids of the PP-RGD domain altered its ability. Ninety-six-well plates were precoated with 250 nM or 1 mM of normal rPP-DSSD and various mutated rPP-DSSD proteins, seeded with MG63 cells in serum-free medium, and incubated for 1 hr. The number of attached cells was evaluated as described above. Each value represents the mean of triplicate determinations; bars mean 6SD. Statistical analysis was performed by a one-way ANOVA, followed by Dunnett's test. *p,0.05, **p,0.01, and ***p, 0.001 indicate significantly higher than rPP-DSSD-coated wells at the same concentration. doi:10.1371/journal.pone.0112490.g009

Anti-PP antibody preparation
An affinity-purified rabbit anti-PP polyclonal antibody was generated using a custom antibody preparation service (Operon Biotechnologies Japan). The carboxyl-terminal amino acid sequences to be used as antigens were selected based on low sequence similarities. Japanese white rabbits were primed by an intradermal injection of the antigen peptide (0.2 mg) 3 times at intervals of one week. After 3 weeks of booster injections, 0.025 mg of the same antigen was primed by an intravenous injection to rabbits 3 times at intervals of one week. These rabbits were killed 1 week later, and antisera were harvested from the carotid artery. Antisera were then loaded onto an antigen-binding column and the rabbit anti-PP antibody was eluted. This eluent was used as an affinity-purified rabbit anti-PP antibody. The column flow-through solution was also collected for ELISA.

ELISA
Ninety-six-well, flat-bottomed microtiter plates (Immulon 4HBX, Thermo Fisher Scientific, Waltham, MA) were coated with either 5 mg/ml antigen peptide in DPBS or DPBS only overnight at 4uC. Plates were washed 3 times with 0.1% Tween 20-DPBS (T-DPBS) and then blocked with 1% BSA in DPBS for 2 hr at room temperature. Plates were washed 5 times with T-DPBS and incubated with the rabbit anti-PP antibody, whole antisera (without antigen-affinity purification), or flow-through solution in triplicate in a serial dilution for 90 min at room temperature. After washing, the wells were incubated with goat anti-rabbit IgG antibodies conjugated with HRP in DPBS (1:3000) for 1 hr at room temperature. They were washed and detected with 3, 3, 5, 59-Tetramethyl Benzidine substrate solution (TMB, Pierce, Rockford, IL) for 30 min at room temperature. After the addition of 2 N HCl to stop the colorimetric reaction, optical density was measured at 450 nm using a microtiter plate reader (Multiskan, Thermo Scientific Japan, Tokyo, Japan).

Dot-blot for the PP antibody
Dentin extracts were collected from wild type and DSPP-null mice as described previously (Suzuki et al., 2009). Mice were euthanized by CO 2 inhalation, and the molars were extracted. Extracted molars were crushed using a mortar and pestle, and incubated with 4 mol guanidine HCl in the presence of Complete mini EDTA-free protease inhibitor (Roche, Alameda, CA) for 24 hr at RT in order to remove proteins not incorporated into the mineralized matrix. The residue was then demineralized with 4 mol guanidine HCl and 0.5 mol EDTA plus protease inhibitors for 2 days at RT. The solution containing proteins incorporated into the mineralized matrix was dialyzed against 4 mol guanidine HCl for 1 day, then against distilled water for another 2 days. The dialyzed solution was lyophilized and dissolved into distilled water containing a protease inhibitor. Five micrograms in 20 ml of the dentin extracts from wild type and DSPP-null mice and BSA were dot-blotted onto a nitrocellulose membrane with a dot-blot apparatus. The membrane was blocked with a solution of 3% non-fat dried milk in DPBS solution, followed by incubation with the rabbit anti-PP antibody (1:2500) for 1 hr at room temperature. After washing the membrane with DPBS buffer, it was incubated with the goat anti-rabbit antibody conjugated with HRP in DPBS (1:10,000) for 1 hr at room temperature, and developed using the chemiluminescent substrate (WesternBright ECL, Advansta, Menlo Park, CA).

SDS-PAGE, gel staining, and western blotting
All protein samples were reduced with DTT and loaded onto NuPAGE Bis-Tris (Life Technologies) in MOPS buffer. Separated proteins were transferred onto a PVDF membrane for immunodetection. The membrane was blocked with a PVDF blocking agent (TOYOBO, Tokyo, Japan) for 1 hr at room temperature, and was then incubated with the anti-PP antibody (1:2500) for 1 hr at room temperature. After washing the membrane with T-DPBS, it was incubated with the HRP-conjugated goat anti-rabbit antibody (1:20,000) as the secondary antibody for 1 hr at room temperature. Immunoreactive proteins were identified using a chemiluminescent substrate (WesternBright ECL). To efficiently detect acidic proteins, the gel was washed and fixed with 25% isopropanol for 3 hr. The fixative solution was replaced every 20 min. The gel was stained overnight with 0.025% Stains-All (Sigma) containing 30 mM Tris-HCl, 7.5% formamide, and 25% isopropanol at pH 8.8 and then washed with deionized water until the protein bands were visible. Coomassie brilliant blue staining was performed with Coomassie brilliant blue R (Sigma) following a general protocol.

Preparation and purification of recombinant proteins
Mouse PP (from the Asp 452 to the stop codon) was amplified from mouse DSPP cDNA obtained from mouse incisors and cloned into a pBS-sk(+) vector (Stratagene, La Jolla, CA). Since the long-repeat nucleotide sequence-coding SSD repeats in PP were easily released from circular DNA during the transformation process, PP nucleotide sequences were cloned following a previously described procedure [6]. PP nucleotide sequences was cloned into the pCEP4-Mul-PURD expression vector (a kind gift from Dr. Yoshi Yamada, NIDCR/NIH) [51] such that the aminoterminus of recombinant proteins was 66 His-tagged.
The PP vector was transfected into 293EBNA cells utilizing X-tremeGENE (Roche, Indianapolis, IN) and incubated for 24 hr. The transfection medium was replaced with growth medium containing puromycin (5 mg/ml), and transfected cells were then cultured for 3 days. Surviving cells were routinely cultured with puromycin (0.5 mg/ml) and used in the expression of the Histagged recombinant PP (rPP) protein. To collect supernatants containing recombinant proteins, these puromycin-resistant cells were cultured with growth medium containing 10% fetal bovine serum (FBS) until confluent. The medium was replaced with new growth medium containing 5% FBS and 16 Insulin-Transferrin-Selenium (ITS) (BD Biosciences) and then cultured for 3 days. The growth medium was replaced with serum-free collection medium containing 16 ITS and then cultured for 5 days. The collection medium was collected and replaced every 24 hr during a 5-day period. Recombinant proteins were purified from the collection medium as follows. All procedures were generally performed in a cold room (4uC). Non-serum medium containing recombinant proteins was added to anion-exchange columns (Q Sepharose Fast

Cell culture
The mouse pre-osteoblastic cell line MC3T3-E1 (subclone 4) and mouse myoblast cell line C2C12 were purchased from the ATCC (Rockville, MD). The human osteosarcoma cell lines MG63 and Saos2, and parental heterogeneous MC3T3-E1 were purchased from the RIKEN Cell Bank (Tsukuba, Japan). Human dental pulp (hDPC) cells from three different patients were isolated from healthy teeth extracted for orthodontic purposes with informed consent. The pulp chamber was opened by minimum drilling and pulp tissue was removed from the chamber using a barbed broach. The extracted tissue was placed onto a culture dish plate and cells that were out-growths from the tissue were used as dental pulp cells. MC3T3-E1 cells, both subclone 4 and heterogeneous cells, were maintained in alpha MEM (Life Technologies). 293EBNA, MG63, 3 different kinds of hDPC, and C2C12 cells were maintained in DMEM (Life Technologies). Saos2 cells were maintained in McCoy's medium (Life Technologies). All media were supplemented with 100 units/ml penicillin, 100 mg/ml streptomycin (Life Technologies), and 10% FBS, except for McCoy's medium, which was supplemented with 15% FBS. All cells were cultivated at 37uC under humidified 5% CO 2 , and 95% air atmospheric conditions.

Cell adhesion assay
Fifty microliters of a 20 or 100 nM solution of recombinant proteins or vitronectin was coated onto a 96-well plate (Immulon 4HBX) in DPBS overnight at 4uC. The wells were then rinsed twice with DPBS and blocked with 1% BSA in DPBS for 1 hr at 37uC, followed by 3 washes with DPBS. Subconfluent cells were rinsed twice with DPBS, harvested using 5 mM EDTA in DPBS for 20 min at room temperature, collected by centrifugation, and then suspended in serum-free RPMI 1640 (Life Technologies) with or without divalent cations as noted. Suspended cells were preincubated with 1 mM of the various peptides for 10 min at 37uC for blocking experiments. These cells were seeded onto wells at a density of 5610 4 cells/well in 50 ml of serum-free medium. After a 1-hr incubation at 37uC, the wells were rinsed 3 times with serum-free medium to remove non-adherent cells. Adherent cells were fixed with 3.7% formaldehyde for 10 min and then stained with 0.2% crystal violet in 10% ethanol for 10 min at room temperature. Wells were rinsed several times with water and dried overnight. The dye was then solubilized with 150 ml of 1% SDS and quantified by measuring absorbance at 570 nm on a microtiter plate reader (Multiskan).

Solid phase binding assay
Ninety-six-well plates (Immulon 4HBX) were coated with rPP, rPP-DSSD, rPP-RGE, rC-DMP-1, and vitronectin and then blocked as described in the Cell adhesion assay section, followed by rinsing twice with binding buffer (50 mM Tris, 150 mM NaCl, 1 mM MnCl2). Various amounts (0,400 ng) of recombinant integrin avb3 and avb5 in 50 ml of binding buffer were then added to the coated wells and incubated overnight at 4uC. To immunologically detect bound avb3 and avb5, the wells were rinsed twice with 0.1% Tween-binding buffer (T-binding buffer) and incubated with the rabbit anti integrin av antibody (C2C3) (1:1000) for 90 min at room temperature. After washing twice with T-binding buffer, the wells were incubated with goat anti-rabbit IgG antibodies conjugated with HRP in binding buffer (1:1000) for 1 hr at room temperature. The wells were then rinsed twice with 0.1% Tween-binding buffer and twice with binding buffer, and detected with TMB for 30 min at room temperature. After the addition of 2N HCl to stop the colorimetric reaction, optical density was measured at 450 nm using a microtiter plate reader (Multiskan).

Integrin competitive binding assay
The wells of 96-well plates (Immulon 4HBX) were coated with 0.05 mg/well recombinant integrin avb3 overnight at 4uC. The coated plates were washed twice with T-binding buffer and blocked with 1% BSA in binding buffer for 1 hr at 37uC. The wells were then washed twice with T-binding buffer and incubated with 5 nM biotinylated-GRGDS peptide with or without various peptides or cyclic-(GRGDSP) for 3 hr at room temperature. The wells were rinsed twice with T-binding buffer, and then incubated with the streptavidin-HRP polymer (Sigma-Aldrich) (1:5,000) for 1 hr at room temperature to detect bound biotinylated-GRGDS. The wells were rinsed twice with T-binding buffer and twice with binding buffer, and detected with TMB for 30 min at room temperature. After the addition of 2N HCl to stop the colorimetric reaction, optical density was measured at 450 nm using a microtiter plate reader (Multiskan).

rPP-DSSD and rPP-(Ala terminal) binding assay
The Alexa Fluor 488 tetrafluorophenyl ester (Life Technologies) was mixed with PP-DSSD and rPP-(Ala terminal) at a dye: protein molar ratio = 30: 1 for 1 hr at room temperature in DPBS containing 0.1 M sodium bicarbonate. The conjugated samples were extensively dialyzed against DPBS to remove unconjugated dye. The protein concentration of Alexa Fluor 488 conjugated-PP-DSSD (Alexa-PP-DSSD) and Alexa Fluor 488 conjugated-rPP-(Ala terminal) (Alexa-rPP-(Ala terminal)) was determined using a BCA kit. One hundred microliters of 15.625, 62.5, and 250 nM of Alexa-PP-DSSD and Alexa-rPP-(Ala terminal) were added onto black 96-well plates and fluorescence intensity was measured by ARVO X One (PerkinElmer Japan, Kanagawa, Japan) to analyze the degree of labeling. In the cytofluorometric assessment, MG63 cells were harvested as described for the cell adhesion assay and then suspended in 0.5% BSA in DMEM with 2 mM MnCl 2. Suspended cells were incubated with soluble Alexa-rPP-DSSD and Alexa-rPP-(Ala terminal) (250 nM) for 1 hr at 4uC. Cell pellets were washed twice with DPBS containing 2% FBS with 200 nM MnCl 2 and the binding of soluble Alexa Fluor 488-rPP-DSSD and rPP-(Ala terminal) was then determined by flow cytometry.

Immunofluorescence analysis for the actin cytoskeleton
Thirty-five-millimeter glass dishes (Matsunami, Japan) were coated with 100 nM solution of rPP, rPP-(Ala terminal), rPP-RGE-(Ala terminal), or rC-DMP1 in DPBS overnight at 4uC. The wells were then rinsed twice with DPBS and blocked with 1% BSA in DPBS for 1 hr at 37uC, followed by 3 washes with DPBS. Cells were harvested as described for the cell adhesion assay and then suspended in serum-free RPMI 1640 with 1 mM MnCl 2 . Cells were seeded in serum-free medium and incubated at 37uC for 1 hr. Cells were then washed twice with DPBS, fixed with 2.5% paraformaldehyde, permeabilized with 0.5% Triton X-100, and then incubated with Actin-stain 488 phalloidin (Cytoskeleton, Inc., Denver, CO) for 30 min. Actin staining was observed using a Leica microscope and images were captured with a digital camera.

Cell migration assay
Haptotaxis cell migration was assayed with Transwell migration chambers, the pore size of which was 8 mm (353097 cell culturetreated, Falcon System Inc., Columbia, MD). The undersides of the membranes were coated at room temperature overnight with 100 ml of a 1 mM solution of recombinant proteins followed by 2 washes with DPBS. MG63 cells were harvested as described for the cell adhesion assay and then suspended in 0.1% BSA in RPMI 1640. These cells were seeded onto the upper chamber at a density of 1610 5 cells/well in 150 ml of 0.1% BSA in RPMI 1640. The lower chamber was filled with 600 ml of 0.1% BSA in RPMI 1640. After a 12-hr incubation at 37uC, the membranes were rinsed twice with warmed DPBS, fixed with 3.7% formaldehyde for 10 min, and then stained with 0.2% crystal violet in 10% ethanol for 10 min at room temperature. Non-migrated cells were then removed from the upper surface of the membranes with a damp cotton swab. Cells that had migrated were counted using 10 randomly selected fields at 2006 magnification.

Statistical Analysis
Statistical analysis was performed by a one-way analysis of variance (ANOVA), followed by Dunnett's test for Figures 3, 4, 5, 6, 7D, 8, and 9 and by Tukey's test for Figures 7A, B, and C. Standard samples of the posthoc test have been described in each of the figure legends.