Negative Regulation of Osteoclastogenesis by Ectodomain Shedding of Receptor Activator of NF-κB Ligand*

Receptor activator of NF-κB ligand (RANKL) is a transmembrane glycoprotein that has an essential role in the development of osteoclasts. The extracellular portion of RANKL is cleaved proteolytically to produce soluble RANKL, but definite RANKL sheddase(s) and the physiologic function of RANKL shedding have not yet been determined. In the present study, we found that matrix metalloproteinase (MMP) 14 and a disintegrin and metalloproteinase (ADAM) 10 have strong RANKL shedding activity. In Western blot analysis, soluble RANKL was detected as two different molecular weight products, and RNA interference of MMP14 and ADAM10 resulted in a reduction of both the lower and higher molecular weight products. Suppression of MMP14 in primary osteoblasts increased membrane-bound RANKL and promoted osteoclastogenesis in cocultures with macrophages. Soluble RANKL produced by osteoblasts from MMP14-deficient mice was markedly reduced, and their osteoclastogenic activity was promoted, consistent with the findings of increased osteoclastogenesis in vivo. RANKL shedding is an important process that down-regulates local osteoclastogenesis.

Receptor activator of NF-B ligand (RANKL), 2 also known as TNF-related activation-induced cytokine, osteoprotegerin ligand, and osteoclast differentiation factor, is a type II transmembrane glycoprotein of the TNF ligand family. RANKL is expressed on the plasma membrane of osteoblasts, bone marrow stromal cells, and T-lymphocytes (1)(2)(3)(4)(5) and exerts its activity through binding to its TNF family receptor RANK, which is expressed on monocyte-macrophage lineage osteoclast precursors (3,4,6). Consequently, RANK activates downstream signaling pathways such as NF-B, p38 mitogen-activated protein kinase, c-Jun N-terminal kinase, and nuclear factor of activated T cells c1 and leads to the differentiation, activation, and survival of osteoclasts (6 -9). Both RANKL-and RANK-deficient mice develop severe osteopetrosis due to a lack of osteoclasts (10,11). Conversely, a deficiency of osteoprotegerin (OPG), a natural inhibitor of RANKL, causes severe osteoporosis due to enhanced osteoclastogenesis (12). These findings suggest a crucial role for the RANKL/RANK/OPG axis in osteoclast development.
A number of transmembrane proteins undergo proteolysis and are released from the plasma membrane, a process called ectodomain shedding. The biologic and pathologic significance of ectodomain shedding varies between substrate proteins. For example, cytokines such as TNF-␣ and epidermal growth factor are released from local environments and exert their activity in paracrine and endocrine signaling by ectodomain shedding (13)(14)(15). In some cases, ectodomain shedding is necessary even for the local effects of growth factors such as epidermal growth factor (16). Although membrane-bound RANKL is converted to a soluble form through ectodomain shedding (17,18) similar to other TNF family members, such as TNF-␣ (13,14) and Fas ligand (19), the RANKL sheddases involved in physiologic and pathologic bone resorption, and the role of RANKL shedding is unknown. We report that RANKL shedding is regulated by matrix metalloproteinase (MMP) 14 and a disintegrin and metalloproteinase (ADAM) 10 in bone marrow stromal cells and osteoblasts. MMP14 is mainly involved in RANKL shedding in osteoblasts, and its deficiency up-regulates osteoclastogenesis by reducing RANKL shedding in the local bone milieu.

EXPERIMENTAL PROCEDURES
Reagents-DNA polymerase, KOD plus, was purchased from TOYOBO (Osaka, Japan). Antibodies for the His-tag were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), antibodies for the V5 tag were from Invitrogen, antibodies for actin were from Sigma-Aldrich, and the antibody for RANKL was from Active Motif (Carlsbad, CA). MMP inhibitors, FR255031 and FR217840, were generous gifts from Astellas Pharma Inc. (Tokyo, Japan).
Cell Culture-The human kidney cell line 293T were cultured in Dulbecco's modified eagle medium (Sigma-Aldrich) and the mouse bone marrow stromal cell line TM8B2 (generous gift from Prof. Noda, Medical Research Institute, Tokyo Medical and Dental University) were cultured with ␣-MEM, and both of them were supplemented with 10% FBS and 1% penicillin-streptomycin solution (Sigma-Aldrich Co.). Cells were incubated at 37°C in a humidified atmosphere containing 5% CO 2 . Primary osteoblasts were harvested from the calvaria of newborn ddy (purchased from Sankyo Labo Service Co., Tokyo Japan) and C57/BL6 (MMP14-deficient mice and wildtype littermates) mice as previously described (27).
Screening of RANKL Sheddases-The alkaline phosphatase assay of the culture media and the Western blot analysis were described previously (20,28). To detect soluble RANKL released into the culture media of TM8B2 cells, 3 ml of the culture media was incubated with 4 l of recombinant protein G-agarose (Invitrogen) and 1 g of recombinant OPG-Fc chimeric protein (R&D Biosystems, Minneapolis, MN) for 16 h at 4°C, then recovered by brief centrifugation. The pellets were suspended in TNE buffer (10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40), and subjected to SDS-PAGE.
Retrovirus Infection-The procedure for preparation of the retrovirus was described previously (29). Primary osteoblasts were incubated with the culture media containing retrovirus supplemented with 5 g/ml polybrane (Sigma-Aldrich) for 16 h.
Western Blot of Primary Osteoblasts-Cells were treated with or without 10 Ϫ8 M 1␣,25-dihydroxyvitamin D 3 (1␣,25(OH) 2 D 3 ), 10 Ϫ6 M prostaglandin E 2 (PGE 2 ), and 10 ng/ml interleukin-1 (IL-1) for 72 h, and 10 ml of culture media were collected. Cell lysates were collected using M-PER Mammalian Protein Extraction Reagent (Pierce) according to the manufacturer's protocol. Culture media and cell lysates were treated using the same procedure as for culture media of TM8B2.
Determination of Cleavage Sites of RANKL-pSG5-RANKL were transfected to TM8B2 cells. Seventy hours after transfection, 200 ml of the culture medium was collected, and soluble RANKL was recovered using an OPG-Fc column. Samples were subjected to SDS-PAGE, transferred to polyvinylidene membrane, and stained with Coomassie Brilliant Blue. The band was excised, and the N-terminal amino acid sequence was determined by Hokkaido System Science Co., Ltd. (Hokkaido, Japan).
ELISA-Primary osteoblasts were seeded in 48-well plates at a concentration of 1 ϫ 10 5 cells/ml. If needed, cells were infected with retrovirus vectors as mentioned above. After 24 h of incubation, some cultures were treated with 10 Ϫ8 M 1␣,25(OH) 2 D 3 and 10 Ϫ6 M PGE 2 for 72 h. The concentration of RANKL in the culture media and cell lysates (homogenized in 200 l of TNE buffer per well) was determined by ELISA system developed using a TRANCE DuoSet ELISA development kit (R&D Biosystems).
Coculture of Primary Osteoblasts and Bone Marrow Macrophages-Bone marrow macrophages were obtained from tibias of 5-8 weeks-old C57/BL6 mice by flushing the bone marrow with ␣-MEM, and cultured in ␣-MEM supplemented with 10% FBS, 1% penicillin-streptomycin solution, and 10 ng/ml M-CSF (Wako Pure Chemicals, Osaka, Japan) for 16 h, and non-adherent cells were collected. For the coculture, primary osteoblasts were seeded in 48-well plates at a concentration of 1 ϫ 10 5 cells/ml. If needed, cells were infected by a retrovirus, as described above. After 24-h incubation, macrophages were seeded on the primary osteoblasts at a concentration of 1 ϫ 10 5 cells/ml and treated with 10 Ϫ8 M 1␣,25(OH) 2 D 3 and 10 Ϫ6 M PGE 2 for 5 days. Osteoclasts were stained with tartrate-resistant acid phosphatase (3), and the cell number was counted. For separate culture, primary osteoblasts were cultured on the cell culture insert (BD Biosciences, San Jose, CA), and bone marrow macrophages were seeded on the bottom of 24-well plates.
Osteoclast Formation Assay, Survival Assay, and Pit Formation Assay of Macrophages of MMP14-deficient Mice-The MMP14-deficient mice were generated and backcrossed to the C57/BL6 strain as previously described (30). Livers and spleens were harvested from newborn MMP14-deficient mice and wild-type littermates, and cells were strained using a cell strainer. The obtained cells were incubated in ␣-MEM supplemented with 10% FBS, 1% penicillin-streptomycin solution, and 200 ng/ml M-CSF (Wako) for 4 days in 10-cm Petri dishes, and then M-CSF-dependent macrophages were recovered by trypsinization. For the osteoclast formation assay, macrophages were seeded in 48-well plates at a concentration of 2 ϫ 10 5 cells/ml and treated with 10 ng/ml M-CSF and 100 ng/ml RANKL (Wako) for 6 days. For the survival assay and the pit formation assay, macrophages (2 ϫ 10 5 cells/ml) were cocultured with primary osteoblasts (1 ϫ 10 5 cells/ml) in ␣-MEM containing 10% FBS, 10 Ϫ8 M 1␣,25(OH) 2 D 3 , and 10 Ϫ6 M PGE 2 on culture dishes precoated with 0.2% collagen gel matrix (Nitta Gelatin, Osaka, Japan). When osteoclasts were formed after 7-day culture, they were released from the dishes by treatment with 0.2% collagenase in phosphate-buffered saline, and collected by centrifugation. For the survival assay, cells were cultured on 48-well plates for 4 h, and osteoclasts were isolated by removing the osteoblasts by 0.1% collagenase and 0.2% dispase in phosphate-buffered saline. Purified osteoclasts were further incubated in ␣-MEM containing 10% FBS for 20 h. For the pit formation assay, cells were seeded on the dentin slices (Wako) precoated with FBS, and incubated in ␣-MEM containing 10% FBS, 10 Ϫ8 M 1␣,25(OH) 2 D 3 , and 10 Ϫ6 M PGE 2 for 24 h. Then the cells were removed by ultrasonication after adding 1 M NH 4 OH. Resorption pits were visualized by staining with 0.5% toluidine blue and analyzed using Adobe Photoshop (Adobe, San Jose, CA).
Histology-Tibias of 2-to 3-week-old male MMP14-deficient mice, and wild-type littermates were fixed in 4% paraformaldehyde and decalcified in 10% EDTA. Paraffin-embedded samples were sectioned at 5 m and stained with hematoxylin and eosin or tartrate-resistant acid phosphatase. Observers who were blinded to the origin of the images counted the number of osteoclasts on the trabecular bone surface and measured total length of the trabecular bone using Image J software (National Institutes of Health, Bethesda, MD).
Statistical Analysis-Statistical analyses were performed using Student's t test for the alkaline phosphatase assay and a two-tailed unpaired Student's t test for the real-time PCR and ELISA. A p value of Ͻ0.05 was considered to be statistically significant.

Screening of MMPs and ADAMs as RANKL Sheddases-We
previously developed a RANKL shedding activity screening system using an expression vector encoding a fusion protein of secreted placental alkaline phosphatase (SEAP) with the C-terminally truncated form of RANKL, which contains the stalk region, transmembrane domain, and intracellular domain of RANKL (tRANKL-SEAP) (20). Expression vectors of various MMPs or ADAMs were transfected to 293T cells together with tRANKL-SEAP plasmids, and the alkaline phosphatase activity of the culture media was measured. In this assay system, increased alkaline phosphatase activity in the medium indicates an increase in RANKL shedding (Fig. 1A). Several MMPs and ADAMs exhibited tRANKL-SEAP shedding activity, and ADAM10, MMP14 (membrane-type 1 matrix metalloproteinase, MT1-MMP), MT2, MT3, and MT5-MMP had relatively stronger activity (Fig. 1B). In contrast, MMP13 (Fig. 1B), MMP1, MMP2, MMP3, MMP9, MMP11, MMP23, and MMP28 (data not shown) had no tRANKL-SEAP shedding activity.
MMP14 and ADAM10 Are Major RANKL Sheddases in TM8B2 Cells-To confirm the ability of these proteinases to cleave full-length RANKL, they were overexpressed in the mouse bone marrow stromal cell line TM8B2 together with full-length RANKL. Soluble RANKL was recovered from the culture media using an OPG-Fc recombinant protein column and subjected to Western blot analysis. When RANKL (C-terminally tagged with V5 and 6ϫ His) was overexpressed in TM8B2 cells (ϳ39 kDa), there were two different molecular mass products (25 and 24 kDa) in the culture media ( Fig. 2A).
The N-terminal sequences of the two bands are shown in Fig.  2B. These two bands were differentially diminished by various metalloproteinase inhibitors (Fig. 2A). When TM8B2 cells were treated with FR255031, which inhibits several MMPs but not TNF-␣ secretases (31), only the lower band disappeared. On the other hand, treatment with FR217840, which not only inhibits several MMPs, but also inhibits TNF-␣ secretion (32), both of  (n ϭ 3). The y-axis represents the relative mRNA levels of the indicated genes expressed as -fold from MMP14 level. E, efficient gene knock-down by siADAM10 and siMMP14. TM8B2 cells were transfected with pcDNA3.1-V5HisA, piGENEmU6-siGFP, -siADAM10 -1, -2, siMMP14 -1, -2, or pcDNA3.1-mMMP14-V5HisA, and the expression of ADAM10 (endogenous) and MMP14 (overexpressed, detected by anti-V5 epitope tag antibody) was analyzed 48 h after transfection. F, Effect of siADAM10 and siMMP14 on RANKL cleavage in TM8B2 cells. TM8B2 cells were transfected with piGENEmU6-siGFP, -siADAM10 -1, -2, siMMP14 -1, -2, or pcDNA3.1-V5HisB, together with pcDNA3.1-RANKL-V5HisB. Seventy-two hours after transfection, culture media and cell lysates were collected and subjected to Western blot analysis. G, effects of siADAM10, siMMP14, and chemical inhibitors on RANKL shedding in TM8B2 cells. Procedures were the same as in A. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 the bands disappeared. These results suggest that the upper band is produced by ADAM family proteinase(s) and the lower band by MMP(s).

RANKL Shedding Regulates Osteoclastogenesis
ADAM9, ADAM10, and ADAM19 cleaved full-length RANKL, but only ADAM10 produced soluble RANKL with the same molecular weight as the upper band observed in TM8B2 cells (Fig. 2C). ADAM17 did not cleave full-length RANKL. MMP14, MT2, MT3, and MT5-MMP generated soluble RANKL with the same molecular weight as the lower band. MT4-and MT6-MMP did not increase the soluble RANKL detected in the culture media (data not shown). Real-time PCR showed that the MMP14 mRNA level was much higher than that of MT2-, MT3-, or MT5-MMP in TM8B2 cells and primary osteoblasts (Fig. 2D), suggesting that MMP14 is mainly involved in the production of the lower band. The mRNA level of ADAM10 was comparable to that of MMP14 in these two cells (data not shown).
To determine the role of ADAM10 and MMP14 in the constitutive shedding of RANKL in TM8B2 cells, we constructed expression vectors of siRNA for ADAM10 and MMP14 (siADAM10 and siMMP14) (Fig. 2E). We constructed two different siADAM10 and siMMP14 vectors, and all of them effectively suppressed expression of the target molecules (Fig. 2E). The upper band of soluble RANKL diminished when the RANKL construct was transfected to TM8B2 cells along with siADAM10 plasmids; the lower band diminished with siMMP14 (Fig. 2F); and only the upper band was faintly observed when both siADAM10 and siMMP14 were transduced (Fig. 2G). The faint band completely disappeared in the presence of FR217840, but not FR255031 (Fig. 2G). These results indicated that ADAM10 and MMP14 are two major RANKL sheddases in TM8B2 cells.
MMP14 Cleaves RANKL in Primary Osteoblasts-We next examined the RANKL sheddases in primary osteoblasts. When primary osteoblasts were stimulated with 1␣,25(OH) 2 D 3 , PGE 2 , and IL-1, two bands appeared in the culture media with the same molecular masses (ϳ22 and 21 kDa) as those observed in TM8B2 cells transfected with full-length RANKL without epitope tag ( Fig. 3A and data not shown), but the lower band was more prominent. The expression of MMP14 was also much higher than that of MT2-, MT3-, or MT5-MMP in osteoblasts, suggesting that MMP14 is the major RANKL sheddase in primary osteoblasts (Fig. 2D). We constructed retrovirus vectors of siMMP14 that efficiently suppressed MMP14 expression in primary osteoblasts (Fig. 3B). When osteoblasts were infected with siMMP14 retrovirus, the lower band was diminished in the culture medium and membrane-bound RANKL was increased in the cell lysate (Fig. 3A). The concentration of soluble RANKL in the culture media and membranebound RANKL in the cell lysates was also measured by ELISA. A significant decrease in soluble RANKL and an increase in membranebound RANKL was observed by MMP14 knockdown (Fig. 3C), consistent with the results of Western blot analysis. Conversely, overexpression of MMP14 in primary osteoblasts increased the lower band of soluble RANKL in the culture media and reduced membranebound RANKL in the cell lysates (Fig. 3, D and E). To further confirm the role of MMP14 in RANKL shedding in osteoblasts, we analyzed primary osteoblasts obtained from MMP14 knock-out mice. When primary osteoblasts of MMP14-deficient mice were stimulated with 1␣,25(OH) 2 D 3 , PGE 2 , and IL-1, only the upper band of soluble RANKL was observed in the culture medium (Fig. 3F). The amount of soluble RANKL was markedly reduced in the medium, and membrane-bound RANKL was increased (Fig. 3G).
RANKL Shedding Negatively Regulates Osteoclastogenesis-To determine the role of MMP14 in the osteoclastogenic activity of osteoblasts, we cocultured osteoblasts infected with siMMP14 retrovirus with bone marrow macrophages. The siMMP14 retrovirus-infected osteoblasts, in which RANKL shedding activity was suppressed and membrane-bound RANKL was increased, stimulated osteoclastogenesis more effectively than control cells (Fig. 4A). Conversely, overexpression of MMP14 in primary osteoblasts suppressed osteoclastogenesis (Fig. 4B). These results suggest that membrane-bound RANKL induces osteoclastogenesis more efficiently than soluble RANKL, and the ectodomain shedding of RANKL by MMP14 negatively regulates osteoclastogenesis.
We then examined the osteoclastogenic activity of the soluble RANKL generated by MMP14. The culture media of 293T cells transfected with RANKL and MMP14 expression vectors were collected 48 h after transfection, and bone marrow macrophages were cultured in media supplemented with 10 ng/ml M-CSF (Fig. 4C). When macrophages were cultured in media recovered from the empty vector or RANKL expression vector-transfected cell cultures, mature osteoclasts did not form, even in the presence of M-CSF. In contrast, the culture media from the cells transfected with both RANKL and MMP14 expression vectors induced osteoclast formation, indicating that the soluble RANKL generated by MMP14 had osteoclastogenic activity. When primary osteoblasts overexpressing MMP14 by retrovirus vectors and bone marrow cells were cultured separately using cell culture inserts, which allow the passage of culture media but prevent cell-cell contact, and were stimulated with 1␣,25(OH) 2 D 3 , PGE 2 , and IL-1, very few osteoclasts were formed (Fig. 4D). These results suggest that the amount of endogenous soluble RANKL generated in primary osteoblasts is not sufficient to induce osteoclastogenesis in our assay system.
We finally analyzed the osteoclastogenic activity of MMP14-deficient osteoblasts. When normal bone marrow macrophages were cocultured with osteoblasts from MMP14-deficient mice, osteoclasts were formed more efficiently than in cocultures with wild-type osteoblasts (Fig. 5A). On the other hand, bone marrow macrophages from MMP14-deficient mice differentiated into mature osteoclasts with the same efficiency as wild-type cells (Fig. 5B), and their bone-resorbing activity and survival rate were comparable (Fig.  5C). Consistent with these in vitro observations, soft x-ray images of MMP14-deficient mice displayed osteoporosis (Fig. 6A). Histologic sections showed thin cortical bone, decreased cancellous bone, and increased osteoclasts on the trabecular bone surface, which agreed with a previous report (33) (Fig. 6B). These data suggested that the increased osteoclast number in MMP14-deficient mice is due to the increase in membrane-bound RANKL in osteoblasts.

DISCUSSION
There is accumulating evidence that protein ectodomain shedding affects a large number of cellular membrane proteins, including TNF family members, but the functional outcome of shedding depends on the particular substrate protein. For example, TNF-␣ is cleaved by some proteinases such as TNF-␣ converting enzyme and released into the circulatory system to exhibit strong systemic effects (13,14). In contrast, Fas ligand, which is a strong apoptosis inducer, has reduced effects in its soluble form (19). RANKL is a member of the TNF family of cytokines, and recent studies indicate that RANKL is essential in skeletal homeostasis for regulating the differentiation and activation of osteoclasts. Membrane-bound RANKL is proteolytically processed to generate the soluble form, but the specific proteinases involved in RANKL shedding and its physiologic and pathologic importance are not known. Western blot analysis revealed that there are two different cleaved products in the supernatants of both TM8B2 bone marrow stromal cells and primary osteoblasts, and the lower band was the major band in primary osteoblasts (Fig. 3A). The N-terminal sequence of the upper band coincides with the constitutive cleavage site previ-ously reported in COS7 cells, and that of the lower band coincides with the putative MMP cleavage site. The lower band was decreased in the presence of an MMP inhibitor FR255031, and both bands were diminished by treatment with FR217840, which inhibits both MMPs and ADAMs. Previous studies also demonstrated that various MMPs and ADAMs have RANKL shedding activity (17, 20, 34 -36). Therefore, we first analyzed the RANKL shedding activity of several MMPs and ADAMs using the previously reported RANKL shedding assay (20). MMP7, MMP19, ADAM9, ADAM10, ADAM17, ADAM19, MMP14, MT2, MT3, MT4, MT5, and MT6-MMP had tRANKL-SEAP shedding activity. Among these proteinases, ADAM9, ADAM10, ADAM19, MMP14, MT2, MT3, and MT5-MMP cleaved full-length RANKL. Although ADAM9 and ADAM19 had RANKL shedding activity, the molecular weights of the cleaved products were different from those con- stitutively generated in TM8B2 cells and primary osteoblasts. Among MT-MMPs, MMP14 was predominantly expressed in TM8B2 and primary osteoblasts. MMP7 was reported to cleave RANKL in vitro or under overexpressing conditions and has an important role in prostate cancer-induced osteolysis (34). Although MMP7 had tRANKL-SEAP shedding activity, it did not cleave full-length RANKL in our assay system. Blobel and co-workers (17) demonstrated that ADAM17 (TNF-␣ converting enzyme) cleaves a partial fragment of RANKL protein, but not full-length RANKL (35), and we confirmed that ADAM17 did not cleave full-length RANKL. Although MMP13 deficiency leads to skeletal abnormalities (37), it did not cleave tRANKL-SEAP or full-length RANKL. These data, in combination with the results of siRNA experiments, led us to speculate that ADAM10 and MMP14 were major RANKL sheddases in TM8B2 cells and osteoblasts, although it is possible that other proteinases are involved in RANKL shedding in some pathologic conditions in which they are up-regulated. The functional difference between MMP14-cleaved RANKL and ADAM10cleaved RANKL is currently under investigation.
In the culture media of primary osteoblasts, the lower molecular weight product was much more predominant than the higher molecular weight product, indicating that MMP14 is the major RANKL sheddase in primary osteoblasts. This was further confirmed by the results obtained from MMP14 knockdown osteoblasts and MMP14-deficient osteoblasts, in which the lower molecular weight band was almost completely diminished and the concentration of soluble RANKL was severely reduced.
Reduced MMP14 expression in primary osteoblasts by siRNA or its deficiency in MMP14 knock-out mouse osteoblasts reduced RANKL shedding and increased membranebound RANKL, which led to increased osteoclastogenic activity in the cells. Although soluble RANKL produced by MMP14 induced osteoclastogenesis from bone marrow macrophages, the culture medium of primary osteoblasts treated with 1␣,25(OH) 2 D 3 and PGE 2 did not induce osteoclastogenesis (38), even when MMP14 was overexpressed. This was probably due to an insufficient amount of soluble RANKL, because the concentration of the soluble RANKL in the culture media was Ͻ1 ng/ml (data not shown), and the amount of recombinant soluble RANKL necessary to induce osteoclastogenesis in vitro was Ͼ10 ng/ml. Still, it is possible that when the expression of RANKL is highly up-regulated, soluble RANKL has substantial effects on general bone metabolism. Gao et al. reported that parathyroid hormone(1-34) not only increases RANKL but also decreases MMP14 production in human osteoblasts, which might result in the efficient up-regulation of bone resorption by increasing membrane-bound RANKL in the cells (39).
In MMP14-deficient mice, generalized osteopenia was observed. Although this might be partially due to the reduced bone formation, as previously reported (33), the osteoclast number was much higher in these mice, possibly due to the increased membrane-bound RANKL in the osteoblasts. In fact, the serum level of soluble RANKL in MMP14 knock-out mice was undetectable, even when they were injected with 1␣,25(OH) 2 D 3 (data not shown). These phenotypes are not due to cell autonomous defects in osteoclast precursors, because there were no abnormalities in osteoclast differentiation from MMP14-deficient macrophages or bone-resorbing function of mature osteoclasts.
We were unable to determine the role of ADAM10 in in vivo RANKL shedding or skeletal homeostasis, because ADAM10deficient animals die at E9.5 due to defects in heart and central nervous system development (40). Therefore, future studies using cell-specific knock-out of ADAM10 are required.
To conclude, we identified MMP14 as a major endogenous RANKL sheddase in primary osteoblasts. RANKL shedding negatively affects local osteoclastogenesis by reducing membrane-bound RANKL in the skeletal environment.