Bone serves several distinct functions, including supporting and protecting the organs and providing sites for hematopoiesis and mineral storage.1, 2 The coordinated cycle of osteoclast bone resorption and osteoblast bone formation is requisite for maintaining these bone functions.1 The equilibrium of osteoclast and osteoblast activity is tightly regulated,3, 4 and excessive osteoclast activation disturbs this equilibrium and culminates in osteopenic disorders.5 Numerous studies, including detailed in vitro analyses and in vivo genetic investigations, have revealed that the receptor activator of the NF-κB (RANK) signaling is crucial for differentiation and activation of osteoclasts.6–8 Most previous studies have focused on the total amount of RANK ligand (RANKL) expressed in whole osteoblastic cells, including studies focused on transcriptional regulation of RANKL.9–11 However, only RANKL molecules expressed at the cell surface actually can bind to RANK, which is located at the cell surface of osteoclast precursor cells.12–14 Therefore, the quantity of RANKL on the osteoblastic cell surface could determine the magnitude of the signal input and the degree of osteoclastogenesis.
In an effort to understand the regulation of subcellular trafficking of RANKL in osteoblastic cells, we previously showed that most newly synthesized RANKL is transferred from the Golgi apparatus to the lysosomal storage compartment via a route involving vacuolar protein sorting 33a (Vps33a) in osteoblastic cells.15 There is also a minor pathway transporting RANKL from the Golgi apparatus to the plasma membrane.15 In addition, we showed that stimulation of osteoblastic cells with RANK extracellular domain–conjugated beads, which mimics contact with osteoclasts, induces translocation of RANKL from lysosomal storage to the bead interface.15 This finding suggests that RANKL is released from secretory lysosomes to the cell surface in a regulated manner. Furthermore, we recently showed that osteoprotegerin (OPG) associates with RANKL in the Golgi apparatus and that the formation of this complex strengthens the interaction between RANKL and Vps33a, which results in the effective transport of newly synthesized RANKL to the lysosomes without transiting through the plasma membrane.16 In OPG-deficient primary osteoblastic cells, the accumulation of RANKL at the Golgi apparatus correlated with an increase in the amount of RANKL that was leaked to the cell surface.16 In vitro analyses using several OPG mutants revealed that the activity of OPG as a traffic regulator for RANKL requires the latter domain of OPG and is as important as the function of OPG as a decoy receptor for RANKL.16 These facts strongly indicate that regulation of RANKL protein trafficking, especially the release of RANKL from secretory lysosomes, has a physiologic effect on osteoclastogenesis.
In this study, we investigated the molecular mechanisms of RANKL release from secretory lysosomes at the cell surface of osteoblastic cells. Previous studies have established that Rab family GTPases are involved in secretion of the contents of many types of intracellular vesicles. Rab proteins in their active form (GTP-bound form) are carried on the surfaces of cargo vesicles and function as markers for the transport and fusion of these vesicles with the acceptor membrane17 by associating with effector molecules that mediate their interaction with motor proteins,18 phospholipids, or soluble NSF attachment protein receptor (SNARE) components.19 Rab27a, which is expressed in many types of secretory organs, is reported to be involved in secretion from lysosomal organelles such as secretory granules in insulin-secreting cells,20 dense granules in platelets,21 dense-core vesicles in neuroendocrine cells,22 melanosomes in melanocytes,23 and lytic granules in cytotoxic T-lymphocytes (CTLs).24 Several Rab27a effector molecules also have been identified. In melanocytes, Rab27a and its effector, melanophilin, are involved in the transport of melanosomes along actin filaments,25 and the subsequent tethering, docking, and fusion steps at the plasma membrane are regulated by Rab27a and another effector, Slp2-a.26 Likewise, the secretion of lytic granules from CTLs is also regulated by Rab27a in concert with the effector molecules Slp1, Slp2-a, and Munc13-4 through a series of events including tethering, docking, and fusion of lytic granules to the plasma membrane.27–31 Previous studies also have shown that Rab27b, a closely related Rab27a homologue, is expressed in many types of cells and shares effector molecules with Rab27a.31, 32 There are some reports showing that both Rab27a and Rab27b are involved in the same vesicular trafficking events. However, the function of these two isoforms is not necessarily redundant, and they may regulate different steps in a series of events in some cases.31, 33
This study focuses on Rab27a and Rab27b as candidate molecules involved in RANKL release from the secretory lysosomes of osteoblastic cells. Measurement of the RANKL release in response to stimulation showed that RANKL release was regulated by Rab27a and Rab27b. Among Rab27a/b effector molecules, Slp4-a, Slp5, and Munc13-4 also were shown to be involved in the RANKL release process. Moreover, the physiologic significance of stimulation-dependent RANKL release in vivo was addressed by analyzing the bone phenotype of Jinx mice, which have a loss-of-function mutation in the Munc13-4 gene.
Materials and Methods
Jinx mice (homozygote) were obtained from the Mutant Mouse Regional Resource Centers (MMRRC, Davis, CA, USA) and propagated. Wild-type control mice (C57/BL6) were supplied by Japan SLC, Inc (Shizuoka, Japan). All procedures involving experimental animals were carried out using protocols approved under the local institutional guidelines for animal care at the University of Tokyo.
ST2 cells, obtained from Riken Cell (Tsukuba, Japan), were cultured in α modified essential medium (α-MEM) containing 10% fetal bovine serum (FBS; Invitrogen, Carlsbad, CA, USA) and penicillin-streptomycin (PCSM, Invitrogen). MC3T3-E1 cells, obtained from American Type Culture (Rockville, MD, USA), were cultured in α-MEM containing 10% FBS and PCSM. Mouse bone marrow (BM) cells were collected from the tibia bone marrow of 8- to 10-week-old C57BL6 mice using previously reported methods,34 and these cells were subjected to coculture experiments. Primary osteoblastic cells (POBs) were collected from the calvaria of 2- to 3-day-old wild-type C57BL6 or Jinx mice using previously reported methods.35
RNA interference was used to suppress Rab27a, Rab27b, and their effector molecules by transient transfection with specific siRNAs directed against each gene using Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. The sequences selected as targets for each gene were designed by BLOCK-iT RNAi Designer on the Web (Invitrogen) and are as follows: 5'-GCU UAA CCA CUG CAU UCU U-3' for Rab27a, 5'-CCA GUC AAC AGA GCU UCU U-3' for Rab27b, 5'-CCA UCU CAG GAG AGG CUU U-3' for Slp1, 5'-GCG UUU ACA GUG GAG ACU U-3' for Slp2-a, 5'-GCU UGC UUC CCA GGU GAA U-3' for Slp4-a, 5'-GCG UUU CAA GCA AGU CAA U-3' for Slp5, and 5'-GGA GAA CUU CAG CAG CCU U-3' for Munc13-4. siRNA duplexes with a dideoxythymine overhang at the 3' terminal of each oligo were synthesized (Sigma Aldrich, St Louis, MO, USA) and used for transfection. The siPerfect Negative Control (Sigma Aldrich), which contains a sequence with a minimum of three mismatches against human, mouse, and rat genes, was used as a negative control.
Construction and transfection of expression vectors
Mouse RANKL, Rab27a, Rab27b, Rab3a, Slp1, Slp2-a, Slp4-a, Slp5, and Munc13-4 were cloned from cDNA from ST2 cells. Each gene was subcloned into pEGFP-C1 vector, pAcGFP-C1 vector, or pDsRed-monomer-Hyg-C1 vector. Plasmids encoding Slp4-a, Slp5, and Munc13-4 fused with His-tag and green fluorescent protein (GFP) at the N-terminal were constructed by site-directed mutagenesis against pEGFP-based constructs. Plasmids encoding RANKL fused with His-tag were subcloned into pcDNA3.1 vector. Each construct was introduced into ST2 cells using the GeneJuice (Novagen, Walkersville, MD, USA) or MultiFectam (Promega, San Luis Obispo, CA, USA) transfection reagent according to the manufacturer's protocol.
Live cell staining
Cells transiently expressing GFP-fused constructs were stained with LysoTracker Red (Cambrex, Walkersville, MD, USA) according to the manufacturer's protocol. Fluorescence was detected using a confocal laser-scanning microscope (Fluoview FV1000; Olympus, Melville, NY, USA) equipped with a UPLSAPO 100 × O objective lens (numerical aperture 1.40; Olympus).
Preparation of lentiviruses
GFP-RANKL was subcloned into the pLenti6.3/V5-DEST vector (Invitrogen). According to the manufacturer's protocol, lentival vectors and other packaging plasmids were transfected into 293FT cells using Lipofectamine 2000. Twenty-four hours after transfection, sodium butyrate was added to the culture medium to a final concentration of 10 mM. Forty-eight hours after transfection, the culture medium was collected as viral solution.
Confirmation of expression and mRNA quantification
To check the endogenous expression of genes, reverse transcription and PCR were performed using cDNA from osteoblastic cells, GoTaq PCR mix (Promega), and the following primers: 5'-TTC CTC AAT GTC CGA AAC TGG AT-3' and 5'-ACC GTA CCA CTC CCT CCG-3' for mouse Rab27a, 5'-TGG ACA AGA GCG GTT CCG-3' and 5'- CCC TTT GGT CTG GCA GGT CT-3' for mouse Rab27b, 5'-ATG TCC TCA AAC GAG ATG CC-3' and 5'-GGA CGA GGT CAG AAC CAA AA-3' for mouse Slp1, 5'- GGC AAT TCT TAA AGA CGC AGA A-3' and 5'- TGC TGT CCC AAT CCC ACG-3' for mouse Slp2-a, 5'-GAA AGT GCA ATC CGT ATG TCA A-3' and 5'-CGA AAG TCG GGT CCA GCG-3' for mouse Slp3-a, 5'-TCC TTC CCA TGA GGA ACA AG-3' and 5'-CCG GTC CCA TAC AGT CAG TT-3' for mouse Slp4-a, 5'-GTC TGT GCC AAA GTC GCG -3' and 5'-CGA CAT CAG TGC CCA GAA CA-3' for mouse Slp5, 5'-GAA TCC AGC AGC AGG CCG-3' and 5'-GGG CAG CAG GCT AGA AGC AC-3' for mouse Munc13-4, and 5'- TTC AAC ACC CCA GCC ATG TAG G-3' and 5'-GTG GTG GTG AAG CTG TAG CC-3' for β-actin. mRNA was quantified using quantitative real-time PCR with SYBR GreenER qPCR SuperMix Universal (Invitrogen), Chromo4 (Bio-Rad, Hercules, CA, USA) and the associated software, and the primers listed above.
Capture of RANKL released on stimulation
Polystyrene beads coated with RANK-Fc (RANK beads) were prepared as described previously.15 ST2 cells were seeded onto 12-well plates at a density of 1 × 104 cells per well and treated with siRNA for each gene. The culture medium was replenished 12 hours after siRNA transfection, and lentivirus encoding GFP-RANKL was added to the medium 24 hours after siRNA transfection. Forty hours after lentiviral infection, 100 µg of RANK beads was added to each well and incubated for 12 hours. The cells were washed with PBS and incubated in PBS containing 10 mg/mL of soluble RANKL for 30 minutes on ice to mask the unbound RANK molecules on the beads. Then cells were lysed in lysis buffer [10 mM Tris/acetic acid (pH 7.8), 0.25 M sucrose, 1 mM EDTA, 0.2% NP-40, and protease inhibitors]. An aliquot of the lysate was sampled to measure RANKL expression in whole-cell lysate. The remainder was subjected to iodixanol gradient centrifugation to purify the beads. The lysate and purified beads were subjected to immunoblotting using rabbit anti-GFP antibody (Invitrogen) and ECL Plus reagent (GE Healthcare Bioscience, Buckinghamshire, UK). Finally, the amount of GFP-RANKL captured by RANK beads was compared.
RANK bead pull-down assay
ST2 cells were seeded onto 12-well plates at a density of 1 × 104 cells per well, and lentivirus encoding GFP-RANKL was added to the medium 12 hours after seeding. Twelve hours after lentiviral infection, the plasmids encoding GFP-fused constructs were transfected into the cells. Forty hours after transfection, 100 µg of RANK beads was added to each well and incubated for 6 hours. The cells were washed with PBS and incubated on ice in PBS containing 10 mg/mL of soluble RANKL for 30 minutes to mask the unbound RANK molecules on the beads. Then cells were collected in lysis buffer [10 mM Tris/acetic acid (pH 7.8), 0.25 M sucrose, 1 mM EDTA, 0.02% NP-40, and protease inhibitors] and homogenized with one freeze-thaw cycle. An aliquot of lysate was taken to measure RANKL expression in whole-cell lysate. The remainder of the lysate was subjected to iodixanol gradient centrifugation to purify the beads. The collected beads were gently washed in PBS containing 0.02% NP-40. The lysate and purified beads were then subjected to immunoblotting using rabbit anti-GFP antibody and ECL Plus reagent. Finally, the GFP-fused proteins captured by RANK beads were detected.
Observation of RANKL and lysosomal contents after stimulation
Silica microspheres coated with RANK-Fc were prepared as follows: One milligram of streptavidin functionalized silica microspheres (Polyscience, Morin Heights, Quebec, Canada) was incubated with 0.67 µg of biotinylated Protein G (Thermo Scientific, Bremen, Germany) in PBS (pH 8.0). The beads were washed after incubation and further incubated with 2.7 µg of RANK-Fc chimera (R&D Systems, Minneapolis, MN, USA) in PBS (pH 8.0). The beads then were washed with PBS and used as RANK beads for confocal observation. ST2 cells were seeded onto 35-mm glass-bottomed dishes at a density of 4 × 104 cells per well and treated with siRNA for each gene. The culture medium was replenished 12 hours after siRNA transfection, and GFP-RANKL was transfected into the cells 24 hours after siRNA transfection. To visualize lysosomal components, 38 hours after lentiviral infection, cells were incubated in medium containing Alexa Fluor 546–labeled dextran (Invitrogen) for 1 hour. Then the cells were incubated in fresh medium without dextran for 1 hour. Finally, about 30 µg of RANK beads for confocal observation was added into the culture medium, and the cells were incubated for 12 hours and observed using a confocal microscope.
Observation of RANKL and other proteins after stimulation
RANK beads for confocal observation were prepared as described earlier. ST2 cells were seeded onto 35-mm glass-bottomed dishes at a density of 3 × 104 cells per well, and DsRed-RANKL and GFP-fused constructs were transfected into the cells the next day. Then 30 µg of RANK beads for confocal observation was added to the medium and incubated for 6 hours, and cells then were observed using a confocal microscope.
His-tag pull-down assay
Slp4-a, Slp5, or Munc13-4 fused with a His-tag and GFP at the N-terminal and Rab27a, Rab27b, or Rab3a fused with GFP at N-terminal were cointroduced into ST2 cells. A cell lysate was prepared using lysis buffer [50 mM HEPES/KOH (pH 7.2), 150 mM NaCl, 1% Triton X-100, and protease inhibitors] and then incubated with the Profinity IMAC Ni resin (Bio-Rad) to purify His-tagged proteins. The captured proteins were eluted by 300 mM imidazole, and the eluate was analyzed by immunoblotting using anti-GFP antibody.
Fluorescence resonance energy transfer (FRET) assay
ST2 cells transiently expressing AcGFP-fused constructs, DsRed-fused constructs, and His-tagged RANKL were prepared by cotransfection of plasmids. The subcellular localization of AcGFP-fused constructs and DsRed-fused constructs was detected using excitation wavelengths of 488 and 543 nm, respectively. To determine the FRET region, the subcellular localization of the DsRed fluorescence excited by the GFP-fused constructs was observed by detecting the DsRed fluorescence excited only by the 488-nm excitation wavelength. The FRET efficiency was determined as a relative ratio of the intensity of DsRed fluorescence excited by 488-nm waves to that of GFP fluorescence. FRET occurrence was confirmed by Acceptor photobleaching method (data not shown). The pixels exhibiting maximum FRET efficiency in a whole cell were set to the maximum of a pseudo–color scale.
The coculture of ST2 cells with BM cells followed by the tartrate-resistant acid phosphatase (TRACP) assay was performed as described previously with some modifications.15 ST2 cells were seeded with simultaneous lipofection of siRNA using the reverse-transcription method described in the manufacturer's protocol. On day 1, ST2 cells treated with siRNA were trypsinized and counted. ST2 cells treated with siRNA (7.5 × 103 cells/well) were cocultured with BM cells (7.5 × 104 cells/well) in the presence of 10 nM 1α,25-dihydroxyvitamin D3 [1α,25(OH)2D3] and 100 nM dexamethasone (Dex) on a 96-well plate. These cells were cultured for 6 days, and the culture medium was replaced every 2 or 3 days. On day 7, a colorimetric TRACP assay was performed using a TRACP and ALP Assay Kit (TaKaRa, Tokyo, Japan) according to the manufacturer's protocol. In this assay, TRACP activity is quantitated by measuring p-nitrophenol production from p-nitrophenyl phosphate in an acidic buffer with tartrate. Coculture of POBs from wild-type or Jinx mice with BM cells from wild-type mice followed by the TRACP assay was performed using the same methods, except that POBs were used without siRNA treatment. Since RANKL expression level in the whole cell is a factor that influences TRACP activity, endogenous RANKL expression level was measured. ST2 cells pretreated with siRNA and POBs from wild-type and Jinx mice were cultured for 4 days in the presence of 10 nM 1α,25(OH)2D3 and 100 nM Dex and then collected and lysed in lysis buffer [PBS (pH 8.0) supplemented with 0.1% NP-40 and protease inhibitors]. The lysate then was subjected to immunoblotting using rabbit anti-RANKL antibody (PeproTech, Rocky Hill, NJ, USA) and ECL Plus reagent.
Osteoclast differentiation assay
BM cells were collected from tibial bone marrow and cultured in α-MEM supplemented with 5 ng/mL of macrophage colony-stimulating factor (M-CSF) for 17 to 18 hours. Nonadherent cells were used as BM macrophages (BMMs). BMMs (7.5 × 103 cells/well) were seeded on a 96-well plate and cultured in α-MEM supplemented with 40 ng/mL of M-CSF and 100 ng/mL of sRANKL for 7 days, with the culture medium replenished every 3 days. After the culture period, a colorimetric TRACP assay was performed using a TRACP and ALP Assay Kit (TaKaRa) according to the manufacturer's protocol.
Analysis of skeletal morphology
All analyses were performed as described previously with minor modifications.36 All histologic analyses were carried out on 8-week-old male C57BL6 mice having wild-type Munc13-4 gene (wild-type mice) or Jinx/Jinx mutation (Jinx mice). Undecalcified sections of tibias were subjected to toluidine blue staining. Using the section stained by toluidine blue, osteoclast number, osteoblast number, erosion surface, and bone surface were measured. 3D micro–computed tomographic (µCT) scans were taken using a composite X-ray analysis system (ScanXmate-A080; Comscantecno, Kanagawa, Japan). Trabecular bone parameters were measured in an area 500 µm in length from 300 µm below the growth plate at the proximal metaphysis of the tibia.
Measurement of serum bone metabolic markers
Serum was collected from 8-week-old male wild-type and Jinx mice (homozygote). Serum type I collagen cross-linked N-telopeptide (NTX) was measured as a marker for bone resorption using an Osteomark NTX Serum ELISA kit (Inverness Medical, Princeton, NJ, USA), as described previously.37
All data are expressed as means ± SD from at least three independent experiments. Statistical analysis was performed using a Student's t test or an analysis of variance (ANOVA) followed by Bonferroni's test where applicable.
Rab27a and Rab27b are involved in stimulation-dependent RANKL release
We first sought to determine the involvement of Rab27a and Rab27b in stimulation-dependent RANKL release in osteoblastic cells. RT-PCR was performed to confirm the expression of Rab27a and Rab27b using mRNA derived from POBs and from two mouse osteoblastic cell lines, ST2 cells and MC3T3-E1 cells (Fig. 1A). Next, we investigated the involvement of Rab27a and Rab27b in stimulation-dependent RANKL release from secretory lysosomes using RNAi gene-suppression experiments. Knockdown efficiency was confirmed by measuring mRNA levels in ST2 cells treated with siRNAs against Rab27a or Rab27b using the real-time PCR system (Fig. 1B). To check the knockdown efficiency at the protein level, the levels of N-terminal GFP-fused Rab27a or Rab27b (GFP-Rab27a or GFP-Rab27b, respectively), which were introduced into ST2 cells pretreated with each siRNA, were assessed (Fig. 1C). The amount of GFP-Rab27a or GFP-Rab27b was decreased to at least less than half by each siRNA treatment, suggesting that these siRNA oligos were functional at the protein level. Cell viability also was assessed to confirm that siRNA treatment did not affect the cell viability (data not shown). We measured RANKL release to the plasma membrane on stimulation with RANK-Fc-conjugated beads (RANK beads) in ST2 cells that were pretreated with siRNA against Rab27a or Rab27b and were infected with lentiviruses encoding RANKL fused with GFP at the N-terminal (GFP-RANKL). The cells were incubated with RANK beads in the culture medium for 6 hours, and then proteins captured by the beads were collected as described under “Materials and Methods.” Under the suppression of Rab27a or Rab27b, the amount of RANKL captured by RANK beads was decreased, indicating that Rab27a and Rab27b are involved in stimulation-dependent RANKL release (Fig. 1D). To confirm these results, we observed the subcellular localization of RANKL in response to RANK stimulation in ST2 cells. First, to confirm that RANKL captured by the RANK beads was derived from lysosomes, the activity of lysosome-associated membrane protein 1 (LAMP-1), a lysosomal marker protein whose trafficking to lysosomes is mediated by an YXXϕ motif and differs from RANKL trafficking, was observed before and after RANK stimulation. GFP-RANKL introduced into ST2 cells was colocalized predominantly with LAMP-1 fused with Kusabira orange at the C-terminal (LAMP-1-KuOr) under the nonstimulated condition (Fig. 1E). When ST2 cells expressing GFP-RANKL and LAMP-1-KuOr contacted the RANK beads, both GFP-RANKL and LAMP-1-KuOr accumulated at the interface (Fig. 1E), indicating that the lysosomal contents were released to the cell surface on RANK stimulation. Subsequently, to examine the effects of Rab27a or Rab27b suppression, the RANKL subcellular localization after stimulation was observed in ST2 cells pretreated with siRNA against Rab27a or Rab27b. Alexa Fluor 546–labeled dextran was preloaded in lysosomes to determine whether lysosomal vesicles were fused to the plasma membrane or gathered beneath the plasma membrane at the interface with the RANK beads. After a 12-hour incubation with RANK beads, GFP-RANKL accumulated at the interface with the beads, and only a small amount of dextran was observed in the same area under control conditions, suggesting that lysosomal contents were released into the medium (Fig. 1F). However, GFP-RANKL-containing vesicles tended to cluster around RANK beads, and dextran was retained inside these vesicles with either Rab27a or Rab27b suppression (Fig. 1F). These results suggest that both Rab27a and Rab27b play an important role in stimulation-dependent RANKL release and especially in the fusion of RANKL-containing lysosomes to the plasma membrane.
Rab27a and Rab27b reside where RANKL is released on stimulation
Next, the subcellular behavior of Rab27a and Rab27b in ST2 cells was investigated. We first confirmed that the stimulation-dependent RANKL release was mediated by Rab27a and Rab27b by the following rescue experiment using GFP-fused Rab27a and Rab27b (GFP-Rab27a and GFP-Rab27b, respectively): ST2 cells treated with siRNAs against Rab27a or Rab27b and infected with lentiviruses encoding GFP-RANKL were transfected with a plasmid encoding GFP-Rab27a or GFP-Rab27b with silent mutations in the target region of siRNA (GFP-Rab27a* and GFP-Rab27b*, respectively). The cells then were incubated with the RANK beads in the culture medium for 6 hours, and the amount of RANKL release was compared. GFP-Rab27a*- or GFP-Rab27b*-introduced cells showed an increase in RANKL release compared with cells that did not express the mutant constructs, suggesting that Rab27a and Rab27b are surely involved in the stimulation-dependent RANKL release (Fig. 2A). It also was confirmed by this result that GFP-Rab27a or GFP-Rab27b retains the function of endogenous Rab27a or Rab27b in stimulation-dependent RANKL release. The subcellular behavior of Rab27a or Rab27b therefore was assessed using GFP-fused constructs. Under nonstimulated conditions, both GFP-Rab27a and GFP-Rab27b were dispersed throughout the cytosol and partially associated with lysosomes, which was confirmed by live-cell staining with a lysosomal marker, LysoTracker Red (Fig. 2B). In addition, the subcellular localization of GFP-Rab27a and GFP-Rab27b was compared with that of RANKL in ST2 cells using a fusion protein construct with DsRed at the N-terminal (DsRed-RANKL). DsRed-RANKL was localized mainly to lysosomes, consistent with our previous results,15 and was colocalized with GFP-Rab27a and GFP-Rab27b in lysosomes (Fig. 2C). Next, the localization of DsRed-RANKL and GFP-Rab27a or GFP-Rab27b after RANK stimulation was investigated. Consistent with our previous results using GFP-RANKL (Fig. 1E, F; see also ref. 15), when ST2 cells were stimulated with RANK beads, DsRed-RANKL accumulated at the interface with the beads (Fig. 2D). At the same time as RANKL accumulated, GFP-Rab27a and GFP-Rab27b also resided at the interface with the beads (Fig. 2D). To confirm these observations, the proteins included in the fraction pulled down by RANK beads were analyzed. Briefly, ST2 cells expressing GFP-RANKL and either GFP-Rab27a or GFP-Rab27b were incubated with RANK beads. Cells were lysed under very mild conditions to collect the proteins that accumulated in a specific membrane domain, such as the area adjacent to the beads, but that did not necessarily form protein complexes. The lysate was subjected to iodixanol gradient centrifugation to recover the RANK beads and collect the proteins residing beneath the beads. Rab3a, one of the Rab GTPase members closely related to Rab27a and Rab27b, was used as a negative control protein to show selectivity to Rab27a and Rab27b in this releasing events. While GFP-Rab3a was not detected in the captured protein fractions, both GFP-Rab27a and GFP-Rab27b were coprecipitated with RANK beads and RANKL (Fig. 2E). These results suggest that both Rab27a and Rab27b reside in the area where RANKL is released on stimulation.
Slp4, Slp5, and Munc13-4 are candidate Rab27a/b effectors involved in stimulation-dependent RANKL release
It has been established that Rab GTPases function in concert with effector molecules. Because the suppression of Rab27a and Rab27b resulted in a marked reduction in RANKL release from lysosomal vesicles accumulating near RANK beads, we focused on the effector molecules involved in the fusion of secretory lysosomes to the plasma membrane. Among the 11 known effector molecules for Rab27a and Rab27b, five Slp family members and Munc13-4 appeared likely to match the preceding criteria.38 Expression of these effector molecules was investigated in POBs, ST2 cells, and MC3T3-E1 cells using RT-PCR, and it was determined that Slp1, Slp2-a, Slp4-a, Slp5, and Munc13-4 were expressed in these cells (Fig. 3A). Then the molecules involved in membrane fusion of secretory lysosomes in stimulation-dependent RANKL release were identified by siRNA gene silencing. Knockdown efficiency was confirmed by measuring the mRNA level of each effector molecule in siRNA-treated ST2 cells using the real-time PCR system (Fig. 3B). To assess the knockdown efficiency at the protein level, the expression level of each introduced effector molecules fused with GFP at the N-terminal (GFP-Slp1, GFP-Slp2-a, GFP-Slp4-a, GFP-Slp5, and GFP-Munc13-4) was measured after pretreatment of ST2 cells with siRNAs specific for each effector (Fig. 3C). The amount of each GFP-fused construct was decreased to less than half by each siRNA treatment, suggesting that these siRNA oligos were functional at the protein level, and they were subsequently used in the experiments described. In addition, siRNA treatment did not affect the cell viability in comparison with untreated control cells (data not shown). RANKL release after stimulation was measured in siRNA-treated ST2 cells. While suppression of Slp1 and Slp2-a did not affect the amount of RANKL captured by the RANK beads, suppression of Slp4-a, Slp-5, and Munc13-4 resulted in a decrease in the amount of RANKL captured by the beads (Fig. 3D). In addition, the subcellular behavior of RANKL was observed after RANK stimulation in ST2 cells pretreated with siRNA against Slp4-a, Slp5, and Munc13-4. Under these conditions, GFP-RANKL-containing vesicles tended to localize near RANK beads, but dextran preloaded in lysosomes was retained in the same vesicles (Fig. 3E). These results are consistent with the results obtained under Rab27a- and Rab27b-suppressed conditions, suggesting that Slp4-a, Slp5, and Munc13-4 function as effector molecules for Rab27a and Rab27b in the fusion of RANKL-containing lysosomes to the plasma membrane.
Slp4, Slp5, and Munc13-4 reside where RANKL is released after stimulation
As further confirmation of the involvement of these effectors in RANKL release, we determined the subcellular localization of Slp4-a, Slp5, and Munc13-4 after RANK stimulation in ST2 cells transfected with DsRed-RANKL and effector molecules fused with GFP at the N-terminal (GFP-Slp4-a, GFP-Slp5, and GFP-Munc13-4, respectively). In each case, the effector molecule was partially localized to the area adjacent to the RANK beads when RANKL was released to the interface with the beads (Fig. 4A). To examine the contents of the fraction pulled down by the RANK beads in ST2 cells transfected with GFP-RANKL and effector molecules, the proteins associated with the lipid membrane just beneath the beads, in addition to the proteins interacting with the RANK beads by protein-protein interaction, were collected. We used N-terminal GFP-fused asialoglycoprotein receptor 1 (ASGR-1), which is expressed mainly at the plasma membrane, as a reference protein to assess nonspecific contamination of the plasma membrane fraction. While ASGR-1 was not detected, Slp4-a, Slp5, and Munc13-4 were detected together with RANKL as proteins captured by the RANK beads (Fig. 4B). These results suggest that Slp4-a, Slp5, and Munc13-4 reside adjacent to the RANK beads. If these effector molecules are involved in stimulation-dependent RANKL release, the interaction between each effector molecule and each Rab27 isoform in osteoblastic cells should be detected. To confirm this, His-tag pull-down assays were carried out. ST2 cells were transfected with plasmids encoding each GFP-Rab and each effector molecule fused with His-tag and GFP (His-GFP-Slp4-a, His-GFP-Slp5, AND His-GFP-Munc13-4). The cells were lysed, and His-tagged proteins in the lysate were purified using an affinity resin. Rab27a and Rab27b, but not Rab3a, copurified with Slp4-a, Slp5, or Munc13-4, suggesting that these effector molecules physically interact with Rab27a and Rab27b in osteoblastic cells. To confirm the interaction between each effector and Rab27a or Rab27b and to identify the region where this interaction occurs, fluorescence resonance energy transfer (FRET) analysis was performed using N-terminal AcGFP-fused effectors (AcGFP-Slp4-a, AcGFP-Slp5, and AcGFP-Munc13-4) and N-terminal DsRed-fused Rab27a/b (DsRed-Rab27a or DsRed-Rab27b). ST2 cells were transfected with AcGFP-fused effectors, DsRed-Rab27a/b, and His-tagged RANKL. Although the FRET signal also was detected at several cellular compartments other than the interface with RANK beads, a clear FRET signal was observed between Rab27a and each effector and was minimal between Rab27b and each effector at the interface with the beads (Fig. 4D). Together, these results suggest that Slp4-a, Slp5, and Munc13-4 are functional in the region where RANKL is released after stimulation.
Suppression of stimulation-dependent RANKL release decreases osteoclastogenic activity in osteoblastic cells
We next investigated the physiologic significance of stimulation-dependent RANKL release in vitro. A coculture assay system for BM cells using ST2 cells pretreated with siRNA for Slp4-a, Slp5, and Munc13-4 was used to determine the effects of suppression of these genes on osteoclastogenesis in osteoblastic cells. First, we confirmed that endogenous RANKL expression was not affected by siRNA treatment (Fig. 5A). TRACP activity was compared in each coculture system as an indicator of osteoclast formation. In cocultures with ST2 cells treated with siRNA against Slp4-a, Slp5, and Munc13-4, TRACP activity was significantly decreased compared with controls (Fig. 5B). Jinx mice, which have a loss-of-function mutation in the Munc13-4 gene and lack mature Munc13-4 protein,39 also were used to investigate the physiologic significance of stimulation-dependent RANKL release. Several experiments using POBs from Jinx mice and wild-type control mice were performed. First, RANKL release was measured in response to stimulation with RANK beads. POBs derived from Jinx mice showed lower RANKL capture than POBs from wild-type mice (Fig. 5C). In addition, we determined the subcellular localization of RANKL after RANK stimulation in POBs from Jinx and wild-type mice. In POBs from Jinx mice, GFP-RANKL-containing vesicles tended to accumulate around RANK beads, and dextran preloaded in lysosomes also was retained inside these vesicles, whereas in POBs from wild-type mice, dextran preloaded in lysosomes was released normally after stimulation (Fig. 5D). The osteoclastogenic ability of POBs from Jinx and wild-type mice also was assessed. We confirmed that endogenous RANKL expression was similar in POBs from Jinx and wild-type mice (Fig. 5E). Both types of POBs were cocultured with BM cells from wild-type mice, and TRACP activity was measured. Consistent with the results from the coculture of ST2 cells treated with siRNA against Munc13-4, TRACP activity was significantly decreased in BM cells cocultured with POBs from Jinx mice compared with cocultures with POBs from wild-type mice (Fig. 5F). To examine whether the Jinx mutation in the Munc13-4 gene affects the autonomous differentiation of osteoclasts, an osteoclast differentiation assay was performed using BMMs derived from wild-type and Jinx mice. After a 7-day culture of BMMs in the presence of soluble RANKL and M-CSF, TRACP activity was measured. The TRACP activity level in BMMs derived from Jinx mice was almost identical to the activity in wild-type mice (Fig. 5G). These results suggest that absence of the molecule involved in stimulation-dependent RANKL release impairs osteoclastogenesis owing to osteoclastogenic ability in osteoblastic cells but does not suppress the differentiation mechanisms of osteoclasts.
Jinx mice exhibit increased bone volume
We next examined the physiologic significance of stimulation-dependent RANKL release in vivo. Since Munc13-4 suppression in osteoblastic cells resulted in decreased osteoclastogenic ability in vitro, the bone phenotypes of Jinx mice were analyzed. We next analyzed the morphology of tibias from 8-week-old male Jinx and wild-type mice using µCT. 3D images of bone sections (500 µm thickness, from 300 to 800 µm from the growth plate) were reconstituted (Fig. 6A). The bone volume per tissue volume in trabecular bone and the number of trabecular bones were significantly higher in Jinx mice, whereas the separation between trabecular bones was lower in these mice (Fig. 6B–E). Undecalcified sections of tibias from 8-week-old male Jinx and wild-type mice were subjected to toluidine blue staining. The number of osteoclasts or osteoblasts per trabecular perimeter and the erosion surface per bone surface were compared. Osteoclast number and erosion surface were decreased in Jinx mice (Fig. 6F, G), whereas osteoblast number was not affected (Fig. 6H). Serum NTX was measured as a marker for bone resorption. Serum NTX was lower in Jinx mice than in wild-type mice (Fig. 6I). These results suggest that Jinx mice exhibit increased bone volume owing to lower bone-resorptive activity compared with wild-type mice in the region close to the tibial metaphysis, as suggested by the in vitro results.
Osteoblasts can support osteoclastogenesis by expressing RANKL molecules on their surface,12–14 suggesting the importance of both RANKL expression levels at the whole-cell level and RANKL subcellular localization. While numerous studies have focused on RANKL transcriptional regulation, contributing to an understanding of the regulatory mechanisms of RANKL expression at the whole-cell level,40–42 the subcellular localization of RANKL in osteoblastic cells has not been widely investigated. Therefore, we have analyzed the subcellular localization of RANKL and found that RANKL is localized predominantly to the secretory lysosomes and is released to the cell surface on stimulation.15 It appears that another route to the cell surface also exists, by which a small amount of newly synthesized RANKL in the Golgi apparatus is transported directly to the cell surface.15 Since more than 90% of RANKL molecules were localized to secretory lysosomes in the absence of stimulation,15 we focused on stimulation-dependent RANKL release and examined the physiologic significance of this pathway by investigating its molecular mechanisms.
The small GTPases Rab27a and Rab27b were considered to be candidate molecules involved in stimulation-dependent RANKL release because they are reported to be involved in the stimulation-dependent release of contents from lysosomal organelles in many other kinds of cells.32, 43, 44 In osteoblastic cells, suppression of either Rab27a or Rab27b inhibited stimulation-dependent RANKL release (Figs. 1 and 2). Previous reports showed that Rab27a and Rab27b are likely to exist in their activated form (GTP-bound form) at the basal condition.45 Therefore, Rab27a and Rab27b may not be responsible for triggering RANKL release in response to the RANK-mediated activation of a still-unknown signaling pathway, but rather osteoblastic cells may use the already-activated machinery of the Rab27-effector complex in stimulation-dependent RANKL release. We also showed that Slp4-a, Slp5, and Munc13-4, which are known effector molecules for Rab27a and Rab27b, are involved in this pathway. The physiologic importance of stimulation-dependent RANKL release in bone homeostasis was addressed by examining Jinx mice lacking functional Munc13-4. The results showed that stimulation-dependent RANKL release was reduced in POBs derived from Jinx mice, resulting in reduced osteoclastogenic ability (Fig. 5). Consistent with these results, bone volume was increased in the proximity of the tibial metaphysis in Jinx mice, and serum NTX was decreased in Jinx mice (Fig. 6). Collectively, these data suggest that stimulation-dependent RANKL release plays a role in osteoclast activation in vivo. However, the phenotype of Jinx mice, which have impaired stimulation-dependent RANKL release, is not as marked as that of RANKL-deficient mice. RANKL-deficient mice show complete loss of the BM cavity.7 In Jinx mice, the BM is formed (Fig. 6B). This difference is probably due to the presence of RANKL molecules remaining at the cell surface despite Munc13-4 disruption, which may occur through the direct-release pathway going from the Golgi apparatus to the cell surface. In order to fully understand the physiologic impact of RANKL subcellular trafficking, it will be necessary to also elucidate the molecular mechanisms of this direct pathway. Another possible explanation for the relatively mild bone phenotype in Jinx mice is a deficiency in the release of specific hormones that are mediated by Munc13-4, which would indirectly affect bone homeostasis.
It has been reported that Rab27a and its effectors are involved in the release of Fas ligand, which is stored in lytic granules in cytotoxic T-lymphocytes.28, 44 The regulation of RANKL in osteoblastic cells is similar to that of FasL in cytotoxic T-lymphocytes; however, there are also some differences.24, 46–49 Both Rab27a and Rab27b are involved in stimulation-dependent RANKL release in osteoblastic cells (Figs. 1 and 2). In contrast, most studies investigating the release of FasL stored in lytic granules in cytotoxic T-lymphocytes have focused on Rab27a,27–30 and involvement of Rab27b has been reported rarely. This may be due to the low expression of Rab27b in hematopoietic cells.50, 51 In osteoblastic cells, there seems to be no difference in the subcellular behavior of Rab27a and Rab27b (Figs. 1 and 2). The functional differences between Rab27a and Rab27b in osteoblastic cells will be discussed in detail later. Another point of comparison is the identity of effector molecules acting in concert with Rab27a/b. Slp4-a, Slp5, and Munc13-4 are involved in stimulation-dependent RANKL release, although Slp1 and Slp2-a are also expressed in osteoblastic cells (Fig. 3A). This is in marked contrast to the mechanism of lytic granule secretion in hematopoietic cells, in which Slp1, Slp2-a, and Munc13-4, but not Slp4-a and Slp5, have been reported to be involved.28, 35
As mentioned earlier, there are functional differences between Rab27a and Rab27b. There are reports suggesting that the Rab27 isoforms are functionally redundant and can compensate for each other.51 However, analysis of mice lacking Rab27a and/or Rab27b function has shown that Rab27a and Rab27b appear to have an additive effect on 5-hydroxytriptamine release from platelet dense granules.52 Also, recent analysis of Rab27a and Rab27b null mice has shown that Rab27a and Rab27b regulate NADPH oxidase release in neutrophils through independent mechanisms.33 In stimulation-dependent RANKL release in osteoblastic cells, the function of Rab27a and Rab27b does not seem to be redundant because individual knockdown of each of these genes resulted in a significant reduction in RANKL release to the cell surface (Fig. 1D). Given that the subcellular localization of Rab27a and Rab27b after RANK stimulation is very similar, it can be assumed that Rab27a and Rab27b are involved in different steps in the same stimulation-dependent RANKL release event, such as vesicle recruitment, docking, and membrane fusion. To determine this involvement of Rab27a or Rab27b in more detail, analysis at the electron microscopic level is necessary. The combination of Rab27a or Rab27b with effector molecules at each step also could determine whether functional compensation between Rab27a and Rab27b occurs. For example, the affinity of Munc13-4 for Rab27a is reported to be much stronger than for Rab27b.21 This trend also was observed in the present results of the His-tag pull-down assay (Fig. 4C); therefore, the contribution of Rab27b to the mechanism involving Munc13-4, Slp5, and probably Slp4-a is lower than that of Rab27a, raising the possibility that the function of Rab27b in stimulation-dependent RANKL release also may be coordinated with other effector molecules.
Next, the role of each effector molecule will be discussed. As shown in this study, Slp4-a, Slp5, and Munc13-4 are involved in stimulation-dependent RANKL release in osteoblastic cells (Fig. 3). The intracellular region in which these molecules are functional also was investigated. According to our FRET results (Fig. 4D), the interaction occurred at the interface with the beads as well as in other intracellular regions, especially the association between Rab27a and each effector. The fact that the interaction occurs at the region where RANKL is released suggests that the system consisting of Rab27a and Slp4-a, Slp5, and Munc13-4 can be functional in the release of RANKL after stimulation. On the other hand, although the interaction between Rab27b and Slp4-a, Slp5, and Munc13-4 was detected at low levels at the interface, the interaction was stronger in other intracellular compartments, suggesting that Rab27b may be weakly functional at the interface with the beads and that Rab27b may have a more important role in other sites in the stimulation-dependent RANKL release. The detection of a FRET signal in intracellular regions other than those in the interface with RANK beads implies the potential involvement of Rab27a and Rab27b in different secretion events. As mentioned earlier, Rab27a and Rab27b are likely to exist in an activated form owing to their involvement in different secretion events in the cell. Therefore, the interaction between Rab27a or Rab27b and Slp4-a, Slp5, or Munc13-4 could correspond with other secretion events not associated with RANKL release. The involvement of Rab27a, Rab27b, Slp4-a, Slp5, and Munc13-4 in the stimulation-dependent RANKL release also was strengthened from the same figure. The exogenous introduction of Rab27a or Rab27b enhanced the accumulation of each effector molecule around the beads (Fig. 4D). However, our results cannot explain the functional differences between these effector molecules. Fukuda proposed the assumed functions38 based on studies using nonosteoblastic cells.19, 27, 53, 54 Referring to Fukuda's proposal, the function of each effector molecule in the stimulation-dependent RANKL release in osteoblastic cells can be hypothesized as follows: Slp5 regulates recruitment, which is the step before the docking of the RANKL-containing lysosomal vesicles to the plasma membrane; Slp4-a is involved in docking of the lysosomal vesicles to the plasma membrane; and Munc13-4 plays a role in vesicle fusion (Fig. 7).
In conclusion, stimulation-dependent RANKL release is mediated by Rab27a and Rab27b and their effectors, Slp4-a, Slp5, and Munc13-4, and the process is physiologically important for osteoclastogenesis. Bone homeostasis may be maintained through stringent regulation of RANKL subcellular localization.
All the authors state that they have no conflicts of interest.
We greatly thank Dr Bruce Beutler and MMRRC for providing the Jinx mice and permission for their use. This work was supported in part by Grant-in-Aid for Scientific Research from the Ministry of Education, Science and Culture of Japan.