In this study, we used the dentin matrix protein 1 promoter-diphtheria toxin receptor (DMP1-DTR) transgenic (Tg) mice established by Tatsumi et al. (32). Injecting these mice with diphtheria toxin (DT) achieves in vivo inducible and specific ablation of osteocytes. Ten-week-old DMP1-DTR Tg mice and wild-type (WT) mice as littermate controls (Cont) were maintained at 23°C on a 12-h light/dark cycle. The mice had unlimited access to water and a standard rodent diet. A metabolic cage was used to measure body weight, and to collect blood, urine, and feces. All animal studies were performed in accordance with the guidelines for the care and handling of laboratory animals and were approved by the animal care committee of Tokushima University.

For the dietary regulation experiments, 10-week-old male DT-injected mice (OCL and Cont mice) were placed on one of the following two isocaloric diets for 7 days: (1) control-Pi (CP) 0.6% Pi and (2) high-Pi (HP) 1.2% Pi (34, 35). For analysis of the osteocyte network-disrupted mice, we used granulocyte colony-stimulating factor (G-CSF)-injected mice. Injecting the mice with G-CSF induced hematopoietic stem/progenitor cell mobilization (36). Disruption of the osteocyte network was induced as described previously (37). Male C57 BL/6 mice were injected with recombinant human G-CSF (Filgrastim, Kyowa Kirin, 250 mg/kg body weight/day subcutaneously every 12 h in 10 divided doses unless otherwise indicated) in phosphate-buffered saline supplemented with 0.1% bovine serum albumin, and maintained for 1 day after administering the last dose of G-CSF for 24-h urine collection (37).

DMP1-DTR Tg mice were created based on the diphtheria TRECK system (32). The mouse DMP1 gene fused to human HB-EGF (DTR) cDNA with a polyadenylation signals were constructed. The DMP1-DTR Tg mice were mated with WT mice and the offspring were genotyped by polymerase chain reaction (PCR) targeting the inserted gene using genomic DNA extracted from the tail. The following primers were used: 5′-GGGTCCCTCTTCTTCCCTAGC-3′ (forward) and 5′-GTATCCACGGACCAGCTGCTAC-3′ (reverse) (200 bp product for the transgene). Injecting DMP1-DTR Tg mice with DT resulted in targeted and inducible ablation of osteocytes. DT (50 µg/kg body weight)-treated DMP1-DTR Tg mice were used as OCL mice (Figure 1A).

Intestinal Ca²⁺ absorption was assessed in 8-week-old mice by measuring serum ⁴⁵Ca²⁺ at early time-points after oral gavage. Mice were fasted 12 h before the test. Animals were hemodynamically stable under anesthesia (urethane 1.4 mg/g body weight) during the entire experiment. The test solution contained 0.1 mM CaCl₂, 125 mM NaCl, 17 mM Tris, and 1.8 g/l fructose, and was enriched with 20 μCi ⁴⁵CaCl₂/ml for oral tests (Cont and OCL mice) (38). For oral tests, 15 µl/g body weight was administered orally by gavage. Blood samples were obtained at the indicated time intervals. Serum (10 µl) was analyzed by liquid scintillation counting (38). Changes in the plasma calcium concentration (Δμmol) were calculated from the ⁴⁵Ca content of the plasma samples and the specific activity of the administered calcium (38).

Total RNA was isolated from bone, liver, proximal intestine, and distal intestine with ISOGEN (Wako, Osaka, Japan). The proximal intestine refers to the duodenum and the proximal part of the jejunum. The distal intestine refers to the distal part of jejunum and ileum (39). For quantitative reverse transcription (RT)-PCR, total RNA (1 µg) was reverse-transcribed using Moloney Murine Leukemia Virus Reverse Transcriptase (Invitrogen), and samples were analyzed using the Applied Biosystems^®StepOnePlus Real-Time PCR system (Thermo Fisher Scientific Inc., Japan). The primers and product size used are shown in Table S1 in Supplementary Material. The amount of target mRNA was normalized to that of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA. RT-PCR was performed using SYBR^® Premix Ex Taq II (Tli RNaseH Plus) (TAKARA, OSAKA, Japan). The PCRs contained 1 µl of cDNA (equivalent to 50 ng of total RNA), 2× SYBR^® Premix Ex Taq II and 400 nM specific primers in a total of 25 µl. Relative expression values were evaluated with the 2^−ΔΔCt method. Data were normalized to GAPDH and pooled from three independent experiments.

Plasma, feces, and urinary Pi and calcium and serum blood urea nitrogen (BUN) were determined by the Phospha-C test, Calcium-E test, or BUN-B test (Wako Pure Chemical Industries, Ltd., Osaka, Japan), respectively (40). To measure the fecal Pi and calcium levels, the collected feces (24 h) were dried at 110°C for 24 h, 250°C for 3 h, and at 350°C for 3 h. FGF23 and PTH concentrations were determined using an FGF23 ELISA kit (Kainos Laboratories, Inc., Tokyo, Japan) and mouse PTH 1-84 ELISA kit (Immutopics, CA, USA), respectively (40). 1,25(OH)₂D levels were measured using a radioreceptor assay (SRL, Inc., Tokyo, Japan). Heparinized mixed arterial–venous blood was collected and analyzed immediately for pH, blood gases, and electrolytes using an OPTI CCA TS blood gas analyzer (Sysmex Corporation, Kobe, Japan).

Eight- to ten-week-old Tg mice and their WT littermates were perfused with 4% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4). Femurs and tibiae were immersed in the same fixative for 12 h prior to decalcification with 10% EDTA (pH 7.4) for 2 weeks. The specimens were embedded in paraffin and subjected to histochemistry with hematoxylin and eosin staining and immunohistochemical analysis using DMP1 antibody (TAKARA, Kyoto, Japan).

Brush border membrane vesicles were prepared from kidney and distal intestine using the Ca²⁺ precipitation method, and used for immunoblot analysis as previously described (34, 41). Levels of leucine aminopeptidase, Na⁺-K⁺-ATPase, and cytochrome c oxidase were measured to assess membrane purity. Uptake of ³²P into the BBMVs was measured by the rapid filtration technique (34, 41).

Protein samples were in denatured with 2-mercaptoethanol and subjected to 8 or 10% SDS-PAGE. The separated proteins were transferred by electrophoresis to Immobilon-P polyvinylidene difluoride (Millipore, Billerica, MA, USA) and then treated with the following diluted antibodies. Immunoblot analyses were performed using the following primary antibodies: affinity-purified anti-Npt2a (1:4,000) (34), anti-Npt2c (1:3,000) (41), and anti-Npt2b (1:2,000), as described previously (42). Anti-Klotho (for mouse total lysate; Trans Genic Inc., Fukuoka, Japan) was used following the manufacturer’s instructions. Mouse anti-actin monoclonal antibody (Chemicon, Temecula, CA, USA) was used as an internal control. Horseradish peroxidase-conjugated anti-rabbit or anti-mouse IgG was utilized as the secondary antibody (Jackson Immuno Research Laboratories, Inc., West Grove, PA, USA), and signals were detected using Immobilon Western (Millipore). Membranes were exposed to standard X-ray film and densitometric quantification was performed using ImageJ software (National Institutes of Health, Bethesda, MD, USA). All experiments were repeated at least five times.

Immunohistochemical analyses of rat kidney sections were performed as described previously with minor modifications (43). Specimens were embedded in paraffin and subjected to immunohistochemistry for affinity-purified Npt2a antibodies or affinity-purified Npt2c antibodies. Sections were then treated with Envision (+) rabbit peroxidase (Dako, Carpinteria, CA, USA) for 30 min at room temperature. Immunoreactivity was detected by incubating the sections with 0.8 mM diaminobenzidine (43). Masson trichrome staining was observed using a Trichrome Stain (Masson) Kit (Sigma-Aldrich, St. Louis, MO, USA) following the manufacturer’s instructions. Von kossa stain performed in OCL and klotho-deficient mice (43). Von Kossa staining for mineral deposition in the kidney was performed by applying 5% silver nitrate to the renal sections and exposing the sections to bright light for 30 min (44). Paraffin sections were counterstained with hematoxylin for evaluation of tissue and cell morphology.

Data are expressed as mean ± SEM. Statistical analysis was performed using the Student’s t-test. We evaluated differences between the two groups by an unpaired t-test and differences among multiple groups were analyzed by ANOVA followed by Tukey’s post-test for multiple comparisons. All computations were performed using GraphPad Prism 5.0 Software. In all experiments, differences were considered statistically significant at P < 0.05.

We investigated the physiologic function of osteocytes in OCL mice. Expression of DT receptors on osteocytes was targeted specifically using the DMP1 promoter (32). A single intraperitoneal injection of DT into DMP1-DTR Tg mice killed osteocytes harboring the DT receptor and resulted in a number of empty lacunae containing no osteocytes, as reported previously (32). The OCL mice exhibited a non-significant tendency toward decreased food intake, consistent with a previous report (45). Histochemical analysis of the cortical bone revealed a low abundance of osteocytes compared with Cont mice (32). Consistent with previous findings (9), OCL mice exhibited a decreased osteocyte number at 7 days after DT injection (Figures 1A,B). After ablated osteocytes, bone resorption was increased, bone formation and mineralization are suppressed, resulting in osteoporosis (32).

We also measured the expression of Pi regulatory genes in the osteocyte-rich fractions as described previously (32). The mRNA levels of DMP1 (P < 0.01), FGF23 (P < 0.01), matrix extracellular phosphoglycoprotein (MEPE) (P < 0.05), and phosphate-regulating endopeptidase homolog X-linked (Phex) (P < 0.05) as Pi regulatory factors were significantly decreased in OCL mice. SOST (P < 0.05) and E11/GP38 (P < 0.05) as osteocyte markers were also significantly decreased in OCL mice (Figure 1C). On the other hand, osteoblast markers were not changed in OCL mice compared with Cont mice. Compared with DT-injected WT mice, DT (50 µg/kg, DT50)-injected Tg mice showed no changes in plasma Pi, calcium, glucose, pH, BUN, Na⁺, K⁺, and Ca²⁺ (Figure 1D). In addition, urinary Ca/Cr, Na/Cr, and K/Cr concentrations were not different between DT50 mice and Cont mice (Figure 1E). By contrast, OCL mice exhibited prominent hyperphosphaturia (P < 0.01) (Figure 1E). Food consumption by the OCL mice was slightly lower than that of Cont mice. Food consumption by the Cont mice was adjusted to the amount in OCL mice. Calcium and Pi levels in the feces were normalized to the amount of diet consumed. The Pi content in the feces was reduced by 56.9% that in the Cont mice (P < 0.01) (Figure 1F). The calcium content in the feces was 45.4% higher than that in Cont mice (P < 0.05) (Figure 1G). These findings suggest that renal Pi excretion and intestinal Pi and calcium absorption may be abnormal in OCL mice. In particular, it is expected that the increased Pi absorption in the small intestine influences systemic Pi metabolism.

To investigate the initial triggers for the increase in renal Pi excretion, we analyzed the time-course of the changes in the Pi excretion levels, plasma PTH, FGF23, and renal Klotho protein after DT treatment. Renal Pi excretion was significantly increased at 5 days after DT injection (3.45-fold, P < 0.01) (Figure 2A). Plasma FGF23 levels first decreased at 1 day (32.4%, P < 0.01) and 3 days (32.4%, P < 0.01) after injection of DT and then recovered (Figure 2B). By contrast, plasma PTH levels were significantly increased at 9 days (1.52-fold, P < 0.01) (Figure 2C). Renal Klotho levels were significantly (75.5%) decreased at 5 days after DT injection (P < 0.01) (Figure 2D). These data suggest that the beginning of the increase in Pi excretion is not consistent with that of the increase in plasma PTH and FGF23 levels. We analyzed renal fibrosis (Figure 2E). High dietary Pi loading increases renal fibrosis (46). In OCL mice, kidney samples were obtained at 7 days after DT injection. The kidney samples obtained from OCL mice were stained with Masson’s trichrome to detect total collagen deposits (Figure 2E), which revealed interstitial fibrosis and increased renal Pi excretion independent of plasma PTH and FGF23 levels. The kidney calcification observed in Klotho-deficient mice was not detected in OCL mice (Figure 2F).

We then investigated if DT injection induced dose-dependent changes in the osteocyte number and renal Pi excretion (32). At 5 days after DT (0, 15, or 50 µg/kg BW) injection, we measured Pi excretion and renal Klotho and 1,25(OH)₂D levels (Figure 2G). At 5 days after DT injection, the number of osteocytes in the cortical bone decreased to 33.5 ± 1.7% (DT15) or 75.3 ± 4.2% (DT50) that in the DT0 mice (P < 0.01) (Figure 2G). The plasma 1,25(OH)₂D levels did not change after DT injection (Figure 2H). Urinary Pi excretion was significantly increased at 5 days after DT50 injection (P < 0.01) (Figure 2C). By contrast, DT50 injection markedly suppressed renal Klotho expression (P < 0.01) (Figure 2D). These data suggest that the increased renal Pi excretion was dependent on the number of osteocytes ablated.

To examine whether the expression of renal Pi transporters was affected in OCL mice, Npt2a and Npt2c mRNA and protein levels were investigated (Figure 3). Npt2a and Npt2c mRNA levels were significantly decreased in OCL mice (P < 0.01) (Figure 3A). PiT2 mRNA levels were not changed in OCL mice compared with Cont mice (Figure 3A). 1α(OH)ase mRNA levels were markedly increased in OCL mice (4.9-fold, P < 0.01), whereas 24(OH)ase mRNA expression levels were not changed (54.4% decreases, P < 0.01) (Figure 3A). Klotho mRNA expression levels were significantly (44.4%) decreased in OCL mice (P < 0.01) (Figure 3A).

The mRNA levels of the transcription factor early growth response 1 (Egr-1) were significantly increased in OCL mice (2.4-fold, P < 0.01), whereas FGF receptor 1 (FGFR1) mRNA levels were not changed. The mRNA expression levels of FGF receptors 3 (FGFR3) were significantly increased (2.1-fold, P < 0.01) and FGF receptors 4 (FGFR4) were significantly decreased (34%, P < 0.01) (Figure 3A). Npt2a and Npt2c protein levels were markedly decreased in OCL mice (P < 0.01, P < 0.05) (Figure 3C). In addition, sodium-dependent Pi cotransport activity in the BBMVs was also significantly reduced in OCL mice compared with Cont mice (P < 0.01) (Figure 3D). Immunohistochemical analysis revealed reduced Npt2a and Npt2c immunoreactivity in the proximal tubular cells (Figure 3B). Intestinal Npt2b protein levels were significantly increased in OCL mice compared with Cont mice (fourfold, P < 0.01) (Figure 3E). Intestinal Pi transport activity in OCL mice was 4.2-fold higher than that in the Cont mice (P < 0.01) (Figure 3F).

We then investigated calcium absorption in the small intestine. In contrast to intestinal Pi absorption, the calcium content in the feces in OCL mice was 45.4% higher than that in Cont mice (P < 0.05) (Figure 1G).

Oral calcium absorption analysis using ⁴⁵Ca²⁺ solution indicated that calcium absorption (plasma ⁴⁵Ca²⁺) levels were significantly (62–69%) decreased in OCL mice (P < 0.01) (Figure 4A). Intestinal transient receptor potential vanilloid 6 (TRPV6), calbindin-D9k (CaBP D9k), and plasma membrane calcium ATPase 1b (PMCA1) mRNA levels, however, were not altered in OCL mice (Figure 4B). On the other hand, mRNA expression levels of claudin 2, which regulates intestinal Ca²⁺ absorption via paracellular pathways, were dramatically (79.1%) decreased in OCL mice (P < 0.01) (Figure 4C). Although claudin 12 mRNA expression levels were not changed in OCL mice, claudin 15 mRNA expression levels were decreased in OCL mice (30.4% decreases, P < 0.05). Thus, intestinal calcium absorption was significantly decreased in OCL mice. On the other hand, renal excretion of calcium was not affected in OCL mice. In the kidney, calcium-sensing receptor mRNA levels were not changed and TRPV5 mRNA expression was significantly (1.7-fold) increased in OCL mice (P < 0.01) (Figure 3A).

We investigated other mouse models of osteocyte network disruption. Eight-week-old C57BL/6 mice were administered recombinant human G-CSF (250 µg/kg body weight/day) and analyzed 1 day after the last dose of G-CSF (Figure 5A). In the G-CSF injected mice, hematopoietic stem cells were removed from the medullary cavity (data not shown). RNA was extracted from the tibias and femurs of vehicle- and G-CSF-treated mice after flushing out the bone marrow. Gene expression was assessed by quantitative RT-PCR and the results were normalized to GAPDH mRNA. In G-CSF injected mice, Col1a1 (64.6%, P < 0.01), osteocalcin (53.8%, P < 0.01), and osteopontin mRNA (58.8%, P < 0.05) as osteoblastic markers were decreased; and FGF23 (83.8%), DMP1 (54.8%), and Phex (62.4%) as osteocytic markers were also decreased (P < 0.01), suggesting that differentiation of osteoblasts into osteocytes was suppressed by G-CSF injection (Figure 5C). Immunohistochemical analysis using DMP1 antibody showed that the length and number of canaliculi were markedly decreased in the femoral bone of G-CSF treated mice (Figure 5B). Remarkably, canaliculi did not reach the surface of the bone (Figure 5C). Therefore, these results indicate that the osteocytic canalicular network was disrupted. The serum Pi concentration did not significantly differ between the G-CSF-injected mice and vehicle-injected mice (Figure 5D) and was similar to that in the OCL mice (Figure 1D). Urinary Pi levels, however, were significantly higher (2.2-fold, P < 0.05) in the G-CSF-injected mice than in the vehicle-injected mice (Figure 5D), same as in OCL mice (Figure 1E). In the renal BBM of G-CSF mice, Npt2a and Npt2c levels were significantly decreased (P < 0.05) (Figure 5E), similar to the renal Npt2a and Npt2c protein levels in OCL mice (Figure 3C). In G-CSF-injected mice, plasma iPTH and FGF23 levels did not differ significantly from those in the vehicle-injected mice (Figure 5F). On the other hand, renal Klotho protein levels were decreased in G-CSF-injected mice (P < 0.05) (Figure 5F). Thus, mice with G-CSF-induced osteocyte network disruption showed hyperphosphaturia independent of the plasma PTH and FGF23 levels.

The increased renal Pi excretion in OCL and G-CSF mice may be due to the increase in Pi released from the bone and intestinal Pi absorption. Therefore, we investigated whether feeding mice a high Pi diet increases urinary Pi excretion. Figures 6A,G show the schedule for dietary phosphate loading in OCL and G-CSF mice. After feeding on a high Pi diet, Pi excretion was significantly increased in the Cont mice (P < 0.01), but not changed in the OCL and G-CSF mice (Figures 6B,H). After feeding on a high Pi diet, plasma Pi and calcium levels were not changed in the OCL mice (Figure 6C).

In Cont mice, feeding on a high Pi diet significantly reduced renal Npt2a and Npt2c protein levels (P < 0.01) and Na/Pi uptake in BBMs (P < 0.01), whereas no changes were observed in OCL mice (Figures 6D,E). In the mice fed a high Pi diet, the levels of plasma intact PTH were not changed in OCL mice, but significantly increased in Cont mice (1.9-fold, P < 0.01) (Figure 6F). The increase in plasma PTH levels and serum FGF23 levels induced by a high Pi diet in Cont mice was not observed in OCL mice. Therefore, for dietary Pi adaptation, signaling between the intestine and parathyroid gland, and between the intestine and kidney may be disturbed in OCL mice.

We performed DNA microarray analysis of the intestine, liver, and kidney in OCL mice (data not shown). In these analyses, we focused on bile acid metabolism, a commonly changing metabolic system, in OCL mice and G-CSF-administered mice, because we did not detect prominent differences in the expression of genes related to glucose, amino acid, other nutrient metabolic pathways. Bile acid metabolism was markedly disturbed in the intestine and liver in OCL mice compared with Cont mice (Figure S1 in Supplementary Material). In a previous study, Sato et al. reported that osteocytes regulate fat metabolism, that OCL mice lack visible white adipose tissue, and plasma leptin levels decrease in association with fat loss (45). Bile acids serve as ligands for the nuclear receptor farnesoid X receptor (FXR) (46, 47). Bile acids act on FXR in ileal enterocytes to induce the expression of FGF15 (47, 48). FGF15 also stimulates gallbladder filling (47, 48). FGF15 inhibits bile acid synthesis by repressing the transcription of Cyp7al, which encodes the first and rate-limiting enzyme in the classic bile acid synthesis pathway (43). FGF15 acts through a cell surface receptor complex composed of the FGF receptor 4 and β-Klotho (49, 50). In liver, transcription of the small heterodimer partner (SHP) gene is induced by bile acids via FXR, and SHP, in turn, binds to the CYP7A1 promoter to repress gene transcription (47). SHP is required for FGF15 to efficiently repress bile acid synthesis (48). In OCL mice compared with Cont mice, transcripts of FXR (liver: P < 0.05, distal intestine: P < 0.01), Cyp7A1 (P < 0.05), FGF receptor 4 (FGFR4) (P < 0.05), β-Klotho (P < 0.01), SHP (distal intestine; P < 0.01), and FGF15 (P < 0.01), the concentrations of serum bile acid (P < 0.05), and feces bile acids (P < 0.05) intestinal bile acids (P < 0005) were significantly increased in the liver and/or distal intestine compared with Cont mice (Figures S1B–D in Supplementary Material). Intestinal vitamin D receptor, calcium-sensing receptor, and parathyroid hormone 1 receptor mRNA levels were not changed in OCL mice (data not shown).

Next we investigated whether OCL mice are in a high Pi state. Lipid accumulation in white adipocytes was decreased in OCL and high Pi diet-fed mice (Figures S1A and S2A in Supplementary Material). We observed alterations in the genes involved in bile acid metabolism (P < 0.01) (Figure S1 in Supplementary Material). Quantitative PCR analysis revealed that, compared with Control Pi-fed mice (CP mice), high Pi-fed mice (HP mice) had significantly lower transcript levels of Cyp7A1 (P < 0.01), Cyp8b1 (P < 0.01), FGFR4 (P < 0.05), and β-Klotho (P < 0.05) in the liver. FXR (P < 0.01), OST-α (P < 0.05), and FGF15 (P < 0.01) levels in the distal intestine were also significantly decreased in HP mice (Figures S2B,C in Supplementary Material). On the other hand, the concentrations of serum bile acid and feces bile acids were not changed in the liver or distal intestine in HP mice compared with CP mice (Figure S2D in Supplementary Material). The serum triglyceride and cholesterol concentrations were lower in HP mice than in CP mice (Figure S2D in Supplementary Material).