Increased osteopontin contributes to inhibition of bone mineralization in FGF23-deficient mice

Abstract

Excessive FGF23 has been identified as a pivotal phosphaturic factor leading to renal phosphate-wasting, and the subsequent development of rickets and osteomalacia. In contrast, loss of FGF23 in mice (Fgf23^−/−) leads to high serum phosphate, calcium and 1,25-vitamin-D levels resulting in early lethality attributable to severe ectopic soft-tissue calcifications and organ-failure. Paradoxically, Fgf23^−/− mice exhibit a severe defect in skeletal mineralization despite high levels of systemic mineral ions and abundant ectopic mineralization, an abnormality that remains largely unexplained. Through use of in situ hybridization, immunohistochemistry and immunogold labeling coupled with electron microscopy of bone samples we discovered that expression and accumulation of osteopontin (Opn/OPN) was markedly increased in Fgf23^−/− mice. These results were confirmed by qPCR-analyses of Fgf23^−/− bones and ELISA measurements of serum OPN. To investigate whether elevated OPN levels were contributing to the bone mineralization defect in Fgf23^−/− mice, we generated Fgf23^−/−/Opn^−/− double-knockout mice (DKO). Biochemical analyses showed that the hypercalcemia and hyperphosphatemia observed in Fgf23^−/− mice remained unchanged in DKO mice, however µCT and histomorphometric analyses showed a significant improvement in total mineralized bone-volume. The severe osteoidosis was markedly reduced and a normal mineral apposition rate was present in DKO mice, indicating that increased OPN levels in Fgf23^−/− mice are at least in part responsible for the osteomalacia. Moreover, the increased OPN levels were significantly decreased upon lowering serum phosphate by feeding low phosphate diet or deletion NaPi2a, indicating phosphate attributes in part to the high OPN levels in Fgf23^−/− mice. In summary, our results suggest that increased OPN is an important pathogenic factor mediating the mineralization defect and the alterations in bone metabolism observed in Fgf23^−/− bones.

Introduction

FGF23, a member of the FGF19 subfamily of fibroblast growth factors, has been shown to play a key role in balancing mineral ion homeostasis (1–4). Consequently, it has been implicated in the pathogenesis of various phosphate-wasting diseases, including autosomal dominant hypophosphatemic rickets (ADHR) (5), X-linked hypophosphatemia (XLH) (6), oncogenic osteomalacia (OOM) (7), chronic kidney disease (CKD) (8), familial tumoral calcinosis (FTC) (9), McCune-Albright syndrome and fibrous dysplasia of bone (10, 11).

FGF23 is mainly secreted by bone cells (11–13) and its function is dependent on an interaction with the co-factor Klotho (14–16). Together, these form a complex with FGF receptor 1c (FGFR1c), thereby converting this otherwise canonical FGF receptor into a receptor specific for FGF23 (15). FGF23 uses the FGFR1c/Klotho complex to directly target the kidney where it induces phosphate wasting by decreasing the expression of the sodium-dependent phosphate co-transporters NaPi2a and NaPi2c (17, 18).

Recent studies (18, 19) have demonstrated that overexpression of FGF23 in mice leads to hyposphosphatemia and hyperphosphaturia. Conversely, mice lacking FGF23 function (Fgf23^−/−) exhibit a phenotype that includes profound growth retardation, muscle wasting, infertility, atherosclerosis, extensive soft tissue calcifications, pulmonary emphysema, general tissue atrophy, severely shortened lifespan, and biochemical disorders including hyperphosphatemia, hypercalcemia, high serum 1,25(OH)₂D levels, and decreased serum PTH levels (20–22). Despite the presence of such a high serum mineral ion content and even the presence of severe soft tissue calcifications (23, 24), Fgf23^−/− mice present with severe defects in skeletal mineralization (osteomalacia/osteoidosis). The reason for this reduced skeletal mineralization occurring in the presence of high serum calcium and phosphate is largely unknown, but such observations suggest an accumulation of a mineralization inhibitor locally within the extracellular matrix of bone.

FGF23 itself has been demonstrated to be an inhibitor of mineralization, but whether it acts directly or indirectly is not yet known. Wang et al (25) showed that adenoviral overexpression of FGF23 in rat calvarial cells in vitro inhibits bone mineralization independent of its systemic effects on phosphate homeostasis. We, and others, have also demonstrated that FGF23 treatment of primary calvarial osteoblast cultures from wild-type (WT) mice or from the osteoblastic MC3T3-E1 cell line leads to an inhibition of mineralization (21, 26), thus showing an effect on mineralization independent of circulating factors. The paradox of these findings inspired us to investigate the underlying mechanisms for the abnormal skeletal mineralization pattern seen in Fgf23^−/− mice. We confirmed the accumulation of osteoid (osteoidosis) in the bone of these mice and discovered that expression of osteopontin (OPN), a well-known inhibitor of bone mineralization (27–30) is substantially elevated, suggesting a possible explanation for the improper mineralization of the bone. We now also demonstrate that ablation of Opn (Spp1) from Fgf23^−/− mice, significantly ameliorates the osteomalacia. Collectively, these findings indicate that increased OPN levels are responsible, in part, for the skeletal mineralization defect seen in Fgf23^−/− mice.

Material and Methods

Animals

Opn (Spp1)-knockout mice were obtained from the Jackson Laboratory (Bar Harbor, ME). Heterozygous Fgf23^+/− animals were interbred with Opn^+/− or NaPi2a^+/− animals to obtain Fgf23^−/−/Opn^−/− or Fgf23^−/−/NaPi2a^−/− double-knockout mice for subsequent analyses. Low-phosphate diet (0.17%) was obtained from PharmaServ Inc. (TestDiet, Cat#: 5857, Framingham, MA) and fed to the mice starting at 3-weeks of age for a duration of 3 weeks before collection of serum. The compound knockout Fgf23^−/−/NaPi2a^−/− mice and the low-phosphate diet were used to examine the effects of phosphate and hypophosphatemia on Opn gene expression. The total body weight of each mouse was measured weekly starting at 3 weeks after birth. All animal procedures used were approved by the Institutional Animal Care and Use Committee at the Harvard Medical School.

Biochemical analyses

Blood was obtained by puncturing the cheek pouch of animals. Serum was isolated by centrifugation at 3000 g for 10 minutes and stored at − 80°C. Total serum calcium and phosphorus levels were determined using Stanbio LiquiColor (Arsenazo III) and LiquiUV kits (Stanbio Laboratory, Boerne, TX), respectively. Serum concentrations of OPN, CTX and 1,25(OH)₂D were measured using commercial ELISA kits from R&D Systems, Inc. (Minneapolis, MN) and IDS (Fountain Hills, AZ), respectively.

Bone histology, µCT analyses and histomorphometry

Micro-computed tomography (µCT) analysis was performed according to the recent guidelines (31) using a Scanco Medical µCT 35 system (Scanco, Southeastern, PA) with an isotropic voxel size of 7 µm to image the distal femur. For histomorphometry processing of undecalcified bone specimens and cancellous bone were performed as described previously (23). Femurs were fixed in 10% buffered formalin at 4°C overnight and stored in 70% ethanol at 4°C before being processed and embedded in methylmethacrylate. Three-µm-thick midsagittal sections of the distal femurs were prepared using a HM 360 microtome (Microm, Walldorf, Germany), and were stained with von Kossa/McNeal. Histomorphometric measurements in the distal femur were made on sections stained with von Kossa/McNeal using a semiautomatic system (OsteoMeasure, OsteoMetrics, Decatur, GA), and a Zeiss Axioskop microscope with a drawing attachment. The area within 0.25 mm from the growth plate was excluded from the measurements. All histomorphometric parameters were calculated and expressed according to the suggestions made by the ASBMR nomenclature committee (32).

In situ hybridization

Animals were dissected and tissues were fixed in 4% paraformaldehyde (PFA)/PBS pH 7.4 at 4°C for several days. Bones were subsequently demineralized for 1–2 weeks in 20% EDTA. All tissues were rinsed in PBS, dehydrated at room temperature through an ethanol series: 70% for 6 h, 80% for 1 h, 96% for 1 h and 100% for 3 h, cleared twice in xylene for 1 h/step, embedded in paraffin, serial sectioned at 6 µm using a Microm HM 360 microtome (Microm, Walldorf, Germany) and mounted on SuperFrost Plus slides. Complementary ³⁵S-UTP-labeled riboprobe OPN was used for performing in situ hybridization on paraffin sections, as described previously (33).

Immunohistochemistry

Immunohistochemistry was performed on paraffin sections using anti-mouse OPN antibody (R&D, Minneapolis, MN) with a working concentration of 0.5 µg/ml overnight at 4 °C. Non-immune immunoglobulin of the same isotype was used as a negative control. Tissue was stained with anti-goat HRP substrate and DAB (Vector, Burlingame, CA), and then counterstained with hematoxylin.

Quantitative real-time PCR

Total RNA from cortical bone of the femurs was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol. For qRT-PCR, cDNA was prepared using QuantiTec reverse transcription kit (Qiagen, Valencia, CA) and analyzed with SYBR GreenMaster Mix (SABiosciences, Valencia, CA) in the iCycler (Bio-Rad, Hercules, CA) using specific primers designed for each targeted gene. Relative expression was calculated using the 2^−ΔΔCt method by normalizing with Gapdh housekeeping gene expression, and presented as fold increase relative to control.

Electron microscopy and immunolabeling of mouse bone sections for OPN

Calvariae from 3-wk-old Fgf23^−/− and WT littermate mice were embedded in plastic for histology by light and transmission electron microscopy, and for high-resolution, immunogold ultrastructural labeling for OPN. Bones were fixed in 4% paraformaldehyde plus 1.0% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.3. Calvariae samples were left undecalcified for embedding in Epon epoxy resin (Cedarlane, Burlington, ON), or were decalcified for immunogold labeling in 8% EDTA over 2 weeks followed by embedding in LR White acrylic plastic (London Resin Company, Berkshire, UK). Samples destined for embedding in Epon for morphological analysis were additionally osmicated for 1 hour in potassium ferrocyanide-reduced 1% osmium tetroxide. Prior to embedding, all samples were dehydrated in a graded ethanol series, infiltrated with the embedding media, placed into mounting molds and the blocks were polymerized at 55°C for 2 days. For light microscopy, 1-µm-thick survey sections were cut from the polymerized blocks on a Leica Ultracut E ultramicrotome (Leica, Wetzlar, Germany) and stained for mineral by von Kossa staining (for the undecalcified samples) followed by counterstaining with toluidine blue. For light microscopy, sections were mounted on glass slides, cover-slipped, and viewed and photographed using a Leitz DMRBE (Leica, Wetzlar, Germany) optical microscope equipped with a 3-CCD Sony DXC-950 camera (Sony, Tokyo, Japan).

For transmission electron microscopy, 80-nm-thick sections were cut on the ultramicrotome followed by conventional staining with uranyl acetate and lead citrate after which the sections were viewed in a FEI Technai 12 transmission electron microscope (Hillsboro, OR) operating at 120 kV and equipped with a 792 Bioscan 1k × 1k wide-angle multiscan CCD camera (Pleasanton, CA). For immunogold labeling of OPN prior to transmission electron microscopy, LR White sections were incubated with polyclonal goat anti-mouse OPN antibody (R&D Systems, Minneapolis, MN), followed by rabbit anti-goat secondary antibody (Sigma Aldrich, St. Louis, MO), and then protein A-colloidal gold (14 nm) conjugate (Dr. G. Posthuma, University of Utrecht, Utrecht, The Netherlands).

Statistics

Statistically significant differences were evaluated by Student’s t-test for comparison between two groups or by one-way analysis of variance (ANOVA) followed by the Tukey’s test for multiple comparisons. All values are expressed as mean ± SD. A p value of less than 0.05 was considered to be statistically significant.

Results

Loss of FGF23 function results in a skeletal mineralization defect

Fig. 1A shows sections of undecalcified femurs obtained from 6-week-old Fgf23^−/− and WT animals. Accumulation of osteoid (indicated by red arrows) is prominent in trabecular and cortical bone of Fgf23^−/− animals. Comparison and quantification of the osteoid (osteoid over bone volume - OV/BV) showed a significant osteoid increase in the mutant mice (55.8±11.5 %) when compared to that of control WT mice (2.6±0.9 %) (Fig. 1B). This observation was confirmed by the analyses of histology survey sections of undecalcified calvariae. The presence of thin osteoid seams in WT bone (Fig. 1C) is visible, but Fgf23^−/− bone had large tracts of well-formed, but unmineralized bone in many locations. Upon closer inspection, even by light microscopy, this prominent osteoidosis in Fgf23^−/− calvariae was unique in that a fine, speckled appearance was present (white boxes, right panel) in the widened osteoid. An ultrastructural investigation by transmission electron microscopy was made at these and other sites as indicated by the white boxes. This analysis revealed that the small speckles observed by light microscopy in Fgf23^−/− bone were in fact numerous aborted mineralization foci in the widened osteoid seam that decreased in abundance towards the periosteal surface of the calvariae. While initial bone mineralization towards the endosteal surface proceeded in a seemingly normal manner to produce a thin continuous layer of mineralized bone, bulk mineralization of the Fgf23^−/− bone matrix towards the periosteal surface terminated as a sharply defined border delimiting an aborted mineralization front.

Figure 1. Defective skeletal mineralization of Fgf23^−/− mice.

(A) Representative images of undecalcified sections of femurs stained with von Kossa reagent for mineral (black) and counter stained with toluidine blue. Accumulation of osteoid (indicated by red arrows) is prominent in trabecular and cortical bone of Fgf23^−/− mice. (B) Quantification of the osteoid volume (OV/BV) in trabecular bone. (C) Light micrographs (upper two panels) and electron micrographs (lower four panels whose approximate locations are framed by white boxes) of 3-wk-old undecalcified calvariae. By light microscopy, the normal bone in WT mice shows full-thickness mineralization with thin unmineralized osteoid seams. Transmission electron microscopy of WT bone shows osteoblasts closely opposed to the osteoid surface, osteocytes embedded within mineralized bone, and the presence of an indistinct mineralization front (red dashed line) separating the osteoid from the mineralized bone. In Fgf23^−/− calvarial bone, greatly widened osteoid seems are present, with a speckling in the innermost regions of the osteoid that were revealed by electron microscopy to be abundant aborted mineralization foci (arrows) in a generally unmineralized osteoid. In deeper regions that were in fact well mineralized, the mineralization front (red dashed line) was sharply delineated as an electron-dense boundary indicating a termination of mineralization at this site. Immediately surrounding osteocytes in this widened osteoid was a peripheral accumulation at the lacunar wall of finely granular mineral deposits (asterisks). Data are presented as Mean ± SD. ***: p<0.001vs WT.

There was a normal progression of bulk matrix mineralization to incorporate osteocytes into lacunae whose walls were well mineralized in WT calvarial bone. However, in Fgf23^−/− calvariae, cells remaining as osteoid osteocytes in the widened osteoid seam were associated with a peculiar peripheral mineralization attempt characterized by finely granular mineral deposits surrounding the cells. This granular appearance again was reminiscent of aborted mineralization, which would otherwise be more confluent in appearance as in the wild-type bone surrounding osteocytes.

Increased OPN levels in Fgf23^−/− mice

We investigated the expression of OPN, which is a well-recognized mineralization inhibitor (27–30), and found significantly elevated levels of this inhibitory protein in Fgf23^−/− mice. Both in situ hybridization (Fig. 2A) and immunohistochemistry (Fig. 2B) were performed on femur sections of 6-week-old animals showing an enhanced OPN signal in bones of Fgf23^−/− mice. This was confirmed by qRT-PCR analysis of cortical bones (femur), indicating that Opn mRNA levels in Fgf23^−/− bone were 2-fold higher than in controls (Fig. 2C). Moreover, serum OPN levels in Fgf23^−/− mice at 3 and 6 wks of age were 5.28±0.65 µg/ml and 4.42±1.41 µg/ml, respectively. These are over 13-fold higher than the levels in of controls (3 wk: 0.39±0.09 µg/ml; 6 wk: 0.28±0.14 µg/ml) (Fig. 2D).

Figure 2. Increased OPN levels in Fgf23^−/− mice.

(A) In situ hybridization and (B) immunohistochemistry staining performed on femur sections of 6-week-old animals. Both showed increased OPN signals in Fgf23^−/− bone. (C) mRNA expression of Opn in the cortical bones (femur) quantified by qRT-PCR analysis. (D) Serum OPN measurements. (E) Transmission electron micrographs after immunogold labeling for OPN in undecalcified calvarial bone at the locations framed by the white boxes. Compared to the moderate extent of immunolabeling of WT bone seen as patches of gold particles dispersed throughout the matrix and surrounding osteocyte lacunae. The bone of Fgf23^−/− is intensely labeled at several locations. Abundant gold-particle labeling is observed over the aborted mineralization foci (arrows) in the osteoid, at the sharply demarcated mineralization front (bracket), and immediately lining the lacunar wall (asterisks) surrounding osteocytes. Osteocytes found in these heavily OPN-labeled regions of osteoid bone matrix show prominent secretory granules intensely labeled for OPN indicating the local secretion of this protein by bone cells. The blue dashed line indicates the cell-matrix interface. Data are presented as Mean ± SD. *: p<0.05, ***: p<0.001vs WT.

Given its potency as a mineralization-inhibiting protein, we next investigated if there is an abnormal accumulation of OPN at these aborted mineralization sites in Fgf23^−/− mice (Fig. 2E). Using the high-resolution immunogold staining method, we found intense labeling for OPN at levels well above those in WT bone at the aborted mineralization front, at the aborted mineralization foci in the widened osteoid seam, and at the edges of bone matrix surrounding the osteoid osteocytes. The locations correlate exactly with the sites described previously in Figure 1. Moreover, OPN derived at least in part locally from nearby osteocytes, as evidenced by secretory granule immune-gold labeling for OPN in these cells.

Deletion of Opn partially rescues the mineralization defect of Fgf23^−/− mice

To investigate whether the increased OPN levels contribute to the skeletal mineralization defect in Fgf23^−/− animals, we generated Fgf23^−/−/Opn^−/− double-knockout mice (DKO) by interbreeding heterozygous Fgf23^+/− and Opn^+/− mice. DKO mice were more active, healthier and larger in size than Fgf23^−/− littermates (Supplementary Fig. 1A). Their body weight was significantly higher than Fgf23^−/− mice, although still lower than that of WT and Opn^−/− mice (Supplementary Fig. 1B).

Biochemical assays of serum were performed for all four genotypes. No differences could be detected between the Fgf23^−/− and DKO groups, with both being similarly hypercalcemic and hyperphosphatemic, and having significantly elevated serum 1,25 vitamin levels (Fig. 3).

Figure 3. Serum biochemical measurements.

Fgf23^−/− and DKO mice show similarly increased levels of serum calcium (A), phosphate (B), and 1,25(OH)₂D (C). Data are presented as Mean ± SD. *: p<0.05, ***: p<0.001vs WT; #: p<0.05, ###: p<0.001vs Fgf23^−/−; and $: p<0.05, $$$: p<0.001 vs DKO.

We then performed µCT analyses and three-dimensional reconstruction of the bones to examine the skeletal phenotype of the mutant mice. Representative images of distal femoral metaphysis and midshaft cortex are shown in Figures 4A, C. Quantification of the µCT data demonstrated that the trabecular bone volume fraction (BV/TV) of the Fgf23^−/− mice (2.0±1.3%) was significantly reduced compared to the other groups (Fig. 4B); it was restored in DKO mice (16.2±3.6%) to a volume exceeding that in WT (10.1±3.3%) and Opn^−/− (14.2±3.2%) mice (Fig. 4B). Moreover, the thickness and architecture of the cortical bone were also significantly improved (Figs. 4C, D). As µCT analyzes only the mineralized part of the bone, these results indicate that the mineralization defect in these mice was partially rescued.

Figure 4. µCT analyses.

(A) Representative images of 3D reconstruction of distal femoral metaphyses. (B) Trabecular bone volume fraction (BV/TV) is significantly increased in DKO. (C) Representative images of 3D reconstructions of midshaft cortical bone. (D) The cortical thickness (C.Th) of DKO bone is improved compared to that of Fgf23^−/− bone. Data are presented as Mean ± SD. *: p<0.05, **: p<0.01, ***: p<0.001vs WT; ###: p<0.001vs Fgf23^−/−; and $$$: p<0.001 vs DKO.

To further support our observation, we then performed histomorphometry of the trabecular area of the femora using undecalcified sections. We were able to confirm that BV/TV in the DKO bone was increased compared to that of Fgf23^−/− (Fig. 5A & B). More importantly, the osteoidosis in the trabecular bone was markedly reduced in DKO bones (Fig. 5A). The osteoid volume (OV/BV) of the DKO mice (25.1±21.4%) was significantly lower than that of the Fgf23^−/− mice (62.7±7.6%) (Fig. 5B). Similar changes were observed for the osteoid surface (OS/BS) and osteoid thickness (O.Th). Meanwhile, the mineralized bone volume (Md.V/TV) was significantly increased in DKO (Fig. 5B). In addition, the mineral apposition rate (MAR) in the DKOmice (1.9±0.3 µm/d) was completely restored to levels similar to the WT mice (2.6±0.7 µm/d). Moreover, there was no significant difference in the mineralization lag time (Mlt) when compared to the wild-type. In contrast, fluorochrome labeling of Fgf23^−/− bones was generally indistinct and unsuccessful which we attribute to their prominent mineralization defect. Furthermore, osteoclast numbers (N.Oc/Md.Pm) were decreased in DKO mice as compared to Fgf23^−/− mice (Fig. 5B). This was confirmed by the measurement of serum carboxy terminal cross-linked telopeptide of type I collagen (CTX), which showed that the increased level in Fgf23^−/− mice was restored to a normal level in the DKO mice (Supplementary figure 2).

Figure 5. Histological and histomorphometric analyses.

(A) Representative images of undecalcified sections of distal ends of femurs from 6wk-old littermates stained with von Kossa and McNeal stains. Red arrows indicate the large amount of unmineralized osteoid in Fgf23^−/− bone. (B) Histomorphometric analysis confirmed that the mineralization defect of Fgf23^−/− bone was partially rescued in DKO mice. BV/TV: bone volume; OV/TV: osteoid volume; OS/BS: osteoid surface; O.Th: osteoid thickness; Md.V/TV: mineralized trabecular bone volume; MAR: mineral apposition rate; and Mlt: mineralization lag time; N.Oc/Md.Pm: osteoclast number/mineralized bone perimeter. (C) Light micrographs of 3-wk-old undecalcified calvarial bone after von Kossa staining for mineral. Whereas in the Fgf23^−/− mice the osteoidosis is severe, in the DKO calvariae there is an obvious decrease of osteoid thickness and with breakthrough isolated “islands” of mineralization (arrows) extending throughout the osteoid. Data are presented as Mean ± SD. *: p<0.05, **: p<0.01, ***: p<0.001 vs WT; and #: p<0.05, ##: p<0.01, ###: p<0.001 vs Fgf23^−/−; $$: p<0.01 vs DKO.

As shown in Figure 5C, the calvariae of the Fgf23^−/− mice exhibited a widened osteoid. This was considerably reduced in the DKO, although some osteoidosis did persist. Breakthrough patches of mineralization were observed in the osteoid of the DKO mice, which indicates a partial recovery of mineralization in this widened osteoid seam (Fig. 5C).

Increased OPN levels were induced by the hyperphosphatemia

Fgf23^−/− mice are severely hyperphosphatemic and we hypothesized that the increased OPN levels were induced by this hyperphosphatemia. We successfully reversed the serum phosphate levels in Fgf23^−/− mice by generating Fgf23^−/−/NaPi2a^−/− double-knockout mice and by feeding Fgf23^−/− mice a low-phosphate diet. In both cases, the serum phosphate levels were significantly lower than those of WT and Fgf23^−/− mice (Fig. 6A). We also found that the OPN levels in Fgf23^−/− mice were significantly decreased when serum phosphate levels were lowered, although they remained significantly higher than in controls (Fig. 6B). These data indicate that phosphate is an important contributor to the high OPN levels in Fgf23^−/− mice. Low-phosphate diet feeding also led to a partial decrease of OPN in WT mice, while it remained unchanged in NaPi2a^−/− mice.

Figure 6. Decreased serum OPN levels upon lowering of serum phosphate.

(A) Serum phosphate levels of mice with corresponding genotypes are shown. The hyperphosphatemia of Fgf23^−/− mice was reversed to hypophosphatemia by deleting NaPi2a (Fgf23^−/−/Napi2a^−/−) or by feeding a low-phosphate diet (Fgf23^−/−; Low Pi Diet). (B) The high serum OPN levels of Fgf23^−/− significantly decreased upon lowering serum phosphate levels in Fgf23^−/−/NaPi2a^−/− and Fgf23^−/−; Low Pi Diet mice, but remained markedly higher than in wild-type (WT), NaPi2a^−/− or wild-type mice fed a low-Pi diet (WT; Low Pi Diet). Data are presented as Mean ± SD. *: p<0.05, ***: p<0.001 vs WT; #: p<0.05, ###: p<0.001 vs Fgf23^−/−.

Discussion

Renal phosphate wasting typically leads to hypophosphatemia and osteomalacia. However, Fgf23^−/− mice having a severe osteomalacic mineralization defect paradoxically have hyperphosphatemia (and hypercalcemia). Here, we have demonstrated that Fgf23^−/− mice have greatly elevated Opn gene expression and OPN protein accumulation in bone, and also highly elevated serum OPN. Our experiments that reduced serum phosphate levels by dietary and genetic means demonstrated that the increase in OPN in the Fgf23^−/− mice was caused at least in part by higher circulating phosphate. Furthermore, we have shown that deletion of Opn from Fgf23^−/− mice partially rescues the mineralization defect such that the osteomalacia is significantly reduced. Taken together, these data suggest that increased OPN levels are at least in part responsible for the skeletal mineralization defect observed in Fgf23^−/− mice.

OPN, also originally known as secreted phosphoprotein 1 (SPP1), is a member of the SIBLING family (small integrin-binding ligands N-linked glycoproteins) of extracellular matrix mineral-binding proteins (34–36). In bone, OPN is produced abundantly by osteoblasts and osteocytes as a phosphorylated, secreted extracellular matrix protein (37, 38). Once in the extracellular matrix, OPN binds avidly to mineral crystal surfaces (39–42) to inhibit their growth as it loads into the bulk phase of the mineralizing bone matrix. Within the extracellular matrix of bone, enzymatic degradation of OPN by PHEX (phosphate-regulating gene with homologies to endopeptidases on the X-chromosome) and removal of phosphates by TNAP (tissue-nonspecific alkaline phosphatase, ALPL) may modulate the mineralization-inhibiting activities of this protein (28, 29, 43, 44). OPN also accumulates at cell- and matrix-matrix interfaces where mineralization is tightly controlled such as at cement lines and at the lamina limitans found at the osteocyte- and bone-lining cell-matrix interface where mineralization is essentially terminated, presumably via the actions of the accumulated inhibitory OPN at these interfacial sites (45), while not excluding similar inhibitory functions by other proteins/peptides at these same locations. Besides non-collagenous proteins, mineralization may likewise be partly regulated in the extracellular matrix by collagen itself and its processing by enzymes (46).

Amongst myriad functions shown for OPN (27, 47, 48), one of its most prominent functions includes direct inhibition of hydroxyapatite crystal growth by binding to lattice calcium exposed at crystal surfaces, regulating crystal dimensions by limiting growth where it has bound (49, 50). Its high negative charge derived primarily from abundant aspartic acid and glutamic acid amino acid residues, along with it having a high level of post-translational phosphorylation, all lead to an extended and flexible protein capable of binding positively charged calcium atoms residing at crystal surfaces (51, 52). Proteins binding in this way to bio-minerals generally act as inhibitors, influencing the number, form, shape and alignment of the mineral crystals (53) in bones and teeth, mineral crystal growth takes place in the collagenous extracellular matrix permeated by numerous noncollagenous proteins and small proteoglycans thought to regulate the mineralization process within the collagen scaffold. OPN is the most widely studied protein within the SIBLING family of bone and tooth noncollagenous proteins, and its properties as a mineralization inhibitor appear to be far greater than the other SIBLING proteins (54). OPN is upregulated locally and accumulates at high levels at sites of ectopic calcification where it is thought to be involved in the host response to limit this pathologic mineralization (55, 56); indeed, deletion of OPN from matrix Gla protein (MGP)-deficient mice showing severe vascular calcification leads to increased vascular calcification and even earlier mortality (57). As another example, OPN-deficient mice are also prone to kidney stone formation (58). OPN also interplays with the other members of SIBLING family. For example, OPN protein expression in bone is dramatically decreased in MEPE overexpressing mice (59).

For bone, OPN-deficient mice show increased mineralization (60) coincident with reduced bone biomechanical properties (61). In recent years, several knockout mouse models have been shown to exhibit elevated OPN that has been suggested to contribute to the osteomalacic phenotype (62, 63). Also of importance in determining OPN levels in the bone is the observation that OPN and its peptides (including the ASARM peptide) can be inactivated by their essentially complete degradation by the enzyme PHEX (29, 43), thus explaining the accumulation of OPN and OPN fragments in the Hyp mouse model of X-linked hypophosphatemia, a human disease with inactivating mutations in PHEX leading to renal phosphate wasting, hypophosphatemia and osteomalacia.

In the present study, we demonstrated that the elevated OPN levels seen in the Fgf23^−/− mice were significantly decreased when serum phosphate levels were lowered – either through diet or deletion of renal phosphate co-transporter activity – indicating that the hyperphosphatemia in Fgf23^−/− mice is at least in part responsible for the increased OPN levels. Indeed, phosphate is not only a potent regulator of mineralization, but it acts as a specific signal for induction of Opn expression in osteoblast lineage cells (64, 65). This latter effect requires the activity of ERK1/2 and protein kinase C, as well as the glucocorticoid receptor and proteosomal/ubiquination pathways (65, 66). Phosphate also induces the expression of Opn in vascular smooth muscle cells relevant to the frequency with which vascular calcification is observed (67). In contrast, however, we have previously reported that mice deficient for both KLOTHO and PTH, which exhibit extremely high serum phosphate levels, have normal OPN expression and serum levels (68) accompanied by normal bone mineralization. PTH is another known regulator of Opn, which increases its transcription and expression in osteoblasts (69, 70); accordingly, this loss of PTH induction of OPN could offset the effect of the high serum phosphate in these mice in terms of modulating OPN levels. Related to these findings, we observed that PTH levels in Fgf23^−/− mice were low-to-undetectable (2, 21), eliminating this PTH-induction pathway option as a potential explanation for the high OPN levels. In addition, even genetic deletion of PTH from Fgf23^−/− mice did not affect the increased OPN level (68) or rescue the skeletal mineralization defects (24), again suggesting that PTH is not a key factor responsible for the increased OPN level observed in Fgf23^−/− mice.

In terms of vitamin D regulation, 1,25(OH)₂D is known to induce gene expression of Opn via binding of the VDRE (vitamin D response element) to regulatory sequences directly upstream of the Opn promoter (71). Lieben et al. (72) demonstrated that 1,25(OH)₂D suppresses mineralization by upregulating Opn expression and PPi levels. Our previous study also showed that ablation of vitamin D signaling rescues bone and mineral homeostasis in Fgf23^−/− mice fed a high-calcium and high-phosphate diet (73). To investigate whether the elevated serum levels of 1,25(OH)₂D are another potential explanation for the high OPN levels in Fgf23^−/− mice, we eliminated all vitamin D signaling by genetic deletion of 1a(OH)ase from Fgf23^−/− mice (Fgf23^−/−/1a(OH)ase^−/−). This deletion did not change the increased OPN levels, suggesting that vitamin D is not responsible for the elevated OPN level observed in Fgf23^−/− mice (Supplementary Figure 3).

Transmission electron microscopy was used to examine the inhibited mineralization process and the cellular secretion and accumulation of OPN in Fgf23^−/− mice at the ultrastructural level. Inhibited mineralization appeared to reside at three locations within the bone matrix, with each inhibited site showing an abundant accumulation of OPN as detected by immunogold labeling. These sites where mineralization was inhibited, and possibly even completely aborted, included small punctate foci throughout the osteoid, a pericellular rim of matrix/mineral at the margins of osteocyte lacunae, and a generally sharply defined mineralization front separating the osteoid from the mineralized bone matrix proper. The tissue morphology at these locations indicates interruption of normal mineralization pathways, which together with the excessive accumulation of OPN at these sites which otherwise show much lesser OPN accumulation in normal WT mice at similar sites (45), all indicate that the mechanism of this inhibition resides in OPN’s mineral-binding and inhibitory function. Normally, OPN is thought to regulate crystal growth as it occurs in the bulk bone matrix at small mineralization foci and at the mineralization front, and surrounding osteocytes where mineralization has to be terminated to maintain the patency of osteocyte lacunae. Indeed, in the Fgf23^−/− mice, OPN was found at its highest levels to inhibit mineralization precisely at these same locations where regulation/inhibition of mineralization is normally ongoing, and removing OPN in FGF23-deficient mice in the double knockout (Fgf23^−/−/Opn^−/−) partially rescued this inhibition.

In addition to advances made in recent years in understanding the roles of phosphorylated SIBLING proteins (and functionally active peptides derived from them) in modulating mineralization, similar progress has been made towards understanding the role of pyrophosphate (PPi) in inhibiting mineralization. In both cases – for proteins and PPi – processing enzymes play key roles in inactivating and activating these molecular determinants of mineralization, as shown by the consequences (osteomalacia) of inactivating mutations in ALPL affecting PPi degradation (hypophosphatasia) (74), and in PHEX affecting OPN degradation (X-linked hypophosphatemia) (75), where PPi and OPN respectively accumulate in the bone matrix to inhibit mineralization. Elegant genetic studies in mice by Millan and colleagues have also clearly shown the combined action of these two determinants to modulate the P:PPi ratio and to elevate inhibitory OPN protein levels (63), both of which may simultaneously contribute to osteomalacia where serum phosphate and calcium levels are within the normal range. Such inhibition of bone mineralization – unlike the systemic effects of renal phosphate wasting – is thought to occur locally at the level of the extracellular matrix, and for OPN, phosphorylation status appears to be particularly important (29, 44, 50, 76, 77). Indeed, phosphorylated OPN may be a more potent inhibitor than PPi in vivo given that elevated extracellular PPi levels found in Opn^−/− mice did not cause osteomalacia (62, 63), where it was proposed that much higher levels of PPi might be needed in the absence of OPN to affect mineralization. Here, in the present study, we provide new evidence in another genetic mouse model (Fgf23^−/−) for this local matrix mineralization inhibitory effect of OPN, and we additionally show by immunogold labeling the ultrastructural accumulation of OPN in the extracellular matrix at precisely the sites where matrix mineralization is inhibited. Taken together, these findings demonstrate that apart from the well-known effects on skeletal mineralization of changes in systemic phosphate and calcium homeostasis, that there exists another level of potent regulation at the local level residing within the activities of the noncollagenous proteins of the extracellular matrix – most notably that of OPN.

Supplementary Material

Supp Fig S1. Supplementary Figure 1.

Macroscopic phenotypes of 6-wk-old mice (A) Complete deletion of Opn from Fgf23^−/− mice results in larger, heavier and more active DKO mice compared to Fgf23^−/− littermates. (B) The body weight of DKO mice is significantly higher than that of Fgf23^−/− mice. Data are presented as Mean ± SD. *: p<0.05, **: p<0.01, ***: p<0.001vs WT; ##: p<0.01, ###: p<0.001vs Fgf23^−/−; and $: p<0.05, $$$: p<0.001 vs DKO.

Supp Fig S2. Supplementary Figure 2.

Serum carboxy terminal cross-linked telopeptide of type I collagen (CTX) in all four genotypes. Increased levels of CTX in Fgf23^−/− mice are normalized in DKO (Fgf23^−/−/Opn^−/−) mice. Data are presented as Mean ± SD. ***: p<0.001vs WT; ###: p<0.001vs Fgf23^−/−; and $: p<0.05 vs DKO.

Supp Fig S3. Supplementary Figure 3.

Measurement of serum OPN levels. Deletion of vitamin D signaling (1α(OH)ase^−/−) from Fgf23^−/− mice does not significantly decrease serum OPN levels. Data are presented as Mean ± SD. (NS: no significant difference)