Functional Involvement of PHOSPHO1 in Matrix Vesicle–Mediated Skeletal Mineralization


  • The authors state that they have no conflicts of interest.


PHOSPHO1 is a phosphatase highly expressed in bone. We studied its functional involvement in mineralization through the use of novel small molecule inhibitors. PHOSPHO1 expression was present within matrix vesicles, and inhibition of enzyme action caused a decrease in the ability of matrix vesicles to calcify.

Introduction: The novel phosphatase, PHOSPHO1, belongs to the haloacid dehalogenase superfamily of hydrolases and is capable of cleaving phosphoethanolamine (PEA) and phosphocholine to generate inorganic phosphate. Our aims in this study were to examine the expression of PHOSPHO1 in murine mineralizing cells and matrix vesicles (MV) and to screen a series of small-molecule PHOSPHO1-specific inhibitors for their ability to pharmacologically inhibit the first step of MV-mediated mineralization.

Materials and Methods: q-PCR and immunohistochemistry were used to study the expression and localization profiles of PHOSPHO1. Inhibitors of PHOSPHO1's PEA hydrolase activity were discovered using high-throughput screening of commercially available chemical libraries. To asses the efficacy of these inhibitors to inhibit MV mineralization, MVs were isolated from TNAP-deficient (Akp2−/−) osteoblasts and induced to calcify in their presence.

Results: q-PCR revealed a 120-fold higher level of PHOSPHO1 expression in bone compared with a range of soft tissues. The enzyme was immunolocalized to the early hypertrophic chondrocytes of the growth plate and to osteoblasts of trabecular surfaces and infilling primary osteons of cortical bone. Isolated MVs also contained PHOSPHO1. PEA hydrolase activity was observed in sonicated MVs from Akp2−/− osteoblasts but not intact MVs. Inhibitors to PHOSPHO1 were identified and characterized. Lansoprazole and SCH202676 inhibited the mineralization of MVs from Akp2−/− osteoblasts by 56.8% and 70.7%, respectively.

Conclusions: The results show that PHOSPHO1 localization is restricted to mineralizing regions of bone and growth plate and that the enzyme present within MVs is in an active state, inhibition of which decreases the capacity of MVs to mineralize. These data further support our hypothesis that PHOSPHO1 plays a role in the initiation of matrix mineralization.


During the process of endochondral bone formation, chondrocytes and osteoblasts are believed to mineralize their extracellular matrix by promoting the initial formation of crystalline hydroxyapatite (HA) in the sheltered interior of membrane-limited matrix vesicles (MVs).(1) This is followed by the modulation of matrix composition to further promote propagation of apatite outside of the MVs.(2,3) Regulation of this biphasic two-step mineralization process depends on a regulated balance of a number of factors such as Ca2+ and inorganic phosphate (Pi) concentrations, the presence of matrix proteins, and the presence of adequate mineralization inhibitors including inorganic pyrophosphate (PPi), matrix gla protein, and osteopontin.(4–8) Three osteoblast molecules have been identified as affecting the controlled deposition of bone mineral by regulating the extracellular levels of PPi, and in turn, of osteopontin (i.e., tissue-nonspecific alkaline phosphatase [TNAP], a nucleotide pyrophosphatase/phosphodiesterase isozyme [NPP1], and the ANK gene product).(9–17)

In bone, TNAP is confined to the cell surface of osteoblasts and chondrocytes, including the membranes of their shed MVs.(18,19) It has been proposed that the role of TNAP in bone matrix is to generate Pi for HA crystallization.(20–22) However, TNAP has also been hypothesized to hydrolyze the mineralization inhibitor PPi(4) to facilitate mineral precipitation and growth.(23–25) A variety of loss-of-function mutations in the human TNAP gene (ALPL) lead to hypophosphatasia, an inborn error-of-metabolism characterized by rickets and osteomalacia.(26) Mice with null mutations in the orthologous Akp2 gene phenocopy infantile hypophosphatasia(27,28) including elevations in the known substrates of TNAP (i.e., pyridoxal-5′-phosphate and PPi). Electron microscopy revealed that TNAP-deficient MVs, both in patients with hypophosphatasia and in Akp2−/− mice, contain apatite crystals, but that extravesicular crystal propagation is retarded.(29,30) This crystal growth retardation, referred to as the second step of MV-mediated calcification, most likely results from the accumulated levels of PPi in the extracellular matrix as a consequence of the lack of TNAP pyrophosphatase function and the concomitant pyrophosphate-induced increase in osteoblast production of osteopontin, another potent inhibitor of calcification.(17) However, why are Akp2−/− mice born with a mineralized skeleton and still contain HA crystals inside their MVs? TNAP is known to sit on the outer surface of the MV membrane, and although it has not been adequately resolved that there is no TNAP inside MVs, it is likely that another enzyme is responsible for either cleaving PPi or elevating the intravesicular concentration of Pi so as to achieve a Pi/PPi ratio conducive for crystallization.

We have previously proposed that PHOSPHO1, a novel phosphatase, plays the important role of increasing the Pi/PPi ratio inside MVs and thus control the first step of HA crystal deposition inside MVs.(31) MVs have long been recognized to contain high levels of Pi.(32) Since its identification in the chicken,(33) PHOSPHO1 orthologs have also been identified in a number of other species including humans, mice, and zebra fish.(34,35) However, to date, no information exists on the expression of PHOSPHO1 in the mammalian skeleton. PHOSPHO1, a member of the haloacid dehalogenase superfamily, is localized to mineralizing surfaces in both bone and cartilage of the chicken, where its expression precedes the deposition of mineral, suggesting that it is involved in the initial events of mineral formation.(31,36) It is a soluble cytosolic enzyme that has specificity for phosphoethanolamine (PEA) and phosphocholine (PCho)(37,38) but not for PPi and a number of other potential substrates.(37) PEA and PCho are the two most abundant phosphomonoesters in cartilage.(39) In addition, the proportions of membrane phospholipids containing these groups decrease in MVs during mineralization, whereas 1,2-diacyl glycerol accumulates, indicative of phospholipase C activity.(40) This gives rise to the possibility of a novel mechanism whereby plasma membrane bound phosphate may be released through the action of PHOSPHO1 and phospholipase C to contribute to the Pi concentration inside the MV.

The critical first step of mineralization mediates the deposition of the initial crystals of HA. We hypothesize that a TNAP-independent step involves PHOSPHO1 functioning to increase the local concentration of Pi inside the MVs. Thus, we would envisage that an inhibition of PHOSPHO1 activity would result in decreased mineralization within MVs. Therefore, to test this hypothesis experimentally, we examined the expression of PHOSPHO1 in murine mineralizing cells and MVs, screened for and characterized a series of small-molecule PHOSPHO1-specific inhibitors, and used these compounds to pharmacologically inhibit the first step of MV-mediated mineralization.


Chemical libraries

The LOPAC1280 (Sigma, St Louis, MO, USA) and Spectrum (Microsource Discovery, Gaylordsville, CT, USA) libraries were used as a source of potential small-molecule PHOSPHO1 inhibitors. The LOPAC library consists of pharmacologically active compounds covering most of the major target classes (i.e., G protein–coupled receptors and kinases), whereas the Spectrum library contains known bioactives, natural products, and their derivatives. The complete libraries (3280 compounds) were screened. The use of these two libraries allowed the evaluation of hundreds of marketed drugs and biochemical standards. Each compound within these collections was dissolved in 10% DMSO at approximate concentrations of 100 μM and tested at a final concentration of ∼10 μM.

Recombinant PHOSPHO1

A cDNA corresponding to Met19-Cys267 of human PHOSPHO1 was amplified and cloned into the pBAD TOPO TA vector (Invitrogen) as previously described.(37) Briefly, the construct was designed to express PHOSPHO1 fused to a V5 epitope and 6 His-tag at the C terminus. A clone containing the PHOSPHO1 fragment in the correct orientation was identified by restriction digestion of plasmid minipreparations. E. coli were grown in Lauria-Bertani broth (10 liters, 37°C), and recombinant protein expression was induced by treatment with 0.1% (wt/vol) l-arabinose for 4 h.

Inhibitor screening

The semiautomated screening used a Beckman Coulter dual-bridge Biomek FX liquid handler, consisting of a 96-tip head bridge for full plate pipetting. The reactions were measured in 96-well plates containing 25 μl 20 mM MES-NaOH, pH 6.7, 0.01% (wt/vol) BSA, 0.0125% (vol/vol) Tween 20, 2 mM MgCl2, 62.5 μM PEA, 10 μM test compound, and 500 ng (15.5 pmol) of purified recombinant PHOSPHO1. Substrate addition was used to initiate the reaction, thus allowing for an enzyme/compound preincubation. Reactions were allowed to proceed for 60 minutes at room temperature and stopped by the addition of 50 μl BIOMOL green reagent (Biomol International, Plymouth Meeting, PA, USA). The absorbance of each well was measured at 620 nm, and the inhibitory effect of each compound calculated as a percentage in relation to controls containing 1% (vol/vol) DMSO, because each compound was dissolved in 10% DMSO, giving 1% in the final reaction. Each compound that exhibited an inhibition of 40% or more using the automated system was reconfirmed manually, in duplicate, to eliminate the possibility of false positives. The assay used included negative controls (i.e., no inhibition, contained both enzyme and PEA) and positive control (i.e., 100% inhibition, contained only PEA).

Characterization of inhibitors

The IC50 for each inhibitor was determined using the phosphatase assay detailed above. The inhibitor concentration was varied between 100 and 0.3 μM, with each individual reaction repeated in triplicate. The data of absorbance versus inhibitor concentration was plotted using SigmaPlot, and a four parameter logistic curve was fitted. The IC50 was calculated using the following equation: y = min + (max − min/1 + 10(logEC50-x)Hillslope). Effect of inhibitors on recombinant mammalian expressed TNAP protein(41) was assessed under optimal conditions; 1 M diethanolamine buffer, pH 9.6, 1 mM MgCl2, 20 μM ZnCl2, 20 μM inhibitor, 0.5 mM p-nitrophenylphosphate (pNPP). This assay was repeated using recombinant PHOSPHO1 under the optimal conditions described below. Absorbances were measured at 405 nm.

The continuous phosphatase assay to determine kinetic parameters of the reactions involved monitoring the dephosphorylation of pNPP that causes an absorbance change at 405 nm. The reactions were measured in 96-well plates containing 20 mM MES-NaOH, pH 6.7, 0.01% (wt/vol) BSA, 0.0125% (vol/vol) Tween 20, 2 mM MgCl2, and 1.5 μM purified recombinant PHOSPHO1 at room temperature. pNPP and inhibitor concentrations were varied accordingly. Absorbances were measured continuously at 405 nm using a VICTOR HTS plate-reader.(37)

Immunolocalization of PHOSPHO1 within long bones

Ten-day-old male mice were killed by cervical dislocation, and tibias were fixed in 4% paraformaldehyde in PBS for 24 h before decalcification in 0.5 M EDTA (pH 8.0) for a further 24 h at 4°C. The fixed tissue was dehydrated and paraffin embedded using standard techniques. Paraffin sections (6 μm) were dewaxed in xylene and rehydrated through a graded series of alcohol solutions, and antigen retrieval was achieved by heating in sodium citrate for 90 minutes at 70°C followed by extensive washing in PBS. Endogenous peroxidases were blocked by incubating the sections with 3% hydrogen peroxide (in methanol), followed by three washes in PBS. Unspecific protein binding was blocked by normal goat serum (1:5) diluted in PBS for 30 minutes. Rabbit antisera to mouse PHOSPHO1 (a generous gift from Professor Ikramuddin Aukhil, University of Florida) was diluted 1:200 in PBS and incubated with the tissue section at 4°C overnight. Control sections received a similar dilution of normal rabbit serum. After this, the sections were washed in PBS and incubated with a 1:100 dilution of goat anti- rabbit IgG-peroxidase (DAKO, Cambridgeshire, UK) for 60 minutes. DAB substrate reagent (0.06% DAB, 0.1% H2O2 in PBS) was incubated for 8 minutes, rinsed in PBS, and counterstained with Mayer's hematoxylin (Sigma) for 5 minutes. The sections were finally dehydrated and mounted in DePeX.

Isolation of MVs from chick growth plate cartilage

All animal studies and protocols were approved by the Institutional Animal Users Committees of both Roslin Institute and Burnham Institute for Medical Research. Under sterile conditions, growth plate cartilage from 3-week-old male broiler chickens was collected and diced. MVs were released by incubating the tissue at 37°C with 0.45% collagenase (Worthington, type II) in 50 mM Tris-HCl, pH 7.6, 120 mM NaCl, and 10 mM KCl for 3 h with constant agitation. The digested tissue was passed through a 40-μm sieve to remove undigested material. MVs were harvested from the digest by differential centrifugation as previously described.(42) Briefly, the digest was centrifuged for 30 minutes at 1500g to collect chondrocytes, at 30,000g to remove subcellular debris, and at 250,000g to pellet MVs.

Isolation of MVs from primary cultured calvarial osteoblasts

Mouse calvarial cells were isolated from 3-day-old mice through sequential collagenase digestion, as previously described.(11,16) Calvarial cells from Akp2−/−, Akp2+/−, and wildtype (WT) mice were pooled separately and plated at a density of 20,000/cm2 in α-MEM (Gibco, Paisley, UK) containing 10% FBS and 50 μg/ml ascorbate for a period of 21 days. The cell monolayer was washed with 50 mM Tris-HCl, pH 7.6, 120 mM NaCl, and 10 mM KCl and incubated with 0.45% collagenase (Worthington, type II) in 50 mM Tris-HCl, pH 7.6, 120 mM NaCl, and 10 mM KCl at 37°C for 120 minutes, with constant agitation. This cell suspension was subjected to differential centrifugation as described above to isolate both cells and MVs.

MV phosphatase activity and mineralization ability

Phosphatase activity within chick and mouse MVs was determined using the standard discontinuous colorimetric assay.(43) In brief, reactions were measured in 96-well plates containing 200 μl of 25% (wt/vol) glycerol, 20 mM TBS, pH 7.2, 25 μg/ml BSA, 2.5 mM PEA, 2 mM MgCl2, and 12 μg MV protein.(37) PHOSPHO1 inhibitors at a final concentration of 1 mM were used where appropriate. The mouse MVs were either left intact or ruptured by sonication to release the cytosolic contents. TNAP activity was determined using the Thermo-line ALP reagent (Melbourne, Australia). Total TNAP activity was expressed as nmoles pNPP hydrolyzed per minute per milligram protein. The in vitro calcification ability of MV was determined by their ability to form calcium phosphate in vitro.(44) In brief, samples of MV protein (15 μg of chick MVs and 20 μg of TNAP null osteoblast-derived MVs) were incubated in calcification buffer in the presence of 0–3 mM phosphoester substrate for 5.5 h at 37°C. The reaction was terminated by centrifugation at 8800g for 30 minutes to pellet both MVs and any calcium phosphate mineral formed during incubation. The pellet was solubilized with 0.6N HCl for 24 h and used directly for calcium quantification using the O-cresolphthalein complexone method (CPC Kit; Thermotrace).

Western blotting of PHOSPHO1 within MVs

Murine MVs were analyzed for the presence of PHOSPHO1 by immunoblotting. MV preparations were lysed in PBS containing 1.6 mg/ml of Complete protease inhibitor cocktail (Roche, Lewes, UK). Samples corresponding to 10 μg total protein were incubated at 70°C for 10 minutes in LDS sample buffer before loading. Samples were run on a 10% Bis-Tris NuPAGE gel and electroblotted to nitrocellulose, which were incubated in blocking solution (5% nonfat milk in Tris buffered saline with 0.1% Tween 20). The membranes were probed with a 1:750 dilution of rabbit-anti-PHOSPHO1 anti-sera in blocking solution and washed three times with PBS. Blots were incubated with goat anti- rabbit IgG-peroxidase (DAKO) diluted 1:2000 in blocking solution. The immune complexes were visualized by enhanced chemiluminescence.

Real-time quantitative PCR of PHOSPHO1 within murine tissues

Tissues from several 10-day-old male mice were pooled, and RNA was isolated by phenol/chloroform extraction and used directly in a quantitative RT-PCR (qPCR) reaction. The Brilliant SYBR Green QRT-PCR Master Mix Kit (Stratagene) method was used to allow quantification by fluorescence during the PCR reaction. Briefly, 25 μl SYBR green mastermix was added to 10 ng RNA along with 0.2 μM forward and reverse primers (forward: GACAATGAGCGGGTGTTTTC; reverse: GGGGATGGTCTCGTAGACAG). The RT-PCR reaction was cycled in a Perkin-Elmer Applied Biosystems Prism 7700 sequence detector as follows: 50°C for 30 minutes (RT step), 95°C for 10 minutes, 40 cycles of 95°C for 30 s, 57°C for 30 s, and 72°C for 1 minute. Each tissue sample tested was tested in triplicate and compared with 18S RNA (classic II primers; Ambion) external control that allowed normalization of results. A dilution series of both gene of interest and external control were carried out and subjected to an identical PCR to allow estimation of PCR efficiency. Relative differences in expression were calculated using the 2−ΔCT method.

Semiquantitative PCR of osteoblast-specific genes

Murine osteoblasts (Akp2−/− and WT) were cultured in α-MEM containing 10% FBS, 50 μg/ml ascorbate, and β-glycerophosphate (10 mM) for up to 14 days. Total RNA was extracted by phenol/chloroform extraction and treated with DNase I (Ambion) according to the manufacturer's instructions. The RT-PCR reaction was carried out using the SUPERSCRIPT-First Strand synthesis system for RT-PCR with Oligo dT (Roche). Primer pairs for β-actin RNA were used as a control. The reaction mixture contained 5 μg RNA, 500 ng Oligo dT, 2 μl 10>RT buffer, 2 mM MgCl2, 10 mM DTT, 0.5 mM DNTPs, and 200 μl Superscript enzyme. The reaction was cycled as follows: 25°C for 10 minutes, 42°C for 50 minutes, and 70°C for 15 minutes. The cDNA was used directly in a PCR reaction containing 0.2 mM dNTP mix (Promega), 5 μl 10>PCR Buffer (Roche), 5 units Taq polymerase (Roche), 0.5 μM of the forward and reverse primers, and 1 μl cDNA. This was cycled in at 94°C for 5 minutes, 30 cycles of 94°C for 30 s, 57°C for 30 s, and 72°C for 1 minute, and finally one step on 72°C for 10 minutes. The primers sets were designed to span at least one intron so that any amplification from contaminating genomic DNA would be identified. The primers were as follows: PHOSPHO1 (forward: GACAATGAGCGGGTGTTTTC; reverse: GGGGATGGTCTCGTAGACAG); TNAP (forward: ACTACCACTCGGGTGAACCA; reverse: TGAGATCCAGGCCATCTAGC); and β-actin (forward: TCCATCATGAAGTGTGACGT; reverse: ACGATGGAGGGGCCGGACTC). Each reaction was analyzed on 1.5% agarose gels run in the presence of ethidium bromide (250 μg/liter).

Statistical analysis

ANOVA was performed to determine the significance of a given result. General linear model analysis incorporating pairwise comparisons using Tukey's test was used to compare groups within the ANOVA models. All data are expressed as the mean ± SD. Statistical analysis was performed using Minitab 14. Statistical significance was accepted at p < 0.05.

Figure Figure 1.

qPCR of PHOSPHO1 in murine tissues. Data show relative expression of PHOSPHO1 with reference to that of the liver that is set at 1 because of the lowest expression level of this tissue. RNA was isolated from mouse tissues by phenol chloroform extraction and used directly in the qPCR reaction using the DNA intercalating dye, SYBR green. Results are mean ± SD (n = 3).


PHOSPHO1 expression in whole tissues

To study the levels of PHOSPHO1 gene expression in different mammalian tissues, we used qPCR to obtain relative expression values. Bone had the highest expression levels, whereas the least amount of PHOSPHO1 transcript was found in the liver (Fig. 1). The difference in expression between these two tissue types was ∼120-fold. Furthermore, low transcript levels were detectable in all tissues examined: heart (5.67), bone marrow (3.57), adipose tissue (3.38), brain (1.96), and gut (1.20; numbers in brackets represent the fold difference compared with PHOSPHO1 expression in liver, which was arbitrarily set as 1 for comparison).

Immunolocalization of PHOSPHO1 to skeletal cells and identification of PHOSPHO1 in MVs

The high expression level of PHOSPHO1 in bone observed by qPCR was confirmed and extended by the immunolocalization of PHOSPHO1 to primary regions of ossification in both growth plate cartilage and trabecular and cortical bone. PHOSPHO1 was restricted to the hypertrophic zone of the growth plate (Fig. 2A). No staining was observed in the proliferating chondrocytes, and in comparison with the prehypertrophic chondrocytes, the terminally differentiated cells displayed little staining. PHOSPHO1 immunoreactivity was also observed within the chondrocytes of the developing secondary ossification center (Fig. 2B). PHOSPHO1 was also present on the surface of the trabecular bone of the metaphysis and within the osteoblasts lining the bone-forming surfaces of the primary osteons within the periosteal region of cortical bone (Figs. 2C and 2D). All control sections were negative. Confirmation that PHOSPHO1 is a strong candidate as a modulator of matrix mineralization was obtained by the detection of PHOSPHO1 in MVs isolated from mouse calvarial osteoblasts (Fig. 2E). A single band of 29 kDa was detected, which agrees with the expected protein size for one of the PHOSPHO1 splice variants (S Roberts and C Farquharson, unpublished observation, 2006).

Figure Figure 2.

Expression of PHOSPHO1 in mouse tibial sections and MV. (A) Strong specific localization within the growth plate was limited to the early hypertrophic chondrocytes (arrows). Chondrocytes of the proliferating zone were negative (*). Trabecular bone surfaces within the metaphyseal region of the tibia were also positive (arrowheads; scale bar = 100 μm). (B) Positive PHOSPHO1 staining in the developing secondary ossification center of the tibia (arrows; scale bar = 100 μm). (C) Higher magnification of positively stained osteoblasts lining the bone forming surfaces of primary osteons in cortical bone (arrows; bar = 50 μm). (D) Trabecular bone surfaces within the metaphysis containing positively stained osteoblasts (arrows; scale bar = 50 μm). Control sections are displayed in the bottom panel of each pair (B–D). (E) Immunoblot showing the localization of PHOSPHO1 to murine MVs isolated from calvarial osteoblasts. Immunoreactivity is compared with that of recombinant PHOSPHO1. The recombinant protein is of greater mass because of the presence of a C-terminal tagged region.

Identification of active PHOSPHO1 in MVs

To determine if PHOSPHO1 present in murine MVs was active, we tested for hydrolase activity within intact and sonicated MVs isolated from cultures of Akp2−/−, Akp2+/−, and WT osteoblasts. This strategy was adopted to eliminate the possibility that TNAP hydrolysis of PEA may mask PHOSPHO1 activity. Cell phenotype was confirmed using ALP histochemistry and mineralization capability by von Kossa staining (data not shown). TNAP activity of intact MVs purified from these cultures mirrored that of TNAP histochemistry (Fig. 3A). MV preparations were used directly in the standard discontinuous colorimetric assay. It was found that the WT and Akp2−/− intact MVs had a hydrolase activity of 3.3 ± 0.4 and 0.07 ± 0.4 nmol/min/mg MV protein, respectively. TNAP heterozygous MVs had an activity of 1.68 ± 0.5 nmol/min/mg MV protein (Fig. 3B). Sonication of the MV preparation, to assess PHOSPHO1 and TNAP combined activity, caused an increase of ∼1 nmol/min/mg in all cases (p < 0.05), indicating that the enzyme responsible for this increased hydrolysis is cytosolic.

Is PHOSPHO1 expression influenced by the presence of TNAP?

To analyze whether PHOSPHO1 expression was altered by TNAP expression, RNA was extracted from cells cultured for up to 14 days under mineralizing conditions. The TNAP genotype was confirmed by PCR, and PHOSPHO1 expression was not altered by the background TNAP status of the cells nor by continuous culture for up to 14 days (Fig. 4).

Identification and characterization of PHOSPHO1 inhibitors

To be able to conduct studies to examine the effects of PHOSPHO1 inhibition on MV-mediated calcification, we first needed to identify and characterize small molecule inhibitors specific for PHOSPHO1. Consequently, we screened two well known chemical libraries (i.e., the LOPAC and the Spectrum libraries) and identified 17 compounds capable of inhibiting recombinant PHOSPHO1 activity with IC50 values of 10 μM or less. From these 17 compounds, 3 were selected based on certain criteria (i.e., solubility, reconfirmed inhibitory potential of 80% or more, and the absence of any reactive groups such as thiols). This led to the identification of SCH 202676, Lansoprazole, and Ebselen as compounds for further study (structures shown in Fig. 5A).

Figure Figure 3.

Hydrolase activity of MVs from WT, Akp2+/−, and Akp2−/− osteoblasts. (A) pNPP and (B) PEA phosphatase activity of MV preparations. Data are presented as the mean ± SD of nine replicates from each genotype. PEA hydrolase activity was significantly higher (p < 0.05) in sonicated vs. intact MVs.

Figure Figure 4.

Analysis of TNAP and PHOSPHO1 expression in WT and Akp2−/− osteoblasts. Comparison of TNAP and PHOSPHO1 expression between WT and Akp2−/− osteoblasts over a 14-day culture period.

Figure Figure 5.

Identification of PHOSPHO1 inhibitors. (A) Structures of three compounds found to inhibit recombinant PHOSPHO1 by >80%. (B) IC50 determination of each PHOSPHO1 inhibitor. Recombinant human PHOSPHO1 was incubated with various concentrations of SCH 202676, Lansoprazole, or Ebseleln at room temperature for 60 minutes, and phosphate released during the reaction was measured. Results are mean ± SD (n = 3).

The calculated IC50 values for each of these compounds when tested against PHOSPHO1-mediated hydrolysis of PEA were 1.97 ± 0.01 (SCH 202676), 4.71 ± 0.1 (Lansoprazole), and 2.81 ± 0.04 μM (Ebseleln) (Fig. 5B). The PHOSPHO1 inhibition displayed by each of these compounds was also studied by analyzing data of initial reaction velocity in the presence of varying concentrations of inhibitor and pNPP. From analysis of the resultant Line-Weaver Burke plots, all four inhibitors displayed lines that intercept after the y-axis close to the x-axis, which is a hallmark of noncompetitive inhibition (data not shown). A Michaeles-Menten curve was constructed under saturating quantities of pNPP, which also indicated that the inhibitors are of a noncompetitive nature (data not shown) The inhibitor constant Ki, for the inhibitor/enzyme complex, was also calculated for each of these reactions, with SCH 202676, Lansoprazole, and Ebseleln displaying Ki values of 1.08 ± 0.83, 71.28 ± 9.21, and 31.00 ± 0.74 μM, respectively.

Effect of PHOSPHO1 inhibitors on MV phosphatase activity

With these novel inhibitors at hand, we used chick MV protein directly as a source of WT PHOSPHO1 in a discontinuous colorimetric assay to assess the effect of inhibiting PHOSPHO1 activity on Pi generation. Each inhibitor was used at a concentration of 1 mM. The uninhibited MV protein exhibits a PEA hydrolase activity of 87.1 ± 1.8 nmol/min/mg, whereas in the presence of Lansoprazole and SCH202676, this activity was reduced to 62.9 ± 1.1 and 73.0 ± 2.7 nmol/min/mg, respectively (Fig. 6A). In comparison with the uninhibited reactions, these activities relate to a reduction in activity of ∼28% for Lansoprazole (p < 0.001) and 16% for SCH202676 (p < 0.001). Neither Lansoprazole nor SCH202676 interfered with the ability of TNAP to catalyze the hydrolysis of pNPP under optimal conditions, but did exhibit inhibition of PHOSPHO1 mediated hydrolysis of pNPP similar to that seen when using PEA as a substrate (Fig. 6B).

Effect of PHOSPHO1 inhibitors on MV calcification

Using the in vitro calcification assay, we found that PEA supports the calcification of chick MVs, purified from growth plate cartilage, which was comparable with that shown by the phosphoester β-glycerophosphate, 1.0 ± 0.01 and 1.3 ± 0.04 units of precipitated calcium, respectively (Fig. 7A). Chick MV preparations were subsequently incubated with 1 mM of Lansoprazole or SCH202676, which resulted in a slight decrease in calcifying potential in the presence of PEA (∼10% with each inhibitor; p < 0.001; Fig. 7B).

To increase the sensitivity of this assay, the inhibitor actions were examined on MVs extracted from cultured Akp2−/− osteoblasts. This strategy eliminated the possibility of TNAP hydrolyzing the phosphoester and potentially masking the effects of the PHOSPHO1 inhibitors. Using this strategy, a much more pronounced inhibitory effect was seen with Lansoprazole and SCH202676. In comparison with uninhibited reactions, Lansoprazole exhibited a 56.8% inhibition of calcification (p < 0.01), whereas SCH202676 inhibited calcification by 70.7% (p < 0.001; Fig. 7C).


The mechanisms and proteins regulating matrix mineralization are not yet fully understood, and we are still unclear whether, under physiological conditions, mineralization is caused by the presence of promoters of mineralization (active process) or a lack of mineralization inhibitors (passive process).(45) Notwithstanding the presence of an optimum balance of both inhibitors and promoters, the mineralization process is clearly dependent on the attainment of sufficiently high concentrations of Pi for the formation of de novo calcium phosphate crystals. Generation of Pi for mineralization has long been attributed to the actions of TNAP,(6,20) but it is now clear that the production of Pi for matrix calcification is not entirely attributable to TNAP activity.

Figure Figure 6.

PEA hydrolase potential of chick MVs in the presence of PHOSPHO1 inhibitors. (A) Chick MVs were isolated from tibial growth plate cartilage and used directly to measure phosphate released, as an indicator of PEA hydrolase activity, in the presence of Ebselen (EBS), Lansoprazole (LAN), or SCH 202676 (SCH). (B) Effect of inhibitors on PHOSPHO1 (open bars) and TNAP (shaded bars) pNPP hydrolase activity under optimal conditions. Results are mean ± SD (n = 3, ***p < 0.001 for inhibited reaction compared with uninhibited reaction).

In newborn Akp2 knockout mice, bone development and mineralization seem normal even though hypomineralization and other abnormalities do subsequently appear.(11,27,28) Although mechanisms involving in utero protection cannot be ruled out, the postnatal mineralization defects in Akp2−/− mice are likely to be caused by a build up of PPi, a known substrate of TNAP(21) and a potent inhibitor of HA crystal formation.(4) Indeed, hypomineralization is greatly reduced in [Akp2−/−; Enpp1−/−] double-knockout mice where extracellular PPi (ePPi) concentrations return to normal after ablating NPP1 enzyme activity (i.e., the enzyme that produces PPi at the surface of osteoblasts and MVs).(10,11) Hypomineralization in these double knockout mice is corrected in a site-specific manner in the calvaria and vertebrae, whereas residual hypomineralization remains in metatarsals and long bones.(6) Mechanistically, the aforementioned build-up of ePPi in Akp2−/− mice in turn leads to upregulated osteopontin expression by Akp2−/− osteoblasts,(17) and it has recently been shown that it is the combined accumulation of ePPi and osteopontin that causes osteomalacia in Akp2−/− mice.(17)

Figure Figure 7.

Potential for MVs to calcify in vitro in the presence of PEA and PHOSPHO1 inhibitors. (A) Ability of chick MVs to induce calcification in the presence of 3 mM PEA and 3 mM β-glycerophosphate (RU, relative units). (B) Calcification of chick MVs in the presence of 1 mM Lansoprazole (LAN) or SCH 202676 (SCH). (C) Calcification of murine TNAP-null MVs in the presence of 1 mM Lansoprazole (LAN) or SCH 202676 (SCH). Results are mean ± SD (n = 3; ***p < 0.001, **p < 0.01 for inhibited reaction compared with uninhibited reactions).

The location of TNAP to the outer surface of MV membrane and the shown role of this enzyme as a pyrophosphatase in vivo suggest that other molecules or mechanisms are responsible for increasing intravesicular Pi levels to achieve a Pi/PPi ratio conducive for crystallization. The sodium-dependent phosphate type III transporter Pit-1 (GLvr-1)(46–48) and other phosphatases, such as pyrophosphatase, AMPase, and ATPase, known to be present in MVs,(6,49,50) may be involved. In this study, we provide the first functional evidence that PHOSPHO1 is a MV phosphatase involved in skeletal mineralization.

The immunolocalization of PHOSPHO1 in the mouse is consistent with its presumed role in matrix mineralization. PHOSPHO1 was localized in chondrocytes in the early hypertrophic zone of the growth plate and to bone-forming surfaces in both long bones (trabecular and cortical) and the site of the developing secondary ossification center. This distribution, together with the absence of strong PHOSPHO1 immunoreactivity in the hypertrophic chondrocytes situated deep in the mineralized zone, suggests that PHOSPHO1 has a pivotal role in the initial stages of the mineralization process and is needed for the de novo formation of the inorganic phase but not for the continued crystal growth of hydroxyapatite. This concurs with studies of the developing chick skeleton, where the expression of PHOSPHO1 in the midshaft of long bones preceded alizarin red staining, indicative of calcium deposition.(31) This localization of PHOSPHO1 to bone-forming surfaces mirrors that of TNAP(51,52) and perhaps indicates synergy between the two enzymes to achieve HA crystal growth. Although likely working in concert, no evidence was obtained to suggest that PHOSPHO1 expression was influenced by the presence/absence of TNAP, possibly indicating that PHOSPHO1 is functioning independently from TNAP.

The matrix staining of PHOSPHO1 on bone-forming surfaces is likely to reflect the presence of the enzyme in osteoblast-derived MVs, which are deposited within newly formed osteoid.(53) Indeed, the presence of PHOSPHO1 in MVs derived from calvarial osteoblasts was confirmed by immunoblotting, as was the ability of TNAP null MV lysates to catalyze the hydrolysis of PEA. This indicates that the enzyme responsible for this degradation is cytosolic and, as according to the KEGG database(54) only TNAP and PHOSPHO1 have the ability to cleave PEA, it is likely that this cytosolic enzyme is in fact PHOSPHO1.

The high expression of PHOSPHO1 in bone noted by qPCR is consistent with the immunolocalization of PHOSPHO1; however, its low basal expression in a number of soft tissues may indicate other putative roles for this enzyme. Phosphocholine is an important regulatory compound for the metabolism of phosphatidylcholine, the major phospholipid in mammalian tissues.(55) In heart muscle, it has been documented that the PCho pool is much lower than that in the liver or HeLa Cells(56); however, the enzyme responsible for this hydrolysis remains unknown.(57) It is interesting to speculate that this step may be controlled by the PCho activity of PHOSPHO1,(37,38) thereby explaining the low expression in the heart and implicating PHOSPHO1 in a possible novel pathway that may be mediated by one of the alternate transcripts from the PHOSPHO1 gene.

From the high-throughput screening of compounds from the two chemical libraries LOPAC and SPECTRUM, 17 inhibitors of PHOSPHO1-mediated hydrolysis of PEA were observed. Many of these inhibitors, however, had undesirable properties, such as cisplatin, which conceivably intercalates between the disulphide bridge in PHOSPHO1. In addition, compounds such as the dimeric disulphide-containing compound thiram would be susceptible to nucleophlic attack and reduction. Other compounds deemed unsuitable included those that possess thiol groups. The functional group of the amino acid cysteine is indeed a thiol; therefore, it is likely that compounds such as mercaptobenzothiazole, which was identified during this screen, would cause an inhibition through chemical modification of the PHOSPHO1 structure. This compound structure screening led to the identification of the three inhibitors, SCH 202676, Lansoprazole, and Ebseleln, deemed suitable for further characterization.

Lansoprazole and SCH 202676 decreased the amount of liberated Pi using isolated chick MV by 28% and 16%, respectively. Ebselen has no effect on PEA hydrolase activity of MVs; thus, it is likely that this inhibitor is not effective against the WT enzyme. Because this is a noncompetitive inhibitor and binds at a site distinct from the active site, it is conceivable that Ebselen is binding recombinant PHOSPHO1 at the C-terminal tagged region, thus inhibiting the recombinant but not the WT enzyme. SCH 202676 is a thiadiazole compound that acts on and inhibits signaling through G protein–coupled receptors.(58) It is thought that this compound is an allosteric modulator of G protein–coupled receptors and may recognize an intracellular regulatory domain of the protein; however, its exact mode of inhibition remains unknown.(59) In contrast, Lansoprazole is an extremely well-characterized compound. This drug belongs to a class of compounds known as the 2-(2-pyridylmethylsulfinyl)-1H-benzimidazoles and is an inhibitor of H+ and K+ (H+/K+)-ATPase of stomach parietal cells. Under acidic conditions, Lansoprazole is converted into an acid-activated cationic sulphenamide form (AG2000), which acts as a proton pump inhibitor,(60) blocking the final step of acid production.

The effect of Lansoprazole and SCH 202676 on Pi release by chick MVs was greater than their inhibitory effect on MV calcification, which was only ∼10% with each inhibitor. This imbalance is possibly caused by TNAP activity of the MV preparations, masking the effects of the PHOSPHO1 inhibitors by promoting PPi hydrolysis and/or maintaining Pi levels at about the minimum threshold for mineralization to occur normally. Alternatively, preformed mineral crystals found within MVs may be acting as local nucleators, allowing mineral propagation, even in the presence of PHOSPHO1 inhibitors. To avoid interference from TNAP in the assay system, we used MVs isolated from calvaria of TNAP null mice. In this case, a more profound inhibition was observed; SCH 202676 and Lansoprazole caused a 71% and 57% decrease in calcification, respectively. This indicates that PHOSPHO1, which is sequestered within the lumen of the MV, has the ability to hydrolyze PEA to increase the intravesicular concentrations of Pi to allow mineralization to occur. This is likely to occur in synergy with phosphate(46–48) and calcium(61–63) transporters to facilitate the production of the initial crystals for hydroxyapatite deposition. SCH202676 is an allosteric antagonist of dopamine receptors, Lansoprazole is a proton pump inhibitor, and both are already being used in the treatment of patients; however, to the best of our knowledge, their use has not been reported to be associated with rickets or osteomalacia.

In conclusion, the Phospho1 gene is highly expressed in bone tissue, and the PHOSPHO1 protein localization is restricted to sites of skeletal mineralization in the mouse. Furthermore, PHOSPHO1 protein is cytosolic and active within murine osteoblast-derived MVs, retaining the ability to hydrolyze the PEA, as previously shown with the recombinant enzyme.(37,38) In addition, the PHOSPHO1 inhibitors SCH 202676 and Lansoprazole have the ability to modulate the in vitro mineralization of MVs. These data further strengthen the hypothesis that PHOSPHO1 has a role in bone mineralization, likely to be linked to the glycerolipid metabolism pathways involving the degradation of phosphatidylethanolamine and phosphatidylcholine and the production of Pi for MV-mediated mineralization.


The authors thank Helen Owen for assistance with the immunohistochemistry. We also thank Dr Marc F. Hoylaerts, Center for Molecular and Vascular Biology, University of Leuven, Belgium, for invaluable advice regarding the inhibitor assays. This work was supported in part by Grants AR53102, AR47908, and DE12889 from the National Institutes of Health and the Biotechnology and Bioscience Research Council (BBSRC) UK. The authors thank Immunodiagnostic Systems Ltd., UK for a Council for Advancement and Support of Education (CASE) award (to SJR).