In Vivo Overexpression of Tissue-Nonspecific Alkaline Phosphatase Increases Skeletal Mineralization and Affects the Phosphorylation Status of Osteopontin



Functional ablation of tissue-nonspecific alkaline phosphatase (TNAP) (Alpl−/− mice) leads to hypophosphatasia, characterized by rickets/osteomalacia attributable to elevated levels of extracellular inorganic pyrophosphate, a potent mineralization inhibitor. Osteopontin (OPN) is also elevated in the plasma and skeleton of Alpl−/− mice. Phosphorylated OPN is known to inhibit mineralization, however, the phosphorylation status of the increased OPN found in Alpl−/− mice is unknown. Here, we generated a transgenic mouse line expressing human TNAP under control of an osteoblast-specific Col1a1 promoter (Col1a1-Tnap). The transgene is expressed in osteoblasts, periosteum, and cortical bones, and plasma levels of TNAP in mice expressing Col1a1-Tnap are 10 to 20 times higher than those of wild-type mice. The Col1a1-Tnap animals are healthy and exhibit increased bone mineralization by micro–computed tomography (µCT) analysis. Crossbreeding of Col1a1-Tnap transgenic mice to Alpl−/− mice rescues the lethal hypophosphatasia phenotype characteristic of this disease model. Osteoblasts from [Col1a1-Tnap] mice mineralize better than nontransgenic controls and osteoblasts from [Col1a1-Tnap+/−; Alpl−/−] mice are able to mineralize to the level of Alpl+/− heterozygous osteoblasts, whereas Alpl−/− osteoblasts show no mineralization. We found that the increased levels of OPN in bone tissue of Alpl−/− mice are comprised of phosphorylated forms of OPN whereas wild-type (WT) and [Col1a1-Tnap+/−; Alpl−/−] mice had both phosphorylated and dephosphorylated forms of OPN. OPN from [Col1a1-Tnap] osteoblasts were more dephosphorylated than nontransgenic control cells. Titanium dioxide-liquid chromatography and tandem mass spectrometry analysis revealed that OPN peptides derived from Alpl−/− bone and osteoblasts yielded a higher proportion of phosphorylated peptides than samples from WT mice, and at least two phosphopeptides, p(S174FQVS178DEQY182PDAT186DEDLT191)SHMK and FRIp(S299HELES304S305S306S307)EVN, with one nonlocalized site each, appear to be preferred sites of TNAP action on OPN. Our data suggest that the promineralization role of TNAP may be related not only to its accepted pyrophosphatase activity but also to its ability to modify the phosphorylation status of OPN.


Tissue-nonspecific alkaline phosphatase (TNAP) is encoded by the ALPL gene in humans and the Alpl (aka Akp2) gene in mice. Mouse TNAP is first expressed in the inner cell mass of blastocysts, primordial germ cells, the neural tube, and placenta during embryogenesis,[1] then in osteoblasts, chondrocytes, odontoblasts, renal tubule cells, macrophages, adipocytes, and endothelial cells in later stages.[2] We have previously generated Alpl knockout mice and demonstrated that they exhibit impaired bone mineralization and pyridoxine-dependent seizures that lead to death before weaning.[2, 3] The phenotype of the Alpl−/− mouse includes: barely detectable plasma alkaline phosphatase (AP) activity; elevated plasma pyridoxal-5-phosphate (PLP; a form of vitamin B6), and inorganic pyrophosphate (PPi); rickets/osteomalacia; and postnatal death, accurately modeling the infantile form of hypophosphatasia (HPP).[4, 5] TNAP is a glycosylphosphatidylinositol (GPI)-anchored ectoenzyme capable of dephosphorylating a broad range of molecules in vitro, such as p-nitrophenylphosphate, β-glycerophosphate (βGP), DNA, phosphoproteins, PPi, and PLP,[6] and there is conclusive evidence to indicate that the latter two molecules represent physiological substrates of TNAP, because the elevated plasma levels of PPi and PLP in both Alpl−/− mice and HPP patients explain the pathophysiology of HPP.[7] Breeding Alpl−/− mice to mice deficient in the production (Enpp1−/−) or transport (ank/ank) of PPi partially corrects the skeletal defect in Alpl−/− mice, confirming that increased PPi levels are responsible for the skeletal disease seen in HPP.[8, 9] Similarly, pyridoxal supplementation of Alpl−/− mice leads to prevention of the epileptic seizures, confirming the role of TNAP in the metabolism of PLP in vivo.[3, 10] Our recent data have shown that enzyme replacement therapy (ERT) using mineral-targeted recombinant TNAP prevents the skeletal and dental abnormalities associated with HPP in the Alpl−/− model.[11-14] The therapeutic principle involves administration of recombinant TNAP fused to a C-terminal polyaspartic acid sequence that confers high affinity for hydroxyapatite. This mineral-targeted TNAP acts at sites of mineralization to prevent the skeletal and dental defects by reducing local PPi concentrations, and increasing levels of absorbable pyridoxal by dephosphorylating PLP to prevent seizures in HPP mice.[11]

Alpl knockout mice also display marked changes in osteopontin (OPN, encoded by Spp1), with elevated expression at both its RNA and protein levels.[15-17] OPN is expressed in a wide variety of cells, such as osteoblasts, chondrocytes, osteocytes, osteoclasts, nephrons, trophoblasts, T-lymphocytes, vascular smooth muscle cells, macrophages, and certain cancer cells.[18-20] Although the biological role of OPN is incompletely understood, one known function is to anchor osteoclasts to the hydroxyapatite surface through its poly-aspartic acid sequences.[21] OPN also binds to CD44 and αvβ3 integrin via its arginyl-glycyl-aspartic acid (RGD) sequence and mediates cell signaling and/or migration.[22] OPN is a phosphorylated glycoprotein, with 36 serine/threonine phosphorylation sites in the human protein.[23] This phosphorylation is functionally important because the inhibitory effect of OPN on mineral deposition was diminished if 84% of covalently bound phosphates were removed from OPN.[24] Phosphorylated OPN inhibits mineralization in vascular smooth muscle cells, whereas dephosphorylated OPN does not.[25] Certain phosphorylated OPN peptides are also capable of inhibiting hydroxyapatite formation in vitro[26] and cause dose-dependent inhibition of mineralization in cultured cells.[27] It is important to determine whether the OPN in Alpl−/− mice is phosphorylated because increased phosphorylated OPN could contribute to the impaired bone mineralization.

Expression of TNAP precedes that of OPN during osteoblast maturation[28] and an interesting crosstalk between TNAP and OPN expression has been recognized in bone cells. Inorganic phosphate, a product of TNAP activity, induces OPN expression in cultured osteoblastic cells.[29] Mutant mice lacking OPN (Spp1−/−) do not show an obvious bone phenotype,[30] but they are resistant to ovariectomy-induced osteoporosis.[31] We have previously shown that high plasma OPN levels accompany the increased extracellular PPi levels in Alpl−/− mice, and that [Alpl−/−; Spp1−/−] double-knockout mice have a partial improvement of the hypomineralization found in Alpl−/− mice. This indicates that increased OPN contributes to the impaired bone mineralization of Alpl−/− mice,[17] although a mechanistic explanation for this effect is still lacking. Here, we have used genetic means to demonstrate that TNAP affects the phosphorylation status of OPN in vivo. In addition we show that in vivo overexpression of TNAP leads to increased bone mass.

Subjects and Methods

Transgenic mice

We established a transgenic (Tg) mouse line, Col1a1-Tnap, expressing the human TNAP cDNA under control of the mouse Col1a1 promoter. The designation Tg (+/−) or (+/+) is used to denote hemizygosity and homozygosity for the transgene, respectively. Col1a1-Tnap+/+ mice were bred to Alpl−/− mice (Mouse Genome Informatics [MGI] strain ID: Alpltm1Jlm),[2] to generate [Col1a1-Tnap+/−; Alpl−/−] mice that express human TNAP under control of the osteoblast-specific Col1a1 promoter in an Alpl null background. Another Tg mouse line expressing human TNAP cDNA under control of a liver-specific apolipoprotein E promoter (ApoE-Tnap) was reported.[32] ApoE-Tnap mice were also crossed to Alpl−/− mice to produce [ApoE-Tnap+/−; Alpl−/−] mice. All animal studies were conducted with approval of the Animal Usage Committee of the Sanford Burnham Medical Research Institute, La Jolla, CA, USA.

Collection of tissue samples, plasma analysis, and histological studies

Mice were anesthetized by intraperitoneal injection of Avertin and blood was collected by cardiac puncture. Plasma levels of AP activity were measured using a previously reported method.[11] Tissue samples for histological analysis were fixed in 4% paraformaldehyde/PBS solution and processed as described.[2] Bone tissues from adult mice were decalcified with 0.125 M EDTA/10% formalin in H2O (pH 7.2) for 5 days after fixation, and processed for paraffin sectioning. Immunostaining was performed using a standard avidin-biotin complex protocol. Undecalcified bone sections were prepared using Acrylosin SOFT Infiltration and Embedding (DHM, Villa Park, IL, USA). PPi levels were measured as we reported,[8] and mouse OPN in plasma was measured with ELISA (Enzo, Plymouth Meeting, PA, USA), following the manufacturer's protocol.

Bone histomorphometric analysis

Bone samples were fixed in 4% paraformaldehyde/PBS and washed in 10%, 15%, and 20% sucrose/PBS prior to cryoembedding in hexane dry-ice bath. Undecalcified sections were prepared following Kawamoto's film method.[33] To compare vertebrae bones from 4-month-old wild-type (WT), Col1a1-Tnap+/− and Col1a1-Tnap+/+ male mice were analyzed. To examine compensation by the transgene expression in TNAP null mice, tibia samples from WT, Col1a1-Tnap+/−, Alpl−/−, and [Col1a1-Tnap+/−; Alpl−/−] mice were analyzed. Postnatal mice at day 16 were used because this stage is the survival limit for most Alpl−/− mice. Von Kossa/Van Gieson–stained sections were scanned by ScanScopeXT system (Aperio, Vista, CA, USA), and images were analyzed by using the Bioquant Osteo software (Bioquant Osteoanalysis Co., Nashville, TN, USA).

Alizarin red (Sigma, St. Louis, MO, USA) and calcein (TCI, Toshima, Tokyo, Japan) were administered to postnatal mice with an interval of 48 hours by subcutaneous injection. Alizarin red was injected at day 7 (20 mg/kg body weight) and calcein was injected at day 9 (20 mg/kg body weight) to WT, Col1a1-Tnap+/−, Alpl−/−, and [Col1a1-Tnap+/−; Alpl−/−] mice, and the injected mice were collected at day 11. Coronal sections of undecalcified calvarial bones were prepared by Kawamoto's film method as same as the bone/osteoid analysis. Pictures of fluorescent images were taken with a microscope TE300 (Nikon Instruments Inc., Melville, NY, USA), and distance between Alizarin red and calcein were measured by SPOT software (Diagnostic Instruments Inc., Sterling Heights, MI, USA). At least 10 independent positions were measured to obtain an average value per each animal.

Western blots

After careful removal of skin, tendons, and muscle, leg bones were snap frozen with dry ice. All bone tissues were from 4-month-old mice except Alpl−/− samples, because these mice die at around postnatal day 16–20. Frozen bones were crushed into powder and suspended in lysis solution containing 4 M guanidine HCl, 50 mM Tris HCl, 0.5 M EDTA, 2 mM phenylmethylsulfonyl fluoride (PMSF), 5 mg/L pepstatin, and 1 mg/L soybean trypsin inhibitor (pH 7.5).[34] This guanidine-EDTA solution was also used to extract protein from mineralized osteoblast cultures. After rotation at 4°C for 48 hours, the solubilized bone samples were dialyzed against Tris-buffered saline (TBS) solution containing 5 mM EDTA and 2 mM PMSF (pH 7.5) at 4°C to remove guanidine salt. Protein concentration was determined using a bicinchoninic acid assay kit according to manufacturer's instructions (Thermo Fisher Scientific Inc., Rockford, IL, USA). SDS-PAGE was performed using a standard protocol, except that 1.7% SDS was used in the running buffer for bone samples to minimize the background, which is likely caused by OPN because of its absence in the samples from Spp1−/− mice. Rat anti-human TNAP monoclonal antibody (R&D Systems, Minneapolis, MN, USA) and goat anti-mouse OPN antibody (Abcam, Cambridge, MA, USA) were used for detection of TNAP and OPN, respectively. Detection was followed by standard procedure using enhanced chemiluminescence (ECL Plus; GE Healthcare, Pittsburgh, PA, USA).

Phosphorylation analysis

Protein samples extracted with the guanidine/EDTA as described earlier in the method section of Western blots were incubated with mouse anti-OPN antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 4°C overnight, and equally divided into two tubes. Protein A Sepharose (Amersham Pharmacia, Uppsala, Sweden) was added to each tube and incubated at 4°C overnight. Washed beads were resuspended in 50 µL of the same buffer. Two microliters (2 mL) of 10 U/µL calf intestinal alkaline phosphatase (IAP) (New England Biolabs, Ipswich, MA, USA) were added into the first tube, whereas 2 µL of 10 mM Tris HCl (pH 8.2) was added to the second tube. After a 2-hour incubation under rotation at 37°C, reactions were stopped with sample buffer and heated at 95°C for 5 minutes prior to electrophoresis. Blotted membranes were stained with goat anti-OPN antibody (Abcam).

Samples extracted with guanidine/EDTA from osteoblast culture and bones were dialyzed against TBS containing 2 mM PMSF to remove EDTA and loaded onto 10% acrylamide gels containing 30 µM Phos-Tag [ie, 1,3-bis[bis(pyridin-2-ylmethyl)amino]propan-2-olato dizinc(II)], a compound that binds to phosphorylated ions of phosphoproteins and lowers the migration speed (Wako Pure Chemical Industries, Ltd, Osaka, Japan).[35] Migration pattern of OPN was compared by Western blots with goat anti-OPN antibody (Abcam).

For proteomics analysis, guanidine/EDTA extracts of leg bones and mineralized osteoblast culture were incubated with a combination of mouse and goat anti-OPN antibodies, and precipitated with Protein G agarose beads (Thermo Scientific, Rockford, IL, USA). Samples eluted from the Protein G beads were subjected to TiO2-based enrichment procedure. OPN phosphopeptides were analyzed by liquid chromatography and tandem mass spectrometry (LC-MS/MS) on Michrom MS2 HPLC-captive spray-LTQ Orbitrap Velos with ETD (Thermo Scientific). Analytical preparations and data processing were conducted by the Proteomics core facility at the Sanford-Burnham Medical Research Institute.

Micro–computed tomography and X-ray analysis

Skull bones, vertebrae and femur from five Col1a1-Tnap+/+ and five wild-type (WT) control mice (4 months old, two males and three females) were dissected and fixed in 4% paraformaldehyde/PBS, then analyzed using micro–computed tomography (µCT) by Numira Biosciences (Salt Lake City, UT, USA) using a high-resolution, volumetric µCT scanner (µCT40; ScanCo Medical, Zurich, Switzerland). The image data was acquired with the following parameters: 10 µm (skull bones and vertebrae) and 6 µm (femur) isotropic voxel resolution at 300 ms exposure time, 2000 views and five frames per view. The µCT-generated Digital Imaging and Communications in Medicine (DICOM) files were used to analyze the samples and to create volume renderings. The raw data files were viewed using Microview (GE Healthcare, Milwaukee, WI, USA) and AltaViewer software (Numira Biosciences). SCIRun (Scientific Computing and Imaging Institute, University of Utah, Salt Lake City, UT, USA) was used for cutaway images. X-ray images were obtained with a Faxitron MX-20 (Faxitron X-ray Corporation, Chicago, IL, USA) using energy of 20 kV.


Correction of the HPP phenotype by tissue-directed expression of TNAP

Two transgenic models of TNAP overexpression were compared for their ability to prevent the HPP phenotype in Alpl−/− mice. We generated a Col1a1-Tnap mouse strain, expressing TNAP under control of the Col1a1 promoter (Fig. 1A, B). The Col1a1-Tnap mice appear healthy and both genders are fertile. Breeding of this mouse line to Alpl−/− HPP mice demonstrated that expression of the Col1a1-Tnap transgene rescued the lethal phenotype of Alpl−/− mice, and [Col1a1-Tnap+/−; Alpl−/−] animals live at least 20 months without evidence of epilepsy; however, their average body weight is reduced compared to WT littermate controls (Fig. 1C). Plasma levels of TNAP were 527 ± 282 µg/L in Col1a1-Tnap+/− or Col1a1-Tnap+/+, 35 ± 0.01 µg/L in WT, and 478 ± 148 µg/L in [Col1a1-Tnap+/−; Alpl−/−] mice (Fig. 1D). Expression of the Col1a1-Tnap transgene is restricted to osteoblasts of long bones and calvaria but not chondrocytes (Fig. 2I, J). Other organs that express the Col1a1-Tnap transgene include the adrenal glands, spinal cord and fore brain (Fig. 2J, L). Expression of Col1a1-Tnap transgene in bone was also confirmed at the RNA level by q-RTPCR (Supplemental Fig. S1).

Figure 1.

Construction and characterization of Col1a1-Tnap mice. (A) The 5.3-kb transgene fragment consists of a 2.3-kb Col1a1 promoter, an intron, a full TNAP cDNA (2.5 kb) and a polyA sequence isolated from the pStec1 backbone vector by Sac I digestion.[36] The 2.3-kb Col1a1 promoter was isolated from the pJ251 vector (a kind gift from Dr. Gerard Karsenty, Columbia University, New York, NY, USA) and the 2.5-kb fragment of human ALPL cDNA was derived from pSV2Aalp (ATCC 59634). The purified DNA construct was used for microinjection into the pronucleus of embryos from FVB/N mice in the transgenic facility at the Sanford-Burnham Medical Research Institute. (B) Integration of transgene was analyzed by Southern hybridization using standard protocols. Four micrograms (4 µg) of tail DNA was digested with Hind III and probed with a 675-bp fragment from intron 2 of Alpl gene together with a 498-bp Bgl I fragment from the human ALPL cDNA. The transgene was transmitted to the offspring in a Mendelian pattern. (C) Body weight at 4 months of age: female Col1a1-Tnap+/− or Col1a1-Tnap+/+ (ColTg) versus WT p = 0.5401, female WT versus [Col1a1-Tnap+/−; Alpl−/−] (ColTg; Alpl−/−), p = 0.0276. The body weight difference is more pronounced in female than male animals. (D) An average TNAP value from triplicate measurement for each animal is plotted (numbers of animals, Col1a1-Tnap+/−, Col1a1-Tnap+/+, WT and Col1a1-Tnap−/−, are n = 18, 7, 11, and 5 respectively). Plasma AP levels display a wide range but are consistently elevated in the transgenic mice. Col1a1-Tnap+/− or Col1a1-Tnap+/+: 504 ± 261 µg/L, WT: 25.8 ± 14.7 µg/L, [Col1a1-Tnap+/− or Col1a1-Tnap+/+; Alpl−/−]: 478 ± 221 µg/L. The difference between Col1a1-Tnap+/− (hemizygous) and Col1a1-Tnap+/+ (homozygous) was not significant (p = 0.3967), whereas hemizygous Tg display a significant increase in AP compared to WT (p < 0.0001).

Figure 2.

AP activity staining on tissues from 11-day-old mice. (AD) Wild type; (EH) [ApoE-Tnap+/−; Alpl−/−]; (IL) [Col1a1-Tnap+/−; Alpl−/−]; (MP) Alpl−/− mice. A, E, I, and M: fibula. B, E, H, and K: calvaria and forebrain. C, G, K, and O: liver. D, H, L, and P: kidney and adrenal gland. Organs for each genotype shown are derived from a single animal. Two to four mice per each genotype were analyzed and typical results are shown (×100).

The second transgenic line expressing TNAP under control of the ApoE promoter, ApoE-Tnap, overexpresses TNAP in long bone (Fig. 2E) and in the liver and kidney (Fig. 2G, H), while expression in calvarial bone was not as significant as Col1a1-Tnap Tg (Fig. 2F, J). It is noteworthy that TNAP is a well-known serum marker for liver dysfunction in humans,[37] while normal mouse liver expresses virtually no TNAP activity as shown in Figure 2C, except in the bile canaliculi in certain strains.[38, 39] The TNAP plasma levels in ApoE-Tnap+/− and ApoE-Tnap+/+ mice were 11,026 ± 2,094 µg/L and 23,102 ± 15,790 µg/L, respectively. Expression of the ApoE-Tnap transgene in the liver and bone was also confirmed by q-RT-PCR (Supplemental Fig. S2).

Col1a1-Tnap+/− and ApoE-Tnap+/− mice did not show marked reduction of PPi levels in the plasma compared to WT mice (Fig. 3A, C). Plasma PPi of [Col1a1-Tnap+/−; Alpl−/−] was slightly elevated, whereas levels from [ApoE-Tnap+/−; Alpl−/−] were normal, most likely due to the high TNAP expression in the liver and bones in ApoE-Tnap+/− mice (Fig. 2E, G).

Figure 3.

PPi and OPN measurement. (A, C) Plasma PPi; (B, D) plasma OPN; (A, B) 4-month-old Col1a1-Tnap+/−, WT, and [Col1a1-Tnap+/−; Alpl−/−] mice; (C) 4-month-old ApoE-Tnap+/−, WT, and [ApoE-Tnap+/−; Alpl−/−] mice; (D) 11-day-old Col1a1-Tnap+/−, WT, [Col1a1-Tnap+/−; Alpl−/−], and Alpl−/− mice; (E) correlation between body weight and plasma OPN in 11-day-old Alpl−/− mice. An average value from duplicated assay for each mouse was plotted in the figures. (AC: n = 5; D: n = 4 except that Alpl−/− was n = 18).

Overexpression of TNAP leads to increased bone mineralization

We evaluated the ability of Col1a1-Tnap calvarial osteoblasts to mineralize in vitro. We tested both low and high concentrations of βGP (2 mM and 10 mM) as the phosphate source because high βGP (10 mM) can cause nonphysiological mineral deposits.[40] At both concentrations, increased mineralization was observed in the osteoblasts overexpressing the Col1a1-Tnap transgene (Fig. 4) compared to control cells. [Col1a1-Tnap+/−; Alpl−/−] cells showed as much mineralization as heterozygous Alpl+/− osteoblasts, whereas Alpl−/− cells failed to mineralize, in agreement with our previous data.[15]

Figure 4.

Mineralization assay. Osteoblasts were isolated from calvarial bones of 2-day-old mice as described.[8] Culture media containing 100 µg/mL ascorbic acid and 2 or 10 mM βGP was renewed every second day and the cells were fixed in 100% ethanol 19 days later for AP staining and mineralization assay.[41] A Western blot to detect TNAP is shown in the top row. The sample from [Col1a1-Tnap+/−; Alpl−/−] (ColTg; Alpl−/−) represents human TNAP protein expressed from the transgene and the sample from Alpl+/− control shows endogenous mouse TNAP. Human and mouse TNAP share 91.5% homology in their peptide sequences, and the rat monoclonal antibody used in this study recognizes both human and mouse TNAP. Cells of each genotype were derived from a single newborn pup. The same experiment was repeated twice to test reproducibility.

Histomorphometric analysis of adult mice showed that vertebral bones from Col1a1-Tnap+/− and Col1a1-Tnap+/+ mice had higher values of bone volume fraction (BV/TV) compared to WT controls, as well as significantly reduced osteoid (Fig. 5A). We conducted TRAP staining on adult tibias and vertebrae bones. TRAP-positive area per tissue volume (%) values were 3.400 ± 1.049, 3.246 ± 0.2671, and 3.279 ± 0.3498 (%) for WT, ColTg(+), and ColTg(++) tibias, respectively, and 1.535 ± 0.6108, 1.566 ± 0.9479, and 1.456 ± 0.7771 (%) for WT, ColTg(+), and ColTg(++) L3 vertebrae, respectively. Alpl−/− mice do not exhibit changes in osteoclasts numbers as we previously reported,[15] and we did not observe decreased osteoclast activity in the Cola1-Tnap transgenic mice. In tibia samples from 16-day-old mice, increase of mineralization was not significant in Col1a1-Tnap+/− mice; however, poor mineralization in Alpl−/− was improved by expression of human TNAP gene because increased BV/TV and reduced osteoid in Col1a1-Tnap+/−; Alpl−/− were observed (Fig. 5B). The low mineral apposition rate in Alpl−/− was improved by the transgene expression as shown in Fig. 5C.

Figure 5.

Bone histomorphometric analysis. (A) Comparison of area of mineralized bone (BV/TV) and unmineralized osteoid (OV/BV) in L2 and L3 vertebrae bones from 4-month-old WT, Col1a1-Tnap+/−, and Col1a1-Tnap+/+ mice (n = 3). (B) Comparison of area of mineralized bone (BV/TV) and unmineralized osteoid (OV/BV) in tibia from 16-day-old WT, Col1a1-Tnap+/−, Alpl−/−, and [Col1a1-Tnap+/−; Alpl−/−] mice (n = 3). (C) mineral apposition rates in parietal bones from 11-day-old WT, Col1a1-Tnap+/−, Alpl−/−, and [Col1a1-Tnap+/−; Alpl−/−] mice (n = 3).

We also compared the degree of bone mineralization of Col1a1-Tnap and ApoE-Tnap mice by µCT analysis. Measurements of bone surface per volume and trabecular thickness indicated increased bone formation in the Col1a1-Tnap+/+ femur (Table 1, Supplemental Fig. S3). Furthermore, significant changes in the BV/TV, bone surface per volume, trabecular number, trabecular thickness, and trabecular separation were observed in the L2 vertebrae (Table 1). The average structure model index of L2 bone in Col1a1-Tnap+/+ and WT controls was 1.1824 and 1.9962, respectively. Similar results were obtained from analysis of femurs and L2 vertebrae of ApoE-Tnap mice (4 months old, two males and three females) (Supplemental Table S1). In comparison to Col1a1-Tnap, ApoE-Tnap mice showed a more significant increase in femur than vertebrae, most likely a result of the higher expression of TNAP in the trabecular osteoblasts.

Table 1. µCT Analysis of Bones From Col1a1-Tnap Tg Mice
  • BMD = bone mineral density; BV/TV = bone volume/total volume; BS/BV = bone surface/bone volume; Tb.N = trabecular number; Tb.Th = trabecular thickness; Tb.Sp = trabecular separation.
  • *p values were obtained by two-way ANOVA between WT and Col1a1-Tnap+/+.
Distal femurBMD905.3 ± 22.64931.3 ± 40.99p = 0.3737
 BV/TV0.05022 ± 0.019590.06546 ± 0.01097p = 0.2552
 BS/BV69.86 ± 4.88658.36 ± 10.01p = 0.0277
 Tb.N1.731 ± 0.61171.885 ± 0.2661p = 0.6493
 Tb.Th0.02876 ± 0.0021130.03504 ± 0.005694p = 0.0401
 Tb.Sp0.6111 ± 0.22350.5043 ± 0.07631p = 0.4079
L2 vertebraeBMD873.5 ± 26.1908.7 ± 17.15p = 0.1241
 BV/TV0.08462 ± 0.023150.1756 ± 0.03187p = 0.0048
 BS/BV63.94 ± 7.52543.57 ± 43.57p = 0.0020
 Tb.N2.649 ± 0.49833.789 ± 0.4503p = 0.0157
 Tb.Th0.03164 ± 0.0037110.04626 ± 0.004501p = 0.0013
 Tb.Sp0.3574 ± 0.079920.2206 ± 0.03194p = 0.0176

TNAP affects the phosphorylation status of OPN in vivo

Expression of the TNAP transgene is significantly lower than endogenous TNAP (Supplemental Fig. S4A and B), whereas OPN was highly expressed in both [Col1a1-Tnap+/−; Alpl−/−] and WT. The unchanged plasma PPi in adult [Col1a1-Tnap+/−; Alpl−/−] animals (Fig. 3A) may indicate that the level of TNAP expressed in these osteoblasts was not high enough to affect the levels of systemic PPi, and does not exclude the possibility of localized reduction of PPi in the mineralizing microenvironment. The corresponding results in PPi and OPN, unchanged in Col1a1-Tnap+/− and increased in [Col1a1-Tnap+/−; Alpl−/−] as shown in Fig. 3A, B, concur with the previously reported strict correlation between plasma PPi and OPN concentrations.[17] OPN is a heavily phosphorylated extracellular matrix protein, and phosphorylated OPN inhibits mineralization.[24-27] We previously reported that OPN is upregulated in Alpl−/− animals, although the phosphorylation status of the increased OPN in these mice was not determined.[14, 17] According to the algorithm NetPro 2, mouse OPN is predicted to have 43 phosphorylation sites. Because the molecular weight of a phosphate group is 94.97, the total molecular weight of the 43 phosphate residues would be approximately 4 kDa. Thus, we expected that fully phosphorylated OPN would display a 4-kDa difference in SDS-PAGE compared to the fully dephosphorylated OPN, although we also anticipated that OPN signal would appear as a broad band as a result of the high degree of posttranslational modifications. First, we tested if OPN from Alpl−/− osteoblasts exhibits larger molecular size than that from control cells (Fig. 6A). OPN from Alpl−/− cells was detected in the larger molecular mass range (highly phosphorylated) whereas OPN from Alpl+/− display a smaller molecular weight range, although all the bands appear quite broad. OPN from [Col1a1-Tnap+/−; Alpl−/−] cells appeared in the similar range as Alpl+/−, and [Col1a1-Tnap+/−; Alpl+/−] cells showed an increased amount of OPN in the smaller range than Alpl+/−. This result suggests that endogenous mouse TNAP dephosphorylates OPN in osteoblasts and that overexpressed human TNAP dephosphorylates OPN, compensating for the lack of endogenous TNAP in [Col1a1-Tnap+/−; Alpl−/−] cells. In bone extracts, the sizes of OPN were less changed, possibly because of coexisting OPN originating from chondrocytes without transgenic expression of TNAP. However, dephosphorylated OPN is visible as a separate band smaller than the 52-kDa marker. [Col1a1-Tnap+/−; Alpl−/−] bones showed this band as well as WT and Col1a1-Tnap samples, whereas it was not observed in the Alpl−/− bones, indicating that OPN is less dephosphorylated in the Alpl−/− samples (Fig. 6B).

Figure 6.

Western blotting. Thirty nanograms (30 ng) of recombinant mouse OPN (Sigma, St. Louis, MO, USA) was digested with 10 U of IAP in AP reaction buffer for 2 hours and used as a control of dephosphorylated OPN (r.mOPN + IAP). (A) Cultured osteoblasts under mineralizing condition. (B) Bone tissue. The same membranes were reprobed for β-actin. (C) Immunoprecipitation experiment. Untreated and IAP-treated samples were compared by Western blot with goat anti-OPN antibody (C, left and right). (D) Comparison of OPN migration in Phos-Tag SDS-PAGE. Left: osteoblast extracts. Right: bone extracts from 11-day-old and 4-month-old. Western blots with anti-OPN antibody are shown along with Ponceau S staining of the same membrane to indicate the total protein blotted. Samples in each lane were derived from a single mouse and these experiments were repeated at least 2 times with samples derived from different animals.

OPN protein precipitated with a mouse monoclonal antibody was detected as a broad band ranging from 52 to 65 kDa; however, OPN precipitated from the Alpl−/− sample is devoid of smaller size forms (Fig. 6C, left). After dephosphorylation by commercial bovine IAP, all signals shifted to the smaller range closer to the 52-kDa marker (Fig. 6C, right). IAP treatment of Alpl−/−-derived OPN reduced the size of the protein, making it similar to the other samples. The pattern of OPN signal from [Col1a1-Tnap+/−; Alpl−/−] was not distinguishable from either WT or Col1a1-Tnap+/−, indicating that human TNAP is dephosphorylating OPN in vivo.

In SDS-PAGE containing Phos-Tag compound, migration speed decreases as the degree of phosphorylation of the protein increases. OPN from Alpl−/− osteoblasts was not detected, suggesting that this OPN form was highly phosphorylated and did not migrate after binding to Phos-Tag, whereas OPN from cells expressing Col1a1-Tnap contained OPN forms migrating faster than those in non-transgenic control cells because of increased dephosphorylation (Fig. 6D, left). Phos-Tag analysis of bones from 11-day-old mice showed only background signal in Alpl−/− whereas a detectable band is seen in [Col1a1-Tnap+/−; Alpl−/−] bones. In adult bones, OPN in [Col1a1-Tnap+/−; Alpl−/−] showed signals relatively similar to the WT (Fig. 6D, right).

OPN phosphopeptides were isolated from bones and cultured cells from each genotype; ie, Col1a1-Tnap+/−, WT, [Col1a1-Tnap+/−; Alpl−/−], and Alpl−/−, and analyzed by TiO2-LC-MS/MS (Table 2). Most phosphorylation sites are nonlocalized except the one that comprises Ser287. The Alpl−/− bones and osteoblasts contained increased numbers of OPN phosphopeptides. [Col1a1-Tnap+/−; Alpl−/−] osteoblasts showed changes in that pattern, whereas bone of [Col1a1-Tnap+/−; Alpl−/−] did not show reduction of phosphopeptides, most likely due to the large amount of the chondrocytes expressing OPN but no TNAP. Two OPN peptides, p(S174FQVS178DEQY182PDAT186DEDLT191)SHMK and FRIp(S299HELES304S305S306S307)EVN, with one nonlocalized site each, likely contain preferred sites for TNAP-mediated dephosphorylation.

Table 2. Phosphopeptide Sequences Obtained From TiO2-LC-MS/MS Analysis
Phosphopeptide sequenceBoneOsteoblasts
  1. All phosphorylation sites are nonlocalized except the one containing Ser 287. Residue numbers were based on the OPN isoform NP_001191131. The bottom row contains the total number of the spectral counts of individual phosphopeptides in each genotype.
  2. LC = liquid chromatography; MS/MS = tandem mass spectroscopy.

Bone phenotype of [Col1a1-Tnap+/−; Alpl−/−]

Postnatal 11-day-old [Col1a1-Tnap+/−; Alpl−/−] mice did not exhibit any apparent bone phenotype. However, in the 4-month-old adult animals, X-ray analysis revealed a widened tibial epiphysis and shorter tibial length (Fig. 7A). Measurements of the greatest length between the medial malleolus and medial condyle in WT and [Col1a1-Tnap+/−; Alpl−/−] were 18.75 ± 0.24 mm and 17.25 ± 0.69 mm, respectively (n = 5, two females and three males, p = 0.0037). Impaired mineralization was observed in the epiphysis of [Col1a1-Tnap+/−; Alpl−/−] mice (Fig. 7B, E, I, J). Immunohistochemistry indicated that expression of the human TNAP in adult [Col1a1-Tnap+/−; Alpl−/−] mice was significantly low in the endochondral ossification sites (Fig. 7H, K, Supplemental Fig. S4). These observations indicate that human TNAP compensated for lack of endogenous mouse TNAP in the cortical bone, but the lack of TNAP in chondrocytes affected normal development during endochondral ossification despite the high circulating levels of TNAP.

Figure 7.

Bone phenotype of [Col1a1-Tnap+/−; Alpl−/−]. (A) X-ray image of hind limbs: [Col1a1-Tnap+/−; Alpl−/−] (left) and WT (right). (B) Higher magnification of X-ray images of knee joints: [Col1a1-Tnap+/−; Alpl+/+] (top), WT (middle), and [Col1a1-Tnap+/−; Alpl−/−] (bottom). (CE) von Kossa staining on undecalcified femur sections from [Col1a1-Tnap+/−; Alpl+/+], WT, and [Col1a1-Tnap+/−; Alpl−/−] mice, respectively. (FH) WT control. (IK) [Col1a1-Tnap+/−; Alpl−/−]. (F, I) Distal femur (×40). (G, H, J, K) Proximal tibia (×400). (F, G, I, J) Goldner's trichrome staining. Immunohistochemistry showing expression of endogenous mouse TNAP in WT control (H) and human TNAP in [Col1a1-Tnap+/−; Alpl−/−] (K). GH and JK are serial sections, respectively. (I) The epiphyseal plate is almost missing from some areas of the tibia and femur of [Col1a1-Tnap+/−; Alpl−/−] mice. (J) The proliferating chondrocyte zone has an abnormal structure (arrows) and formation of trabecular bones is also reduced in the adult [Col1a1-Tnap+/−; Alpl−/−] mice. Endogenous TNAP is highly expressed near the hypertrophic zone (H), whereas the expression of transgenic TNAP is limited (K).


We reported the association between OPN expression and the pathophysiology of murine HPP in 2006,[17] but a mechanistic interpretation of this association is still lacking. Our current data provide compelling evidence for the role of TNAP in determining the phosphorylation status of OPN, which in turn regulates its mineral binding properties. Our earlier work had shown that OPN expression is upregulated in the plasma and the skeleton of Alpl−/− mice,[15] in agreement with the observation that OPN expression is increased in cultured fibroblasts from HPP patients.[42] Spp1−/− mice show increased bone mineral density (BMD), despite the elevated extracellular PPi that results from downregulated expression of Alpl and upregulated expression of Enpp1 and Ank.[17] The HPP phenotype in Alpl−/− mice was partially corrected in [Alpl−/−; Spp1−/−] double-knockout mice, even though their elevated PPi concentrations were unresolved, suggesting that increased OPN also contributes to the Alpl−/− bone phenotype independently of the mineralization inhibitory action of PPi.[17] Several studies have shown that phosphorylated OPN exhibits an inhibitory effect on mineralization.[24-26] Our current data show that OPN in the long bones of Alpl−/− mice is highly phosphorylated and that the phosphorylation status of OPN is decreased by overexpression of human TNAP in the Alpl−/− background. This is clear evidence that OPN represents a natural substrate for TNAP and extends the role of TNAP in skeletal mineralization to include not only Pi generation from ATP[43, 45] and PPi hydrolysis,[8, 9, 32] but also modulation of OPN function, an important matricellular protein, by dephosphorylation.

Previous reports have shown that PPi induces OPN expression in osteoblast cultures, indicating that increased local PPi causes upregulation of OPN in Alpl−/− bone.[9] Our observations that both plasma PPi and plasma OPN were unchanged in Col1a1-Tnap mice but elevated in [Col1a1-Tnap+/−; Alpl−/−] (Fig. 3A, B, D) is in agreement with those earlier findings and support the strict correlation between plasma PPi and OPN concentrations.[17] In addition, Beck and colleagues[29] reported that free phosphate produced by TNAP, using βGP as substrate, induced OPN expression in culture. OPN expression in [Col1a1-Tnap+/−; Alpl+/−] was markedly higher than Alpl+/− in an osteoblast culture containing excess βGP, in agreement with that earlier report.[29] Thus, OPN expression is under control of the local Pi/PPi ratio, which is known to influence skeletal mineralization in a skeletal site-specific manner.[5, 32]

An interesting sideline observation made during our work is that whereas most Alpl−/− mice exhibit normal or only slightly smaller weight at age 11 days, we have occasionally observed some exceptionally small Alpl−/− animals with poor response to vitamin B6 administration.[3] As shown here, those Alpl−/− mice with small body sizes tend to have highly elevated plasma OPN (Fig. 3E). These data indicate that Spp1 may act as a modifier gene for HPP, a disease that is notorious for the variable penetrance and severity of presentation.[44] Thus SPP1 joins ENPP1 and PHOSPHO1 as genes with the potential to modify the severity of presentation of HPP.[45]

Another observation worth noting is that the [Col1a1-Tnap+/−; Alpl−/−] mice were smaller and their average body weight was lower than WT controls. This can be explained by their shortened long bones. Histological analysis revealed abnormal growth plate morphology in the [Col1a1-Tnap+/−; Alpl−/−] mice (7F, I), indicating that despite TNAP being expressed in osteoblasts and plasma TNAP levels being elevated in these mice, the growth plates were still affected by the lack of TNAP. This observation is important because it reinforces the rationale for using mineral-targeting TNAP for enzyme replacement therapy of HPP[11-14] rather than soluble recombinant TNAP, which had proven ineffective in the past.[46-49] Even though the transgenic expression of TNAP in osteoblasts rescues the lethal phenotype of Alpl deficiency, residual disease can be demonstrated in these animals. This is in sharp contrast to the robustness of the treatment with mineral-targeting TNAP, which reaches notoriously avascular sites such as the enamel organ.[14] These data provide additional evidence to indicate that indeed TNAP must be able to act locally at sites of initiation of mineralization, and that increasing the levels of circulating TNAP are not sufficient for a complete therapeutic response.


All authors state that they have no conflicts of interest.


This work was supported by NIH grants DE12889 and AR047908. We thank the Sanford-Burnham Medical Research Institute Animal Facility for careful husbandry of mice, Dr. Ling Wang for microinjections, Mr. John Shelley in the Histology Core for preparation and staining of undecalcified bone tissue, and Dr. Laurence M. Brill in the Proteomics Core for the mass spectrometry analysis of phosphopeptides. We also thank Ms. Brittney Russell for her help for genotyping of the mice and Dr. Campbell Sheen for valuable suggestions and help during manuscript preparation.

Authors' roles: SN and MCY performed the experiments. SN, MCY, and JLM analyzed and interpreted the data and wrote the manuscript.