Pyrophosphate Inhibits Mineralization of Osteoblast Cultures by Binding to Mineral, Up-regulating Osteopontin, and Inhibiting Alkaline Phosphatase Activity*

Inorganic pyrophosphate (PPi) produced by cells inhibits mineralization by binding to crystals. Its ubiquitous presence is thought to prevent “soft” tissues from mineralizing, whereas its degradation to Pi in bones and teeth by tissue-nonspecific alkaline phosphatase (Tnap, Tnsalp, Alpl, Akp2) may facilitate crystal growth. Whereas the crystal binding properties of PPi are largely understood, less is known about its effects on osteoblast activity. We have used MC3T3-E1 osteoblast cultures to investigate the effect of PPi on osteoblast function and matrix mineralization. Mineralization in the cultures was dose-dependently inhibited by PPi. This inhibition could be reversed by Tnap, but not if PPi was bound to mineral. PPi also led to increased levels of osteopontin (Opn) induced via the Erk1/2 and p38 MAPK signaling pathways. Opn regulation by PPi was also insensitive to foscarnet (an inhibitor of phosphate uptake) and levamisole (an inhibitor of Tnap enzymatic activity), suggesting that increased Opn levels did not result from changes in phosphate. Exogenous OPN inhibited mineralization, but dephosphorylation by Tnap reversed this effect, suggesting that OPN inhibits mineralization via its negatively charged phosphate residues and that like PPi, hydrolysis by Tnap reduces its mineral inhibiting potency. Using enzyme kinetic studies, we have shown that PPi inhibits Tnap-mediated Pi release from β-glycerophosphate (a commonly used source of organic phosphate for culture mineralization studies) through a mixed type of inhibition. In summary, PPi prevents mineralization in MC3T3-E1 osteoblast cultures by at least three different mechanisms that include direct binding to growing crystals, induction of Opn expression, and inhibition of Tnap activity.

Mineralization of bones, and tooth dentin and cementum, occurs in a collagen-rich extracellular matrix. Coincident with establishing an extracellular collagenous network in these tissues, osteoblasts, chondroblasts, odontoblasts, and cementoblasts all secrete additional noncollagenous matrix proteins that integrate with the collagen fibrils and provide additional functionality to the matrix (1). Although much of this functionality relates to cell adhesion and signaling, cell-free in vitro assays and in vivo studies using transgenic mice indicate that certain noncollagenous proteins (some tissue-specific) regulate the induction of mineralization and subsequently control hydroxyapatite crystal growth by binding to mineral surfaces (2).
In addition to the organic, mineral-binding noncollagenous proteins of bones and teeth, inorganic molecules in these tissues have also been proposed as molecular determinants of mineralization. Pyrophosphate (PP i ) is a potent, mineral-binding small molecule inhibitor of crystal nucleation and growth (3), and recent studies using transgenic and naturally occurring mutant mice have identified the proteins/enzymes involved in producing and handling PP i (see references below). PP i is a by-product of many intracellular metabolic reactions (4). PP i is also present in the extracellular matrix of most tissues and bodily fluids including plasma (5, 6), where it acts as a potent inhibitor of mineral nucleation and growth at micromolar concentrations (3,7,8). Although the mechanism by which PP i inhibits hydroxyapatite crystal growth is not entirely known, PP i is thought to adsorb specifically to crystal growth sites, thus preventing further apposition of mineral ions (7). Although some physicochemical studies have been conducted on the actions of PP i on crystal growth in cell-free systems, there is limited information available on putative direct effects of PP i on cells in general and, more specifically, on mineralizing cell culture models (9). Extracellular matrix mineralization in bone is tightly linked to the P i /PP i ratio found in this tissue (10). Extracellular PP i deficiency leads to excess hydroxyapatite formation in the skeleton (11), whereas PP i elevation results in decreased skeletal mineralization (6,(12)(13)(14) and the formation of calcium pyrophosphate dihydrate crystals in joints (15,16). The homeostatic regulation of local PP i levels by bone cells is thus an important part of their function related to mineralization. The major ectoenzyme responsible for generation of extracellular PP i is Enpp1 (ectonucleotide pyrophosphatase/phosphodiesterase 1; also known as NPP1 and NPPS) (17); this enzyme is expressed at particularly high levels in bone, cartilage, and teeth (17)(18)(19). Intracellular PP i can also be exported into the extracellular compartment by the membrane transporter Ank (progressive ankylosis or ANKH), which is also highly expressed in bone, cartilage, and teeth (20 -22). Deletion of the Ank gene in mice results in ectopic mineral formation in joint spaces and, eventually, complete joint ankylosis (23). Given the high expression levels of these two genes in mineralized tissues, removal of PP i is expected to be a prerequisite for physiologic mineralization, and thus regulation of the expression and activities of these two proteins is likely central to the control of mineralization. Tnap (tissue-nonspecific alkaline phosphatase; also known as Akp2), a well characterized marker of the osteoblast lineage (24), is capable of hydrolyzing PP i into P i (25). In osteoblast cultures, Tnap is also responsible for the generation of P i from phosphate esters such as ␤-glycerophosphate (␤GP), 4 which is commonly used as a source of organic phosphate for mineralization (26).
Another key inhibitor of mineralization found in bone is Opn (osteopontin), a highly phosphorylated glycoprotein with strong mineral binding properties (27). A link between the enzymes regulating extracellular PP i levels and nonenzymatic protein inhibitors of mineralization such as Opn has been demonstrated in Enpp1-deficient mice that show a decrease in both PP i and Opn levels and consequential hypermineralization particularly related to joints and ligaments (11). This phenotype is also observed in the naturally occurring mouse mutant ttw (tiptoe-walking mice; Ref. 28).
In view of recent findings linking PPi generators (Enpp1), transporters (Ank), and degraders (Tnap) with extracellular matrix protein inhibitors of mineralization (Opn), we have used the MC3T3-E1 murine preosteoblast cell line to examine whether there is a direct effect of PP i on Opn gene expression and the specific signaling pathways involved in this response. We report that PP i up-regulates Opn in osteoblasts via MAPK pathways, thus providing a dual mechanism for inhibition/control of hydroxyapatite crystal growth during mineralization.

EXPERIMENTAL PROCEDURES
Cell Culture-MC3T3-E1 murine calvarial osteoblasts (subclone 14) cells were a gift from Dr. R. T. Franceschi (University of Michigan, Ann Arbor, MI). The cultures were maintained in modified ␣-minimum essential medium (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin-streptomycin (Invitrogen) at 37°C in a humidified atmosphere of 5% CO 2 . All of the experiments were carried out at a seeding density of 50,000 cells/cm 2 . Cell differentiation and mineralization was initiated 24 h after plating by replacing the medium with ␣-minimum essential medium supplemented with 10% fetal bovine serum, 50 g/ml ascorbic acid (Sigma), and 10 mM ␤GP (Sigma) or 5 mM sodium phosphate (Sigma). The medium was changed every 48 h. For inhibition experiments, sodium pyrophosphate tetrabasic, levamisole, kidney tissue-nonspecific alkaline phosphatase, and phosphonoformic acid (foscarnet) were purchased from Sigma. Signaling pathway inhibitors U0126 and SB202190 were obtained from VWR, and inhibitor SP600125 was purchased from SuperArray. Cytotoxicity of inhibitors and phosphate ester treatments were determined to be negligible by the 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide assay (MTT assay, data not shown).
Quantification of Mineralization-Mineral was visualized by von Kossa staining using a 5% silver nitrate solution (Sigma) combined with exposure of the culture dishes to bright light. For quantification of insoluble calcium in the cell/matrix layer, the cultures were decalcified with 0.5 N HCl, and calcium in the supernatant was determined colorimetrically using a calcium assay kit (Diagnostic Chemicals). Following decalcification, the cultures were solubilized with 0.1 N NaOH and 0.1% SDS, and total protein content was measured with the BCA protein kit (Pierce). The calcium content of the cell/matrix layer was normalized to the protein content.
Assay for Alkaline Phosphatase and Nucleotide Pyrophosphatase Phosphodiesterase Activity-The cultures were washed three times with phosphate-buffered saline (PBS) and solubilized in 10 mM Tris, pH 7.4, 0.2% Igepal (Sigma), and 2 mM phenylmethylsulfonyl fluoride. After sonication and centrifugation, alkaline phosphatase activity in the supernatant was determined colorimetrically in a reaction mixture containing 50 mM Tris-HCl, pH 8.8, 10 mM MgCl 2 , and 20 mM p-nitrophenylphosphate. Calf intestinal alkaline phosphatase (Sigma) was used as a standard. One unit will hydrolyze 1 mol of p-nitrophenylphosphate/min at 37°C. Nucleotide pyrophosphatase phosphodiesterase activity was determined colorimetrically using p-nitrophenylthymidine monophosphate as a substrate (17).
Enzyme Inhibition Kinetics-All of the assays were performed in 50 mM Tris-HCl buffer, pH 8.8, with 10 mM MgCl 2 at 37°C and were repeated in triplicate. All of the enzymatic rates obtained are initial rates. Quantification of substrate reaction product (phosphate) was performed colorimetrically by the ammonium molybdate method. Inhibitor constants were obtained from averages of Dixon and s/v plots (29,30).
PCR efficiency was optimized, and melting curve analysis of products were performed to ensure specificity. All of the primers were used at 0.5 M except for those of Opn, which were used at 0.25 M. PCR was carried out in an Applied Biosystems 7500 system using Power SYBR Green PCR Master Mix (Applied Biosystems). The sequential reaction conditions were 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Gene expression was quantified using the sequence detection software of the system (Applied Biosystems). Gapdh was used as the housekeeping gene for data normalization. Triplicate RNA samples from at least two separate experiments were used for these experiments.
Opn Enzyme-linked Immunosorbent Assay-Opn protein levels were measured using a mouse Opn assay kit (Assay Designs) according to the manufacturer's instructions. Recombinant mouse Opn was used as a standard.
Immunofluorescence-The cells were cultured on 8-well chamber slides and then fixed in 3.7% formaldehyde. The cells were then first permeabilized with 0.25% Triton X-100 in PBS for 5 min or directly blocked with 2% BSA for 30 min. Actin cytoskeleton labeling was performed using phalloidin Alexa Fluor 564, and nuclei were observed using 4Ј,6-diamidino-2phenylindole staining of DNA (Molecular Probes). Opn was immunolabeled using goat anti-mouse Opn from R & D Systems and chicken anti-goat Alexa Fluor 488 secondary antibody (Molecular Probes). The s1lides were mounted with Prolong Antifade (Molecular Probes).
Phosphorylated OPN and Dephosphorylation-Phosphorylated bovine milk OPN (Arla Foods, Denmark; prepared according to methods similar to those developed by E. S. Sørensen (33)) with 24 phosphorylations (of 28 potential sites) either was added directly to the cultures or was dephosphorylated at a concentration of 10 mg/ml in 50 mM Tris-HCl, pH 8.8, containing 10 mM MgCl 2 by passing through an immobilized-alkaline phosphatase column (Mobitec) according to the manufacturer's protocol. Eluted protein was separated from the reaction mixture by precipitating with 12% trichloroacetic acid and washing with acetone. Dephosphorylation of OPN was confirmed using a phosphoprotein assay kit (Pierce).
Immunogold Labeling-Cells cultured in 35-mm dishes were fixed with 2% glutaraldehyde, dehydrated through a series of graded ethanol dilutions, and embedded in LR White acrylic resin (London Resin Company). Thin sections (80 nm) were then placed on Formvar-coated nickel grids. Post-embedding immunolabeling was performed on the grid-mounted sections as described previously (34)   Effect of PP i on MC3T3-E1 osteoblast culture mineral deposition, enzyme activity, and gene expression. A, MC3T3-E1 cultures were treated with 50 g/ml ascorbic acid and 10 mM ␤GP for 12 days to induce osteoblast differentiation and mineralization. Starting on day 6, the cultures were supplemented with the indicated dose of PP i with, or without, TNAP (1 unit/ml) for the remainder of culture period. The medium was replaced every 48 h. On day 12, calcium content in decalcified extracts were measured colorimetrically by the Arsenazo method, and duplicate cultures were stained for mineral using 5% silver nitrate staining (von Kossa method). The extracts were normalized to cellular protein content, and the data are presented as the means Ϯ S.E. B, PP i bound to hydroxyapatite is TNAP hydrolysis-resistant. 1.8 mg of free PP i or 1.8 mg of PP i adsorbed onto 2 mg of hydroxyapatite were incubated at 37°C in a reaction mixture containing 50 mM Tris-HCl (pH 8.8), 10 mM MgCl 2 and 0.125 unit/ml TNAP, and P i release was measured colorimetrically by the ammonium molybdate complex method. Hydroxyapatite-bound PP i was predetermined using the EnzChek PP i assay kit (Invitrogen). The data are presented as the means Ϯ S.E. ***, p Ͻ 0.001, Student's t test. C, MC3T3-E1 cultures were treated as in A following which RNA was extracted and realtime RT-PCR analysis was performed to examine expression of the indicated genes (relative to Gapdh). D, MC3T3-E1 cultures were treated as in A following which Tnap and Enpp1 activities were measured using p-nitrophenylphosphate or p-nitrophenylthymidine monophosphate as substrates, respectively. The data are presented as the means Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001, Student's t-test relative to the 0 M treatment control.
Briefly, the grids were incubated for 10 min with 1% BSA in PBS, followed by incubation with primary antibody for 1 h, blocking with 1% BSA in PBS, and then incubation with protein A-gold for 30 min. The grids were then rinsed with distilled water, air-dried, and stained conventionally for transmission electron microscopy using uranyl acetate and lead citrate. The images were recorded using a Philips FEI Tecnai 12 transmission electron microscope operating at 120 kV.
To visualize OPN adsorption to synthetic hydroxyapatite crystals, 2 mg of synthetic hydroxyapatite (courtesy of Dr. B. Fowler, National Institute of Standards and Technology (NIST), Bethesda, MD) was incubated with gentle agitation for 1 h at room temperature in the presence of 100 l of 20 M OPN. OPN-incubated crystals were then centrifuged and washed three times with distilled water, followed by incubation with anti-Opn (R & D Systems) for 2 h and then a 1-h incubation with protein A-gold conjugate as described above.

Pyrophosphate Inhibits Mineralization of MC3T3-E1 Osteoblast Cell
Cultures-To examine the role of PP i on osteoblast mineralization, we cultured MC3T3-E1 cells in PP i -supplemented medium, and mineral was visualized by von Kossa staining and quantified by a biochemical assay for calcium. As demonstrated in Fig. 1A, PP i inhibited mineralization in a dosedependent manner with maximum inhibition occurring at a dose of 2.5 M. Because PP i potently inhibits mineralization and because osteoblasts actively produce PP i , then enzymatic removal of this inhibitor from the extracellular matrix is likely a requirement for mineralization (18). Supplementing PP i -inhibited cultures with exogenous TNAP (1 unit/ml) rescued the PP i -mediated abrogation of mineralization (Fig. 1A), suggesting that TNAP is a valid candidate for the degradation of PP i in the extracellular matrix. The addition of TNAP alone also increased mineralization as compared with the control, con-sistent with its ability to degrade endogenous PP i and provide P i for mineralization.
Because PP i has previously been shown to bind to growing crystals (3), we next examined whether mineral-bound PP i (versus free PP i ) could also be degraded by TNAP. PP i adsorbed onto synthetic hydroxyapatite crystals were incubated with TNAP, and P i release was assayed biochemically. As shown in Fig. 1B, free (unadsorbed) PP i was hydrolyzed to release P i , whereas mineral-bound PP i was resistant to TNAP hydrolysis and yielded only negligible amounts of P i. Together, these results indicate that PP i inhibits mineralization of MC3T3-E1 osteoblast cultures, and this inhibition can be reversed by TNAP, but not if PP i has bound to mineral.
Given that PP i /P i ratios are regulated by the activities of Tnap, Enpp1, and Ank, we examined the effect of the above PP i treatments (Fig. 1A) on their gene expression and activity. Quantitative real time PCR showed that PP i caused a slight increase in Tnap, which was significant only at the 2.5 M dose (Fig. 1C). Elevation of Tnap levels was anticipated because PP i would be expected to up-regulate the enzyme responsible for its degradation in an attempt to normalize its concentration. Furthermore, any PP i remaining unadsorbed to mineral could be hydrolyzed to P i , which would then stimulate Tnap expression as has been shown previously (35).
The same 2.5 M treatment of PP i led to a decrease in Enpp1 expression (Fig. 1C), likely again reflecting an attempt by the cells to lower extracellular PP i . The addition of exogenous of TNAP resulted in increased Enpp1, presumably representing an attempt by the cells to compensate for the increased PP i degradation. These data are consistent with the increased Enpp1 activity observed in Tnap-transfected osteoblasts (36).
Unexpectedly, we found that PP i dose-dependently increased Ank expression. That Ank, a gene whose encoded transporter protein outwardly pumps PP i out into the extracellular matrix, would be up-regulated in response to increased extracellular PP i is surprising. Although the reason for this is not immediately obvious, we hypothesize that elevated P i produced by TNAP hydrolysis of free, unadsorbed (to mineral) PP i could be responsible because P i has been shown to elevate Ank expression in cementoblasts (31). Furthermore, the structure and function of Ank is not entirely clear (20,37), and Wang et al. (21) have demonstrated that blocking Ank expression led to increases in both intra-and extracellular PP i levels, whereas Ank overexpression led to decreases in both intra-and extracellular PP i levels, suggesting a level of complexity in PPi handling and homeostasis that needs to be better clarified. Increased Ank expression as a result of increased exogenous PP i could simply be an attempt by the cells to return extracellular PP i levels to normal by as yet an unknown mechanism.
The observed changes in Tnap and Enpp1 levels were reflected in corresponding activity assays, with PP i causing a general increase in Tnap activity and suppression of nucleotide pyrophosphatase phosphodiesterase activity (includes Enpp1). The addition of exogenous TNAP appeared to normalize the elevated Tnap levels, whereas it greatly increased nucleotide pyrophosphatase phosphodiesterase activity (Fig. 1D).
Pyrophosphate Up-regulates Opn Expression-Recent data have shown that mineralizing cell culture systems (primarily osteoblasts and cementoblasts) are sensitive to local levels of P i and PP i (22,31,35,38). The role of anions such as PP i in skeletal and dental tissues appears not to be limited to their crystal binding properties. High doses of P i (4 -10 mM) have been shown to regulate levels of the crystal growth inhibitor protein Opn (39). To determine whether PP i could elicit a similar response, we examined levels of Opn by enzyme-linked immunosorbent assay and Western blotting in PP i -treated osteoblast cultures. 0.5 mM PP i treatment for 4 days led to a marked increase in total Opn levels ( Fig. 2A). Most of the Opn was found in the cell/matrix layer, and Opn in the conditioned medium was also slightly (but not significantly) increased by the PP i treatment ( Fig. 2A).
To determine whether the effect of PP i was reflected at the RNA level, we examined the PP i -treated cultures by real time PCR and observed that Opn mRNA was slightly, but significantly, up-regulated (Fig. 2B). Using Western blotting to further examine the up-regulation of Opn, we observed that PP i treatment led to a gradual cumulative increase in Opn levels over 4 days of PP i treatment (Fig. 2C).
We next examined whether this up-regulation of Opn was in fact directly attributable to PP i and not to its hydrolysis end product, P i . PP i treatment of MC3T3-E1 cultures in the presence of a specific Tnap inhibitor, levamisole, did not prevent induction of Opn, suggesting that PP i regulates Opn levels by a hydrolysis-independent mechanism (Fig. 2D). Furthermore, treatment of cultures with an equivalent dose of 0.5 mM P i failed to increase Opn levels. This is consistent with previous work by Beck et al. (39) showing that Opn up-regulation by P i in the MC3T3-E1 cell line requires a minimum dose of 4 mM and is maximal at 10 mM.
To further confirm that PP i and P i regulate Opn by different mechanisms, we then treated MC3T3-E1 cultures with 0.5 mM PP i and with 10 mM P i in the presence of a phosphate uptake inhibitor, foscarnet. 10 mM P i was used here as a positive control for the phosphate transport-dependent (foscarnet-sensitive) system. Both Western blotting and immunofluorescence microscopy confirmed that whereas P i -mediated up-regulation of Opn was sensitive to foscarnet, PP i -mediated induction of Opn was not (Fig. 3, A and B). Foscarnet failed to block Opn induction by PP i , and although this inhibitor caused a slight attenuation of the Opn signal in PPi-treated cultures, this was negligible as compared with its effects on the P i -treated cultures where it completely abolished the strong Opn up-regulation by P i . Localization of Opn to the extracellular matrix in the PP itreated cultures is shown in Fig. 3C. Taken together, these results suggest that up-regulation of Opn by PP i is dependent on neither PP i hydrolysis nor P i uptake.
Pyrophosphate Regulates Opn via MAPK Signaling Pathways-Osteoblast-specific gene expression has previously been shown to be regulated by the MAPK pathways (40 -42). To elucidate the mechanism by which PP i regulates Opn expression, we next examined the involvement of major cell signaling pathways using specific inhibitors. MC3T3-E1 cells were treated for 4 days with 0.5 mM PP i in the presence of the selective MAPK inhibitors U0126 (Erk1/2 inhibitor), SB202190 (p38 inhibitor), or SP600125 (JNK inhibitor). Inhibition of Erk1/2 and p38 prevented induction of Opn by PP i (Fig. 4A), but Jnk inhibition had no significant effect. We then tested the ability of PP i to activate Erk1/2 and p38. Western blot analysis of Erk1/2 and p38 fol- lowed by densitometric analysis of replicate blots from three separate experiments revealed that significant activation of Erk1/2 occurred within 5 min of PP i treatment, whereas p38 activation occurred at 60 min (Fig. 4, B and C) and continued thereafter (data not shown). The samples were then analyzed by Western blotting for Opn and the same blot reprobed for Gapdh. B, activation of MAPK signaling pathways by PP i . MC3T3-E1cells were treated with 0.5 mM PP i for the indicated times, and the protein was extracted and analyzed by Western blotting using phospho-specific antibodies for the activated forms of Erk1/2 and p38. Densitometries are expressed as the means Ϯ S.E. of the integrated densities obtained using Quantity One software adjusted to the total Erk1/2 or p38 bands. *, p Ͻ 0.05; **, p Ͻ 0.01, Student's t test relative to the 0-min time point.

FIGURE 5. Effect of OPN phosphorylation status on MC3T3-E1 osteoblast mineral deposition.
A, MC3T3-E1 osteoblast cultures were treated with 50 g/ml ascorbic acid and 10 mM ␤GP for 12 days to induce cell differentiation and mineralization. Starting on day 6, the cultures were supplemented with the indicated dose of exogenous OPN or TNAP-dephosphorylated OPN (dephos. OPN) for the remainder of culture. The medium was replaced every 48 h. On day 12, the calcium content in decalcified extracts was measured colorimetrically by the Arsenazo method, and duplicate cultures were stained for mineral using 5% silver nitrate staining (von Kossa method). The extracts were normalized to cellular protein content. The data are presented as the means Ϯ S.E. ***, p Ͻ 0.001, Student's t test relative to the 0 M control treatment. B, transmission electron micrograph showing post-embedding, colloidal-gold immunolocalization of endogenous Opn in an 8-day MC3T3-E1 culture using anti-Opn followed by protein A-colloidal gold conjugate. Opn is localized to the margins of small mineralization foci dispersed among uncalcified collagen fibrils (Coll). LR White resin section counterstained with uranyl acetate and lead citrate. C, transmission electron micrograph of OPN binding to hydroxyapatite. Synthetic hydroxyapatite crystals were incubated with OPN followed by colloidal-gold immunolabeling showing localization of the protein at the surface of the hydroxyapatite crystals.

Opn Inhibits Mineralization via Its Phosphate
Residues-Upregulation of Opn by PP i is noteworthy in that although the former is a protein and the latter is an anionic small molecule, they nevertheless apparently share some remarkable similarities including mineral inhibition. As demonstrated in Fig. 1A, the inhibitory action of PP i is inactivated by the actions of TNAP. We thus reasoned that if the levels of one inhibitor are coupled to another, modulation of their activities might also be similarly coupled. To examine this, we added exogenous bovine OPN to MC3T3-E1 cultures that dose-dependently inhibited mineralization, with complete inhibition occurring at a dose of 1.2 M (Fig. 5A). Dephosphorylation of the OPN with immobi-lized TNAP removed its ability to block mineralization at 1.2 M (Fig. 5A). Another similarity between OPN and PP i is their high affinity for mineral. Using immunogold labeling, we demonstrated at the ultrastructural level that endogenous mouse Opn in mineralizing MC3T3-E1 cultures co-localized with mineral deposits (Fig. 5B). Furthermore, incubation of bovine OPN with synthetic hydroxyapatite crystals, where the protein was then visualized by immunogold labeling, likewise demonstrated that OPN can bind to hydroxyapatite (Fig. 5C).
Pyrophosphate Inhibits Phosphate Release from ␤GP and OPN-␤GP is a monophosphoester commonly used as an organic source of phosphate in cell cultures (43, 44) whose FIGURE 6. PP i inhibition is more potent when mineralization is Tnap-dependent. A, MC3T3-E1 osteoblast cultures were treated with 50 g/ml ascorbic acid and either 10 mM ␤GP or 5 mM P i . Starting on day 6, the cultures were supplemented with the indicated dose of PP i for the remainder of culture. The medium was replaced every 48 h. On day 12, the cultures were stained for mineral using 5% silver nitrate staining (von Kossa method). B, MC3T3-E1 cultures were treated as in A, and on day 12 calcium content in decalcified extracts was measured colorimetrically by the Arsenazo method and normalized to protein content. The data are presented as the means Ϯ S.E. C, MC3T3-E1 cultures were treated as in A, following which RNA was extracted, and real time reverse transcription-PCR analysis was performed to examine expression of the indicated genes (relative to Gapdh). D, MC3T3-E1 cultures were treated as in A following which Tnap and Enpp1 activities were measured using p-nitrophenylphosphate or p-nitrophenylthymidine monophosphate as substrates, respectively. The data are presented as the means Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001, Student's t-test relative to the 0 M treatment control. cleavage by TNAP yields free P i for apatite crystal growth during mineralization (26). Mineralization of MC3T3-E1 cells can therefore be initiated with either ␤GP or P i , with the main difference being that ␤GP-induced mineralization is Tnapdependent. Although PP i inhibited mineralization in both culture systems, the dose of PP i required was differed. Whereas 5 M PP i was sufficient to inhibit ␤GP-induced mineralization, surprisingly, 50 M PP i was required to block P i -induced mineralization (Fig. 6, A and B). This 10-fold difference in potency suggests that PP i might be interfering with Tnap-mediated P i release from ␤GP in addition to blocking mineral growth by binding to crystal surfaces. We thus monitored gene expression levels to determine whether cultures induced by both phosphate sources behaved in a similar manner when exposed to PP i (Fig. 6C). Under both conditions, the effect of PP i was more or less analogous, with PP i causing a slight elevation of Tnap, Opn, and Ank. Enpp1 expression decreased at low doses of PP i and increased at the higher doses. This unexpected increase in Enpp1, as well as the increase in Ank, might be attributable to the P i released from the hydrolysis of PP i that is not bound to mineral. Regardless of the source of added phosphate for mineralization, there were no major differences in gene expression between the P i -and ␤GP-treated cultures. The changes in Tnap and Enpp1 expression were reflected at the protein level in phosphatase and nucleotide pyrophosphatase phosphodiesterase enzyme activity assays (Fig. 6D). Although these assays reflect TNAP and Enpp1 protein levels, actual in vitro activity levels may differ because cell lysates were necessarily assayed here in the absence of culture medium, extracellular matrix, and endogenous inhibitors, which could affect activity in situ.
To test the hypothesis that PP i might also be interfering with Tnap-mediated P i release in the ␤GP-induced cultures, causing decreased mineralization of the cultures, we examined the direct effect of PP i on the ␤GPase activity of Tnap in vitro. Cell lysates from mineralized MC3T3-E1 cultures contain high Tnap activity. Incubation of these extracts with ␤GP results in release of free P i as measured colorimetrically (Fig. 7A). The addition of PP i to the cell extract/␤GP reaction mixture led to an inhibition of ␤GP hydrolysis and P i release (Fig. 7A). To confirm that PP i was having a direct effect on the enzymatic activity of Tnap, we performed the above experiment using purified TNAP in place of the cell lysate. As is shown in Fig. 7B, TNAP independently hydrolyzes both PP i and ␤GP, but when the two substrates were combined, there was a reduction in P i release from ␤GP (see also below, Fig. 8). These results therefore suggest that inhibition of MC3T3-E1 mineralization by PP i also involves altered Tnap-mediated hydrolysis of ␤GP.
The ability of PP i to inhibit Tnap activity may provide some insight into why Tnap present in fully differentiated osteoblasts does not enzymatically cleave the phosphates from Opn, which would in turn decrease binding of Opn to mineral. To test the hypothesis that phosphorylated Opn is at least partly protected by PP i from the phosphatase actions of TNAP, we examined whether PP i could inhibit TNAP hydrolysis of bovine OPN. As was demonstrated for ␤GP in Fig. 7 (A and B), hydrolysis of OPN phosphate by TNAP was significantly reduced in the presence of PP i (Fig. 7C). To avoid the potential for substrate end product inhibition, all of the experiments were performed in FIGURE 7. PP i inhibits TNAP hydrolysis of ␤GP and OPN. A, Cell extract from day 12, differentiated and mineralized MC3T3-E1 cell cultures was incubated either alone or with 5 mM PP i , or with 5 mM ␤GP in a cell-free reaction mixture containing 50 mM Tris-HCl, pH 8.8, and 10 mM MgCl 2 . P i released as a result of hydrolysis was measured colorimetrically by the ammonium molybdate complex method. B, TNAP was incubated either alone, with 5 mM PP i , or with 5 mM ␤GP in the reaction mixture described in A, and P i release was measured accordingly. C, TNAP was incubated either alone or with 0.5 mM OPN or with 5 mM PP i in the reaction mixture described in A. The data are presented as the means Ϯ S.E. **, p Ͻ 0.01; ***, p Ͻ 0.001, Student's t test.
the linear range of enzymatic activity as determined by our measuring of the enzyme kinetics (data not shown).
Kinetic Parameters of TNAP Inhibition by Pyrophosphate-To investigate the mode of TNAP inhibition by PP i , a Lineweaver-Burk plot of the initial velocity of ␤GP hydrolysis at various PP i concentrations was determined. Both the slope and intercepts increase with increasing inhibitor concentration, and thus the curves intersected to the left of the y axis and above the x axis indicate a mixed type of inhibition (Fig. 8A). The changes in both apparent K m and V max imply that PP i may bind to both free enzyme and to the enzyme-substrate complex.
To calculate the dissociation constant (K i ) of the PP i -TNAP complex, a Dixon plot of reciprocal velocities against PP i concentration was constructed (Fig. 8B), resulting in a K i of 0.73 mM. The dissociation constant of the PP i -TNAP-␤GP complex (K i Ј) was calculated from a plot of s/v against PP i concentration (Fig. 8C). The intersection of the curves results in a K i Ј of 1.97 mM. Kinetic parameters obtained from the plots are summarized in Fig. 8D. In summary, a molecule of PP i may bind to either the TNAP active site (competitive inhibition) or to the TNAP-␤GP complex at a secondary site that is only available after ␤GP binding (noncompetitive inhibition).

DISCUSSION
PP i , an anion consisting of two P i molecules linked by an ester bond, is the simplest of the condensed phosphates. PP i inhibits hydroxyapatite crystal growth in cell-free systems in vitro (3,7), and recent reports have documented PP i generation, transport, and activity in vivo, most notably for chondrocytes and the tissues of joints (10). Here, we have extended this understanding of the role of PP i in skeletal tissues, and more specifically its effects on biomineralization, by examining its influence on mineralizing MC3T3-E1 osteoblast cultures. We show that PP i directly affects osteoblast activity, and we present evidence that PP i inhibits mineralization of osteoblast cultures by at least three distinct mechanisms that include direct binding to mineral, up-regulation of Opn production, and inhibition of alkaline phosphatase activity.
Consistent with inhibition of crystal growth by PP i , in this study, we show that PP i inhibits mineralization of extracellular matrix in vitro produced by a well characterized osteoblast cell line, the MC3T3-E1 subclone M14 (45). We also show that this inhibition of mineralization by PP i is abrogated by the actions of tissue-nonspecific TNAP, which hydrolyzes PP i into its constituent P i ions. The importance of PP i as a physiologic inhibitor of mineralization is evident from the skeletal and dental hypomineralization observed in patients with hypophosphatasia attributable to impaired TNAP activity. These patients, as well as Akp2 Ϫ/Ϫ knock-out mice, demonstrate elevated plasma PP i levels in the absence of TNAP enzymatic activity (6,13). Indeed, the bone and tooth hypomineralization of Akp2 Ϫ/Ϫ mice is normalized in Akp2 Ϫ/Ϫ ;Enpp1 Ϫ/Ϫ double-knock-out mice (36,46), suggesting that homeostatic regulation of PP i levels by antagonistic Tnap enzyme activity modulates hydroxyapatite formation in these tissues. However, our results also indicate that there is rapid adsorption of PP i onto hydroxyapatite crystals and that once adsorbed, it is resistant to hydrolysis by TNAP (Fig. 1B). Furthermore, the dose of PP i required to inhibit ␤GP-induced mineralization is only 5 M, and thus any hydrolysis would yield a negligible amount of P i relative to the 10 mM amount of ␤GP used here and commonly by others to provide a source of P i for mineralization. The dose of PP i that we have used is in the physiologic range as reported for human plasma (3.5 M average PP i ) by Russell et al. (6).
Our results also demonstrate that PP i is a specific signal for the induction of Opn expression in osteoblasts. Previous studies have demonstrated that PP i deficiency in Enpp1 Ϫ/Ϫ mice also results in Opn deficiency (11). In our study, we have identified the involvement of specific MAPK signaling pathways responsible for Opn regulation by PP i . Recently, others have shown that P i itself can regulate osteoblast and cementoblast gene expression (31,35,38,39). However, whereas P i regulation of Opn is dependent on intracellular uptake via Na ϩ /P i cotransporters (39), most notably Pit-1 (47), we demonstrate here that PP i directly regulates Opn in a phosphate uptake-independent manner. Moreover, whereas previous studies by Beck et al. (39), have shown that P i regulation of Opn is not dependent on the p38 MAPK pathway in osteoblasts (48), we report here that this pathway is activated by PP i to regulate Opn expression. It should be noted however, that although foscarnet is widely used as an inhibitor of P i uptake, there are indeed other mechanisms for P i transport that may be not be foscarnet-sensitive. Nevertheless, because the inhibitor levamisole prevents Tnap-mediated PP i hydrolysis to P i , and this treatment had no effect on the induction of Opn, then these data, plus the failure of an equivalent dose of P i to induce Opn, collectively indicate that PP i regulates Opn by a mechanism distinct from that of P i .
The addition of exogenous PP i resulted in an increase in Opn mRNA, which led to a gradual cumulative increase in Opn protein levels over 4 days of treatment, with 4 days representing roughly one-third of the 12 days of culture typically used to assay for full MC3T3-E1osteoblast differentiation and mineralization of the extracellular matrix. Indeed, induction of Tnap activity during differentiation by the addition of ascorbic acid and ␤GP typically takes up to 6 days in control cultures. Compared with this 6-day induction time for Tnap, Opn induction at 4 days is consistent with this same time frame. Furthermore this might explain why a relatively high dose of PP i (0.5 mM) might be needed to induce Opn over 4 days compared with the low dose (5 M) required to inhibit mineralization after 12 days. The 0.5 mM concentration of PP i required to induce Opn is far less than the P i concentration required (4 -10 mM) to have a similar induction (39).
The mechanism by which extracellular PP i leads to activation of intracellular signaling pathways remains unknown. Based on its size and charge, PP i likely cannot passively cross the cell membrane. In prokaryotes, chloroplasts, and mitochondria, an ATP-ADP translocator has been reported to import PP i (49 -51), but there is no known importer or extracellular receptor for PP i in mammalian cells (10). The possibility that PP i acts through a plasma membrane receptor is not unreasonable.
Other reported examples of small molecule signaling through osteoblast receptors include the ion-sensing Ca 2ϩ -sensing receptor (52), the nucleotide P2-purinoreceptor family (53), and PP i analogues (bisphosphonates) acting to open connexin 43 hemi-channels (54), all of which in turn activate intracellular signaling pathways. Work is ongoing in our laboratory to clarify upstream effectors of the signaling pathways involved in PP i action as well as to identify transcription factors and genes regulated by PP i .
Based on the known inhibitory actions of Opn on hydroxyapatite crystal growth in cell-free systems (55)(56)(57), in cell cultures models (this study Fig. 5A, and Wada et al. (58)), and in vivo in mineralized tissues (59), the up-regulation of Opn by PP i would likely act to further inhibit and thus regulate mineralization. Recent evidence for this in vivo is demonstrated by the elevated Opn levels in Akp2 Ϫ/Ϫ mice that have elevated PP i in the absence of Akp2 enzymatic activity and by the fact that in Akp2 Ϫ/Ϫ ;Opn Ϫ/Ϫ double-knock-out mice, there is partial rescue of the bone hypomineralization phenotype otherwise observed in the Akp2 Ϫ/Ϫ knock-out alone (60). We also propose here, as an alternative (or simultaneous) activity for this elevated Opn when PP i levels are high, that increased Opn might prevent calcium pyrophosphate crystal deposition in fluids and tissues as is commonly seen in joints in pseudogout (61). Studies are currently in progress to assess the ability of Opn to inhibit calcium pyrophosphate crystal growth and to stabilize mineral ions and putative amorphous precursor phases in solution.
Phosphorylation of Opn appears to be ordered in triplet clusters that favor its interaction with calcium in the hydroxyapatite crystal lattice (62,63). Opn binds directly to hydroxyapatite crystal surfaces both in vivo (64) and in vitro (34), and dephosphorylation of Opn by Tnap prevents much of its mineral binding and crystal growth inhibiting activity (56,65,66) (Fig. 5A). Our high resolution, immunogold localization of Opn at crystal growth sites in the extracellular matrix of MC3T3-E1 osteoblast cultures is consistent with previous data demonstrating a similar, electron microscopic localization of Opn in primary rat osteoblast cultures (34). Dose-dependent inhibition of calcification in vitro by the addition of exogenous Opn has been shown previously for vascular smooth muscle cell cultures (58).
Our study in osteoblast cultures confirms this finding and further demonstrates that Tnap modulation of Opn phosphorylation levels is important for inhibition of mineralization. Using a gelatin gel crystal growth system, Gericke et al. (67) likewise demonstrated the importance of phosphorylation status of Opn in inhibiting hydroxyapatite formation, and similar phosphorylation influences were observed for inhibition of hydroxyapatite and also calcium oxalate crystal growth by peptides of Opn (55,57).
Although differentially phosphorylated forms of Opn have been observed in rat bone cell cultures (68), the amount of Opn dephosphorylation resulting from Tnap activity in vivo remains unknown. Indirect in vivo data supporting a link between Tnap activity and Opn and its phosphorylation status is evidenced by the altered distribution of Opn in bone from Akp2 Ϫ/Ϫ mice (69), although direct measurement of phosphorylation level of Opn in this case has not been determined.
Another finding in our study is that the inhibition of mineralization by PP i that is abrogated by the actions of Tnap can be reduced when another Tnap substrate is present. In vitro, this second substrate is ␤GP, but in vivo, other substrates would be present including mineral-inhibiting phosphoproteins such as Opn. Mineralization of osteoblast cultures can be initiated in vitro with a variety of phosphate esters with ␤GP being the organic phosphate source most commonly used (26, 70 -72). Although high doses of P i have also been demonstrated to inhibit Tnap activity (73), our kinetic experiments clearly demonstrate that PP i may function as a mixed type of inhibition system. It is possible that with substrate (␤GP, Opn) already occupying the active site in Tnap, PP i might bind to a second site on the enzyme, causing conformational change that inhibits its activity. This possibility has previously been suggested for the Tnap activity associated with matrix vesicle-mediated mineralization (72) and is also supported by recent data demonstrating that PP i induces conformational changes in Escherichia coli alkaline phosphatase leading to inhibition of its enzymatic activity (74). Our data showing that PP i acts as an inhibitor of TNAP when another substrate such as ␤GP (in vitro) or Opn (or other phosphoproteins, in vivo) is present also provides a hypothesis to explain why osteoblast Tnap does not inactivate the mineral-inhibiting potential of endogenous Opn in the extracellular matrix (Fig. 7). Recent evidence (75) that PP i also inhibits PHEX activity (an enzyme whose inactivating mutations cause X-linked hypophosphatasia in humans) implicates multiple additional factors affecting the P i /PP i balance regulating mineralization, and more work in this area is required to elucidate their respective contributions.
In summary, the in vitro data presented here for PP i , Tnap, and Opn, together with our recent in vivo data in transgenic mice overexpressing Tnap in collagen extracellular matrices (18), suggest the following partial scenario for induction and regulation of extracellular matrix mineralization. Mineralization initially involves a collagen matrix and enzymatic removal of PP i . The enzymatic hydrolysis of PP i by Tnap also produces additional P i potentially available for hydroxyapatite mineral deposition. Once hydroxyapatite crystal growth is initiated by an as yet unknown mechanism, the enzymatically (Tnap) maintained P i /PP i balance regulates further mineralization in part by inhibiting crystal growth via multiple mechanisms that include direct binding to crystals, up-regulation of mineral-inhibiting Opn, and inhibition of Tnap enzymatic activity. Thus, mineralization is not only inhibited by enzymatic inactivation by Tnap of these two molecular determinants of mineralization, one a protein and the other a small anion inhibitor, but is further enhanced by the fact that PP i also up-regulates Opn. Both inhibitors are found at high levels in bone, both bind to mineral via negatively charged phosphate residues, and both are inactivated by Tnap. Inhibition of Tnap by PP i (in the presence of other Tnap substrates, e.g. Opn and potentially other matrix phosphoproteins) also then maintains the phosphorylated form of Opn to further regulate mineral growth. Given that skeletal and dental mineralization occurs to at least some extent in transgenic mice deficient in these and other mineral-regulating proteins, it is likely that other molecular determinants also are important in regulating crystal nucleation and growth within the extracellular matrix.