SLPI is a critical mediator that controls PTH-induced bone formation

Osteoclastic bone resorption and osteoblastic bone formation/replenishment are closely coupled in bone metabolism. Anabolic parathyroid hormone (PTH), which is commonly used for treating osteoporosis, shifts the balance from osteoclastic to osteoblastic, although it is unclear how these cells are coordinately regulated by PTH. Here, we identify a serine protease inhibitor, secretory leukocyte protease inhibitor (SLPI), as a critical mediator that is involved in the PTH-mediated shift to the osteoblastic phase. Slpi is highly upregulated in osteoblasts by PTH, while genetic ablation of Slpi severely impairs PTH-induced bone formation. Slpi induction in osteoblasts enhances its differentiation, and increases osteoblast–osteoclast contact, thereby suppressing osteoclastic function. Intravital bone imaging reveals that the PTH-mediated association between osteoblasts and osteoclasts is disrupted in the absence of SLPI. Collectively, these results demonstrate that SLPI regulates the communication between osteoblasts and osteoclasts to promote PTH-induced bone anabolism.

S keletal tissue undergoes dynamic remodeling throughout the life course. Bone formation by osteoblasts and bone resorption by osteoclasts occur in spatially and temporally discrete units on the bone surface, and bone resorption is followed by bone formation 1 . The activities of osteoblasts and osteoclasts are linked by intercellular signaling for balanced bone remodeling. Given that bone formation and resorption are tightly coupled processes in the bone microenvironment, information concerning the spatiotemporal distribution of the cells involved in these processes would provide insight into bone remodeling. Previously, we visualized osteoblasts and osteoclasts by intravital multi-photon microscopy and revealed that direct contact with osteoblasts inhibited the bone-resorbing activity of osteoclasts in vivo 2 . Moreover, osteoblasts and osteoclasts were segregated in a homeostatic state, whereas intermittent parathyroid hormone (PTH) treatment dynamically altered their distribution and increased their direct cell-cell interactions, thereby inhibiting bone resorption by osteoclasts.
PTH is a major endocrine inducer and stimulator of bone remodeling. PTH (1-34) (teriparatide), a biologically active Nterminal fragment of human PTH, is used clinically to increase bone mass in osteoporosis 3,4 . PTH stimulates osteoblasts, and indirectly activates osteoclasts by regulating receptor activator of NF-κB ligand (RANKL) and osteoprotegerin in osteoblast-lineage cells 5,6 . It has been proposed that the anabolic action of PTH requires the presence of osteoclasts. In a murine model of osteopetrosis caused by a lack of osteoclasts, the anabolic effect of PTH treatment was blunted 7,8 . In the clinic, concomitant treatment of osteoporosis with alendronate, an inhibitor of osteoclasts, significantly decreased the anabolic response to PTH 9,10 . These results imply that the shift from osteoclasts to osteoblasts contributes to the osteoanabolic effect of PTH 11 . However, the mechanisms that coordinately regulate these two cell types in PTH anabolism are unknown.
Secretory leukocyte protease inhibitor (SLPI) is a serine protease inhibitor, secreted in response to tissue damage that plays multiple roles in inflammation in vivo. SLPI inhibits inflammation and initiates the healing of damaged tissue 12 . These functions are dependent on its protease inhibition 13,14 and transcriptional regulation 15 . Therefore, SLPI serves as a molecular switch from inflammation to tissue proliferation 16 .
Although expression of SLPI in osteoblasts has been reported 17 , the functional role of SLPI in bone metabolism in vivo, particularly in bone resorption, has not been described. In the present study, we found that PTH strongly induced Slpi expression in osteoblasts, and the effect of PTH on bone was dependent on SLPI. Moreover, SLPI expressed by osteoblasts promoted direct cell-cell contact, thereby shifting the balance from osteoclastic activity to osteoblastic activity. Therefore, SLPI is an essential mediator that can promote a PTH-mediated switch from bone resorption to bone formation.

Results
Osteoblast Slpi expression is induced by PTH in vivo. The anabolic effect of PTH depends on its binding to the PTH receptor (PTH1R) on osteoblast-lineage cells 11 . Three weeks of PTH administration induced merged distribution of osteoblasts and osteoclasts ( Supplementary Fig. S1a). In addition, the responses of osteoblasts and osteoclasts to PTH treatment in femoral trabecular bone were compatible to those in their calvarial counterparts ( Supplementary Fig. S1b). To assess gene expression changes in response to PTH, we first isolated osteoblasts with high purity from adult bone tissues of transgenic mice, in which mature osteoblasts specifically express enhanced cyan fluorescent protein (ECFP) driven by the type I collagen promoter (i.e., Col2.3-ECFP mice) 2 . Using flow cytometry, we detected the lineage (Lin − ) CD45 − ECFP + after sequential enzymatic digestion, but not from non-adherent bone marrow cells ( Supplementary Fig. S1c). The ECFP + cells expressed high levels of Sp7 and Bglap, and so were identified as osteoblasts ( Supplementary Fig. S1d).
We next evaluated transcripts in osteoblasts by RNA-Sequencing (RNA-Seq), using a PTH mouse model 18 treated intermittently with PTH or vehicle for 3 weeks. We considered genes showing a twofold or greater change to be significantly differentially expressed (p < 0.05); 405 genes were upregulated by PTH, while 883 were downregulated. The expression levels of the osteoblast differentiation markers Sp7 and Bglap were not changed ( Supplementary Fig. S1e), and thus were not affected by the PTH-induced change in cell population. The mRNA level of Fgf2, an important growth factor for PTH anabolism 19 , significantly increased ( Supplementary Fig. S1f, h).
In the Gene Ontology (GO) enrichment analysis, many of the upregulated GO pathways were related to the regulation of proteolysis ( Supplementary Fig. S1g). Among such genes, Slpi was the highly-ranked upregulated gene induced by PTH ( Fig. 1a; Supplementary Fig. S1h). Quantitative polymerase chain reaction (qPCR) confirmed that PTH upregulated the expression of Slpi in osteoblasts in vivo ( Supplementary Fig. S1i). We analyzed the time-course of the effect of PTH and found that to induce an increase in the Slpi mRNA levels required 1-3 weeks of repetitive injection (Fig. 1b). PTH can have anabolic or catabolic effects, depending on its mode of administration: the anabolic effect occurs by intermittent administration of PTH, but not by continuous infusion 20 . Slpi upregulation was not induced by continuous infusion of PTH for 3 weeks (a catabolic regimen; Supplementary Fig. S1j).
SLPI controls the PTH-induced increase in bone mass. The above results prompted us to explore the importance of SLPI in physiological bone remodeling in vivo. Initially, we evaluated baseline bone modeling and remodeling in Slpi-knockout (KO) mice 23 . In a homeostatic state, 5-week-old female mice were injected with vehicle for 6 weeks; distal femoral metaphysis was analyzed by micro-computed tomography (µCT) when the mice were 11 weeks of age. Slpi-KO mice displayed no overt trabecular bone phenotype, compared with wild-type (WT) mice (Fig. 1d, f, h; Supplementary Movies 1, 3). In vitro, WT and Slpi-KO primary osteoblasts showed identical bone-forming potential, as demonstrated by Alizarin Red S staining ( Supplementary Fig. S3a). In addition, the numbers of osteoclasts differentiated from Slpi-KO macrophages were similar to the numbers differentiated from WT macrophages ( Supplementary Fig. S3b). Together, these data suggest that Slpi-KO mice in a homeostatic state show normal bone remodeling with no aberrant phenotype.
To test the function of Slpi expression in PTH anabolism, we next examined the effect of PTH on Slpi-KO mice. Eleven-weekold WT mice subjected to intermittent PTH treatment for 6 weeks showed a significant increase in trabecular content, compared with vehicle-treated mice (Fig. 1d, e; Supplementary Movies 1, 2). This was apparent from the cancellous bone volume (BV/TV), and trabecular thickness (Tb.Th) (Fig. 1h). In contrast, the bone anabolic effect of PTH was severely disrupted in Slpi-KO mice; specifically, the rate of increase in BV/TV in Slpi-KO mice was less than half that in WT mice (Fig. 1e, g, h; Supplementary Movies 2,4). Similarly, PTH-treated WT mice had a significantly higher serum levels of type I procollagen N-terminal propeptide (P1NP), a biochemical marker of bone formation, compared to vehicle-treated WT control mice, which was disrupted in Slpi-KO mice (Fig. 1i). Assessment of the serum levels of C-terminal telopeptide of collagen (CTX), a biochemical marker of resorption, revealed that intermittent PTH treatment did not increase bone resorption in WT mice, but did increase it in Slpi-KO mice (Fig. 1j). Therefore, SLPI affects both bone formation and bone resorption in PTH anabolism, which results in significantly different trabecular bone phenotypes in the two genotypes. In addition, we analyzed trabecular bone in male mice to investigate the effect of sex on PTH anabolism. Five-week-old male mice were injected with vehicle or PTH for 6 weeks, and distal femoral metaphysis was analyzed by µCT. The effect of PTH was weaker in WT male mice than in WT female mice, while PTH treatment decreased trabecular bone mass in Slpi-KO male mice ( Supplementary Fig. S4). Therefore, these data demonstrate that the anabolic effect of PTH is also dependent on SLPI in male mice.
Given that the anabolic action of PTH results from enhanced bone formation, we examined the effect of ablation of Slpi on the cellular response to PTH via histomorphometric analysis. Twelveweek-old female WT or Slpi-KO mice were treated daily with vehicle or PTH for 4 weeks and their proximal tibias were analyzed. The effects of PTH treatment on bone formation rate (BFR/BS) were significantly reduced in Slpi-KO mice compared with WT controls (Fig. 1k-m; Supplementary Table S1). Collectively, bone histomorphometric analysis supported the notion that PTH-induced bone formation is impaired in Slpi-KO mice, consistent with the microstructural and bone metabolic marker analyses.
Induction of Slpi expression enhances osteoblast differentiation and proliferation. To examine the biological role of SLPI in osteoblasts, we constructed a retrovirus vector for constitutive Slpi expression (pMX-Slpi-IRES-Puro), which was engineered to express both Slpi and a puromycin-resistance gene. Induction of Slpi with pMX-Slpi-IRES-Puro enhanced alkaline phosphatase (ALP) activity, and accelerated calcification in MC3T3-E1 cells and primary osteoblasts, as determined by Alizarin Red S staining (Fig. 2a). As SLPI is secreted from cells into cervical mucus or bronchial and nasal secretions 26 , MC3T3-E1 cells also secreted SLPI into culture medium (Fig. 2b). To examine an autocrine effect of SLPI, we added conditioned medium from Slpi-overexpressing MC3T3-E1 cells to control (mock) cells. The results showed that conditioned medium from Slpi-overexpressing cells had no effect on osteoblast differentiation (Fig. 2c). Consistently, conditioned medium from Slpi-overexpressing cells did not upregulate Runx2 (Cbfa1/AML3), the osteogenic master gene for bone formation 27,28 , the osteoblast-specific transcription factor Osterix (Sp7), or the early osteoblast marker gene Col1a1 (Supplementary Fig. S6a). The late osteoblast marker gene osteocalcin (Bglap) was slightly increased by treatment with conditioned medium (Supplementary Fig. S6a). The addition of recombinant human SLPI to the culture medium also did not enhance ALP activity in MC3T3-E1 cells (Fig. 2d). We confirmed that overexpression of Slpi upregulated Runx2, Sp7, Col1a1, and Bglap in MC3T3-E1 cells ( Fig. 2e; Supplementary Fig. S6b).
β-catenin protein were detected by immunoblotting in Slpioverexpressing cells ( Supplementary Fig. S6d). Consistent with regulation of intracellular β-catenin levels by the proteasomal degradation pathway, the mRNA levels of β-catenin were not increased by SLPI (Supplementary Fig. S6e). Therefore, canonical Wnt signaling is a candidate downstream target of SLPI, although the mechanistic relationship between SLPI and the ubiquitinproteasome pathway is unclear.
The above findings show that SLPI promotes osteoblast proliferation and differentiation by targeting key transcription factors such as β-catenin, Runx2, and Osterix.
SLPI is associated with adhesion between osteoblasts and osteoclasts. To determine how SLPI affects bone resorption, we explored the paracrine role of SLPI in signaling between osteoblasts and osteoclasts. Because SLPI is a regulator of nuclear factor kappa B (NF-κB) in monocyte-lineage cells, we examined whether osteoclast differentiation via NF-κB signaling 33 was modulated by SLPI secreted from osteoblast-lineage cells. Treatment with recombinant human SLPI did not affect RANKLinduced osteoclast differentiation in vitro (Fig. 3a). Osteoclastic bone degradation is mediated by cysteine proteases and matrix metalloproteinases, which are not considered mechanistic targets of the serine proteinase inhibitor SLPI [34][35][36] . Consistent with this, we found that osteoclast pit-forming activity in vitro was not affected by recombinant human SLPI (Fig. 3b). Therefore, SLPI secreted from osteoblasts does not directly affect osteoclast differentiation or bone-resorbing activity.
Proteases and protease inhibitors are associated with cellular detachment from surrounding components. Indeed, SLPI in osteoblasts is reportedly associated with focal cell adhesion 37 . Therefore, we examined the effect of SLPI on adhesion between osteoblasts and osteoclasts by coculture experiments. When control (mock) MC3T3-E1 cells were cultured with primary osteoclasts, the duration of osteoblast-osteoclast colocalization was generally less than 1 h (Fig. 3c, d; Supplementary Movie 5). In contrast, induction of Slpi in MC3T3-E1 cells slightly increased the ratio of long cell-cell contact (Fig. 3c, d; Supplementary Movie 6). Consistent with these findings, quantitative analysis revealed that contact duration was significantly increased by Slpi induction (Fig. 3e). Also, Slpi mRNA was expressed in EGFP + osteoblastic cells, but not in tdTomato + osteoclasts, in coculture experiments (Fig. 3f).
Given that SLPI promotes leukocyte chemotaxis in vitro 38 , we examined whether SLPI secreted from osteoblasts directly attracts osteoclasts. We evaluated the effects of SLPI inhibition by commercially available neutralizing antibodies on osteoclast mobility. Time-lapse imaging was performed to assess cocultures of MC3T3-E1 cells and primary osteoclasts until 12 h after treatment with an SLPI neutralizing antibody or an IgG isotypematched control antibody. However, treatment with an SLPI neutralizing antibody did not inhibit the osteoblast-osteoclast contact enhanced by SLPI ( Supplementary Fig. S7).
Next, to investigate the molecular mechanisms by which Slpi induces colocalization of osteoblasts with osteoclasts further, we performed an RNA-Seq analysis using control (mock) MC3T3-E1 cells and Slpi-overexpressing cells (Supplementary Fig. S8a). Chemokines and growth factors have key roles in physiological bone remodeling by controlling migration, and localization of osteoclasts and their precursors 39,40 . However, these factors were overall not remarkably changed by SLPI in osteoblasts (Supplementary Fig. S8a). Similarly, we did not observe reduced expression of the repulsive guidance factor Plexin-B1, which is known to contribute to the spatial segregation of osteoblasts and osteoclasts 41,42 (Supplementary Fig. S8a).
Notably, among the canonical pathways affected by SLPI, Wnt/β-catenin signaling was one of the most significantly affected ( Supplementary Fig. S8b), consistent with the regulation of β-catenin by SLPI ( Supplementary Fig. 6d). RhoA signaling was another pathway significantly affected by SLPI (Supplementary Fig. S8b). Rho-associated coiled-coil kinases (ROCKs) are pivotal downstream effectors of RhoA. Immunoblot analysis revealed that ROCK2, but not ROCK1, was upregulated at the protein level in Slpi-overexpressing cells ( Supplementary  Fig. S8c). ROCK inhibitor Y-27632 significantly prevented the osteoclast-osteoblast contact that had been promoted by SLPI, suggesting that ROCK is associated with Slpi-induced osteoblastosteoclast adhesion (Supplementary Fig. S8d). Moreover, Slpioverexpressing MC3T3-E1 cells had comparably enhanced transmembrane glycoprotein ICAM-1 expression (Supplementary Fig. S8e). These results were consistent with previous studies showing that RhoA signaling is associated with heterotypic cell-cell adhesion, such as hematopoietic stem cell-niche cell adhesion 43 , and osteoblast-osteoclast adhesion 44 , through the expression of transmembrane cell adhesion molecules 44,45 .
Next, we analyzed how interactions between osteoblasts and osteoclasts influenced the activities of both cell types. With respect to bone formation, coculture of osteoblasts and osteoclasts reportedly promoted mineralization in osteoblasts 46 . Consistent with this observation, the presence of differentiated osteoclasts led to markedly increased ALP activity in primary osteoblasts ( Supplementary Fig. S9a, b). Alizarin Red S staining revealed that the presence of mature osteoclasts led to markedly increased mineralization in primary osteoblasts ( Supplementary Fig. S9a, b). Notably, direct coculture with mature osteoclasts for 24 h significantly increased Slpi expression in MC3T3-E1 cells (juxtacrine interaction in Fig. 3g). In addition, the expression of Fgf2, which was upregulated in PTH-treated osteoblasts, was increased in the presence of osteoclasts. We also performed experiments using the Transwell system, which enables sharing of culture medium between osteoblasts and osteoclasts without direct contact. However, Slpi and Fgf2 expression in osteoblasts did not increase under this condition (paracrine interaction in Fig. 3g), suggesting that direct osteoblast-osteoclast contact plays a role in increasing the levels of these anabolic factors. With respect to bone resorption, the presence of differentiated MC3T3-E1 cells inhibited bone marrow macrophages from undergoing osteoclastogenesis ( Supplementary Fig. S9c). Resorption pit formation by mature osteoclasts was also reduced during coculture with differentiated MC3T3-E1 cells ( Supplementary Fig. S9d). Collectively, these results suggest that the close contact between osteoblasts and osteoclasts would result in increased bone formation and decreased bone resorption. Therefore, SLPI-induced colocalization presumably promotes interactions between osteoblasts and osteoclasts, modifying the activities of both cell types to facilitate the anabolic action of PTH.
The PTH-induced direct association between osteoblasts and osteoclasts was disrupted in the absence of Slpi. Cell-to-cell attraction and cell stickiness promote the intermingling of different cell types, whereas tension between cells results in their segregation 47 . Therefore, we used intravital multi-photon microscopy to determine whether SLPI-mediated heterophilic interaction impacted PTH-induced bone cell distribution in vivo. We developed a system enabling simultaneous visualization of osteoblasts and osteoclasts using double transgenic Col2.3-ECFP/tartrate-resistant acid phosphatase (TRAP)-tdTomato reporter mice 2 . We crossed these mice with Slpi-KO mice, and compared the living bone tissue of anabolic PTH-treated mice.
ECFP + osteoblasts and tdTomato + osteoclasts were increased after 3 weeks of PTH treatment (Fig. 4a-c). The densities of ECFP + osteoblasts and tdTomato + osteoclasts between Slpi-KO and WT mice were similar after PTH treatment (Fig. 4a-c). However, the degrees of osteoblast and osteoclast gathering both increased, and their intermingling was disrupted by knockout of Slpi. Quantification of the cell mixture index (CMI), a cell mixture parameter calculated mathematically by hierarchical clustering 2 , demonstrated that the intermingling enhanced by PTH was partially but significantly disrupted in Slpi-KO mice, compared to WT mice (Fig. 4a, d; Supplementary Fig. S10). Consistent with this, the number of dynamic osteoblast-osteoclast contact events normalized to the surface area of osteoblasts or osteoclasts was significantly decreased in Slpi-KO mice, according to three-dimensional co-localization analysis of time-lapse images ( Fig. 4e-g). These results demonstrate that SLPI changes bone cell distribution, resulting in altered cell-cell communication in the setting of PTH anabolism in vivo.

Discussion
Bone remodeling is achieved by the orchestration of two cell types that have opposite functions, i.e., bone-resorbing osteoclasts and bone-forming osteoblasts. PTH has been used clinically to stimulate bone formation by enhancing the activities of osteoblasts and osteoclasts in a coordinated manner, although the precise underlying mechanisms remain unclear 11,48 . In this study, we showed that Slpi is one of the genes most highly upregulated by PTH in osteoblasts and is critical for the osteoanabolic action of PTH. SLPI exerts two distinct functions: (1) it acts within osteoblasts themselves and enhances bone formation by controlling gene expression, and (2) it enhances adhesive force to neighboring osteoclasts, thereby increasing cell-cell communication between osteoblasts and osteoclasts 2 (Fig. 5). Therefore, SLPI expression by osteoblasts concurrently regulates the activity of osteoblasts themselves, and their surrounding microenvironment.
Similar bifunctional activity has also been reported for some molecules related to innate immunity and tissue regeneration, such as high mobility group box-1 (HMGB1) and thymidine phosphorylase (TYMP). HMGB1 is a nuclear DNA-binding protein that is involved in transcriptional regulation by p53 or NF-κB. On the other hand, upon its release from dead or damaged cells, HMGB1 also acts as a ligand that activates macrophages and dendritic cells through damage-associated molecular patterns 49 . TYMP, also known as platelet-derived endothelial cell growth factor, catalyzes the hydrolysis of thymidine to ensure the availability of a pyrimidine nucleotide pool 50 . TYMP is secreted and promotes angiogenesis by inducing proliferation and migration of endothelial cells 51,52 . While the functions seem to be independent in these examples, the function of SLPI is characteristic because both functions contribute to the same consequence: an increase in bone mass. SLPI is a reasonable candidate mediator for finely controlling bone metabolism.
In addition to the interaction between osteoblasts and mature osteoclasts, the recruitment of osteoclast precursors (or monocytes/macrophages) might support the anabolic activity of PTH. Indeed, some types of macrophages are associated with PTH anabolism in bone 53 . Previous studies have shown that the migration of osteoclast precursors is regulated by chemokines (e.g., CCR1 chemokines 39 ), and growth factors (e.g., platelet derived growth factor BB 40 ). In our model, the expression levels of Ccl5 and Pdgfa were increased in Slpi-overexpressing osteoblasts, implying that these chemotactic factors may contribute to the attraction of osteoclast precursors (Supplementary Fig. S7a). Moreover, given the partial contribution of SLPI to PTH-induced change in CMI in the present study, there might be an additional, Slpi-independent pathway for regulation of bone cell re-distribution. Particularly, cysteine-X-cysteine family chemokine ligand 1 (Cxcl1) is highly upregulated in osteoblasts by PTH treatment, and direct osteoclast chemotactic migration 54 . Although the analytical method to evaluate osteoblast-osteoclast precursor communication has not been established yet, the roles of such chemokines in PTH anabolism, as well as migration of osteoclast precursors, should be investigated in the future.
In general, matrix metalloproteases have been considered to play roles in cellular detachment from surrounding components 55 , while serine proteases may also be involved in cell adhesion 56,57 . SLPI plays a role in focal adhesion of cancer cells to the surrounding environment 58 . In our analysis, we focused on the fact that RhoA signaling and ROCK2 protein were affected by the expression of SLPI, as RhoA signaling is associated with cell-cell adhesion through the expression of transmembrane cell adhesion molecules. Given such multiple roles for RhoA signaling, further studies are needed to determine the in vivo role of ROCK proteins in the anabolic state of PTH.
Finally, we should also consider the limitations of this study. Bone histomorphometry showed that intermittent PTH treatment did not significantly change the osteoclast parameters, osteoclast surface/bone surface (Oc.S/BS) or eroded surface (ES/BS), in either WT or Slpi-KO mice (Supplementary  Table S1). This was partially because bone histomorphometry works well for evaluation of bone formation, but it is less reliable when used to evaluate the dynamism of bone resorption 59 . Oc.S/BS and ES/BS are static measures of resorption surfaces, but provide no direct information concerning the dynamic status of osteoclast activity. In particular, bone histomorphometry did not necessarily reflect PTHactivated bone resorption both in our data and in previous reports by another group 60 . To evaluate the dynamic feature of osteoclast activity influenced by PTH, we analyzed serum CTX levels 61 , and showed that intermittent PTH treatment did not increase bone resorption in WT mice, but did increased it in Slpi-KO mice, suggesting a substantial effect of SLPI on bone resorption in PTH anabolism.
In conclusion, based on in vivo screening of gene expression in osteoblasts, we analyzed the bone phenotype of Slpi-KO mice and found that SLPI was strongly associated with PTH anabolism. Analysis of the molecular and cellular networks related to PTH anabolism will provide effective information in terms of the optimal usage of this drug and may facilitate the development of therapies for intractable bone diseases.

Methods
Mice. C57BL/6J mice were obtained from CLEA Japan (Tokyo, Japan). Slpi-KO mice were kindly provided by A. Nakamura's laboratory (Tohoku Medical and Pharmaceutical University, Sendai, Japan) 23 . The generation of Col2.3-ECFP and TRAP-tdTomato mice was described previously 2,62 . Mice were randomized to the treatment or control group. The mice were bred and maintained under specific pathogen-free conditions at the animal facilities of Osaka University (Osaka,  Quantitative reverse transcription-polymerase chain reaction. Total RNA and cDNA were prepared using a Maxwell Simply RNA kit (Promega, Madison, WI, USA) and Superscript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA). Real-time PCR was performed with a Thermal Cycler Dice Real-Time System (TaKaRa Bio, Shiga, Japan) using SYBR Premix EX Taq (TaKaRa Bio). The relative levels of the mRNAs of interest were calculated using the 2 −ΔΔCt method. The mRNA level of Gapdh was used as the internal control. The primer sequences are listed in Supplementary Table S2.
Bone histomorphometric analysis. Female mice were injected subcutaneously with 20 mg/kg body weight calcein (Sigma-Aldrich, St. Louis, MO, USA) at 72 and 24 h before sacrifice 65 . Tibias were collected and fixed in 70% ethanol. For histomorphometric analysis, tibias were embedded in glycolmethacrylate and toluidine blue staining was performed on sagittal sections. Calcein labeling was visualized using an inverted light microscope (DM-IL-LED-3; Leica Microsystems, Wetzlar, Germany). Quantitative bone histomorphometric measurements were performed in the secondary spongiosa using a Histometry-RT light microscope (System Supply, Nagano, Japan).
Measurement of bone metabolic markers. Female mice were fasted for 24 h before blood sampling. Blood was drawn via the inferior vena cava at 24 h after the final injection of PTH or vehicle. Blood specimens were centrifuged at 2000 × g for 15 min in gel-separation tubes (TS-701; Sato Kasei Kogyosho Co. Ltd., Tochigi, Japan) to obtain serum. An enzyme-linked immunosorbent assay (ELISA) was used for quantitation of P1NP (Immunodiagnostics Systems, Fountain Hills, AZ, USA) and CTX (RatLaps TM; Immunodiagnostics Systems) in serum.
Cell culture of osteoblasts. MC3T3-E1 cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). Primary osteoblasts were isolated from 2-day-old neonatal mouse calvaria as described previously 66 . Briefly, calvaria were subjected to five sequential digestions in 0.25% trypsin, 0.1% EDTA, and 0.32% collagenase type II (Worthington Biochemical Corp., Freehold, NJ, USA) solution for 25 min at 37°C. Cell fractions 2-5 were collected as primary osteoblasts. Adult primary osteoblasts derived from bone marrow were prepared as described previously 67 . Briefly, bone marrow cells from mouse long bones were seeded (2.5 × 10 5 cells per cm 2 ) and expanded to obtain bone marrow stromal cells (BMSCs). Before differentiation, BMSCs were purified by negative selection with MACS CD45 microbeads (Miltenyi Biotec., Bergisch Gladbach, Germany).
ALP staining was performed using naphthol-ASMX as the substrate and Fast Blue RR salt as the coupler. ALP activity was detected using pNPP (4-nitrophenyl phosphate) with a TRACP & ALP Assay Kit (TaKaRa) as a colorimetric substrate, then measured by spectrophotometry at 405 nm. Staining and quantification of ALP activity were performed at 3 and 7 days after differentiation of MC3T3-E1 cells and primary osteoblasts, respectively.
Alizarin Red S staining and its quantification were performed using a Mineralization Quantification Kit (PG-Research, Tokyo, Japan). Staining was performed at 14 days after differentiation of MC3T3-E1 osteoblasts and primary osteoblasts. Quantification of Alizarin Red S staining was via extraction with formic acid. The amount of dye released was quantified by spectrophotometry at 405 nm.
For the generation of conditioned medium, supernatants from subconfluent cultures of MC3T3-E1 cells were collected and passed through a 0.45 µm filter to eliminate cells, diluted with one volume of α-MEM, and used as conditioned medium.
For analysis of PTH1R signaling, osteoblasts were cultured under serum-free conditions for 12 h and topically treated with the following stimulators or inhibitors at the indicated concentrations: Human  69 . Two days after inoculation, MC3T3-E1 cells were incubated with lentiviral supernatant from virus-producing HEK293T cells, together with polybrene (8 µg/mL), for 12 h. Stable transformants were selected using an SH800 cell sorter (Sony).
Evaluation of osteoblast proliferation. The proliferation of osteoblasts was assessed by counting cultured osteoblasts using a TC10 Automated Cell Counter (Bio-Rad, Hercules, CA, USA). Briefly, osteoblasts were seeded into six-well plates at 1 × 10 4 /cm 2 . After 3 and 5 days, cells were suspended by incubation with 0.25% trypsin-EDTA (Sigma-Aldrich), and live cells were enumerated by trypan blue staining (Sigma-Aldrich).
Culture of primary osteoclasts. Nonadherent bone marrow cells derived from mice were seeded (1.5-2.0× in a 10 cm Petri dish) and cultured in the presence of 30 ng/mL M-CSF (R&D Systems, Minneapolis, MN, USA). After 3 days, adherent cells were collected in enzyme-free cell dissociation buffer (Millipore, Burlington, MA. USA) at 37°C and used as bone marrow macrophages (BMMs). BMMs were further cultured in the presence of 50 ng/mL soluble RANKL (Oriental Yeast Co., Ld., Tokyo, Japan) and 10 ng/mL M-CSF to generate osteoclasts 70 . Three days later, TRAP staining was conducted using an Acid Phosphatase, Leukocyte (TRAP) Kit (Sigma-Aldrich). TRAP + multinucleated (>3 nuclei) cells were regarded as mature osteoclasts.
To evaluate bone resorptive activity, osteoclasts were cultured on RepCell dishes for 5 days. Next, the cells were transferred to an Osteo Assay Surface (Corning, Corning, NY, USA). After 48 h, resorption pits were observed under an inverted light microscope (DM-IL-LED-3; Leica Microsystems), and the binarized areas of resorption pits were measured using NIS Elements integrated software (Nikon, Tokyo, Japan).
Mass spectrometry. Protein samples in buffer solution were precipitated by methanol-chloroform addition and then dissolved in RapiGest (Waters, Milford, MA, USA). The samples were reduced with 10 mM dithiothreitol followed by alkylation with iodoacetamide, then digested by trypsin and purified using a C18 tip (GL Science, Tokyo, Japan). Trypsin-digested peptides were subjected to liquid chromatography-tandem mass spectrometry, using an Aurora column (Ionoptics) on a Nano Elute nanoLC system coupled to a trapped ion mobility spectrometrytime-of-flight mass spectrometer (Bruker, Billerica, MA, USA). The column temperature was set to 50.0°C. The mobile phase consisted of water with 0.1% formic acid (solvent A) and acetonitrile containing 0.1% formic acid (solvent B). Peptides were eluted at a gradient of 4-30% B for 20 min with a flow rate of 400 nL/min. The mass-scanning range was set from 100 to 1700 m/z and a custom ion mobility revolution mode was used (range: 0.6-1.50 Vs/cm 2 ). The ion spray voltage was set at 1.6 kV in the positive-ion mode. The MS/MS spectra were acquired by automatic switching between MS and MS/MS modes. Data Analysis software (Bruker) was used to process the mass data. Peptides were identified by database searching using our in-house MASCOT Server (Matrix Science, London, UK). The precursor and fragment-searching mass tolerances were set to 20 ppm and 0.05 Da, respectively. Carbamidomethylation of cysteine was set as a static modification, while the oxidation of methionine, acetylation of protein N-term, and deamination of N,Q were set as variable modifications. The MASCOT Server search results were analyzed by Scaffold (Proteome Software Inc., Portland, OR, USA) and the results were limited to those with a GO annotation of "osteoblast." Finally, the quantitative values (normalized total spectra) of mock and immunoprecipitation samples were compared.
Immunoblotting analysis. Cells were lysed in RIPA buffer supplemented with protease inhibitor cocktail (Sigma-Aldrich). Conditioned media were mixed with four volumes of chilled acetone. Proteins were precipitated overnight at −30°C and dissolved in RIPA buffer. Protein concentrations were measured by bicinchoninic acid protein assay (Thermo Scientific Inc., Waltham, MA, USA). Cell lysates were boiled in sodium dodecyl sulfate loading dye. Equal amounts of total protein were separated on 4-20% mini-Protean TGX gradient gels (Bio-Rad) and transferred to polyvinylidene difluoride membranes (Immobilon P). The membranes were incubated with the primary and secondary antibodies. Finally, bands were visualized with the electrochemiluminescence reagents (GE Healthcare, Chicago, IL, USA) and using an ImageQuant LAS-4000 mini system (GE Healthcare). The antibodies for immunoblotting were as follows: anti-β-catenin osteoclasts. To assess the effect of SLPI antibody on cell-cell contact, SLPI antibodies (AF1274, R&D Systems; S0635, US Biological Life Sciences, Salem, MA, USA) 12.5 µg/mL, or control goat IgG (Santa Cruz Biotechnology) were added before time-lapse imaging. After 3 days of coculture, time-lapse images were acquired using an A1 confocal microscope (Nikon). To assess the effect of Y27632 on cell-cell contact, MC3T3-E1 cells were exposed to 10 µM Y27632 for 12 h. The reagent was removed before coculture to prevent it from acting on osteoclasts. For tracking of cell-cell contact, images were first automatically binarized by Otsu's thresholding method using NIS Elements integrated software (Nikon). The surface tool of Imaris software (Bitplane, Zürich, Switzerland) was used to track cell movement. Only EGFP-stained surface objects ≥100-µm 2 and tdTomato-stained surface objects ≥1000-µm 2 were analyzed, to exclude extracellular vesicles from osteoblasts and tdTomato + mononuclear cells. A new channel representing the colocalized area of EGFP and tdTomato voxels was created automatically by the surface tool. Finally, the contact duration was determined by tracking this new channel using Imaris software.
To assess the effects of cell-cell interaction on the differentiation and activity of each cell type, osteoblasts (or MC3T3-E1 cells) and primary osteoclasts were first cultured separately. For osteogenic differentiation, primary osteoblasts and primary osteoclasts were each seeded at a density of 1.5 × 10 4 cells/cm 2 . Osteoblasts were differentiated for 14 days in the presence of RANKL and M-CSF. Osteoclasts were added at 3-day intervals during osteoblast differentiation. For osteoclast differentiation, osteoblasts (or MC3T3-E1 cells) and BMMs were seeded at densities of 1.5 × 10 4 /cm 2 , and 3.75 × 10 4 /cm 2 respectively. Then, osteoclast differentiation was induced in the presence of RANKL and M-CSF for 5 days as described above. For the resorption pit assay, BMMs were differentiated into mature osteoclasts on 6 cm 2 RepCell dishes for 5 days. Then osteoclasts were cocultured with MC3T3-E1 cells on an Osteo Assay Surface (Corning) for 48 h in the presence of RANKL and M-CSF to evaluate resorption activity.
To evaluate gene expression in coculture, MC3T3-E1-EGFP cells were differentiated for 14 days. BMMs derived from TRAP-tdTomato + mice were differentiated into mature osteoclasts on 6-cm 2 RepCell dishes for 5 days. For juxtacrine interactions, MC3T3-E1 cells and osteoclasts were mixed at a ratio of 2:1 and cocultured for 24 h at a density of 3.75 × 10 4 cells/cm 2 . For paracrine interactions, osteoclasts were cocultured for 24 h, using a Millicell Hanging Cell culture insert with a 1-µm pore PET membrane (Millipore), with MC3T3-E1 cells at a density of 2.5 × 10 4 cells/cm 2 . Inserts lacking cells were used as controls. Cells were collected using collagenase type II, and the EGFP + tdTomato -(MC3T3-E1 cells) and EGFP -tdTomato + (primary osteoclasts) fractions were sorted using an SH800 cell sorter (Sony).
The imaging system consisted of an upright multiphoton microscope (A1R MP+; Nikon) with a 25× water-immersion objective (CFI75 Apo 25XC W 1300; Nikon). The system was driven by lasers (Chameleon Vision II Ti: Sapphire; Coherent, Inc, Santa Clara, CA, USA); the main laser was tuned to 860 nm to detect ECFP and second harmonic generation (SHG), and the sub-laser was tuned to 1040 nm to detect tdTomato. Using a Nikon upright microscope, multi-fluorescence images were acquired directly using four external non-descanned detectors equipped with dichroic and emission filters including an infrared-cut filter (DM685), three dichroic mirrors (DM458, DM506, and DM605), and three emission filters (417/60 for the SHG image, 480/40 for ECFP, 583/22 for tdTomato).
Cell mixture analysis. Cell mixtures were evaluated using the CMI 2 . To obtain the tiling images, snapshot image stacks of 16 continuous visual fields were collected across the sagittal suture at depths of 50-200 μm below the skull bone surface (5μm vertical steps) using 1.0× zoom and 512 × 512 X-Y resolution. Maximumintensity projection (MIP) images of tdTomato + mature osteoclasts and ECFP + mature osteoblasts were stitched to produce a tiling image. The threshold for binarization was determined automatically using Otsu's method for a tiling image with gamma correction. Next, the MIP images were divided to create 16 images; only images in which the ratio of red area to cyan area ranged between 0.5 and 2 were analyzed. CMI indicates the degree of mixing of two cell types in the range from 0 to 1, and was calculated by hierarchical clustering. Euclidean distances were calculated between each pixel in the binarized cell area, and Ward's method was used for the hierarchical clustering algorithm. Each pixel was initially considered a cluster of its own. Small clusters were combined sequentially into larger clusters, until all pixels converged into one cluster regardless of the color of individual pixels. The impurity of the clusters was calculated using a Gini-like formula, and CMI was defined as the area under the curve. Cell mixture analysis was performed using R software (ver. 3.6.0; R Development Core Team, Vienna, Austria) and the EBImage package (ver. 4.20.0).
Three-dimensional colocalization analysis. The numbers of osteoblast-osteoclast contacts were determined using Imaris. A Sobel filter was used to detect cell edges, in all depth slices and channels (3 × 3; four directions for images acquired by the Nikon upright microscope with a 1.0× zoom). We next adjusted the raw images to yield edge-enhanced images. The surface tool in Imaris was used to perform automatic cell-surface segmentation of each cyan-positive and tdTomato-positive cell evident on intravital time-lapse bone imaging. Cyan-stained surface objects ≤125 µm 3 in volume, and tdTomato-stained surface objects ≤1000 µm 3 in volume, were not included in the analysis, as these objects were unlikely to represent cells. The surface tool was then used to detect osteoblast-osteoclast contacts, and to automatically create a new channel (yellow) for colocalized cyan and tdTomato voxels.
Statistical analysis. Statistical analysis was performed using unpaired two-tailed ttests for comparisons between two groups, and using analysis of variance (ANOVA) with the Dunnett's, Tukey's, and Šidák's post hoc test for comparisons among three or more groups. Welch's ANOVA followed by Bonferroni's multiple comparisons test was used when the data demonstrated unequal variance. Statistical significance was recorded when p < 0.05. Biological replicates comprised samples from different mice. The results are representative of more than three independent experiments. The minimum possible numbers of samples were used. Mice were excluded if they exhibited any abnormalities in size, weight, or apparent symptoms of disease, before performing the experiments; however, no such phenomena were observed.