Skeletal mineralization deficits and impaired biogenesis and function of chondrocyte-derived matrix vesicles in Phospho1^−/− and Phospho1/Pit1 double knockout mice

Abstract

We have previously shown that ablation of either the Phospho1 or Alpl gene, encoding PHOSPHO1 and tissue-nonspecific alkaline phosphatase (TNAP) respectively, lead to hyperosteoidosis but that their chondrocyte- and osteoblast-derived matrix vesicles (MVs) are able to initiate mineralization. In contrast, the double ablation of Phospho1 and Alpl completely abolish initiation and progression of skeletal mineralization. We argued that MVs initiate mineralization by a dual mechanism: PHOSPHO1-mediated intravesicular generation of P_i and phosphate transporter-mediated influx of P_i. To test this hypothesis, we generated mice with col2a1-driven cre-mediated ablation of Slc20a1, hereafter referred to as Pit1, alone or in combination with a Phospho1 gene deletion. P_it1^col2/col2 mice did not show any major phenotypic abnormalities, while severe skeletal deformities were observed in the [Phospho1^−/−; P_it1^col2/col2] double knockout mice that were more pronounced than those observed in the Phospho1^−/− mice. Histological analysis of [Phospho1^−/−; P_it1^col2/col2] bones showed growth plate abnormalities with a shorter hypertrophic chondrocyte zone and extensive hyperosteoidosis. The [Phospho1^−/−; P_it1^col2/col2] skeleton displayed significantly decreases in BV/TV%, trabecular number and bone mineral density, as well as decreased stiffness, decreased strength, and increased post-yield deflection compared to Phospho1^−/− mice. Using atomic force microscopy we found that ~80% of [Phospho1^−/−; P_it1^col2/col2] MVs were devoid of mineral in comparison to ~50 % for the Phospho1^−/− MVs and ~25% for the WT and P_it1^col2/col2 MVs. We also found a significant decrease in the number of MVs produced by both Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] chondrocytes. These data support the involvement of P_iT-1 in the initiation of skeletal mineralization and provide compelling evidence that PHOSPHO1 function is involved in MV biogenesis.

INTRODUCTION

Mineralization of cartilage and bone occurs by a series of physicochemical and biochemical processes that together facilitate the deposition of hydroxyapatite in specific areas of the extracellular matrix (ECM). Tissue-nonspecific alkaline phosphatase (TNAP) plays a crucial role in restricting the concentration of the mineralization inhibitor inorganic pyrophosphate (PP_i) to maintain a P_i/PP_i ratio permissive for normal propagation of mineral in the extracellular matrix.^(1–5) However, chondrocyte and osteoblast-derived matrix vesicles (MVs) from Alpl^−/− mice are still able to initiate mineralization,^{(6, 7)} indicating that other enzymes or mechanisms are involved in the intravesicular initiation of mineralization. We showed that PHOSPHO1, an enzyme that uses phosphocholine and phosphoethanolamine as substrates, is required for MV-mediated initiation of mineralization⁽⁸⁾ as the lack of PHOSPHO1 (Phospho1^−/− mice) leads to skeletal and dental hypomineralization⁽⁹⁾ and decrease in the mechanical properties of bones⁽¹⁰⁾ but not to elimination of intravesicular mineral formation. Importantly however, the [Alpl^−/−; Phospho1^−/−] double null mice are embryonic lethal and the E16.5 embryos show complete absence of skeletal mineralization and MVs devoid of mineral.^{(9, 11)} We hypothesized that MV-mediated initiation of mineralization results from a dual mechanism, i.e. PHOSPHO1-mediated intra-vesicular production and transporter-mediated influx of P_i.⁽¹²⁾ Two related type III Na/P_i co-transporters, P_iT-1/Glvr1 and P_iT-2/Ram, are expressed by chondrocytes and osteoblasts, but P_iT-1 is the most abundant in these cell types. ^(13−15) To test the hypothesis that P_i-influx into MVs contributes to the initiation of mineralization, we generated mice with a conditional ablation of the P_iT-1 gene (Slc20a1here referred to as Pit1) alone or in the Phospho1^−/− background and we used atomic force microscopy (AFM) to quantify the number of MVs produced by Phospho1^−/− and Phospho1/Pit1 doubly deficient primary chondrocytes and Raman spectroscopy to assess the presence or absence of mineral in the MVs. Our data support the involvement of P_iT-1 function in the initiation of endochondral ossification and also points to PHOSPHO1 as an enzyme involved in MV biogenesis.

MATERIALS AND METHODS

Mice

Phospho1-R74X mutant mice (Phospho1^m1Jlmhere referred to as Phospho1^−/−) were previously described.⁽⁹⁾ The generation and characterization of the P_it1^flox/flox mice (Slc20a1^tm1.1bek) was reported earlier.⁽¹⁶⁾ To generate mice lacking both PHOSPHO1 and P_iT-1, Phospho1^−/− and P_it1^flox/flox mice were crossed and double heterozygote mice were bred to generate [Phospho1^−/−; P_it1^flox/flox] double mutant mice. These mice were then bred with Col2a1-cre mice to generate [P_it1^flox/flox; Col2a1-cre], here named P_it1^col2/col2 and [Phospho1^−/−; P_it1^col2/col2] mice. Phospho1^−/− genotypes were determined using genomic DNA, PCR and restriction digestion by BsrD1 restriction enzyme.⁽⁹⁾ P_it1^col2/col2 and [Phospho1^−/−; P_it1^col2/col2] mice were genotyped by PCR. The primer sequences were: Phospho1: F 5’ TCCTCCTCACCTTCGACTTC-3’, R 5’-ATGCGGCGGAATAAACTGT-3’, P_it1^flox/flox: F 5’ AAGGCATTTGTCAGCCCAGTC-3’, R 5’ ATCGATCCACTCAGTCTAGTGC-3’ and Col2a1-cre: F 5’-GCCTGCATTACCGGTCGATGCAACGA-3’ and R 5’-GTGGCAGATGGCGCGGCAACACCATT-3’. All mice were housed at the Sanford Burnham Prebys Medical Discovery Institute’s vivarium following the guide standards (e.g. contact bedding, temperature and humidity control, etc.). All experiments reported in this manuscript were approved by the Institute under Animal Usage Form #13-058. Animal care technicians conducted routine husbandry procedures (e.g., cage cleaning, feeding and watering) and daily health checks to assess the animals’ condition. To collect blood, mice were anesthetized with Avertin (0.017 ml/g body weight) and after confirmation of deep sleep, blood was obtained by cardiac puncture. The Institute’s Animal Care & Use Program is accredited by AAALAC International and a Multiple Project Assurance A3053-1 is on file in the OLAW, DHHS.

Histological and radiographic studies

Tissues were collected at either 1-month or 4-months of age as indicated and perfused with 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) and fixed for 7 days and decalcified in 10% formalin-0.125 M EDTA (pH 7.3) for 10 days prior to regular processing for paraffin sectioning. Deparaffinized sections were treated with 20 µg/ml Proteinase K in 50 mM Tris HCl-1mM EDTA-0.5% Triton X100 (pH 8.0) for 10 min at 37°C and 10 min at room temperature for unmasking. Immunostaining was processed by the ABC method (Vector Labs, Burlingame, CA) with slight modifications. SuperBlock in PBS (Life technologies, Waltham, MA) was used for blocking of the primary antibody, which was raised in rabbits against amino acids 251–380 of human P_iT-1 (Santa Cruz Biotechnology, Dallas, TX). In order to reduce non-specific staining, 30 min incubation with 2% normal goat serum-1% BSA-PBS was included prior to the secondary antibody as well as the ABC reagent.

Whole-body, long bones and spine radiographic images were taken using an MX20 Specimen Radiograph System (Faxitron X-ray Corporation, Chicago, IL). Paraffin or plastic sections were stained with Von Kossa/Van Gieson stain as before.^{(17, 18)} Von Kossa/Van Gieson–stained slides were used for quantification of osteoid volume using the Bioquant Osteo Software (Bioquant Osteoanalysis Co., Nashville, TN).

P_it1 gene expression

Primary chondrocytes from WT and P_it1^col2/col2 mice were isolated from the knee joint growth plates of 5 day-old pups by collagenase digestion, as described previously.⁽⁹⁾ RNA was extracted using RNAeasy Pus Kit (Qiagen, Valencia, CA). The P_it1 primers and probe were designed using the mouse genome sequences available in the UCSC genome browser and the Genescript online primer/probe design tool (https://www.genescript.com/ssi-bin/app/primer). A Primer BLAST search was also performed for the primers and probe sequences to check for any cross-reactivity for any other gene. These primers/probes were designed at exon/exon junctions so that they amplify only the cDNA and not the genomic DNA. P_it1 primers and probe sequences (Operon (San Diego, CA) are as follows: F-5’GGCTCAGGTGTAGTGACCCT3’, R-5’ CACATCTATCAAGCCGTTCC3’ and FAM-TAMRA Probe-5’CGAAACTGTGGGCTCCGCC3’. Approximately 500 ng-1 µg of RNA were used for reverse transcription. For a 20 µl reaction, reverse transcription was carried out using the superscript kit (Life technologies, San Diego, CA) for 1 h at 42 °C followed by 70 °C for 5 min to inactivate the RT reagents. RNase H (New England Biolabs, Beverly, MA) treatment was given at 37 °C for 20 min. The RT product was then diluted with an equal volume of RNAse and DNAse free water. Specific RNA transcript (mRNA) for Pit1 was quantified by real-time PCR using dual-labeled hydrolysis probes (FAM-TAMRA). Two µl of the (1:10) dilution of the cDNA was then used for qPCR. The reaction utilized 12.5 µl of platinum qPCR UDG Supermix (Life technologies, San Diego, CA) yielding 0.75 U Taq DNA polymerase, 20 mM Tris-HCl, 50 mM KCl, 3 mM MgCl₂and 200 µM of deoxynucleoside triphosphate. Real time reaction was performed in 96-well plates on a Stratagene MX3000p real time machine (Stratagene, La Jolla, CA). The primer and probe combination that gave the lowest Ct and best amplification plots was used for the final analysis. The reaction was run at an initial temperature of 95 °C for 10 min and then at 95 °C for 30 sec, 55 °C for 1 min followed by 72 °C for 30 sec for 45 cycles. The optical signal was recorded at the end of every 72° C extension step. Ct values were determined by the software according to the optimization of the baseline. For computing the relative amounts of the target genes, the average Ct of the primary signal for 18S was subtracted from that of the target gene to give changes in Cts (dCt, a log2 scale). Amplification of the target gene was normalized to that of 18S RNA. The primers and probe sequences for mouse 18S were: F (CGGCTACCACATCCAAGGAA, 0.6 µM), R (GCTGGAATTACCGCGGCT, 0.6 µM), probe (TGCTGGCACCAGACTTGCCCTC, 0.2 µM).

Biochemical assays

Blood was collected by cardiac puncture into lithium heparin tubes and plasma was collected by centrifugation at 5000 rpm for 10min. Total alkaline phosphatase activity in plasma was measured using a previously reported method.⁽¹⁷⁾ PP_i levels were measured using activated charcoal and ³H method as we previously reported.^{(9, 19)}

Micro–computed tomography (µCT) and histomorphometry

Sixteen mice (4 per group) were euthanized at 1 month of age, the tibias and femurs dissected and fixed in 4% paraformaldehyde. Samples were imaged on a Skyscan 1076 µCT scanner (Kontich, Belgium). Samples were wrapped in tissue paper that was moistened with PBS, and scanned at 9 µm voxel size, applying an electrical potential of 50 kV and current of 200 µA, using a 0.5 mm aluminum filter. Mineral density was determined by calibration of images against 2 mm diameter hydroxyapatite (HA) rods (0.25 and 0.75 gHA/cm³). Additionally, a beam hardening correction algorithm was applied prior to image reconstruction.

We used the software Dataviewer, CTAn and CTVox (Skyscan) to visualize and determine bone histomorphometric parameters. Cortical bone analysis was performed on the midshaft of the femurs and tibias. The volumes of interest were selected in reference to an identified landmark.⁽²⁰⁾ Since all animals were 1 month of age, the volumes of interest were (1) 3600–4500 µm proximal to the distal femur growth plate and (2) 3600–4500 µm distal to the tibia proximal growth plate. The cortical bone in this region was selected by automatic contouring of the periosteal tissue excluding the marrow cavity. A global threshold was used to identify cortical bone and an erosion of one pixel was performed to eliminate partial volume effects. From these regions of femoral and tibial cortical bone, the following parameters were determined: cross-sectional tissue area (T.Ar), cross-sectional cortical bone area (B.Ar), cortical bone area fraction (B.Ar/T.Ar), cross-sectional bone thickness (Cs.Th) and tissue mineral density (TMD).

Trabecular bone analysis was performed at the distal femoral metaphysis and proximal tibial metaphysis. The regions of interest were (1) 360–2160 µm proximal to the distal femoral growth plate, and (2) 360–2160 µm distal to the proximal tibial growth plate. The trabecular region was selected by automatic contouring. An adaptive threshold (using the mean maximum and minimum pixel intensity values of the surrounding ten pixels) was used to identify trabecular bone and an erosion of one pixel was performed to eliminate partial volume effects. From these regions of femoral and tibial trabecular bone the following parameters were determined: tissue volume (TV), trabecular bone volume (BV), trabecular bone volume fraction (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), trabecular number (Tb.N), structure model index (SMI), trabecular pattern factor (Tb.Pf), and bone mineral density (BMD).

3-point bending for the determination of bone stiffness and breaking strength

An Instron 3342 materials’ testing machine (Instron, Norwood, MA, USA) fitted with a 2 kN load cell was used to determine bone stiffness and breaking strength.⁽²¹⁾ The span was fixed at 5.12 mm for femurs. The cross-head was lowered at 1 mm/min and data were recorded after every 0.2 N change in load and every 0.1 mm change in deflection. Each bone was tested to fracture. Failure and fracture points were identified from the load-extension curve as the point of maximum load and where the load rapidly decreased to zero, respectively. The maximum stiffness was defined as the maximum gradient of the rising portion of this curve, and the yield point was defined as the point at which the gradient was reduced to 95% of this value. Both values were calculated from a polynomial curve fitted to the rising region of the load-extension curve in Mathcad (Mathsoft Engineering and Education Inc., Cambridge, MA).⁽²¹⁾

Atomic force microscopy (AFM)

MVs were isolated from differentiating chondrocytes by differential centrifugation as described earlier.⁽⁹⁾ A drop (5 µL) of each MV solution in Tris-buffered-saline was spotted on freshly cleaved mica substrates (Ted Pella, Redding, CA) and allowed to stand for 5 min. Next, 5 µL of glutaraldehyde solution (8% in H₂O, Sigma-Aldrich, St. Louis, MO) was dropped onto the samples. The substrates were stored inside a desiccators at room temperature for 24 h. AFM images of dried samples were recorded in air by means of an 5500 atomic force microscope (Agilent Technologies, Santa Clara, CA) equipped with an open-loop probe working in non-contact (AAC) mode. Silicon-nitride cantilevers having a nominal resonance frequency of ~190 kHz (NanosensorsTM, Neuchatel, Switzerland) were used. Tridimensional AFM images were generated by PicoView software (Agilent Technologies). AFM images were used to gather information about the morphology, height, volume and number of MVs in each sample. AFM phase images were also recorded on samples prepared without the use of glutaraldehyde.

Raman spectroscopy

MVs, suspended in Tris buffered saline, were drop-casted onto a glass slide and scanned by means of an alpha300 Raman spectrometer system (WITec GmbH, Ulm, Germany) with a laser of 532 nm wavelength and integration time of 1s. Scans were mostly performed in the central region of the drop because AFM imaging on samples prepared using a similar procedure showed a very high density of MVs and low concentration of salt crystals, which may produce background fluorescence and cover the signal from mineral aggregates inside the MVs, in that region. Six random spectra of 10 accumulations from the WT sample and 5 random spectra of 10 accumulations for the Phospho1^−/− MV samples were collected at 60× magnification. Raman spectra were averaged for both WT and Phospho1^−/− MV samples.

Statistical analysis

All measurements were performed at least in triplicate. Results are expressed as mean ± SEM. The data were analyzed using Student’s t test. For microCT analysis, results are expressed as mean ± SD, statistical differences between experimental groups were analyzed using Kruskal-Wallis (K-W) test by SPSS Statistics, P values less than 0.05 were considered significant. For AFM, statistical differences between samples were calculated by non-parametric Mann–Whitney U analysis performed by SPSS Statistics (IBM Corporation, Armonk, NY).

RESULTS

Skeletal phenotype of [Phospho1^−/−; Pit1^col2/col2] mice

Immunohistochemistry demonstrated reduced P_iT-1 expression in the proliferative and hypertrophic chondrocyte area of the vertebral bones of 4 month-old P_it1^col2/col2 mice compared to WT mice (Fig. 1A). There was visible residual P_iT-1 expression that was estimated to be 30% by qPCR (Fig. 1B). P_it1^col2/col2 mice were otherwise comparable in size to WT mice. The [Phospho1^−/−; P_it1^col2/col2] skeletons, however, showed exacerbated abnormalities compared to the Phospho1^−/− mice, including multiple fractures and callus formation in the ribs, increased bowing of the long bones and increased prevalence of fractures in these bones (Fig. 2). Histology revealed narrower growth plates in the [Phospho1^−/−; P_it1^col2/col2] bones (Fig. 3A). Histomorphometric analyses in 15-days-old mice confirmed the osteomalacia previously reported in 10-days-old and 1-month-old Phospho1^−/− mice⁽⁹⁾ (Fig. 3B and C) as compared to the WT mice and showed worsening of this phenotype in [Phospho1^−/−; P_it1^col2/col2] mice with even complete absence of mineralization in areas of the trabecular bone (Fig. 3C). The secondary ossification centers also show increased amount of osteoid (Fig. 3D). The vertebral sections also showed the presence of widespread hyperosteoidosis in the Phospho1^−/− mice, which further increased in [Phospho1^−/−; P_it1^col2/col2] mice (Supplemental Figure 1). µCT analysis concurred with the radiographic and histology data showing enhanced bowing of the long bones in tibias and femurs of the [Phospho1^−/−; P_it1^col2/col2] compared to Phospho1^−/− mice (Fig. 4A, B). Trabecular parameters (Table 1) of the femurs of [Phospho1^−/−; P_it1^col2/col2] compared to Phospho1^−/− mice showed significantly decreased BV/TV% (p=0.021), decreased trabecular number (p=0.031), increased structure model index (p=0.006) and decreased bone mineral density (p=0.009). At the tibia, the [Phospho1^−/−; P_it1^col2/col2] mice showed a significant decrease in bone mineral density (p=0.023) compared to Phospho1^−/− mice. Also, the cortical parameters (Table 2) of femurs in the [Phospho1^−/−; P_it1^col2/col2] compared to Phospho1^−/− mice show significantly decreased relative bone area (p=0.017) and cross-sectional thickness (p=0.021), and in [Phospho1^−/−; P_it1^col2/col2] compared to P_it1^col2/col2 mice, there was a significant decrease in relative bone area (p=0.045) cross-sectional thickness (p=0.045) and tissue mineral density (p=0.012). The cortical parameters of tibias in these mice show significant decreases in tissue mineral density (p=0.038) in [Phospho1^−/−; P_it1^col2/col2] compared to P_it1^col2/col2 mice. Three-point bending analysis showed that P_it1^col2/col2 femurs break like WT, but that [Phospho1^−/−; P_it1^col2/col2] bones take even longer deflection than Phospho1^−/− bones and do not break (Supplemental Figure 2). Reduced total plasma alkaline phosphatase activity was observed in 1-month-old [Phospho1^−/−; P_it1^col2/col2] mice compared to WT mice (p=0.012) (Supplemental Figure 3A). Consistent with the measured alkaline phosphatase levels we found increased plasma PP_i levels in [Phospho1^−/−; P_it1^col2/col2] compared to WT (p=0.012) and Phospho1^−/− (p=0.04) mice (Supplemental Figure 3B).

Fig. 1.

Conditional ablation of Pit1 in chondrocytes. (A) Immunohistochemistry using anti-P_iT-1 antibody on the vertebral bones of 4-month-old WT and P_it1^col2/col2 mice showed reduced P_iT-1 expression in the hypertrophic chondrocyte region of P_it1^col2/col2 mice while P_iT-1 expression in skeletal muscle and hematopoietic cells remained unchanged. BM: bone marrow; SM: smooth muscle. Bar=100 µm. B) qRT-PCR of RNA extracted from 5-day-old chondrocytes (n=3) revealed ~30% residual P_it1 gene expression compared to WT chondrocytes (n=3).

Fig. 2.

Phenotypic abnormalities in the skeleton of 1-month-old [Phospho1^−/−; P_it1^col2/col2] mice. Radiographic images of representative mice showing worsening of the skeletal abnormalities (arrows) in [Phospho1^−/−; P_it1^col2/col2] mice as compared to the Phospho1^−/− mice (n = 10). Arrows show highly bowed long bones and multiple fractures in the spine and limbs in [Phospho1^−/−; P_it1^col2/col2] mice.

Fig. 3.

Histomorphometric analyses of 15-day-old tibias of Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] mice. Von Kossa/van Gieson staining of the tibial section at the knee joint in WT and P_it1^col2/col2 mice showed no statistically significant difference in the growth plate (A) as well as the BV/TV and OV/BV ratios in the trabecular bone (B and C) and the secondary ossification center (D). This analysis showed trabecular bone surrounded by widespread, extended osteoid in Phospho1^−/− mice (arrows show smaller growth plate (A) and the areas where the osteoid is present). The [Phospho1^−/−; P_it1^col2/col2] mice show even smaller growth plate (A) and more unmineralized bone in tibia. A-10X, B,C and D-20X magnification (n = 3).

Fig. 4.

µCT analysis of 1-month-old femurs and tibias of WT, Phospho1^−/−, P_it1^col2/col2 and [Phospho1^−/−; P_it1^col2/col2] mice. (A) 3D volume renders of the samples (anterior view-femur, anterior view-tibia, side view–full leg, ,) (B) 2D coronal and transaxial cross-sections of femurs and tibias. Scale bar = 1 mm. Both volume renders and cross-sections show highly bowed/ twisted long bones in [Phospho1^−/−; P_it1^col2/col2] mice compared to WT, P_it1^col2/col2 and even Phospho1^−/− mice.

Table 1.

Trabecular bone morphometric parameters of the femur and tibia. Kruskal-Wallis (mean±SD, n=4).

Sample	Site	Tissue volume TV mm3	Bone Volume BV mm3	Relative Bone Volume BV/TV%	Trabecular Thickness Tb.Th mm	Trabecular Separation Tb.Sp mm	Trabecular Number Tb.N 1/mm	Structure Model Index SMI	Trabecular Pattern Factor Tb.Pf 1/mm	Bone Mineral Density BMD g/cm3
WT	Femur	1.77 ±0.23	0.17 ±0.081	9.33 ±3.48	0.039 ±0.002	0.21 ±0.061	2.36 ±0.76	1.97 ±0.24	27.17 ±6.64	0.22 ±0.08
Phospho1−/−	Femur	2.17 ±0.32	0.14 ±0.04	6.50 ±0.88	0.035 ±0.002	0.21 ±0.02	1.86 ±0.23	2.13 ±0.05	33.75 ±2.48	0.17 ±0.02
Pit1col2/col2	Femur	a2.57 ±0.28	0.15 ±0.03	5.74 ±0.92	a0.033 ±0.001	0.24 ±0.04	1.72 ±0.25	2.18 ±0.05	35.76 ±1.87	0.14 ±0.02
[Phospho1−/−; Pit1col2/col2]	Femur	c1.56 ±0.44	a,b,c0.05 ±0.03	a,b3.11 ±1.35	a0.033 ±0.003	0.30 ±0.06	a,b,c0.93 ±0.37	a,b2.44 ±0.11	a,b43.81 ±6.90	a,b0.10 ±0.02
K-W p-values	Femur	0.016	0.048	0.018	0.029	0.072	0.023	0.015	0.009	0.015
WT	Tibia	1.10 ±0.25	0.05 ±0.03	4.57 ±1.75	0.029 ±0.003	0.23 ±0.05	1.53 ±0.47	2.18 ±0.10	39.37 ±5.46	0.17 ±0.04
Phospho1−/−	Tibia	1.64 ±0.33	0.07 ±0.02	4.49 ±0.39	0.029 ±0.001	0.22 ±0.01	1.56 ±0.13	2.15 ±0.05	39.63 ±2.13	0.16 ±0.01
Pit1col2/col2	Tibia	a1.87 ±0.39	0.07 ±0.02	3.52 ±0.62	0.028 ±0.001	0.26 ±0.03	1.24 ±0.20	2.23 ±0.09	41.52 ±2.83	0.12 ±0.01
[Phospho1−/−; Pit1col2/col2]	Tibia	b,c1.09 ±0.29	0.04 ±0.01	3.30 ±0.41	0.03 ±0.004	0.26 ±0.01	1.10 ±0.10	b2.34 ±0.10	43.31 ±1.8	a,b0.11 ±0.00
K-W p-values	Tibia	0.024	0.089	0.20	0.94	0.12	0.10	0.076	0.334	0.017

Differences at p<0.05 are indicated by

^avs WT,

^bvs Phospho1−/−,

^cvs P_it1^col2/col2

Table 2.

Cortical bone morphometric parameters of the femur and tibia Kruskal-Wallis (mean±SD, n=4).

Sample	Site	Cross-sectional Tissue Area T.Ar mm2	Cross-sectional Bone Area B.Ar mm2	Relative Bone Area B.Ar/T.Ar %	Cross-sectional Thickness Cs.Th mm	Tissue Mineral Density TMD g/cm3
WT	Femur	1.50 ±0.19	0.65 ±0.18	42.95 ±6.74	0.16 ±0.03	0.89 ±0.03
Phospho1−/−	Femur	a1.82 ±0.13	0.73 ±0.08	40.19 ±3.99	0.15 ±0.02	0.87 ±0.05
Pit1col2/col2	Femur	1.67 ±0.07	0.65 ±0.07	38.97 ±2.89	0.14 ±0.01	0.90 ±0.01
[Phospho1−/−; Pit1col2/col2]	Femur	b1.53 ±0.17	0.48 ±0.11	a,b,c30.84 ±4.57	a,b,c0.11 ±0.02	a,c0.85 ±0.02
K-W p-values	Femur	0.041	0.079	0.029	0.027	0.049
WT	Tibia	1.50 ±0.26	0.64 ±0.21	42.04 ±7.72	0.16 ±0.03	1.01 ±0.06
Phospho1−/−	Tibia	1.61 ±0.17	0.67 ±0.09	41.60 ±3.37	0.15 ±0.02	a0.92 ±0.03
Pit1col2/col2	Tibia	1.70 ±0.11	0.61 ±0.07	35.82 ±1.82	0.13 ±0.01	0.93 ±0.01
[Phospho1−/−; Pit1col2/col2]	Tibia	1.45 ±0.22	0.51 ±0.12	34.67 ±3.61	0.12 ±0.01	a,c0.87 ±0.03
K-W p-values	Tibia	0.434	0.270	0.117	0.106	0.020

Differences at p<0.05 are indicated by

^avs WT,

^bvs Phospho1−/−,

^cvs P_it1^col2/col2

Analyses of MVs

Isolated MVs imaged by atomic force microscopy (AFM), appeared as flattened globular features either individually dispersed or connected to form several micron-long chains (Supplemental Figure 4A). First, we calculated the number of MVs by counting the globular features in N=20 scan fields with an area of 1 µm² for each sample. The number of MVs was different among samples with the order WT = P_it1^col2/col2 > Phospho1^−/− ≥ [Phospho1^−/−; P_it1^col2/col2] (Fig. 5A). Next, we modeled MVs as oblate spheroids and calculated their heights (i.e. the polar diameters, 2×a in Supplemental Figure 4B) by recording the cross sections of the globular features in the AFM images and measuring the values of the cross sections’ peaks. The distribution of heights was different among samples. Both WT and P_it1^col2/col2 MVs showed broad height distributions centered at ~7 nm, whereas Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] MVs showed narrow distributions centered at 4.8 nm and 3.8 nm, respectively (Fig. 6A). We also calculated MV volume by calculating the equatorial diameter as the cross sections' width at half height (2×b in Supplemental Figure 4B). WT and P_it1^col2/col2 MVs showed broad and right-skewed volume distributions centered at ~140×10³ nm³whereas Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] MVs showed narrow distributions centered at 37×10³ nm³ and 16×10³ nm³respectively Fig. 6B). We also found that MVs with heights smaller than ~5 nm showed a smooth surface, whereas bigger vesicles showed a non-uniform surface with irregularities that were few to several angstroms tall (Fig. 6C and 6D, respectively). Surface irregularities were interpreted as caused by the presence of mineral aggregates underneath the vesicles’ membranes. These observations suggested dividing the MVs in two groups, mineral aggregate-devoid and mineral aggregate-filled MVs. In order to simplify the subsequent calculations, we considered a height threshold of ~5 nm to distinguish between the two groups of MVs for all samples. We found that the percent of aggregate-filled MVs was different among samples with the order WT = P_it1^col2/col2 ≥ Phospho1^−/− > [Phospho1^−/−; P_it1^col2/col2](Fig. 5B).

Fig. 5.

Determination of the number of MVs and percentage of filled MVs isolated from chondrocytes of each genotype. A) The Phospho1^−/− MV preparations show a statistically significant decrease in the total number of produced MVs compared to those isolated from WT chondrocytes. The [Phospho1^−/−; P_it1^col2/col2] MV preparations also showed a significant decrease in the number of MVs compared to those isolated from WT and P_it1^col2/col2 chondrocytes. The difference between the number of MVs isolated from Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] chondrocytes was borderline significant (p=0.056). B) We observed a statistically significant decrease in the number of filled MVs in the [Phospho1^−/−; P_it1^col2/col2] MV samples compared to the WT, Phospho1^−/− and P_it1^col2/col2 MV samples.

Fig. 6.

Atomic force microscopy (AFM) images were recorded in non-contact (AAC) mode. Height (A) and volume (B) distributions were calculated for MVs isolated from WT, Phospho1^−/−, P_it1^col2/col2 and [Phospho1^−/−; P_it1^col2/col2] mice. MV volumes were calculated by assuming MVs as spheroidal structures (see Supplemental Figure 4B). All peaks were fitted by Gaussian curves. C-F) Images of mineral aggregate-unfilled (C and E) and filled (D and F) WT (C, D and F) and Phospho1^−/− (E) MVs. From left to right: topography image, amplitude image and three-dimensional reconstruction of topography image (C and D); three-dimensional reconstructions of topography and phase images (E and F). Scale bars = 100 nm.

Next, we used AFM phase analysis to gather qualitative information about the internal composition of MVs.^{(22, 23)} These studies were performed on WT and Phospho1^−/− MV samples prepared without the use of glutaraldehyde, in order to avoid the contribution on phase variations as a result of the interaction of the tip with the polymeric shell. MVs with heights smaller than ~5 nm exhibited phase angle (φ) values that were almost constant and similar to those of the mica substrate (Fig. 6E and Supplemental Figure 5A). In contrast, AFM phase images for MVs with heights greater than ~5 nm showed a great heterogeneity of φ values with bright spots that had φ values similar to those of the mica substrate surrounded by regions with negative φ values (Fig. 6F and Supplemental Figure 5B). These spots corresponded to height irregularities in topographic images. The bright spots in MV phase images were interpreted as caused by the presence of highly viscoelastic aggregates of minerals, protein, lipids and other biomolecules – i.e. the nucleational core (NC)^{(24, 25)} – under the MV membrane. These aggregates were surrounded by less crowded regions which made the membrane cover deformable thus leading to negative values of φ due to viscous damping of tip vibration. The constant values of φ in phase images for MVs with heights smaller than ~5 nm suggested the absence of the NC within these MVs.

Finally, we recorded the Raman spectra for WT and Phospho1^−/− MV samples to further validate the differences between the material inside MVs with different sizes and surface morphologies. Raman spectra of WT MVs showed several distinguishable peaks, which are characteristic of both the inorganic and organic matrix phases (red spectrum curve in Supplemental Figure 5). The broad peak at ~1070 cm⁻¹ was attributed to carbonate inorganic phase. Peaks at ~1245 cm⁻¹ and ~1460 cm⁻¹ were attributed to protein amide III and CH₂ deformation, respectively, whereas peaks at ~890 cm⁻¹~1297 cm⁻¹ and ~1437 cm⁻¹ (shoulder) were assigned to CH, CH₂ and CH₃respectively, deformation modes of lipid acyl side chains.^(26,27) Finally, the sets of bands at ~1120 cm⁻¹ were attributed to the P-O-C stretching modes of the phospholipid ester bond.²⁶ In contrast, Raman spectra of Phospho1^−/− MVs showed very weak peaks (blue spectrum curve in Supplemental Figure 6).

DISCUSSION

The double ablation of PHOSPHO1 and TNAP function lead to embryos completely devoid of skeletal mineralization.⁽⁹⁾ This double genetic experiment indicated that initiation of MV-mediated mineralization is dependent not only on intravesicular generation of P_i by PHOSPHO1 but also on P_i generated extravesicularly by TNAP.^{(9, 12)} This conclusion was counterintuitive as we had largely assumed that the mM concentrations of P_i present in all biological fluids would be sufficient to support initiation of mineralization. Instead, this double genetic experiment suggested that TNAP was not only required for the hydrolysis of PP_i to allow for propagation of HA deposition in the extracellular matrix, but also for the perivesicular generation of P_i from µM concentrations of ATP, PP_i or both needed to initiate mineralization inside the MVs.(2, 28–33) This premise implied the participation of P_i-transporters to incorporate the P_i generated perivesicularly by TNAP.^{(9, 12)} Type III P_i transport systems, namely P_iT-1/Glvr1 (SLC20a1 and P_iT-2/Ram, are involved in P_i handling by mineralizing cells (i.e. chondrocytes and osteoblasts)^{(34, 35)} with P_iT-1 being predominantly expressed in these cells.^{(13–15, 35, 36)} In the present study we set out to test the hypothesis that ablating P_iT-1 function in the context of a PHOSPHO1 deficiency would prevent skeletal mineralization to an extent analogous to what we had observed in the Phospho1/Alpl double genetic experiment. A previous report⁽¹⁶⁾ examining an allelic series of P_it1 mutations in mice expressing from 0% to 100% of P_iT-1 showed that a complete knockout of Pit1 (0% expression) lead to a lethal phenotype at E12.5, embryos with 6% residual P_it1 expression were able to live until E15.5, and 15% P_it1 expression level was sufficient to bypass the embryonic lethality resulting instead in significant perinatal lethality.⁽³⁷⁾ Here, we resorted to a conditional ablation strategy aimed at affecting P_iT-1 expression in chondrocytes. Our immunohistochemical staining and qPCR data indicate that Cre-mediated ablation of P_it1driven by the Col2 promoter, led not to the complete ablation of P_iT-1 expression in chondrocytes but to a 70% reduction. The skeleton of the Pit1^col2/col2 animals only showed mild defects in agreement with earlier observations⁽³⁷⁾ but the [Phospho1^−/−; Pit1^col2/col2] double deficiency led to a compounding of the mineralization deficits caused by ablation of Phospho1 function. However, the resulting compounded phenotype, while more severe than the Phospho1^−/− phenotype alone, did not approach in severity the dramatic absence of skeletal mineralization observed in the global [Alpl^−/−; Phospho1^−/−] double deficiency. We attribute this milder phenotype to the fact that expression of P_iT-1 was only reduced by 70% in our conditional ablation strategy. However, the possibility that other P_i transporters (e.g. P_iT-2 or other yet unknown transporters) may be compensating for the downregulation of P_iT-1 or acting as alternative pathways cannot be ruled out. Nevertheless, the fact that ~80% of [Phospho1^−/−; P_it1^col2/col2] MVs were devoid of mineral in comparison to ~50 % for the Phospho1^−/− MVs and ~25% for the WT and P_it1^col2/col2 MVs provides compelling proof that P_iT-1 acts in synergy with PHOSPHO1 during MV-mediated initiation of mineralization in chondrocytes. Figure 7 shows a schematic representation of our current understanding of the mechanisms of initiation of skeletal mineralization⁽¹²⁾, integrating the functional roles of PHOSPHO1⁽⁸⁻¹⁰⁾, NPP1^{(3–5, 31)}, TNAP^(30–33), P_iT-1^{(16, 34–37)} (and data presented in this paper) and osteopontin^(38–41).

Fig. 7.

Schematic detailing our current understanding of the biochemical bases for the steps of MV-mediated initiation of skeletal mineralization. Current data are compatible with the following interpretation: MVs initiate mineral deposition by accumulation of P_i generated intravesicularly by the action of PHOSPHO1 on phosphocholine and also via P_iT-1-mediated incorporation of P_i generated extravesicularly by TNAP or NPP1. As shown in the current paper, PHOSPHO1 function also appears to be implicated in the biogenesis of MVs. The extravesicular propagation of mineral onto the collagenous matrix is mainly controlled by the pyrophosphatase activity of TNAP that restricts the concentration of this potent mineralization inhibitor to establish a PP_i/P_i ratio conducive for controlled calcification. Additionally, osteopontin (OPN), another potent mineralization inhibitor that binds to HA mineral as soon as it is exposed to the extracellular fluid, also restricts the degree of extracellular matrix mineralization. ECM: extracellular matrix; HA: hydroxyapatite OPN: osteopontin.

To enable the quantification of the number of MVs and to discern empty versus filled MVs, we used atomic force microscopy (AFM), a method that has been extensively used to characterize natural (e.g. exosomes)^{(42, 43)}, synthetic lipid vesicles (e.g. liposomes),⁽⁴⁴⁾ and bacterial capsules⁽⁴⁵⁾ and that we recently adapted to the study of MVs.⁽⁴⁶⁾ Here, we used AFM analysis to characterize the number, surface morphology and filling of isolated MVs. MVs appeared as flattened globular features (modeled as oblate spheroids) with different sizes and surface morphologies among samples. WT and P_it1^col2/col2 showed heights up to twenty nanometers and volumes ranging from tens to hundreds of cubic nanometers, whereas Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] showed heights and volumes smaller than 10 nm and 100×10³ nm³respectively. AFM topographic and phase analyses showed that MVs with heights smaller than ~5 nm had a smooth surface and phase angles that were constant across vesicle surface, whereas bigger MVs showed a non-uniform surface with several angstrom tall irregularities that corresponded to spots with phase angles equal to or slightly smaller than that of mica substrate. These spots were surrounded by regions with negative phase angles. Raman spectra of WT MVs showed peaks that corresponded to vibrational modes of carbonate inorganic phase as well as proteins and lipids of organic phase, whereas those of Phospho1^−/− MVs showed very weak peaks. Significantly, Raman spectra of WT MVs did not show the phosphate υ₁ vibrational band at ~950 cm⁻¹ of amorphous calcium phosphate. Possible explanations for this phenomenon are that the phosphate groups were present as acidic HPO₄whose Raman peak at ~1010 cm⁻¹ was covered by the intense bands around ~1070 cm⁻¹and/or were tightly bound to proteins and lipids such that the phosphate υ₁ vibrational band was shifted to higher wavelengths or significantly reduced in intensity.⁽²⁶⁾ Taken together, AFM and Raman spectroscopy analyses are compatible with our working assumption that MVs with heights greater than ~5 nm were filled with highly viscoelastic calcium and phosphate-rich nucleational cores (NCs).^{(24, 25)} Previous studies have found that the NCs are aggregates of particles stabilized by lipids (e.g. phosphatidylserine and cholesterol) and proteins (mostly Annexin V).^{(24, 25)} Our Raman data for WT MVs are in good agreement with the presence of lipid- and protein-rich aggregates inside these vesicles. Additionally, previous studies have described single NC particles having a lateral dimension of ~7 nm and as composed by clusters of ~1 nm-in diameter dot-like subunits of Ca²⁺ and HPO²⁻₄/PO³⁻₄ ions, which is in good agreement with the lateral dimensions of the WT MVs' surface irregularities. These considerations enabled us to formulate a relationship between MV biochemical make-up, size, development stage and amount of mineral aggregates within. MVs with heights smaller than ~5 nm were either devoid of a NC or filled with an amount of aggregates that did not lead to detectable changes in phase. These vesicles were at an early stage of their development. On the other hand, MVs with heights greater than ~5 nm were loaded with a NC, as suggested by the irregularities in their topographic and phase images, thus suggesting their progression to a more advanced stage of the mineralization. Additionally, these analyses allowed us to quantify the number of empty versus filled MVs and document the decreased ability of Phospho1^−/− MVs to initiate mineralization, which become significantly more impaired in [Phospho1^−/−; P_it1^col2/col2] doubly deficient MV. These experimental data provide validation for the involvement of P_iT-1 function in the initiation of MV-mediated mineralization.

Another major finding reported in this paper is that Phospho1 deficiency leads to a significant decrease in MV biogenesis. Both the Phospho1^−/− and [Phospho1^−/−; P_it1^col2/col2] chondrocytes produced a greatly reduced number of MVs, but because Phospho1 deficiency is the common denominator in both of these genotypes we conclude that PHOSPHO1 function is involved in MV biogenesis. The findings described in this paper enable a mechanistic explanation for three earlier experimental observations: 1) the reduced mineralization and reduced number of MVs in the dentin of [Phospho1^−/−; Alpl^+/−] mutant mice⁽¹¹⁾; 2) the reduced mineralization and reduced number of MVs observed in cultured pre-odontoblastic cells deficient in the Trps1 transcription factor, where both PHOSPHO1 and TNAP are downregulated⁽⁴⁶⁾ and 3) the synergistic effect of the combined use of PHOSPHO1 and TNAP inhibitors in suppressing vascular smooth muscle cell calcification.⁽⁴⁷⁾ All those earlier observations can now be mechanistically understood as follows: the genetic^{(11, 46)} or pharmacological⁽⁴⁷⁾ ablation of PHOSPHO1 function impairs MV biogenesis while also diminishing intravesicular P_i-generation needed to initiate mineralization as well as causing a downregulation in TNAP expression.⁽⁹⁾ In turn, the genetic^{(4, 9, 46)} or pharmacological ablation of TNAP function^{(47, 48)} leads to an increase in extracellular PP_iwhich suppresses extracellular matrix calcification, and also a reduction in perivesicular P_i-generation needed for P_iT-1 influx (Figure 7). The finding reported in this paper that PHOSPHO1 function is involved in MV biogenesis can be exploited pharmacologically for the treatment of arterial calcification and other forms of ectopic ossification.