Research ArticlePhysiology

Protein kinase N3 promotes bone resorption by osteoclasts in response to Wnt5a-Ror2 signaling

See allHide authors and affiliations

Sci. Signal.  29 Aug 2017:
Vol. 10, Issue 494, eaan0023
DOI: 10.1126/scisignal.aan0023

Bone breakdown

To resorb bone, osteoclasts must remodel the cytoskeleton to form actin rings. Thus, understanding the pathways that control actin ring formation in osteoclasts could reveal potential therapeutic targets for reversing losses in bone density in diseases such as osteoporosis. Using genetically manipulated mice and cells derived from these mice, Uehara et al. characterized a pathway that stimulated bone resorption by osteoclasts. Binding of Wnt5a to one of its receptors, Ror2, activated the small GTPase Rho, its effector Pkn3, and c-Src, a kinase that is required for actin ring formation. The authors note that several of these proteins are implicated in tumor metastasis and that this signaling pathway may operate in metastasizing cancer cells.

Abstract

Cytoskeletal reorganization in osteoclasts to form actin rings is necessary for these cells to attach to bone and resorb bone matrices. We delineated the pathway through which Wnt5a signaling through receptor tyrosine kinase–like orphan receptor 2 (Ror2) promoted the bone-resorbing activity of osteoclasts. Wnt5a binding to Ror2 stimulated Rho, a small GTPase involved in cytoskeletal reorganization. Subsequently, the Rho effector kinase Pkn3 bound to and enhanced the activity of c-Src, a nonreceptor tyrosine kinase that is critical for actin ring formation. Mice with an osteoclast-specific deficiency in Ror2 (Ror2ΔOcl/ΔOcl) had increased bone mass. Osteoclasts derived from these mice exhibited impaired bone resorption and actin ring formation, defects that were rescued by overexpression of constitutively active RhoA. These osteoclasts also exhibited reduced interaction between c-Src and Pkn3 and reduced c-Src kinase activity. Similar to Ror2ΔOcl/ΔOcl mice, mice with a global deficiency of Pkn3 (Pkn3−/−) had increased bone mass. The proline-rich region and kinase domain of Pkn3 were required to restore the bone-resorbing activity of osteoclasts derived from Pkn3−/− mice. Thus, Pkn3 promotes bone resorption downstream of Wnt5a-Ror2-Rho signaling, and this pathway may be a therapeutic target for bone diseases such as osteoporosis, rheumatoid arthritis, and periodontal disease.

INTRODUCTION

The actin cytoskeleton regulates the shape and polarity of cells and also mediates various biological functions such as cell movements and division in all eukaryotic cells (13). The cytoskeleton in osteoclasts is highly organized to resorb the bone matrix (4, 5). A clearer understanding of the regulation of the actin cytoskeleton has important implications for diseases including osteoporosis. Bone mass is maintained by a balance between the activity of bone-resorbing osteoclasts and bone-forming osteoblasts (6, 7). Excessive bone-resorbing activity of osteoclasts causes postmenopausal osteoporosis and inflammatory bone diseases such as rheumatoid arthritis and periodontal disease (8).

The differentiation of osteoclasts is stimulated by receptor activator of nuclear factor κB ligand (Rankl, encoded by Tnfsf11) and colony-stimulating factor 1 (Csf1, encoded by Csf1), both of which are expressed in osteoblast-lineage cells such as osteoblasts (9, 10) and osteocytes (11, 12). Binding of Rankl and Csf1 to their receptors [Rank (encoded by Tnfrs11a) and Csf1 receptor (Csf1r)] triggers the differentiation of osteoclast precursors into osteoclasts. Osteoblast-lineage cells also produce osteoprotegerin (encoded by Tnfrsf11b), which inhibits osteoclast formation by interfering with the Rankl-Rank interaction.

Activated osteoclasts attach to bone surfaces through sealing zones (4, 13, 14), which are ringed-like structures of F-actin dots (also called actin rings). Protons, chloride ions, and several proteases including cathepsin K are secreted into the resorption lacunae surrounded by the sealing zone (4), thereby acidifying the resorption lacunae to resolve mineralized and degrade nonmineralized bone matrices. The formation of the sealing zone in osteoclasts requires cytoskeletal reorganization, which is promoted by small guanosine 5′-triphosphatases (GTPases) such as Rac (15) and Rho (16, 17). Vav3, a Rho family guanine nucleotide exchange factor, is involved in the formation of the sealing zone in osteoclasts (18). Vav3-deficient osteoclasts exhibit impaired bone-resorbing activity in vivo and in vitro, a phenotype attributed to defects in the Csf1-induced activation of Rac and in αvβ3 integrin–induced c-Src activity. These findings suggest that Rac is crucially involved in the bone-resorbing activity of osteoclasts. In addition to Rac, Rho is involved in the formation of actin rings in osteoclasts. Addition of the Clostridium botulinum C3 exoenzyme, an inhibitor of Rho proteins, in osteoclast cultures disrupts actin rings and impairs bone-resorbing activity of osteoclasts (19). The transduction of TAT-fusion constitutively active (CA)–RhoA into osteoclasts stimulates their podosome assembly, motility, and bone-resorbing activity (20). In contrast, a microinjection of CA-RhoA-GFP (green fluorescent protein) complementary DNA (cDNA) into osteoclasts causes podosomes to disappear from the cell periphery (21). Thus, how these small GTPases, especially Rho, are activated and form the sealing zone in osteoclasts has not yet been elucidated.

Wnt proteins promote cell differentiation, migration, and polarization through small GTPases such as Rho, Rac, and Cdc42 (22, 23). Wnt proteins bind to a receptor complex of Frizzled and low-density lipoprotein receptor–related protein 5/6 and activate Wnt/β-catenin signals. On the other hand, a ligand-receptor complex of Wnt, Frizzled, and receptor tyrosine kinase–like orphan receptor 1/2 (Ror1/2) activates β-catenin–independent signaling to promote cell migration and polarization (2426). We have previously shown that Wnt5a, a noncanonical Wnt ligand, enhances Rankl-induced osteoclastogenesis through Ror2 (27). However, the role of Wnt5a-Ror2 signaling in the bone-resorbing activity of osteoclasts remains unclear.

Using genetic approaches, we demonstrated that Wnt5a-Ror2 signals promoted osteoclastic bone-resorbing activity through the Daam2 [dishevelled (Dvl)–associated activator of morphogenesis 2]–Rho–Pkn3 signaling axis. Pkn3 phosphorylated through Ror2 signaling formed a complex of c-Src and proline-rich tyrosine kinase 2 (Pyk2). This complex facilitated the activation of c-Src, thereby inducing the formation of actin rings and the bone-resorbing activity of osteoclasts. Thus, this signaling axis not only helps ensure proper bone homeostasis but also represents a therapeutic target for bone diseases such as osteoporosis and inflammatory bone diseases.

RESULTS

Wnt5a secreted from osteoclasts cell autonomously promotes bone-resorbing activity

To further understand the roles of Wnt signaling in bone resorption, we examined the expression of genes encoding Wnt ligands during Rankl-induced osteoclast formation in bone marrow–derived macrophage (BMM) cultures using real-time polymerase chain reaction (RT-PCR) (Fig. 1A). Culturing BMMs in the presence of Rankl and Csf1 increased the expression of the osteoclast marker Cathepsin K (Ctsk) and of Wnt5a, Wnt6, Wnt10a, Wnt10b, Wnt11, and Wnt16. Wnt5a mRNA was particularly abundant among these Wnt ligand–encoding mRNAs in osteoclast cultures. Immunoblotting analysis confirmed that Wnt5a was abundant in BMM-derived osteoclasts (Fig. 1B). Because Wnt5a−/− mice died before bone marrow is formed (27), we cultured macrophages from the livers of wild-type and Wnt5a−/− mice on dentin slices in the presence of Rankl and Csf1 to induce differentiation into osteoclasts in vitro. No significant differences were observed in the number of osteoclasts between wild-type and Wnt5a−/− macrophage cultures (fig. S1). The area of resorption pits on the dentin slices was significantly lower in Wnt5a−/− osteoclast cultures than in wild-type cultures, suggesting an impairment in the bone-resorbing activity of Wnt5a−/− osteoclasts (Fig. 1C). Rhodamine-labeled phalloidin staining revealed that the number of actin rings was lower in Wnt5a−/− osteoclast cultures than in wild-type cultures (Fig. 1D). The addition of recombinant Wnt5a to Wnt5a−/− osteoclast cultures rescued the impaired actin ring formation and bone-resorbing activity (Fig. 1, C and D). These results suggested that Wnt5a secreted from osteoclasts cell autonomously promoted their bone-resorbing activity.

Fig. 1 Wnt5a secreted from osteoclasts regulates osteoclast bone-resorbing activities.

(A) RT-PCR analysis of Ctsk and Wnt expression in BMM cultures treated with or without Rankl and Csf1. n = 5 dishes for each time point. BMMs were prepared from five mice. ND, not detected. (B) Immunoblotting analysis of Wnt5a abundance in BMM cultures treated with glutathione S-transferase (GST)–Rankl plus Csf1. n = 3 biological replicates. (C and D) Effects of recombinant Wnt5a on the formation of resorption pits (hematoxylin staining) (C) and actin rings (D) by osteoclasts derived from wild-type (WT) and Wnt5a−/− liver macrophages on dentin slices. n = 5 slices for each genotype. Osteoclasts were prepared from three mice for each genotype. Scale bars, 100 μm. In (A), (C), and (D), error bars represent SD. *P < 0.05, **P < 0.01. n.s., not significant. For statistical analyses, Kruskal-Wallis and Steel-Dwass test (A) or analysis of variance (ANOVA) and Scheffé test (C and D) were used.

Ror2-mediated signaling induces the bone-resorbing activity of osteoclasts in vivo

We next examined the expression of Ror1 and Ror2 in osteoclasts using RT-PCR (Fig. 2A). Ror2, but not Ror1, was strongly expressed in osteoclasts, suggesting that Ror2-mediated signaling may have some roles in mature osteoclasts. We generated osteoclast-specific Ror2 conditional knockout mice (Ror2ΔOcl/ΔOcl, Ror2fl/fl: CtskCre/+) by crossing Ror2-floxed mice with CtskCre/+ mice. RT-PCR analysis showed that Ror2 expression was lower in osteoclasts derived from BMMs from these mice than in those derived from Ror2fl/fl mice (control) BMMs (Fig. 2B). The abundance of Ror2 transcripts was significantly lower in Ror2ΔOcl/ΔOcl osteoclasts than in Ror2ΔOcl/ΔOcl BMMs and in osteoclasts derived from control mice. Immunoblotting confirmed that Ror2 protein abundance was reduced in Ror2ΔOcl/ΔOcl osteoclasts (Fig. 2C).

Fig. 2 Ror2-mediated signals are required for bone-resorbing activity of osteoclasts.

(A) Reverse transcription PCR analysis of Ror1 and Ror2 in osteoclasts. n = 3 biological replicates. (B) RT-PCR analysis of Ror2 mRNA in BMMs and osteoclasts from Ror2fl/fl (Control) and Ror2fl/fl: CtskCre/+ (Ror2ΔOcl/ΔOcl) mice. n = 5 cultures of BMMs and osteoclasts for each genotype. (C) Immunoblotting of Ror2 in osteoclasts. n = 3 biological replicates. (D) Micro-CT of distal femurs from male control (Ror2fl/fl) and Ror2ΔOcl/ΔOcl mice. n = 8 mice for each genotype. Scale bar, 1 mm. (E) TRAP and hematoxylin staining of the distal femurs. Scale bar, 50 μm. Osteoclast number per bone perimeter from control and Ror2ΔOcl/ΔOcl mice. n = 8 mice for each genotype. (F) Erosion depth and eroded surface per bone surface. Erosion depth: 120 resorption lacunae assessed from seven control and nine Ror2ΔOcl/ΔOcl mice. Eroded surface per bone surface: n = 8 mice for each genotype. (G) Serum collagen type I cross-linked CTX in control and Ror2ΔOcl/ΔOcl mice. n = 8 mice for each genotype. (H) Alkaline phosphatase activity in serum. n = 8 mice for each genotype. In (B) and (D) to (H), error bars represent SD. ***P < 0.001, **P < 0.01, *P < 0.05. For statistical analyses, Kruskal-Wallis and Steel-Dwass test (B) or two-tailed Student’s t test (D to H) was used.

We observed no gross abnormalities in the skeletal development of Ror2ΔOcl/ΔOcl mice. Micro–computed tomography (CT) of distal femurs from male and female Ror2ΔOcl/ΔOcl mice showed that the bone volume, number of trabeculae, and trabecular thickness were higher, but trabecular separation was lower compared to femurs from control mice (Fig. 2D and fig. S2A). Similar bone phenotypes were observed in the lumbar vertebrae of male and female Ror2ΔOcl/ΔOcl mice (fig. S2B). Histomorphometric analysis revealed that the numbers of osteoclasts in the distal femurs were similar in Ror2ΔOcl/ΔOcl and control mice (Fig. 2E), but that the depth and surface area of bone erosions were decreased in Ror2ΔOcl/ΔOcl mice (Fig. 2F). Serum collagen type I cross-linked C-terminal telopeptide (CTX), a marker of bone resorption, was lower in Ror2ΔOcl/ΔOcl mice (Fig. 2G). Serum alkaline phosphatase activity, a marker for bone formation, was normal in these mice (Fig. 2H).

To confirm the bone-resorbing activity of Ror2ΔOcl/ΔOcl osteoclasts, BMMs from Ror2ΔOcl/ΔOcl mice were cultured on dentin slices in the presence of Csf1 and Rankl (fig. S2, C and D). No significant difference was observed in the number of osteoclasts between control and Ror2ΔOcl/ΔOcl osteoclast cultures (fig. S2C), but the area of resorption pits on dentin slices was smaller in Ror2ΔOcl/ΔOcl cultures (fig. S2D). These results suggested that Ror2ΔOcl/ΔOcl mice exhibited a high bone mass phenotype due to defects in the bone-resorbing activity of osteoclasts.

Wnt5a-Ror2 signaling induces the bone-resorbing activity of osteoclasts through Rho

The small GTPases Rac and Rho are involved in osteoclastic bone resorption by promoting actin ring formation (16, 17). Therefore, we determined whether Wnt5a-activated Rac and Rho in osteoclasts formed from Ror2ΔOcl/ΔOcl and control mice (Fig. 3A). Wnt5a activated both small GTPases in control osteoclasts within 5 min, but not in Ror2ΔOcl/ΔOcl-derived osteoclasts. This result suggested that Wnt5a activated Rac and Rho in osteoclasts through Ror2-mediated signaling. To establish whether Rho or Rac activity was involved in bone-resorbing activity under Ror2 signaling, we assessed the bone-resorbing activity of Ror2ΔOcl/ΔOcl osteoclasts transduced with adenoviruses encoding CA-RhoA or CA-Rac1 (Fig. 3B). The overexpression of CA-RhoA, but not CA-Rac1, rescued the impaired formation of actin rings and resorption pits in Ror2ΔOcl/ΔOcl osteoclasts (Fig. 3, C and D). These results suggested that the Wnt5a-Ror2 signal stimulated osteoclast function by activating Rho.

Fig. 3 Daam2 is a critical scaffold molecule linking Ror2 and Rho.

(A) Wnt5a-induced Rac and Rho activity in osteoclasts. n = 5 dishes for each genotype. a.u, arbitrary units. (B) Immunoblotting of RhoA and Rac1 in osteoclasts expressing CA-RhoA or CA-Rac1. n = 3 biological replicates. (C and D) Effects of CA-RhoA and CA-Rac1 on actin ring formation (C) and resorbing pits (D). n = 5 dentine slices for each genotype. Scale bars, 100 μm. (E) RT-PCR of Daam1 and Daam2 expression. n = 5 dishes for each genotype. (F) RT-PCR of Daam2 expression in osteoclasts transfected with shDaam2. n = 5 dishes for each condition. (G) Effects of shRNA-mediated knockdown of Daam2 on Wnt5a-induced Rho activity in osteoclasts. n = 5 dishes for each group. (H and I) Effects of the knockdown of Daam2 and overexpression of CA-RhoA on actin ring formation (H) and resorbing pits (I) in osteoclasts. n = 5 dentine slices for each group. Scale bars, 100 μm. In (A) and (C) to (I), error bars represent SD. **P < 0.01, *P < 0.05. For statistical analyses, Kruskal-Wallis and Steel-Dwass test (A and G), ANOVA and Scheffé test (C, D, H, and I), two-tailed Student’s t test (F), or two-tailed Welch’s t test (E) was used.

Daam2 is involved in the bone-resorbing activity of osteoclasts

In the Wnt signaling pathway, Daam mediates Rho and β-catenin signaling (28, 29). We investigated whether Daams were involved in the Wnt5a-induced bone-resorbing activity of osteoclasts. RT-PCR analysis revealed greater expression of Daam2 in osteoclasts than in BMMs (Fig. 3E). The expression of Daam1 was not detected in BMMs or osteoclasts (Fig. 3E). Therefore, we focused on the role of Daam2 in Wnt5a-induced Rho activities in osteoclasts. Knockdown of Daam2 by short hairpin RNA (shRNA) abrogated recombinant Wnt5a-induced Rho activity in osteoclasts (Fig. 3, F and G, and fig. S3, A and B). The knockdown of Daam2 in osteoclasts suppressed actin ring formation and pit-forming activity without affecting their differentiation (Fig. 3, H and I, and fig. S3, C to E). In contrast, the overexpression of CA-RhoA rescued the impaired osteoclast function in Daam2-deficient osteoclasts (Fig. 3, H and I, and fig. S3, C and D). These results suggested that Daam2 mediated Rho activity under Ror2 signaling in osteoclasts.

Pkn3 is a Rho effector involved in osteoclast function

Rho-associated kinase (ROCK) (30) and mDia2 (31) act as Rho effectors in osteoclast function. RT-PCR analysis of 13 Rho effectors (32) demonstrated that the expression of Pkn3 (which encodes protein kinase N3) in osteoclasts was significantly higher than that in BMMs (Fig. 4A). The shRNA-mediated knockdown of Pkn3, but not that of Pkn1, Pkn2, or mDia2, inhibited actin ring formation and pit-forming activity of osteoclasts without affecting osteoclast differentiation (Fig. 4, B and C, and fig. S4, A to D). We also determined whether ROCK was involved in the bone-resorbing activity (Fig. 4, D and E). Inhibition of ROCK with Y27632 suppressed stress fiber formation in bone marrow–derived stromal cells but did not affect the formation of actin rings and resorption pits by osteoclasts. These results suggested that Pkn3, but not mDia2 or ROCK, was the Rho effector largely involved in osteoclast function.

Fig. 4 Pkn3 acts as a Rho effector for the bone-resorbing activity of osteoclasts.

(A) RT-PCR analysis of the expression of mRNAs encoding Rho effectors in BMMs and osteoclasts. n = 5 dishes of BMM and osteoclast cultures. (B) Effect of shRNA-mediated knockdown of Pkn family members on the bone-resorbing activity of osteoclasts. n = 5 slices for each group. Osteoclasts were prepared from five mice. Scale bar, 100 μm. (C) Effects of the shRNA-mediated knockdown of mDia2 on the bone-resorbing activity of osteoclasts. n = 5 slices for each group. Osteoclasts were prepared from three mice. Scale bar, 100 μm. (D) Effects of Y27632 on stress fiber formation in bone marrow stromal cells. n = 5 wells for each treatment. Bone marrow stromal cells were prepared from two mice. Scale bar, 100 μm. (E) Effects of Y27632 on the formation of actin rings (the left two images and the left bar graph) and resorbing pits (the right two images and the right bar graph) in osteoclasts cultured on dentin slices. n = 5 slices for each treatment. Osteoclasts were prepared from two mice. Scale bars, 100 μm. In (A) to (E), error bars represent SD. **P < 0.01. For statistical analyses, Mann-Whitney U test (A), Kruskal-Wallis and Steel-Dwass test (B), or two-tailed Student’s t test (C to E) was used.

To further clarify the roles of Pkn3 in bone resorption in vivo, we examined the bone phenotypes of 8-week-old Pkn3−/− mice. Pkn3−/− mice were born at the expected Mendelian ratio, and no gross abnormalities were observed in skeletal development. Immunoblotting analysis showed that the production of Pkn3 was not detected in osteoclasts formed from Pkn3−/− mice (Fig. 5A). Micro-CT analysis revealed that bone volume/tissue volume, trabecular thickness, and trabecular number were increased in the distal femurs and lumbar vertebrae of 8-week-old male Pkn3−/− mice, whereas trabecular separation was lower (Fig. 5B and fig. S5A). Similar bone phenotypes were observed in femurs as well as lumbar vertebrae of female Pkn3−/− mice (fig. S5, A and B). Histomorphometric analysis showed that eroded surface per bone surface was lower in Pkn3−/− mice; however, the number of osteoclasts in bone tissues was similar in Pkn3−/− mice and wild-type mice (Fig. 5C). Furthermore, the erosion depth was shallower in Pkn3−/− mice (Fig. 5D). Serum CTX was lower in Pkn3−/− mice (Fig. 5E). On the other hand, bone formation parameters including osteoblast numbers remained unchanged in Pkn3−/− mice (Fig. 5F).

Fig. 5 Impaired bone-resorbing activity of osteoclasts in Pkn3−/− mice.

(A) Immunoblotting of Pkn3 in osteoclasts formed from Pkn3−/− mice. n = 3 biological replicates. (B) Micro-CT analysis of femurs. n = 7 mice for each genotype. Scale bar, 1 mm. (C) TRAP and hematoxylin staining images and bone histomorphometric analysis of femurs. n = 7 mice for each genotype. Scale bar, 50 μm. (D) Erosion depth and the frequency distribution of the erosion depth in femurs. n = 7 mice for each genotype. Erosion depth: 200 resorption lacunae were assessed. (E) Serum CTX in WT and Pkn3−/− mice. n = 7 mice for each genotype. (F) Bone histomorphometric analysis of bone formation parameters in distal femurs. n = 7 mice for each genotype. (G) Ex vivo analysis of actin ring formations in osteoclasts. (H) Ex vivo analysis of resorbing pits in WT and Pkn3−/− mice–derived osteoclasts. In (G) and (H), n = 5 slices for each genotype. Osteoclasts were prepared from three mice for each genotype. Scale bars, 100 μm. In (B) to (H), error bars represent SD. ***P < 0.001, **P < 0.01, *P < 0.05. For statistical analyses, two-tailed Student’s t test (B to F) or two-tailed Welch’s t test (G and H) was used.

Ex vivo analysis showed that Rankl-induced osteoclast formation was normal in BMM cultures from Pkn3−/− mice (fig. S6A). RT-PCR analysis also confirmed that the expression of osteoclast marker genes such as Tnfrs11a and Csf1r was normal in Pkn3−/− osteoclasts (fig. S6B). However, the formation of actin rings and resorption pits by Pkn3−/− osteoclasts was impaired (Fig. 5, G and H). A podosome belt was observed at the cell periphery of wild-type osteoclasts (fig. S6C). In contrast, podosome dots, but not a podosome belt, were present in Ror2ΔOcl/ΔOcl or Pkn3−/− osteoclasts, suggesting that assembly of podosomes was impaired in these osteoclasts. The overexpression of CA-RhoA failed to rescue the impaired pit-forming activity of Pkn3−/− osteoclasts (fig. S6D). Furthermore, alkaline phosphatase activity and mineralized nodule formation were normal in calvaria-derived osteoblastic cell cultures from Pkn3−/− mice (fig. S6E). These results suggested that Pkn3 acted as a Rho effector in the bone-resorbing activity of osteoclasts.

The proline-rich region of Pkn3 is necessary for bone-resorbing activity

The phosphorylation of Thr718 in human PKN3 is required for full kinase activity (33). Therefore, we investigated the involvement of Ror2 signaling in the phosphorylation of Pkns. Immunoblotting analysis showed that phosphorylated Pkns were markedly decreased in osteoclasts from Ror2ΔOcl/ΔOcl mice (Fig. 6A). RT-PCR analysis confirmed that the expression of Pkn3 mRNA remained unchanged between Ror2ΔOcl/ΔOcl and control osteoclasts (Fig. 6B). These results suggested that the Wnt5a-Ror2 signal was involved in the phosphorylation of Pkn3 in osteoclasts.

Fig. 6 Pkn3 forms complexes with c-Src and Pyk2 to promote bone-resorbing activity.

(A) Immunoblotting of phosphorylated Pkns in osteoclasts. n = 3 biological replicates. (B) RT-PCR of Pkn3 mRNA in osteoclasts. n = 5 dishes for each genotype. (C) Confocal microscopic images in osteoclasts expressing Venus-Pkn3 and DsRed-actin proteins. n = 3 biological replicates. Scale bar, 50 μm. (D) Interactions between Venus-Pkn3 and c-Src in Ror2ΔOcl/ΔOcl osteoclasts. n = 3 biological replicates. (E) c-Src kinase activity in Ror2ΔOcl/ΔOcl osteoclasts. n = 5 dishes of BMM and osteoclast cultures for each genotype. (F) Effects of Daam2 knockdown on interactions between Pkn3, c-Src, and Pyk2. n = 3 biological replicates. (G) Immunoprecipitation (IP) analysis of osteoclasts expressing full-length Pkn3-Venus (full), Pkn3-Venus lacking the PRR domain (ΔPRR), and Pkn3-Venus lacking the kinase domain (Δkinase). n = 3 biological replicates. (H and I) Effects of the enforced expression of Pkn3 full, Pkn3-ΔPRR, and Pkn3-Δkinase on the actin ring formation (H) and resorbing pits (I) of Pkn3−/− osteoclasts. n = 5 slices for each group. Osteoclasts were prepared from three mice. Scale bars, 100 μm. In (B), (E), (G), and (H), error bars represent SD. **P < 0.01. (J) Proposed pathway in our current study. For statistical analyses, Mann-Whitney U test (B), Kruskal-Wallis and Steel-Dwass (E), or ANOVA and Scheffé test (H and I) were used.

We then examined the localization of F-actin and Pkn3 in osteoclasts on calcium phosphate–coated plates. DsRed-fusion β-actin and Venus-fusion Pkn3 (Venus-Pkn3) were coexpressed in osteoclasts using adenovirus-mediated gene transfer. Venus-Pkn3 was observed in a perinuclear region and in the cell periphery of osteoclasts. A part of Venus-Pkn3 was colocalized with actin rings visualized by DsRed-fusion β-actin (Fig. 6C, white arrow). Thus, Pkn3 was localized with actin rings.

Pkn2 interacts with the Src family kinase Fyn and plays a role in the Rho-induced activation of Fyn in keratinocytes (34). Furthermore, Rho activates c-Src in Fyn−/− keratinocytes (34). c-Src is abundant in osteoclasts, and the activation of c-Src is essential to form actin rings in osteoclasts (3537). These findings prompted us to clarify whether the Rho-Pkn3 pathway promotes actin ring formation in osteoclasts through the activation of c-Src. Immunoprecipitation assays using an antibody specific for enhanced GFP (EGFP) revealed that in control osteoclasts, Venus-Pkn3 associated with c-Src and Pyk2, a kinase that promotes osteoclast adhesion and forms a complex with c-Src to activate osteoclast function (Fig. 6D) (38, 39). In contrast, these associations were abolished in Ror2ΔOcl/ΔOcl osteoclasts. The kinase activity of c-Src in osteoclasts from Ror2ΔOcl/ΔOcl mice was lower than that in control osteoclasts (Fig. 6E); however, the abundance of c-Src in Ror2ΔOcl/ΔOcl osteoclasts was similar to that in control osteoclasts (fig. S7A). The kinase activity of c-Src was also lower in Pkn3−/− osteoclasts (fig. S7B). Knockdown of Daam2 suppressed the phosphorylation of Pkn3 and complex formation of Pkn3, c-Src, and Pyk2 (Fig. 6F and fig. S7C). We also determined the domains of Pkn3 that were needed for the interaction between Pkn3 and c-Src by overexpressing deletion mutant forms of Pkn3 in osteoclasts (fig. S7D). Immunoprecipitation assays showed that Pkn3 lacking the kinase domain (Δkinase), but not the form lacking the proline-rich region (ΔPRR), interacted with c-Src or Pyk2 (Fig. 6G). Notably, the expression of full-length Pkn3 (Pkn3 full) rescued impaired actin ring formation and pit-forming activity of Pkn3−/− osteoclasts, whereas that of Pkn3-ΔPRR and Pkn3-Δkinase did not (Fig. 6, H and I). Together, these results indicated that Pkn3 and its kinase activity promoted osteoclast function through an interaction with c-Src.

DISCUSSION

Wnt5a secreted from osteoblast-lineage cells enhances the production of Rank in osteoclast precursors, which, in turn, promotes Rankl-induced osteoclastogenesis (27). Using genetic and biochemical approaches, we showed that Wnt5a also promotes the bone-resorbing activity of osteoclasts through Ror2-mediated signaling. Although Rho is involved in the bone-resorbing activity of osteoclasts as demonstrated by the use of the C3 exoenzyme and CA-RhoA (1921), it currently remains unclear how Rho promotes the formation of actin rings in osteoclasts. Rho promotes the formation of stress fibers in cells including fibroblasts. In contrast, transformation of fibroblasts by active c-Src disrupts actin stress fibers and triggers the alternative formation of podosome belts without a decrease in Rho activity, but rather by localizing active Rho in podosomes (40). Furthermore, inhibition of Rho activity using C3 exoenzyme and transfection of dominant negative forms of Rho disrupts podosome belts (40). These findings indicate that Rho cooperates with c-Src to form podosome belts in c-Src–transformed cells. Therefore, Rho preferentially forms podosome belts, but not stress fibers in c-Src–positive cells including osteoclasts, supporting our model that Rho activated by Wnt5a-Ror2 signals and c-Src are involved in actin ring formation in osteoclasts.

Daam1 binds the PDZ and DEP domains of Dvl2 and Rho and mediates the Wnt-induced activation of Rho (28). However, Rac directly binds to DEP domains of Dvl2 to activate downstream signaling in a Daam-independent manner (41). These findings indicate that Dvl plays an important role as a hub molecule in the activation of Rho and Rac by Wnt signals and that Daam1 is required for the activation of Rho by Fzd-Dvl2 signaling. Osteoclasts strongly expressed Daam2, but not Daam1 (Fig. 3E), suggesting that Rho was activated by Daam2, rather than Daam1. Daam2 reportedly binds not only to the DIX domain of Dvl3 to activate Wnt/β-catenin signaling but also to the PDZ and DEP domains of Dvl3 (29). On the basis of the knockdown experiments of Daam2 in osteoclasts and the overexpression of CA-RhoA in Daam2-deficient osteoclasts, we propose that Daam2 acts as a link between Ror2/Fzd and Rho-mediated signals in osteoclasts.

Rho signaling activates several Rho effectors, including ROCK. The ROCK inhibitor Y27632 did not inhibit the formation of actin rings and resorption activities of osteoclasts under our experimental conditions, suggesting that ROCK is primarily involved in the formation of stress fibers but not podosomes. Although ROCK may be activated in osteoclasts by Wnt5a-Ror2-Rho signals, the activated c-Src may inhibit the stress fiber formations as described above. Further studies are needed to clarify how c-Src inhibits ROCK activity in osteoclasts.

Our results showed that Pkn3 played a critical role in osteoclastic bone resorption stimulated by Wnt5a-Ror2-Rho signaling. SH3 domains bind to proline-rich domains to mediate protein-protein interactions (42). Similarly, Pkn3 may associate with the SH3 domain of c-Src through its proline-rich domain. It is unlikely that Pkn3 directly associates with Pyk2 because Pyk2 does not have an SH3 domain (43). c-Src binds to Pyk2 and phosphorylates tyrosine residues in Pyk2 in osteoclasts in a cell adhesion–dependent manner (44). These previous findings and our present study suggest that c-Src primarily forms a complex with Pyk2, and then the complex associates with Pkn3 in response to Wnt5a-Ror2 signals. The kinase activity of Pkn2 promotes the activity of Fyn (34) and that Pkn1 is autophosphorylated in response to RhoA signals (45). We showed that Daam2 was required for the phosphorylation of Pkn3 in response to Wnt5a-Ror2 signals. These findings also suggest that Pkn3 is autophosphorylated, and then phosphorylated Pkn3 promotes the activation of c-Src. Further experiments are needed to determine how Pkn3 promotes the c-Src activity.

We found that Pkn3 was abundant in osteoclasts, but not in BMMs as osteoclast precursors. However, Pkn1 expression was higher in BMMs than in osteoclasts (Fig. 4A). These findings suggest that Rankl-Rank signaling could induce Pkn3 expression. However, the expression of Pkn3 was lower than that of Pkn1 in spleen and intestine (46), tissues in which Rankl or Rank are abundant (4749). Together, these results suggest that Rankl-Rank signals do not directly induce Pkn3 expression during osteoclast formation.

Vav3 plays a role in the Csf1-induced activation of Rac and αvβ3 integrin–mediated activation of c-Src. The Csf1-induced activation of Rho is normal in Vav3−/− osteoclasts (18). These findings suggest that Vav3 mainly activates Rac under Csf1r-mediated signals, and Rho may be activated by Vav3-independent signals including Wnt5a-Ror2 signals. The activation of c-Src by αvβ3 integrin–mediated signals is impaired in Vav3−/− osteoclasts and also in osteoclasts formed from Ror2ΔOcl/ΔOcl mice. The overexpression of CA-RhoA, but not CA-Rac1, rescued the impaired pit-forming activity of these osteoclasts. These results indicated that Rac activity in response to Csf1r- and αvβ3 integrin–mediated signals and Rho activity in response to Wnt5a-Ror2 signals promote c-Src activity and bone-resorbing activity in osteoclasts. Further studies are needed to clarify the mechanism by which Pkn3 activates c-Src.

The importance of Wnt5a-Ror2 signaling in the activation of c-Src does not appear to be limited to bone resorption. Ror2 knockdown in SaOS-2 cells, a human osteosarcoma cell line, suppresses the activation of c-Src and the production of matrix metalloproteinase-13 (MMP-13), thereby limiting their invasive properties (50). The treatment of SaOS-2 cells with Wnt5a enhances c-Src activation and MMP-13 production. These findings suggest that Wnt5a-Ror2 signaling is involved in the metastasis of malignant neoplasms through the activation of c-Src. Thus, the Wnt5a-Ror2 signaling axis promotes not only bone resorption but also tumor invasion.

We showed that Wnt5a-Ror2 signals promoted the bone-resorbing activity of osteoclasts through a Daam2-Rho-Pkn3 signaling axis (Fig. 6J). This signaling pathway was critical for the bone mass under physiological conditions. This signaling pathway may also be crucial under pathological conditions such as rheumatoid arthritis, a condition in which Wnt5a secretion is increased from synovial cells, osteoblast-lineage cells, and osteoclasts. Thus, Wnt5a-Ror2 signals may represent a therapeutic target for bone diseases such as osteoporosis, rheumatoid arthritis, and periodontitis.

MATERIALS AND METHODS

Animals and reagents

Wnt5a−/− (51) and CtskCre/+ (52) mice were generated and maintained as described previously. Mice harboring the floxed Ror2 gene were maintained as described previously (27). Pkn3−/− mice (46) were generated by H. Mukai and appropriately maintained. All procedures for animal care were approved by the Animal Management Committee of Matsumoto Dental University. GST-Rankl and Csf1 were purchased from Oriental Yeast and Kyowa Kirin, respectively. Rhodamine-conjugated phalloidin was from Molecular Probes. All other reagents were from Sigma.

Analysis of bone phenotypes

Micro-CT analysis (ScanXmate-A080, Comscan Tecno) was performed to measure morphological indices in the distal metaphysis of femurs (27). These indices were calculated in trabecular bones located between 0.5 and 1.5 mm from the growth plates using image analysis software (TRI/3D-BON, Ratoc System Engineering). Morphological indices of fifth lumbar vertebrae were calculated in trabecular bones located between 0.76 and 0.9 mm from the ventral surface of the vertebral body. For histomorphometric analysis, the distal half of the femurs was stained with Villanueva bone stain and embedded in glycol methacrylate (PolyScience). The frontal plane sections of the femurs were subjected to histomorphometric examinations as described previously (27). Collagen type I cross-linked CTX and bone-specific alkaline phosphatase activity in serum were measured using enzyme-linked immunosorbent assay (RatLaps, Immunodiagnostic Systems) and a TRACP & ALP Assay kit (Takara Bio), respectively.

Osteoclast formation and bone-resorbing activity assays in vitro

Bone marrow cells (1 × 107 cells) were cultured in α-MEM (minimum essential medium) containing 10% fetal bovine serum (FBS) in the presence of Csf1 (100 ng/ml) on dishes 60 mm in diameter. Nonadherent cells containing osteoclast precursors were harvested and seeded on dentin slices (5.0 × 104 cells per dentin slice). Cells were cultured for 3 days in the presence of Csf1 (50 ng/ml) and cultured further in the presence of Csf1 (50 ng/ml) and GST-Rankl (200 ng/ml). Macrophages as osteoclast precursors were prepared from the liver cells of Wnt5a−/− and wild-type fetuses as described previously (27). After osteoclast formation, the dentin slices were fixed with 3.7% formaldehyde and then treated with 0.1% Triton X-100 in phosphate-buffered saline (PBS). They were incubated in PBS containing rhodamine-conjugated phalloidin to visualize F-actin. The number of actin rings was counted and adjusted by the area of each dentin slice (53, 54). To visualize podosome belts, osteoclasts were cultured on vitronectin-coated cover slide. To count the number of osteoclasts, cells on dentine slices and culture plates were stained for tartrate-resistant acid phosphatase (TRAP) activity, and TRAP-positive cells with more than three nuclei were counted (27). In some experiments, osteoclasts were treated with 10 μM ROCK inhibitor Y27632. After removing cells from the dentin slices, the slices were stained with Mayer’s hematoxylin (Sigma) to visualize resorption pits (53, 54). The area of resorption pits was measured using ImageJ (National Institutes of Health).

Adenovirus gene transfer

Adenoviruses with CA-Rac1 and CA-RhoA were obtained from Cell Biolabs Inc. Adenoviruses expressing shRNA were prepared as follows. Double-stranded oligonucleotides of the target were inserted into RNAi-Ready pSIREN-Shuttle vectors (Takara Clontech). PCR fragments containing a U6 promoter and the target sequences were ligated into pAdenoX-PRLS-ZsGreen vectors (Takara Clontech). The linearized vectors were transfected into human embryonic kidney (HEK) 293T cells to produce adenoviruses according to the manufacturer’s instructions. The target sequences are as follows: Daam2#1, 5′-GGATGAATTGGACCTCACA-3′; Daam2#2, 5′-CTCTCATTGGCTGCATCAA-3′; Daam2#3, 5′-CTCTCATTGGCTGCATCAA-3′; Pkn1#1, 5′-AGGACAGTAAGACCAAGAT-3′; Pkn1#2, 5′-AGGACAGTAAGACCAAGAT-3′; Pkn2#1, 5′-TCCGGATGCAGATTCTTCA-3′; Pkn2#2, 5′-GTCCAAGTGACAACAGATC-3′; Pkn3#1, 5′-TGAGGACTTCCTGGACAAT-3′; Pkn3#2, 5′-CGTTGAAGAAGCAGGAAGT-3′; mDia2#1, 5′-GCACAAAGTCATCCAGTGT-3′; and mDia2#2, 5′-GGCATAACTCAGTGAACCT-3′. Purified adenoviruses were infected into cells at a dose of 50 to 100 multiplicity of infection.

Construction of adenovirus vectors for the expression of DsRed-actin and Venus-Pkn3

pDsRed-Monomer-Actin and Adeno-X Adenoviral System 3 were purchased from Clontech Laboratories. Mouse Pkn3 cDNA was from Thermo Scientific Open Biosystems. The vector for Venus (an EGFP mutant) expression was provided by RIKEN (55). DsRed-fusion human β-actin (DsRed-actin), Pkn3, and Venus were amplified by PCR. To prepare deletion mutants of Pkn3, PCR fragments containing 1 to 1365 nucleotides and fragments containing 1630 to 2637 nucleotides in the open reading frame of Pkn3 were amplified for Pkn3-ΔPRR. Similarly, PCR fragments containing 1 to 1641 and 2422 to 2637 nucleotides in the open reading frame of Pkn3 were amplified for Pkn3-Δkinase. The fragments were linked with pAdenoX vectors using an In-Fusion enzyme (Takara Clontech) according to the manufacturer’s instructions. The ligated plasmid was linearized with Pac I and then transfected into HEK293T cells using the X-tremeGENE 9 (Roche). After amplification of the adenovirus, the adenoviruses were purified using the Adeno-X Maxi Purification Kit (Clontech) according to the method described previously (56).

Rac and Rho activities

Osteoclast precursors (5 × 105 cells) were seeded on 12-well plates and differentiated into osteoclasts with Csf1 and GST-Rankl as described above. Osteoclasts were starved for 8 hours with 2% FBS containing α-MEM without Csf1 or GST-Rankl. Osteoclasts were then stimulated with Wnt5a (100 ng/ml) or vehicle for 5 min and lysed. Lysates were collected according to the manufacturer’s manual. Active Rac and Rho in cell lysates were assayed by G-LISA Rac Activation Assay Biochem Kit and G-LISA Rho Activation Assay Biochem Kit (Cytoskeleton), respectively, according to the manufacturer’s manual.

Src kinase activity

Cell lysates were immunoprecipitated with an antibody recognizing c-Src (Cell Signaling Technology). Tyrosine kinase activity of precipitated proteins was assayed with Universal Tyrosine Kinase Assay Kit (Takara).

RT-PCR analysis

cDNA was synthesized from total RNA and amplified using a PCR thermal cycler (TP600, Takara). RT-PCR was performed using SYBR Green Master Mix (Life Technologies) with the StepOnePlus System (Life Technologies) as described previously (56). Fold-change ratios were calculated between the test and control samples. PCR primers were purchased from Takara Bio Inc.

Immunological analysis

Cell lysates were subjected to an immunoblot analysis using the following antibodies: goat antibody specific for Wnt5a (R&D Systems, AF645, 1:800), mouse antibody specific for RhoA (Millipore, clone 55, 1:1000), mouse antibody specific for Rac1 (Millipore, clone 23A8, 1:1000), rabbit antibody specific for Pkn3 (Abcam, ab155076, 1:1000), rabbit antibody specific for phosphorylated Pkns (Abcam, ab124709, 1:1000), goat antibody specific for EGFP (Abcam, ab111258, 1:1000), mouse antibody specific for Pyk2 (BD Biosciences, 610548, 1:1000), mouse antibody specific for c-Src (Abcam, ab16885, 1:1000), and goat antibody specific for Daam2 (Santa Cruz Biotechnology, sc-68297, 1:500). Antibody specific for Ror2 (57) was from Y. Minami. All immunoblotting analyses were repeated three times.

For immunoprecipitation assays, cell lysates (250 μg) were incubated with 10 μg of antibody specific for EGFP (Abcam) and protein A sepharose (Amersham Biosciences) at 4°C for 16 hours. Then, protein A sepharose was pelleted and samples were eluted by SDS sample buffer. Twenty micrograms (Fig. 6, D and F) or 10 μg (Fig. 6G) of total cell lysates was loaded as input.

Osteoblast-lineage cell differentiation

Osteoblast-lineage cells were obtained from the calvaria of 1-day-old Pkn3−/− mice and wild-type litters according to a previously described method (56). Cells were cultured in α-MEM containing 10% FBS in the presence of β-glycerophosphate (10 mM) and ascorbic acid (50 μg/ml) for 2 weeks (for alkaline phosphatase staining) or 5 weeks (for Alizarin Red S staining).

Statistical analysis

Statistical analyses were performed using the two-tailed Student’s t test and Welch’s t test, when the number of groups was two. If the number of groups was larger than three, an ANOVA and post hoc test (Scheffé test) were used for statistical analyses. Nonparametrical analysis, such as Kruskal-Wallis, Mann-Whitney U test, and Steel-Dwass test, was also used for statistical analysis of normalized data. Each in vitro experiment was repeated at least three times, and similar results were obtained. No sample was excluded from the statistical analysis.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/10/494/eaan0023/DC1

Fig. S1. Rankl-induced formation of osteoclasts from Wnt5a−/− liver macrophages on dentin slices.

Fig. S2. Micro-CT analysis of femurs and lumbar vertebrae in Ror2ΔOcl/ΔOcl mice and bone-resorbing activity of osteoclasts formed from Ror2ΔOcl/ΔOcl mice.

Fig. S3. Effects of suppression of Daam2 on osteoclasts.

Fig. S4. Effects of shRNA-mediated knockdown of Pkns and mDia2 on osteoclasts.

Fig. S5. Micro-CT analysis of femurs and lumbar vertebrae of Pkn3−/− mice.

Fig. S6. Osteoclast and osteoblast differentiation in cultures prepared from Pkn3−/− mice.

Fig. S7. The expression of c-Src, phosphorylation of Pkn3, and schematic of Pkn3.

REFERENCES AND NOTES

Acknowledgments: We thank A. Kikuchi (Osaka University) for providing recombinant Wnt5a and reading the manuscript. We also thank A. Miyawaki (RIKEN Brain Science Institute) for providing pCS2-Venus vectors and T. Ara (Matsumoto Dental University) for providing advice on statistical analysis. Funding: This study was supported by Japan Society for the Promotion of Science KAKENHI grants JP25462904 and JP16K11494 (S.U.), JP24390417 and JP16H05508 (N.U.), JP25221310 and JP16H05144 (N.T.), and JP25293423 and JP16H02691 (Y.K.) and by the Japanese Association for Oral Biology Grant-in-Aid for Young Scientists (S.U.). Author contributions: S.U. and Y.K. performed experiments and prepared the manuscript. A.I., K. Maeda, T.Y., and K. Murakami contributed to in vitro and in vivo experiments and data interpretations. H.M., M.N., Y.M., T.N., and S.K. supported the generation of genetically modified mice and contributed to data interpretations. N.U. and N.T. contributed to data interpretations and preparation of the manuscript. Y.K. directed the project and wrote the manuscript. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: The use of Pkn3−/− mice requires a materials transfer agreement (MTA) from Kobe University. The use of Ror2fl/fl mice requires an MTA from Kobe University. The use of CtskCre/+ mice requires an MTA from Institute of Molecular and Cellular Biosciences, University of Tokyo. The use of pCS2-Venus vectors requires an MTA from RIKEN Brain Science Institute.
View Abstract

Navigate This Article