The mouse pre-osteoblast MC3T3-E1 and mouse bone marrow-derived stromal ST2 cells (both from Riken Cell Bank, Japan) were maintained in standard growth medium [minimum essential medium alpha (αMEM) supplemented with 10% FBS, 100 units/ml penicillin and 100 µg/ml streptomycin, all from Thermo Fisher Scientific].

Primary calvarial osteoblasts were isolated from newborn mice as described previously (15). To investigate transcriptional induction of Nrp2 by 1,25(OH)₂D₃, cells were incubated with vehicle (0.001% ethanol) or 1,25(OH)₂D₃ (10^-8 M) in the presence or absence of actinomycin D [30 min pretreatment with actinomycin D (5 µg/ml) and 6 h treatment with 1,25(OH)₂D₃ (10^-8 M) and actinomycin D] or cycloheximide (100 ng/ml, 12 h) (Merck). To assess mineralization, primary osteoblasts were cultured until confluency in standard growth medium and then switched to osteogenic medium [α-MEM with 15% FBS, 50 µg/ml ascorbic acid (Merck), 10 mM β-glycerophosphate (Merck), 10^-8M dexamethasone (Merck), 100 units/ml penicillin and 100 µg/ml streptomycin]. After 14 days, cultures were stained with alizarin red to evaluate mineralization.

Osteoclast differentiation was examined in in vitro cultures of hematopoietic bone marrow cells incubated with exogenous macrophage colony-stimulating factor (M-CSF, R&D Systems) and receptor activator of nuclear factor κB ligand (RANKL, Peprotech). Bone marrow cells, from hind- and forelimbs of 8- to 10-week-old mice, were centrifuged on a Ficoll-Pague gradient (Stem Cell Technologies) to isolate mononuclear cells, which were plated overnight in αMEM containing M-CSF (10 ng/ml), penicillin (100 units/ml), and streptomycin (100 µg/ml). The non-adherent cells were harvested at day 1 and plated at a density of 120,000 cells per cm² in αMEM containing M-CSF (20 ng/ml), RANKL (100 ng/ml), penicillin (100 units/ml), and streptomycin (100 µg/ml). Medium was replaced after 3 days and tartrate-resistant acid phosphatase (TRAP) staining was performed at day 6 to assess osteoclastogenesis as described previously (9). At least two technical replicates were included for each independent measurement.

For cocultures of primary osteoblasts and osteoclasts, mouse bone marrow cells were isolated from femurs and tibias of 7- to 9-week-old mice by flushing the bones with standard growth medium, passing the cells through a 70 μm cell strainer and pelleting them by centrifugation at 1330 rpm for 5 min. Osteoclasts were generated by coculturing the freshly isolated bone marrow cells (200,000 cells/cm²) with primary osteoblasts (20,000 cells/cm²) in the presence of 10^-9 M 1,25(OH)₂D₃ and prostaglandin E₂ (10^-6 M). Culture medium was replaced every 2 days and after 6 days the cultures were stained for TRAP to visualize and count the osteoclasts or RNA was isolated. At least two technical replicates were performed for each independent measurement.

ChIP-seq analyses were previously performed in MC3T3-E1 cells (GSE51515 and GSE41955) (16, 17) and performed as described previously (11).

A genomic DNA fragment (mm9) of Nrp2 (Chr1: 62,768,609-62,768,872) was cloned in the pGL3-Basic luciferase reporter vector (Promega). Four nucleotides of the DR3-type VDRE in this genomic fragment were mutated with the QuickChange II Site-Directed mutagenesis kit (Stratagene) and the resulting fragment was cloned in the pGL3-Basic luciferase reporter vector. Exponentially growing MC3T3-E1 cells were transfected with Fugene 6 transfection reagent (Roche) with one of these reporter vectors (200 ng) together with the pcDNA3.1(-)/myc-His/lacZ vector (20 ng, Thermo Fischer Scientific) to assess transfection efficiency. After transfection, cells were cultured overnight and stimulated the following day with vehicle or 1,25(OH)₂D₃ (10^-8 M). After 24 h, luciferase activity was assessed and normalized to β-galactosidase activity. At least 2 technical replicates were performed for each independent measurement.

Primary calvarial osteoblasts were cultured for 48 h in standard growth medium. Prior to the experiments, medium was switched for 4 h to αMEM containing 0.1% BSA.

For adhesion experiments, 96-well dishes were coated overnight with 10 µg/ml fibronectin, washed with αMEM containing 0.1% BSA, and blocked for 2 h with αMEM containing 0.5% BSA. Primary osteoblasts (30,000 cells/cm²) were seeded on fibronection-coated wells in αMEM containing 0.1% BSA, and allowed to adhere for 30 min at 37°C. Afterwards, the 96-well dish was vigorously shaken for 10 sec on a rocking platform at 2000 rpm, and wells were washed with αMEM containing 0.1% BSA. Adherent cells were fixed for 10 min with 4% paraformaldehyde (PFA), washed with PBS, and stained for 10 min with crystal violet (5 mg/ml in 2% ethanol). Wells were washed with distilled water, air dried, incubated with 2% sodium dodecyl sulphate (SDS, Thermo Fisher Scientific) at room temperature (RT) for 30 min, and absorbance was measured at 590 nm.

For migration experiments, the underside of polycarbonate membranes of Boyden chambers (8 µm pores, Corning) was coated overnight with 2.5 µg/ml fibronectin. Afterwards, primary osteoblasts (30,000 cells/well) were seeded in the upper wells of the coated Boyden chambers in αMEM containing 0.1% BSA. Cells were allowed to migrate for 4 h to the lower compartment of the Boyden chambers, which were filled with standard growth medium. Thereafter, cells at the upper side of the carbonate filters were removed and cells at the lower side were fixed for 10 min with 4% PFA, washed with PBS, and stained for 10 min with crystal violet (5 mg/ml in 2% ethanol). Carbonate filters were washed with distilled water, air dried, incubated with 2% SDS at RT for 30 min, and absorbance measured at 590 nm. At least two technical replicates were performed for each independent measurement.

Nrp2 heterozygous and Nrp2^lox/lox mice on C57Bl/6 background were kindly provided by D. Ginty (The Johns Hopkins University School of Medicine, Baltimore, MD) and were maintained in a conventional animal facility (18). B6.Cg-Tg(Prrx1-cre)1Cjt/J and B6.129P2-Lyz2^tm1(cre)Ifo/J mice were obtained from the Jackson Laboratory. Mice with a conditional deletion of Nrp2 in osteoblast precursors and mature osteoblasts were generated by crossing paired related homeobox 1 (Prrx1)-Cre mice with Nrp2^lox/lox mice, yielding osteoblast-specific Nrp2 knockdown mice (Nrp2^Ob- mice). Mice with a conditional deletion of Nrp2 in osteoclast precursors and mature osteoclasts were generated by crossing Lysozyme 2 (Lyz2)-Cre mice with Nrp2^lox/lox mice, yielding osteoclast-specific Nrp2 knockdown mice (Nrp2^Oc- mice). Genotyping was performed by polymerase chain reaction (PCR) on genomic DNA from toe clips obtained from preweanling mice. All animal experiments were approved by the ethical committee of the KU Leuven (P090/2019).

Serum N-terminal propeptide of type I procollagen (PINP) levels and degradation products of C-terminal telopeptides of type II collagen (CTX) were determined by ELISA according to the manufacturer’s recommendations (ImmunoDiagnostic Systems).

Total RNA was isolated, cDNA prepared with oligo d(T)16 primers (Thermo Fisher Scientific), and quantitative real-time PCR (qPCR) reactions performed with 1/10 diluted cDNA templates with the Fast SYBR Green Master Mix (Thermo Fisher Scientific) or the TaqMan Fast Universal Master Mix (Thermo Fisher Scientific) as described previously (19). Relative gene expression was calculated using the 2^-ΔCt method and the reference genes β-actin and/or Hprt were used to normalize gene expression. Primers and probes were obtained from Integrated DNA Technologies and sequences are available upon request.

Total protein was isolated from primary osteoblast and osteoclast cultures. Protein isolation and Western blotting was performed as described in Verlinden et al. (9) Briefly, total protein extracts were isolated and 20 µg of protein was separated by SDS-polyacrylamide gel electrophoresis (PAGE) using 4-12% polyacrylamide gels (Thermo Fisher Scientific) and transferred to a nitrocellulose membrane (GE Health care). Membrane blocking was done in TBS (10 mM Tris-HCl, pH 7.6; 150 mM NaCl) with 5% non-fat dry milk and 0.1% Tween 20 (Merck). The rabbit monoclonal (D39A5) anti-mouse NRP2 antibody (CST) was diluted 1/500, while the mouse anti-actin (Merck) was diluted 1/5000. Incubation with a secondary antibody (Dako) was performed for 1 h at RT, and enhanced chemoluminescence (Perkin Elmer) was used to visualize protein bands.

The high resolution Skyscan 1272 system (Bruker) was used for µCT analysis with the following parameters: 60 kV as source voltage, 83 μA as source current, 0.5 mm aluminum filter, 5 µm pixel size, 0.4° rotation step. Data reconstruction was performed with NRecon (Bruker) and data analysis with CTAn (Bruker). Trabecular bone parameters were defined in a volume of interest (VOI) between 1 and 2.5 mm below the growth plate, whereas cortical bone parameters were quantified in a VOI between 2.5 to 3 mm below the growth plate. Analysis was performed according to the guidelines of the American Society for Bone and Mineral Research (20). 3D models were constructed with CTvox software (Bruker).

PFA-fixed bones were decalcified in 0.5 M ethylenediamine tetraacetic acid (EDTA) in PBS for 2 weeks at 4°C, dehydrated in graded ethanol concentrations, and embedded in paraffin. Paraffin sections (4 µm) were cut, dewaxed in xylene, and rehydrated before staining. For general morphological analysis, paraffin sections were stained in hematoxylin (Prosan) and eosin (H&E; 0.6% eosin Y, 1% phloxine B, 2% orange G, Merck) for 30 sec and 2 min, respectively. Paraffin sections were stained for TRAP activity and counterstained with Light Green SF Yellowish (Merck) to quantify osteoclasts (15). NRP2 immunostaining was performed as described previously by Verlinden et al. (9)

Bones were fixed for 24 h in 2% PFA, rinsed and stored in PBS. After decalcification, bones were embedded in paraffin and 4 µm thick sections were used for stainings. Osteoclasts and osteoblasts covering the trabecular surfaces were measured on 3 TRAP- and H&E- stained sections, respectively, and expressed relative to the total bone surface (21). In each section, three consecutive fields (0.8 mm width en 0.4 mm height) were measured along the vertical axis of the central metaphysis, starting 0.2 mm from the distal end of the growth plate as described previously by Verlinden et al. (9).

Results from in vitro experiments are expressed as mean and SEM, whereas data from in vivo measurements are expressed as mean and SD. To assess differences between two experimental groups, two-tailed Student’s t-tests were performed. To determine the effect of selective Nrp2 deletion in male and female mice, two-way ANOVA was performed with genotype and sex as two independent variables, followed by Sidak’s multiple comparisons test. To evaluate the effect of treatment with 1,25(OH)₂D₃ or the vitamin D analog, MV2, in osteoblast- or osteoclast-specific Nrp2^-/- mice, we performed two-way ANOVA analysis, with genotype and treatment as independent variables, followed by Sidak’s multiple comparisons test.

We previously demonstrated that transcript levels of the majority of SEMA3 family members are directly transcriptionally regulated by 1,25(OH)₂D₃ (11). Therefore, we investigated the expression of their co-receptor Nrp2 in osteoblasts. Nrp2 transcript levels were present in the mouse (pre-) osteoblast MC3T3-E1 and ST2 cell lines as well as in primary mouse calvarial osteoblasts and its transcript levels were induced approximately 2-fold after a 24 h incubation with 1,25(OH)₂D₃ (10^-8 M) ( Figure 1A ). Interestingly, in the presence of the transcriptional inhibitor actinomycin D, incubation with 1,25(OH)₂D₃ did not result in elevated Nrp2 transcript levels, whereas transcript levels were induced upon co-incubation of 1,25(OH)₂D₃ with the protein synthesis inhibitor cycloheximide (CHX) ( Figure 1B ). ChIP-seq analysis demonstrated direct VDR and RXR binding as well as elevated H4 acetylation in the second intron of the Nrp2 gene ( Figure 1C ). Further ChIP-qPCR analysis confirmed 1,25(OH)₂D₃-inducible association of VDR and RXR with this genomic region ( Figure 1D ). In silico analysis revealed the presence of a DR3 type VDRE in this VDR- and RXR-bound genomic region under the peak maxima aligning where the H4K5 acetylation was vacated. The functionality of this VDRE was demonstrated in transient transfection experiments. Indeed, transcription of a luciferase reporter vector driven by the VDRE-containing genomic region was significantly elevated after treatment with 1,25(OH)₂D₃, and mutation of the putative VDRE completely abrogated the 1,25(OH)₂D₃-mediated transcriptional induction of Nrp2 ( Figure 1E ).

In conclusion, these data demonstrate that osteoblasts express NRP2 and that its transcription is directly increased by 1,25(OH)₂D₃ through elevated VDR/RXR binding to a newly identified VDRE.

Our previous findings indicated that NRP2 expression is involved in normal bone homeostasis both in male (9) and female mice ( Supplementary Figure 1 ). To investigate whether the effects of NRP2 were mediated systemically or rather by NRP2 expression specifically in bone cells, we deleted Nrp2 in osteoblasts and its progenitors (Nrp2^Ob- mice) by crossing Prrx1-Cre mice with Nrp2^lox/lox mice. NRP2 expression was significantly lower in primary calvarial osteoblasts isolated from Nrp2^Ob- than from Nrp2^Ob+ mice, confirming an efficient deletion of NRP2 in osteoblasts ( Figure 2A ). In addition, Nrp2 transcript levels were significantly lower in whole bone RNA extracts from Nrp2^Ob- mice, compared to Nrp2^Ob+ mice in both male and female mice ( Figure 2B ). Because the Prrx1 promoter targets osteochondroprogenitors, we also evaluated NRP2 expression in chondrocytes and adipocytes. Immunohistochemical analysis did not demonstrate NRP2 protein expression in chondrocytes (9). qPCR analysis of white adipose tissue (WAT) indicated that Nrp2 is expressed at low levels in adipocytes, however, no difference in Nrp2 expression was observed in WAT derived from Nrp2^Ob+ and Nrp2^Ob- mice (data not shown).

Tibia length was not different between 8-week-old Nrp2^Ob- mice and their wildtype littermates (data not shown). Interestingly, µCT analysis of the tibia of 8-week-old mice demonstrated a significant reduction of trabecular bone mass (BV/TV) in male, but not in female mice ( Figure 2C ). The reduced trabecular bone mass in male mice mainly resulted from a reduction in trabecular number and a tendency towards a decreased trabecular thickness ( Figures 2C, E ). Remarkably, tibias of female and, especially, male mice at 8 weeks of age were characterized by a reduction in mean total cross-sectional tissue area and medullary area, whereas cortical thickness ( Figures 2D, E ) and cortical porosity (data not shown) were not altered. This decrease in bone diameter progressed with older age as illustrated by µCT analysis and 3D reconstructions of tibias of 26-week-old females, while no significant changes in trabecular bone parameters were observed at this age ( Supplementary Figure 2A ).

Despite the reduced trabecular bone mass in male Nrp2^Ob- mice, bone histomorphometry revealed no differences in osteoblast number or in osteoclast surface between Nrp2^Ob- mice and their wildtype littermates ( Figure 3A ). Also in female mice, histomorphometric analysis did not reveal any effect of Nrp2 deletion on osteoblast or osteoclast numbers ( Figure 3B ). Serum P1NP levels, a marker for osteoblast activity, tended to be slightly lower in male Nrp2^Ob- mice, however, this difference did not reach significance ( Figure 3C , left panel), whereas serum levels of CTx, a marker for osteoclast activity, were significantly increased in Nrp2^Ob- mice ( Figure 3C , right panel). Transcript analysis of whole bone homogenates from Nrp2^Ob- mice and their wildtype littermates did not show any differences in expression of osteoblast (osteocalcin, osteopontin) or osteoclast (cathepsin K) markers in male or female mice ( Supplementary Figures 2B, C ). To investigate whether Nrp2 deletion in calvarial osteoblasts affected their intrinsic differentiation capacity, we performed in vitro mineralization studies. However, alizarin red staining and qPCR analysis did not show any differences in mineralization efficacy of osteoblasts derived from Nrp2^Ob- mice or from their wildtype littermates ( Supplementary Figure 3 ). Finally, as NRP2 can act as a coreceptor for different semaphorin ligands, which are involved in the regulation of cell adhesion and migration, we evaluated whether osteoblastic Nrp2 deletion affected their adhesive and migratory capacity. As can be observed in Figure 3D , fibronectin coating enhanced cell adhesion and induced cell migration, however, to the same extent for Nrp2^+/+ and Nrp2^-/- osteoblasts. Similar results were obtained for vitronectin coating (data not shown).

In conclusion, these data indicate that there is a sex-specific role for osteoblastic Nrp2 expression in the control of trabecular bone mass. However, both in vitro and in vivo analyses failed to demonstrate an effect of osteoblastic Nrp2 expression on osteoblast adhesion, migration, or mineralization.

Previous research from our group demonstrated that systemic Nrp2 deficiency is associated with an increased osteoclast number in bone and that hematopoietic cells derived from Nrp2^-/- mice differentiate more efficiently into osteoclasts (9). We therefore examined whether Nrp2 deletion specifically in osteoclasts and its precursors (Nrp2^Oc- mice) affected osteoclast formation and bone homeostasis by crossing Nrp2^lox/lox mice with LysM-Cre mice. Osteoclasts, generated by culturing hematopoietic cells from Nrp2^Oc- mice in the presence of M-CSF and RANKL, did not express NRP2 transcript or protein, confirming an efficient NRP2 deletion in osteoclasts ( Figure 4A ). In addition, Nrp2 transcript levels were significantly reduced in RNA extracts of whole bone homogenates from Nrp2^Oc- mice ( Figure 4B ). However, this substantial reduction in osteoclastic Nrp2 expression did not affect bone homeostasis in male nor female mice as µCT analysis did not reveal any differences in trabecular or cortical bone mass, nor in cross-sectional bone area between 8-week-old Nrp2^Oc+ and Nrp2^Oc- mice ( Figure 4C ). Also at 16 weeks of age, no differences in bone mass between female Nrp2^Oc+ and Nrp2^Oc- mice were observed ( Supplementary Figure 4 ). Concordantly, bone histomorphometry did not show differences in osteoblast number or osteoclast surface between Nrp2^Oc+ and Nrp2^Oc- mice in females ( Figure 4D ). Rather surprisingly, transcript levels of osteocalcin were significantly elevated in female Nrp2^Oc- mice compared to their control littermates, whereas the expression of other osteoblast (osteopontin) or osteoclast (cathepsin K) markers was not different ( Figure 4E ). In contrast to the in vivo findings, in vitro osteoclast cultures showed that osteoclast numbers and expression of osteoclast markers were increased when hematopoietic cells of Nrp2^Oc- origin were used, whether cultured in the presence of M-CSF and RANKL or cocultured with calvarial osteoblasts ( Figures 5A, B ).

Collectively, these data illustrate that Nrp2 deletion from osteoclast precursors results in an elevated in vitro but not in vivo osteoclastic differentiation potential.

Since we have shown that Nrp2 is a direct vitamin D target gene, we aimed to investigate whether treatment with the vitamin D analog WY 1048, which is able to prevent bone loss in an ovariectomy-induced model of osteoporosis (22), increases bone mass to the same extent in osteoblast-specific Nrp2 knockout mice as in their wildtype littermates. Therefore, 8-week-old male Nrp2^Ob- and Nrp2^Ob+ mice were injected daily for 2 consecutive weeks with the analog WY 1048 (0.4 µg/kg/d). In both genotypes, short-term treatment with WY 1048 resulted in a significant, but comparable increase in trabecular bone mass, which was accompanied by non-significant increases in trabecular number and thickness ( Figure 6A ). Cortical thickness was not altered upon treatment with WY 1048 (data not shown). In agreement with the elevated BV/TV after treatment with WY 1048, there was a trend towards an increased bone calcium content in WY 1048-treated mice ( Figure 6B ). In addition, serum calcium levels as well as renal fractional excretion of calcium were similarly elevated in Nrp2^Ob- and Nrp2^Ob+ mice after treatment with WY 1048 ( Figure 6B ), illustrating that short-term treatment with the vitamin D analog WY 1048 has similar bone anabolic effects in Nrp2^Ob- and Nrp2^Ob+ mice, which are accompanied by a limited increase in serum and urinary calcium levels.

Because of the apparent discrepancy between the elevated in vitro osteoclast differentiation of Nrp2^Oc– derived osteoclast precursors and the normal osteoclast numbers in Nrp2^Oc- mice, we aimed to investigate bone homeostasis of Nrp2^Oc- mice in conditions of high bone resorption, induced by daily administration of supraphysiological doses of 1,25(OH)₂D₃ (1 week, 0.5 µg/kg/d). As expected, trabecular bone volume and trabecular thickness as well as cortical thickness tended to decrease upon administration of 1,25(OH)₂D₃, however these parameters were reduced to the same extent in Nrp2^Oc+ and Nrp2^Oc- male mice ( Figure 7A ). Also systemic calcium handling was similar in both genotypes as evidenced by the highly elevated serum calcium levels and urinary calcium excretion ( Figure 7B ). Similar results were obtained in female mice (data not shown). These data indicate that even in conditions of high bone remodeling, osteoclastic Nrp2 deletion does not affect bone homeostasis.