VDR in Osteoblast‐Lineage Cells Primarily Mediates Vitamin D Treatment‐Induced Increase in Bone Mass by Suppressing Bone Resorption

Long‐term treatment with active vitamin D [1α,25(OH)2D3] and its derivatives is effective for increasing bone mass in patients with primary and secondary osteoporosis. Derivatives of 1α,25(OH)2D3, including eldecalcitol (ELD), exert their actions through the vitamin D receptor (VDR). ELD is more resistant to metabolic degradation than 1α,25(OH)2D3. It is reported that ELD treatment causes a net increase in bone mass by suppressing bone resorption rather than by increasing bone formation in animals and humans. VDR in bone and extraskeletal tissues regulates bone mass and secretion of osteotropic hormones. Therefore, it is unclear what types of cells expressing VDR preferentially regulate the vitamin D–induced increase in bone mass. Here, we examined the effects of 4‐week treatment with ELD (50 ng/kg/day) on bone using osteoblast lineage‐specific VDR conditional knockout (Ob‐VDR‐cKO) and osteoclast‐specific VDR cKO (Ocl‐VDR‐cKO) male mice aged 10 weeks. Immunohistochemically, VDR in bone was detected preferentially in osteoblasts and osteocytes. Ob‐VDR‐cKO mice showed normal bone phenotypes, despite no appreciable immunostaining of VDR in bone. Ob‐VDR‐cKO mice failed to increase bone mass in response to ELD treatment. Ocl‐VDR‐cKO mice also exhibited normal bone phenotypes, but normally responded to ELD. ELD‐induced FGF23 production in bone was regulated by VDR in osteoblast‐lineage cells. These findings suggest that the vitamin D treatment‐induced increase in bone mass is mediated by suppressing bone resorption through VDR in osteoblast‐lineage cells. © 2017 The Authors. Journal of Bone and Mineral Research Published by Wiley Periodicals Inc.


Introduction
V itamin D 3 was discovered as an anti-rachitic agent that improved bone mineralization. (1) The biologically active metabolite of vitamin D 3 , 1a,25-dihydroxyvitamin D 3 [1a,25 (OH) 2 D 3 ], exerts a variety of actions on target tissues through the vitamin D receptor (VDR), a member of the nuclear receptor superfamily. (1,2) VDR is expressed in specific cell types in various organs. (2)(3)(4) Tissues expressing high levels of VDR, such as the intestines, kidneys, parathyroid glands, and bone, communicate with each other to regulate bone and mineral metabolism. (2)(3)(4) Therefore, it is not clear how much of the effect of vitamin D on bone is attributed to the direct actions of VDR in bone.
Three hormones, 1a,25(OH) 2 D 3 , parathyroid hormone (PTH), and fibroblast growth factor 23 (FGF23), regulate calcium and phosphate metabolism as an endocrine system. (2,5,6) The production and clearance of these hormones are regulated by VDR and cell-surface receptors for PTH (PTH/PTHrP-R) and FGF23 (FGFR-Klotho complexes) in a ligand-dependent manner. (5,6) 1a,25(OH) 2 D 3 increases skeletal secretion of FGF23, which reduces 1a,25(OH) 2 D 3 production in the kidneys by suppressing 1a-hydroxylase activity. (2,5,6) In contrast, PTH enhances 1a,25 (OH) 2 D 3 production by stimulating 1a-hydroxylase activity in the kidneys, which in turn represses PTH transcription through VDR in the parathyroid glands. (7)(8)(9) In addition, serum calcium levels per se directly regulate PTH secretion through the calcium-sensing This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. receptor (CaSR) in the parathyroid glands. (1,4) Thus, several negative and positive feedback loops maintain serum concentrations of these hormones, calcium, and phosphate. (5,6) Hence, a part of VDR-mediated actions in bone is included in the vitamin D-FGF23-PTH axis as the endocrine system.
The importance of VDR in bone has been validated by genetic studies of humans and mice. (10,11) VDR KO mice recapitulated rachitic skeletal phenotypes in patients with hereditary vitamin Dresistant rickets. Bone formation and mineralization were impaired in VDR KO mice. (12) However, feeding VDR KO mice with a highcalcium diet abolishes rachitic changes in bone. (13) These findings suggest that the positive effects of 1a,25(OH) 2 D 3 on bone are a consequence of upregulation of intestinal calcium absorption.
In the present study, we established two VDR cKO mouse lines in which VDR was ablated in osteoblast-lineage cells (Ob-VDR-cKO mice) and in osteoclasts (Ocl-VDR-cKO mice). Using these mouse lines, we examined the in vivo effects of ELD on bone resorption and bone formation. We found that VDR in osteoblast-lineage cells, but not osteoclasts, is critically involved in the ELD administration-induced increase in bone mass, and that VDR in osteoblast-lineage cells is involved in the regulation of serum calcium and FGF23 levels.

Animals
Osterix (Osx)-Cre Tg/0 were purchased from the Jackson Laboratory. (23) Cathepsin K-Cre Knock-in (Ctsk Cre/+ , Ctsk-Cre) mice were generated by the authors (TN, YY, and SK). (24) Ctsk-Cre mice are established mice that exhibit highly specific expression of Cre in osteoclasts. (24) VDR-floxed mice were generated by the authors (YY, YN, TN, and SK). (25) The global VDR-KO (VDR D/D ) mice bearing a germline null mutation were generated by crossing VDR-floxed mice with CMV-Cre transgenic mice. (25) Genotyping primers for the VDR gene (Fig. 1A, B) are listed in Supporting Table 1. To maintain the health of the global VDR KO mice, weaned global VDR KO mice were supplied minerals by feeding with a high-calcium diet (a standard diet [CLEA CE-2; Japan CLEA, Tokyo, Japan] supplemented with 2.0% calcium, 1.25% phosphorus, and 20% lactose). The other mice were fed a normal diet containing 1.1% calcium and 1.1% phosphorus, and 220 IU/100 g vitamin D 3 (CLEA CE-2). For mating, one male and one or two female mice were housed in a cage (177 Â 285 Â 140 mm in size). Delivered pups were raised in the same cage until they were weaned. Then, the prepubertal mice were transferred to a different cage and housed by the samegender siblings at maximum five mice per cage until they were subjected to ELD treatment. All mice were housed in a specificpathogen-free facility in Matsumoto Dental University at 24°C AE 2°C and 50% to 60% humidity with a 12-hour light/ dark cycle, and were provided with sterilized water and diets ad libitum. Our facility is not approved by the Association for Assessment and Accreditation for Laboratory Animal Care.

Administration of agents
A solution of 20 ng/mL ELD (Chugai Pharmaceutical, Tokyo, Japan) in 0.1% ethanol/medium-chain triglyceride (Nisshin Oillio, Tokyo, Japan) was prepared. The ELD solution was administered to 10week-old male mice (50 ng/kg) daily by oral gavage for 4 weeks. Researchers knew what treatment was given to the mice. The mice were randomly assigned to either vehicle treatment group or ELD treatment group. Mice given ELD or the vehicle were doublelabeled with tetracycline (20 mg/kg; Sigma, St. Louis, MO, USA) and calcein (10 mg/kg; Sigma) by subcutaneous injections at 5 days and 2 days, respectively, before being euthanized. During the entire treatment period, test animals were individually housed in a small cage (122 Â 192 Â 110 mm in size). Blood and tissue samples were collected at 24 hours after the last ELD treatment. Endpoint analysis of radiological bone morphometry and bone histomorphometry were performed in a blinded fashion.

Radiological analysis of bone
Left femurs were collected and fixed in 70% ethanol. Threedimensional (3D) reconstructions in an upside-down position were obtained for measuring the trabecular bone volume/tissue volume (BV/TV), trabecular number (Tb.N), and trabecular thickness (Tb.Th) of distal femurs with micro-CT (mCT) (ScanXmate-A080; Comscan Tecno, Yokohama, Japan) and integrated software (TRI/3D-BON; Ratoc System Engineering, Tokyo, Japan). The region of interest (ROI) for metaphyseal trabecular bone was set by hand to between 0.5 mm and 1.5 mm from the distal growth plate of femurs with deleted cortical shell. The distal femurs were scanned by mCT in air at 10.9 mm voxel resolution, 250 mA current, and 28.5 keV energy. No density calibration was performed for the mCT. Tomographic measurements of BMD and cortical morphology were performed by peripheral quantitative CT (pQCT) (XCT Research SAþ; Stratec Medizintechnik, Birkenfeld, Germany). The bone was placed horizontally inside a tube and scanned in air using following conditions: 0.07 mm voxel size, 300 mA current, and 40 keV energy. Density calibration phantoms were performed separately in pQCT analysis. The scan line was adjusted using the scout view. The image analysis was carried out using integrated XCT 620C software (Stratec Medizintechnik). The metaphyseal pQCT slices of the distal femurs at 1.5 mm and 6 mm (0.46 mm slice thickness) from the growth plate were collected for the measurement of trabecular and cortical BMD, respectively. In pQCT and mCT analyses, one sample was placed in each scan.

Bone histomorphometry
Left tibias were fixed with ethanol, undecalcified, embedded in methyl methacrylate, frontally sectioned into 5-mm slices, and subjected to Villanueva bone staining. (27) Histomorphometric analyses were performed by Ito Bone Histomorphometry Institute (Niigata, Japan). Standard bone histomorphometric nomenclature, symbols, and units, were used as described by the American Society for Bone and Mineral Research Histomorphometry Nomenclature Committee. (28) Serum biochemistry Calcium and phosphorus concentrations in serum were determined by a Calcium E-test kit (Wako, Osaka, Japan) and a Phospha-C test kit (Wako). Serum concentrations of PTH were measured using a mouse PTH 1-84 ELISA kit (Immonotopics, San Clemente, CA, USA). Serum FGF23 concentrations were quantified by a FGF23 ELISA kit (KAINOS Laboratories, Tokyo, Japan). Serum tartrate-resistant acid phosphatase 5b (TRAP5b) activities were measured using a mouse TRAP assay kit (Immunodiagnostic Systems, Boldon, UK). Serum C-terminal cross-linked telopeptide of type I collagen (CTx) concentrations were determined using a RatLaps EIA kit (Immunodiagnostic Systems). Serum procollagen type I N-terminal propeptide (P1NP) concentrations were measured using a rat/mouse P1NP EIA kit (Immunodiagnostic Systems). Serum 1a,25(OH) 2 D levels were determined using a 1,25-dihydroxyvitamin D ELISA kit (IBL international, Hamburg, Germany). ELD was not measurable in the 1a,25(OH) 2 D assay (data not shown).

Real-time RT-PCR
Tissue samples were collected, immediately soaked in TRIzol (Invitrogen, Carlsbad, CA, USA), and homogenized with TissueLyser II (Qiagen, Venlo, Netherlands). Total RNA was extracted with a Purelink RNA mini kit (Invitrogen). First-strand cDNA was synthesized from total RNA with an oligo (dT) [12][13][14][15][16][17][18] primer (Invitrogen) and ReverTra Ace reverse transcriptase (ToYoBo, Osaka, Japan) according to the manufacturers' protocols. For preparation of bone samples, tibias were isolated and the epiphysis and adherent soft tissues were cut away and rubbed off with Kimwipe papers. The cleaned tibias containing bone marrow were subjected to the total RNA extraction. Realtime RT-PCR for the quantification of mRNA expression was performed using the Fast SYBR Green (Applied Biosystems, Waltham, MA, USA) and StepOnePlus System (Applied Biosystems). The following temperature profile was used: 95°C for 20 s, followed by 40 cycles of 95°C for 3 s and 60°C for 30 s. Each expression level was calculated using a relative standard curve. Gene expression was normalized to Gapdh. Mouse primers for Vdr, 24-hydroxylase (Cyp24a1), Fgf23, osteoprotegerin (Opg), receptor activator of NF-kB ligand (Rankl), M-csf, IL-34, Sost, Gapdh, phosphate-regulating neutral endopeptidase, X-linked (Phex), and dentin matrix protein-1 (Dmp-1), are listed in Supporting Table 2. Immunohistochemical staining Harvested tissues from 3-month-old male mice were fixed in 4% paraformaldehyde in PBS, processed with serial dehydration, embedded in paraffin, and sectioned into 4-mm slices. Tibias were decalcified with 10% EDTA for 2 weeks prior to serial dehydration. The sections were subjected to staining for VDR (aVDR mAb, Clone D2K6W; Cell Signaling Technology, Danvers, MA, USA). VDR protein was visualized with aratIgG-horseradish peroxidase (HRP) (GE Healthcare, Little Chalfont, UK) and diaminobenzidine solution (Dako, Glostrup, Denmark). Some sections were also stained for TRAP and counterstained with hematoxylin. Images were obtained using an Axioplan-2 imaging microscope (Carl Zeiss, Oberkochen, Germany) and AxioVision LE64 software (Carl Zeiss).

Immunofluorescence staining
Tibias from 3-month-old male were undecalcified, frozen in hexane at -80°C using a cooling apparatus (PSL-1800; Tokyo Rikakikai, Tokyo, Japan), and embedded in a cryoembedding medium (Section-lab, Hiroshima, Japan). Eight-mm-thick sections were prepared using Kawamoto's film method (cryofilm; Section-lab) and fixed in ice-cold 5% acetic acid in ethanol. (29) The sections were subjected to staining for VDR (aVDR mAb, Clone D2K6W). VDR protein was visualized with aratIgG-HRP and a Tyramide Signal Amplification kit (PerkinElmer, Waltham, MA, USA). Nuclei were stained with Hoechst (Thermoscientific, Waltham, MA, USA). Images were obtained using a confocal laser scanning microscope (LSM 510; Carl Zeiss). Images were reconstructed from four slices of z-stacks with a step size of 1.15 mm using ZEN lite software (Carl Zeiss).

Statistics
The statistical analysis was performed using GraphPad Prism 7 statistical software (GraphPad Software, San Diego, CA, USA). The data obtained were analyzed by the D'agostino-Pearson test (n ! 8) or the Shapiro-Wilk test (n < 8) to assess normal distribution. To compare two groups, equality of the two variances was assessed using an F-test. When the data sets met both the test requirements for distribution and variance, a Student's t test was used. When the data did not meet one of the test requirements, non-parametric, Mann-Whitney U test was used. Actually, some data sets of the Ob-VDR-cKO line exhibited non-normal distribution (

Study approval
All animal studies were conducted in accordance with the guidelines for studies with laboratory animals of the Matsumoto Dental University Experimental Animal Committee.

Deletion of the VDR gene in osteoblast-lineage cells
We generated Ob-VDR-cKO (VDR fl/fl ; Osx-Cre Tg/0 ) mice by crossing VDR-floxed mice (25) with Osterix-Cre (Osx-Cre) transgenic mice. (23) The efficiency of osteoblast lineage-specific VDR inactivation in Ob-VDR-cKO mice was evaluated by DNA, RNA, and protein levels ( Fig. 1). Genomic DNA extracted from major target tissues of vitamin D was subjected to PCR analysis (Fig. 1A,  B). A 338-bp amplicon indicating correct excision of the floxed VDR exon was observed in the calvaria and long bone, but not in the duodenum, skin, or kidneys (Fig. 1B). Real-time RT-PCR analysis revealed that the expression levels of Vdr mRNA in the duodenum, skin, and kidneys were comparable between VDR +/+ ; Osx-Cre Tg/0 mice (control) and Ob-VDR-cKO mice (Fig. 1C). In contrast, the expression of Vdr mRNA in bone was much lower in Ob-VDR-cKO mice than in control mice ( Fig. 1C; also see Fig. 3A). Immunohistochemical analysis showed that VDR protein was expressed in intestinal epithelial cells, hair follicle keratinocytes and kidney tubular cells similarly in both control and Ob-VDR-cKO mice, but not in the global VDR-KO mice (Fig. 1D). In bone tissues, VDR protein was expressed in osteoblasts and osteocytes, but was not appreciably in osteoclasts or hypertrophic chondrocytes in control mice (Fig. 1E-I). Hematopoietic cells in bone marrow (BM) hardly expressed VDR protein (Fig. 1E, F, H). VDR protein in bone largely disappeared in Ob-VDR-cKO mice ( Fig. 1E). Thus, the establishment of osteoblast lineage-specific ablation of VDR was confirmed in Ob-VDR-cKO mice.
Bone phenotypes and responses to ELD in Ocl-VDR-cKO mice Both 1a,25(OH) 2 D 3 and ELD were reported to suppress M-CSF and RANKL-induced osteoclast differentiation in vitro. (30,31) This suggests that though the levels of VDR in bone marrow cells and osteoclasts are quite low, vitamin D compounds may directly act on osteoclasts in vivo. We then generated Ocl-VDR-cKO (VDR fl/fl ; Ctsk Cre/+ ) mice by crossing VDR-floxed mice with Cathepsin K-Cre Knock-in (Ctsk-Cre) mice (24) (Fig. 2). The functional impairment of VDR in Ocl-VDR-cKO mice was confirmed by assessing 1a,25 (OH) 2 D 3 -induced transcription of 24-hydroxylase (Cyp24a1) in osteoclasts generated from BM cells of Ocl-VDR-cKO mice ( Fig. 2A). Appreciable expression of VDR in osteoclasts was not observed in immunohistochemistry, but the level of 24-hydroxylase mRNA was up-regulated by 1a,25(OH) 2 D 3 in osteoclasts prepared from VDR +/+ ; Ctsk Cre/+ (control) mice ( Fig. 2A). Such upregulation, however, was markedly blunted in osteoclasts from Ocl-VDR-cKO mice ( Fig. 2A). Femurs and tibias of control and Ocl-VDR-cKO mice were subjected to radiographic and histomorphometric analysis (Fig. 2B-F; Tables 1 and 2;  Supporting Table 3). The 3D mCT images of distal femurs revealed that Ocl-VDR cKO mice had no obvious abnormalities in bone tissues (Fig. 2B). Then, ELD (50 ng/kg body weight) was orally administered every day to control and Ocl-VDR cKO mice for 4 weeks. The mCT analysis revealed that ELD treatment increased trabecular bone mass similarly in control and Ocl-VDR cKO mice (Fig. 2B). Peripheral quantitative CT (pQCT) analysis showed that the ELD treatment increased trabecular BMD similarly in both control and Ocl-VDR cKO mice ( Fig. 2C; Supporting Table 3). Consistent with a previous report, (22) cortical BMD was not affected by ELD treatment (Supporting Fig. 1; Supporting Table 3). Morphometric analysis using mCT revealed that trabecular bone volume (BV/TV) ( Fig. 2D; Supporting Table 3), and trabecular number (Tb.N) ( Table 1; Supporting  Table 3) were increased by ELD administration in control and Ocl-VDR cKO mice in a similar fashion. Trabecular thickness (Tb. Th) tended to be increased by ELD treatment, although the difference was not statistically significant (Table 1;  Supporting Table 3). Trabecular separation (Tb.Sp) was significantly decreased by ELD treatment in both genotypes  Table 1; Supporting Table 3). We previously reported that ELD treatment decreased osteoclast number (N.Oc/BS) and bone formation rate (BFR/BS). (22) N.Oc/BS and BFR/BS were suppressed by ELD treatment similarly in both genotypes (Fig. 2E, F; Supporting Table 3). Responses to ELD in other histomorphometric parameters, such as osteoblast number (N.Ob/BS), mineral apposition rate (MAR), and double-labeled surface (dLS/BS), and single-labeled surface (sLS/BS) were similar between control and Ocl-VDR-cKO mice ( Table 2; Supporting  Table 3). These results suggest that VDR in osteoclasts does not play an important role for normal bone homeostasis and vitamin D-induced increase in bone mass.

Bone phenotypes and responses to ELD in Ob-VDR-cKO mice
We then investigated the role of VDR in osteoblast-lineage cells for bone metabolism using Ob-VDR-cKO mice treated with ELD or the vehicle. The bone Vdr mRNA level in Ob-VDR-cKO mice was markedly lower than that in control mice (Fig. 3A). ELD administration did not change the expression levels of VDR in both genotypes. The 3D mCT and pQCT data of the vehicletreated groups indicated that VDR ablation in osteoblast-lineage cells affected neither BMD nor radiographic morphology (Fig. 3B-G; Supporting Table 4). However, ELD treatment increased trabecular bone mass in control but not Ob-VDR-cKO mice (Fig. 3B). ELD treatment also increased trabecular BMD by approximately 20% in control but not Ob-VDR-cKO mice ( Fig. 3C; Supporting Table 4). The cortical BMD, thickness, and area measured by pQCT were comparable among all groups (Supporting Fig. 1A-C). The pQCT data showed that marrow area was slightly decreased and cortical area/total area was slightly increased by ELD treatment in control mice (Supporting Fig. 1D, E). The morphometric analysis using mCT revealed that parameters of trabecular bone structure, such as BV/TV, Tb.N, and Tb.Th, were significantly increased and Tb.Sp was significantly decreased by ELD administration in control but not Ob-VDR-cKO mice (Fig. 3D-G; Supporting Table 4). These results suggest that VDR in osteoblast-lineage cells is largely dispensable for homeostatic control of bone structure, but is essential for the ELD-induced increase in bone mass.
ELD treatment slightly increased serum calcium levels in control but not Ob-VDR-cKO mice (Fig. 3H). Serum phosphorus levels were not altered by the ELD treatment in both genotypes (Fig. 3I). Serum PTH concentrations were significantly downregulated by the ELD treatment in both control and Ob-VDR-cKO mice (Fig. 3J). Serum levels of 1a,25(OH) 2 D were comparable between control and Ob-VDR-cKO mice, though the levels were similarly suppressed by ELD treatment in both genotypes (Fig. 3K). FGF23 levels of serum and mRNA in bone were significantly lower in Ob-VDR cKO mice than in control mice (Fig. 3L). The FGF23 levels were upregulated by ELD treatment, but the upregulation was markedly blunted in Ob-VDR-cKO mice. These results suggest that VDR in osteoblast-lineage cells is involved in FGF23 production.

Bone resorption and formation in response to ELD in
Ob-VDR-cKO mice Effects of ELD treatment on bone resorption were also assessed histologically and biochemically. TRAP-positive cells were similarly observed in both control and Ob-VDR-cKO mice given the vehicle (Fig. 4A, B; Supporting Table 4). N.Oc/BS was reduced by approximately 40% by the ELD treatment in control but not Ob-VDR-cKO mice (Fig. 4A, B; Supporting Table 4). Circulating bone resorption markers, TRAP5b and CTx, were significantly decreased by ELD treatment in control but not Ob-VDR-cKO mice (Fig. 4C, D; Supporting Table 4). We then examined the effects of ELD on mRNA expression of osteoclastogenesis-related factors in bone (Fig. 4E). RANKL and M-CSF are cytokines responsible for osteoclastogenesis. (32)(33)(34) OPG is a decoy receptor of RANKL, which inhibits osteoclastogenesis by blocking the RANKL-RANK interaction. (35) IL-34 plays a key role in osteoclastogenesis in M-CSF-deficient op/op mice, because IL-34 and M-CSF share the same receptor, c-Fms. (26,36,37) ELD treatment failed to affect the expression of Rankl, M-csf, and IL-34 mRNA (Fig. 4E). ELD   Values represent the mean AE SE. Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001; by two-tailed Student's t test except for C (one-way ANOVA followed by Dunnett's test), G (nonparametric, one-tailed Mann-Whitney U test); compared to vehicle; K (one-way ANOVA followed by Tukey's test).
increased the expression of Opg mRNA, thereby decreasing the Rankl/Opg ratio in control mice (Fig. 4E). These results suggest that VDR in osteoblast-lineage cells is required for suppression of bone resorption by ELD treatment. Effects of ELD treatment on bone formation were assessed by dynamic histomorphometry and serum biochemical parameters ( Fig. 5; Supporting Table 4). No obvious differences in the morphology, distribution, number, or function of osteoblasts were observed between control or Ob-VDR-cKO mice given the vehicle (Fig. 5A-G). Histological examination using Villanueva bone staining can identify the bone phenotype between the mature mineralized bone (unstained) and osteoid (purplish red) (Fig. 5A). ELD administration reduced the number of active osteoblasts with cuboidal shape (Fig. 5A, B, arrows) and osteoid surfaces (Fig. 5A, C, purplish red) in control but not Ob-VDR cKO mice (Fig. 5A-C; Supporting Table 4). N.Ob/BS was reduced approximately 50% by ELD treatment in control but not Ob-VDR-cKO mice ( Fig. 5B; Supporting Table 4). We then estimated bone formation by double-fluorescence labeling of the mineralized matrix (Fig. 5D-H). MAR was not affected by ELD treatment in both control and Ob-VDR-cKO mice ( Fig. 5E; Supporting Table 4). dLS/BS, sLS/BS, and BFR/BS were significantly reduced by the ELD treatment in control mice (Fig. 5F-H; Supporting Table 4). No appreciable changes in these parameters were observed by ELD administration in Ob-VDR-cKO mice. Coincidentally, the serum indicator for bone formation, P1NP, was reduced significantly by the ELD treatment in control mice only ( Fig. 5I; Supporting Table 4). The expression of Sost mRNA encoding sclerostin, an inhibitor of the Wnt signaling pathway, was upregulated by the ELD treatment in control but not Ob-VDR-cKO mice (Fig. 5J). Phex and Dmp-1 are early osteocyte markers and are regarded as upstream factors in suppressing FGF23 expression. (38,39) The expression of Phex and Dmp-1 mRNA was comparable among all groups (Fig. 5K, L). Taken The ratio of Rankl/Opg (middle) was calculated in each mouse and the data of each vehicle-treated group was set as 1. n ¼ 5 mice per group. Values represent the mean AE SE. Ã p < 0.05, ÃÃ p < 0.01; by two-tailed Student's t test except for E (nonparametric, one-tailed Mann-Whitney U test); compared to vehicle. together, these results suggest that ELD treatment increases bone mass by suppressing osteoclastic bone resorption through osteoblast-lineage VDR.

Discussion
Not only osteoblast-lineage cells, but also osteoclasts and their precursors were shown to be the targets of 1a,25(OH) 2 D 3 . (30,31) It was also reported that 1a,25(OH) 2 D 3 acted on hypertrophic chondrocytes and promoted endochondral ossification. (40,41) By using a specific and sensitive antibody, we showed here that VDR is expressed preferentially in osteoblasts and osteocytes in bone tissues. Immunohistochemical signals of VDR were faint in osteoclasts, hematopoietic cells, and several differentiation stages of chondrocytes. Osteoblast-specific deletion of the VDR gene largely abolished VDR immunoreactivity in bone. These results suggest that osteoblast-lineage cells are the major target cells of 1a,25(OH) 2 D 3 in bone. Although immunohistochemical signals were faint in osteoclasts in vivo, we have shown that 24-hydroxylase in osteoclasts was induced by 1a,25(OH) 2 D 3 . These results suggest that VDR in bone cells other than osteoblast-lineage cells plays a role in bone formation and resorption under certain conditions such as osteoporosis, osteoarthritis, or bone metastasis.
In spite of the anti-rachitic activity of vitamin D in vivo, 1a,25 (OH) 2 D 3 inhibited mineralization in cultures of osteoblastlineage cells. (42,43) 1a,25(OH) 2 D 3 was also reported to stimulate or inhibit the production of type I collagen by osteoblast-lineage cells in culture. (44,45) These observations suggest that 1a,25 (OH) 2 D 3 directly controls differentiation of osteoblast-lineage cells and mineralization of bone. The present study, however, showed that Ob-VDR cKO mice have no discernible defects in bone formation, mineralization, or serum P1NP levels in homeostasis. In agreement with our results, two different osteoblast-lineage VDR cKO mouse lines utilizing Col1a1-Cre (25) and Dmp1-Cre mice (42) showed normal bone formation. Collectively, these findings suggest that 1a,25(OH) 2 D 3 is unlikely to be a positive regulator for bone formation. The positive effect of 1a,25(OH) 2 D 3 on bone mineralization in vivo appears to be mediated by VDR in extraskeletal tissues such as the intestines, kidneys, and parathyroid glands.
Long-term treatment with 1a,25(OH) 2 D 3 , as well as ELD, suppressed bone formation in control mice. (22) This suppressive effect on bone formation appears to be an indirect action of 1a,25(OH) 2 D 3 and ELD, that is, a consequence of coupling bone resorption to bone formation. Our laboratory has recently proposed that sclerostin may be a key mediator for the coupling process. OPG KO mice exhibit stimulated bone formation coupled with excessive bone resorption (27) and coincidentally show abrogated expression of sclerostin. (46) When excessive bone resorption occurring in OPG KO mice was normalized by the treatment with bisphosphonates or anti-RANKL neutralizing antibody, expression of sclerostin was regained and the increased bone formation returned to the normal level. Indeed, ELD treatment increased Sost mRNA expression in bone of control but not Ob-VDR-cKO mice. However, ELD failed to increase Sost mRNA expression in primary cultures of osteoblastlineage cells (Supporting Fig. 2), suggesting that the upregulation of Sost mRNA expression in vivo was mediated by factors other than ELD. These findings suggest that the suppression of bone formation by long-term treatment with 1a,25(OH) 2 D 3 or ELD is due to the consequences of coupling (Fig. 6). Studies are ongoing in our laboratory to elucidate the mechanism by which bone resorption suppresses Sost mRNA expression in osteocytes.
Suppression of bone resorption induced by long-term treatment with ELD occurred through VDR in osteoblast-lineage cells but not in osteoclasts (Fig. 6). Among osteoclastogenesisrelated factors (RANKL, OPG, M-CSF, and IL-34) examined, Opg expression was increased by the ELD treatment in control but not Ob-VDR cKO mice. The RANKL/OPG ratio is believed to be one of the most important factors that regulate osteoclastogenesis in vivo. Actually, ELD treatment decreased the RANKL/OPG ratio in control mice. However, we could not reproduce in vitro downregulation of the RANKL/OPG ratio in cultures of osteoblast-lineage cells. Therefore, the change of the RANKL/OPG ratio by the ELD treatment does not appear to be a direct action of VDR on OPG or RANKL promoters. Bone microenvironment and endocrine systems may be required for appropriate osteoclastogenesis. Expression of RANKL is upregulated by PTH, as well as 1a,25(OH) 2 D 3 . (32) The reduced PTH levels by the ELD treatment may decrease the RANKL/OPG ratio in osteoblast-lineage cells of control mice. Besides RANKL and OPG, several factors produced by osteoblast-lineage cells, such as semaphorin 3A, (47) Wnt 5A, (48) and EphB4, (49) modulated differentiation and function of osteoclasts. Expression of these factors may be changed by the vitamin D treatment. Further studies are required to assess the possible involvement of such modulators in the actions of vitamin D on bone resorption.
In this study, we showed genetic evidence that VDR in osteoblast-lineage cells has a pivotal role for FGF23 production. Upregulation of serum levels of FGF23 by ELD treatment was much greater in control than Ob-VDR-cKO mice. Nevertheless, Fig. 6. Hypothetical mechanisms of long-term vitamin D treatmentinduced increase in bone mass. Long-term treatment with vitamin D compounds decreases PTH production in parathyroid glands and increases FGF23 production in osteocytes through VDR. FGF23 but not PTH secretion is affected by VDR ablation in osteoblast-lineage cells. The persistent working of the vitamin D-FGF23-PTH axis likely affects bone microenvironment and decrease the RANKL/OPG ratio and bone resorption. Reduced bone resorption increases sclerostin expression in osteocytes through an unknown coupling factor secreted from osteoclasts and gives rise to suppression of bone formation, eventually leading to a net increase in bone mass. These sequential events are conceivably caused by active vitamin D-induced persistent activation of VDR in osteoblast-lineage cells. serum phosphate levels were unchanged in control and Ob-VDR-cKO mice treated with ELD or the vehicle. PTH, the production of which is regulated by FGF23 and vitamin D, is also known to reduce serum phosphate levels by decreasing renal phosphate reabsorption. (5,6) Therefore, the vitamin D-FGF23-PTH interaction may be regulated to maintain phosphate homeostasis.
There are four limitations in our study. The first is the limitation of statistical assessment. Unequal variances and nonnormal distribution were observed in some data sets. That is probably due to an insufficient number of samples. The second is radiological assessment. Due to our insufficient skills for the mCT analysis, the morphology of cortical bone was evaluated using pQCT. Biomechanical data of cortical bone were not shown in this study. The third is immunohistochemistry. VDR was clearly detected in osteoblasts and osteocytes but not other cells in bone. However, it is possible that cells expressing VDR at levels below the detection threshold respond to ELD. The fourth is experiments using cKO mice. Ob-VDR-cKO mice expressed Vdr in bone at 10% transcript level of control mice. This may be due to the incomplete deletion of VDR in osteoblast-lineage cells or VDR expression by bone cells other than osteoblast-lineage cells. In addition, VDR expression in tissues that we did not examine may be downregulated in Ob-VDR-cKO mice. Therefore, we admit the possibility that some of our conclusion may include overestimation or underestimation of roles of VDR in osteoblast-lineage cells and other bone cells. Further experiments using different VDR cKO lines would be essential to exclude this possibility.
This study raises a new question about mechanisms underlying the lack of calcemic effect of ELD in Ob-VDR-cKO mice. Bone formation was suppressed by ELD in control mice but not in Ob-VDR-cKO mice. Accordingly, consumption of blood calcium for bone formation may be decreased by ELD treatment in control mice and leads to mild hypercalcemia in control mice only. Dysregulation of calcemic action and FGF23 production in Ob-VDR-cKO mice may be related to different responsiveness to ELD in sclerostin expression and the RANKL/OPG ratio from control mice (Fig. 6). An unbiased approach, comparison of metabolome between the two genotypes treated with ELD, would be required for understanding the overall mechanisms of the ELD actions.
In conclusion, VDR in osteoblast-lineage cells is a key mediator of ELD actions for increasing bone mass. 1a,25(OH) 2 D 3 and ELD act on VDR in multiple cell types in extraskeletal tissues as well and possibly induce unfavorable side effects. There are clinically and experimentally used agents that are delivered specifically to bone. (50,51) The development of a method for bone-specific delivery of active derivatives of vitamin D will bring a promising strategy for prevention and treatment of osteoporosis.

Disclosures
NT has received a grant from Chugai Pharmaceutical Co., Ltd. The remaining authors state that they have no conflicts of interest.