Actin-binding protein coronin 1A controls osteoclastic bone resorption by regulating lysosomal secretion of cathepsin K

Osteoclasts degrade bone matrix proteins via the secretion of lysosomal enzymes. However, the precise mechanisms by which lysosomal components are transported and fused to the bone-apposed plasma membrane, termed ruffled border membrane, remain elusive. Here, we identified coronin 1A as a negative regulator of exocytotic release of cathepsin K, one of the most important bone-degrading enzymes in osteoclasts. The modulation of coronin 1A expression did not alter osteoclast differentiation and extracellular acidification, but strongly affected the secretion of cathepsin K and osteoclast bone-resorption activity, suggesting the coronin 1A-mediated regulation of lysosomal trafficking and protease exocytosis. Further analyses suggested that coronin 1A prevented the lipidation-mediated sorting of the autophagy-related protein LC3 to the ruffled border and attenuated lysosome–plasma membrane fusion. In this process, the interactions between coronin 1A and actin were crucial. Collectively, our findings indicate that coronin 1A is a pivotal component that regulates lysosomal fusion and the secretion pathway in osteoclast-lineage cells and may provide a novel therapeutic target for bone diseases.


Results
Expression patterns of coronin 1A and its localization during osteoclast differentiation. We hypothesized that the actin-binding protein coronin 1A might play a crucial role in osteoclast differentiation and/ or function. Therefore, we first analysed the expression and localization of coronin 1A during osteoclast differentiation. Coronin 1A was highly expressed in osteoclast precursor cells (bone marrow macrophages; BMMs) and was decreased during osteoclastogenesis in both mRNA and protein levels (Fig. 1a,b). Coronin 1A expression was decreased in BMMs from 24 h after RANKL stimulation and continued to decrease until 48 h. Coronin was first identified in Dictyostelium and was characterized as an actin-binding protein. Consistent with the previous reports 22,24,30 , coronin 1A was distributed near the plasma membrane and co-localized to F-actin (Fig. 1c). Of note, coronin 1A was detected in osteoclast precursor cells (Fig. 1c: small cells), but to a lesser extent in mature osteoclasts ( Fig. 1c: multinucleated cells [MNCs, > 3 nuclei]), confirming that coronin 1A is reduced during osteoclastogenesis.
Role of coronin 1A in RANKL-dependent osteoclastogenesis. We next investigated whether coronin 1A is required for RANKL-induced osteoclastogenesis. Primary BMMs infected with empty lentiviral vector or coronin 1A-expressing lentivirus were cultured with M-CSF and RANKL ( Supplementary Fig. 1a,b), following which they were stained for TRAP, a marker of osteoclasts. Despite the importance of coronin 1A in calciumcalcineurin signalling 25,27,29 , which had been shown to be important for osteoclastogenic NFATc1 autoamplification 3,4,31 , the enforced coronin 1A expression did not affect RANKL-induced osteoclast differentiation (Fig. 2a,b) and NFATc1 induction (Fig. 2c) as well as precursor cell proliferation or survival ( Supplementary Fig. 2a,b). Similar to the effects of overexpression of coronin 1A, its knockdown in BMMs did not affect osteoclastogenesis ( Fig. 2d-f, Supplementary Fig. 1c) and precursor cell proliferation or survival ( Supplementary Fig. 2c,d). Subsequently, we examined the role of coronin 1A in RANKL-dependent signalling pathways, including extracellular signal-regulated kinase (ERK), p38, c-Jun N-terminal kinases (JNK), NF-κ B and v-akt murine thymoma viral oncogene (AKT), which are activated via the RANKL-TNF receptor-associated factor 6 (TRAF6)-dependent pathway 32 . As shown in Supplementary Fig. 3a and b, neither coronin 1A overexpression nor knockdown affected RANKL-dependent NF-κ B, MAPK and AKT activation. In addition, mRNA expression levels of the osteoclastogenic markers Nfatc1, Dcstamp, ATPase H + transporting V0 subunit d2 (Atp6v0d2), Acp5 and Ctsk were not significantly affected by coronin 1A overexpression or its knockdown (Fig. 2c,f). These results indicated that the role of coronin 1A in osteoclast differentiation is minute.
Osteoclastic bone resorption activity regulated by coronin 1A. Bone resorption requires lysosomal trafficking to and fusion with ruffled border [8][9][10] . Since coronin 1A has been identified to inhibit lysosomal fusion 25-28 , we assessed whether coronin 1A is involved in osteoclastic bone resorption activity. When osteoclasts were cultured on the bone biomimetic synthetic surface, resorption pits formed by coronin 1A-overexpressing osteoclasts were significantly smaller than those formed by control cells (Fig. 3a,b). Previous studies showed that cathepsin K is a key player in bone resorption via degradation of bone-matrix proteins, type I and type II collagen [33][34][35][36] . Thus, we focused on cathepsin K and assessed the amount of cathepsin K in whole cell lysates and culture supernatants by immunoblotting. We found that the amount of cathepsin K in the culture supernatants of coronin 1A-overexpressing osteoclasts was significantly smaller than that of control cells, whereas the total expression levels of cathepsin K in whole cell lysates did not differ between coronin 1A-overexpressing cells and control cells (Fig. 3c). In line with these results, the bone resorption area and the amount of cathepsin K in the culture supernatants tended to increase in osteoclasts transfected with coronin 1A specific small interfering (si)RNA, without any changes in the expression of cathepsin K (Fig. 3d-f), suggesting that coronin 1A inhibits cathepsin K secretion from osteoclasts. Although cytokines are also contained in small secretary vesicles and are transported to the plasma membrane and released out into the extracellular space 37 , the secretion of proinflammatory cytokines, such as tumor necrosis factor (TNF), interleukin 6 (IL-6) and IL-12/IL-23p40 from BMMs was not affected by coronin 1A overexpression or its knockdown ( Supplementary Fig. 4a,b). The extracellular acidification of osteoclast resorption lacunae, which is mediated by vacuolar H + -ATPases (V-ATPases), is a prerequisite for efficient degradation of bone matrix proteins by cathepsin K 10,38,39 . Thus, we investigated whether coronin 1A is involved in the extracellular acidification of osteoclasts. Using Acridine orange staining assay, which has been widely used to reveal the extracellular acidification and/or proton-pump activity in osteoclasts [40][41][42][43] , we found that the extracellular acidification was not affected by the coronin 1A-overexpression or -knockdown compared with control cells ( Supplementary Fig. 5a,b), suggesting that coronin 1A regulation of the osteoclast proton secretion is not likely. These results together suggested that coronin 1A plays an important and specific role in the secretion of lysosomal enzymes such as cathepsin K but not in the secretion of cytokines and osteoclast extracellular acidification.
Coronin 1A inhibits the localization of cathepsin K at ruffled border. We then focused on the influence of coronin 1A overexpression on spatiotemporal regulation of cathepsin K. We first investigated the actin-ring formation during osteoclast differentiation. Coronin 1A-overexpressing osteoclasts normally formed actin rings in response to RANKL (Fig. 3g,h), suggesting that cytoskeleton organization and the signals that activate this process are intact. However, the amount of cathepsin K localized within actin rings was greatly reduced by coronin 1A overexpression (Fig. 3g,i), indicating that coronin 1A prevents the sorting and/or fusion of cathepsin K-containing lysosomes to ruffled border membrane surrounded by actin rings. To obtain a more direct evidence of the regulatory role of coronin 1A in the fusion of lysosomes with the plasma membrane of ruffled border, we employed widely used LAMP1 (a lysosome marker) staining to assess the lysosomal fusion and found that coronin 1A overexpression impaired RANKL-induced localization of LAMP1 within osteoclast actin rings, which normally occurs after the lysosomal fusion to ruffled border membrane ( Supplementary Fig. 6).
Coronin 1A inhibits LC3-I-to-II conversion and LC3 localization at ruffled border. Next, we attempted to clarify which mechanisms underlie the coronin 1A suppression of cathepsin K secretion. The previous study indicated that autophagic component LC3, one of eight mammalian Atg8 family of proteins, is converted from its cytosolic LC3-I form into LC3-II, which is a C-terminally phosphatidylethanolamine-conjugated form; and this post-translational modification process, called lipidation, is essential for both LC3 localization at ruffled border and lysosomal secretion from ruffled border 21 . Therefore, we quantified the production of LC3-II and examined LC3 localization in osteoclast-lineage cells using immunoblotting and confocal microscopy, respectively. LC3-II levels in osteoclasts were decreased by coronin 1A overexpression and were increased by the coronin 1A knockdown (Fig. 4a,b), suggesting the importance of coronin 1A for LC3 lipidation in osteoclasts. Moreover, enforced coronin 1A expression inhibited the localization of LC3 within actin ring (Fig. 4c,d). These results indicate that coronin 1A regulates LC3 lipidation and its localization at the osteoclast ruffled border.
As previously reported, small guanosin-5′ -triphosphatase (GTPases) Rab7 and Rab27a are necessary for ruffled border formation and/or lysosomal secretion in osteoclasts 12,13 . Furthermore, the localization of Rab7 and LC3 are abnormal, and cathepsin K secretion is significantly decreased in Atg5-deficient osteoclasts 21 . Here, we asked whether coronin 1A overexpression could affect Rab7 or Rab27a in osteoclasts. The expression levels of Rab7 and Rab27a were not affected by coronin 1A overexpression ( Supplementary Fig. 7a). Rab7 was observed principally at ruffled border in coronin 1A-overexpressing osteoclasts, and the co-localization with F-actin was evident both in coronin 1A-overexpressing and control osteoclasts ( Supplementary Fig. 7b). In addition, Rab27a localization at cytoplasm was not also changed by coronin 1A overexpression ( Supplementary Fig. 7b). These results indicated that coronin 1A specifically regulates LC3 lipidation and localization at ruffled border, but not Rab7 and Rab27a at least in osteoclasts.

Coronin 1A regulates osteoclastic bone resorption activity via the actin-related mechanism(s).
Coronin 1A has been known to interact with cytoskeleton 24,30 . Therefore, we hypothesized that coronin 1A regulates LC3 lipidation and localization through actin-related modulation. We first constructed a mutant of coronin 1A (with an R29D mutation) that is devoid of actin-binding ability 44,45 and then examined its effect on osteoclastogenesis and bone resorption. Coronin 1A overexpression by lentivirus ( Supplementary Fig. 8), regardless of the R29D mutation, did not affect osteoclastogenesis (Fig. 5a,b). However, interestingly, the R29D mutation in coronin 1A reversed the inhibitory effect of coronin 1A overexpression on osteoclastic bone resorption activity (Fig. 5c,d). Of note, suppression of LC3 lipidation and cathepsin K secretion by coronin 1A overexpression were also abolished with the actin-binding deficient R29D mutant (Fig. 5e,f). These results strongly indicated that the interaction between coronin 1A and actin is crucial for coronin 1A-mediated suppression of bone resorption.

Effect of enforced expression coronin 1A on lipopolysaccharide (LPS)-induced bone resorption in vivo.
Finally, we assessed whether enforced expression of coronin 1A would affect the severity of LPS-induced bone resorption in a mouse model. Mice were injected with LPS and adenovirus expressing LacZ (Ad LacZ: control) or coronin 1A (Ad coronin 1A), and we found that although the number of TRAP-positive cells on the bone surface was only slightly affected (Fig. 6a,b), Ad coronin 1A-injected mice showed less susceptibility to LPS-induced bone loss compared to Ad LacZ-injected mice; i.e., histological analysis (Fig. 6c), micro computed tomography (Fig. 6d) and the levels of serum CTx-I (collagen type 1 cross-linked C-telopeptide: a marker for bone resorption) (Fig. 6e) showed a significant decrease in bone resorption in the Ad coronin 1A-injected mice, suggesting that the coronin 1A-mediated machinery also operates in vivo and plays a crucial role in regulating bone resorption.

Discussion
In the context of osteoclastic secretion of bone-degrading enzymes, the precise mechanism of the trafficking and fusion of lysosomal cargo to the ruffled border membranes remains poorly understood. In the present study, we revealed that coronin 1A is a negative regulator of cathepsin K secretion in osteoclast-lineage cells. Interestingly, coronin 1A prevented LC3 lipidation and LC3 localization to the ruffled border membranes, and the R29D mutation in coronin 1A reversed the inhibitory effects, indicating that coronin 1A attenuates the lipidation-mediated LC3 sorting through its binding capacity to actin.
Autophagy-related proteins have been shown to be responsible for lysosomal secretion in osteoclasts and melanocytes 21,46,47 . Actually, the autophagy-related proteins Atg5, Atg7, Atg4B and LC3, a member of the Atg8 family proteins, have been shown to be important in osteoclastic bone resorption, and LC3 lipidation and LC3 localization at ruffled border membrane were shown to be critical for efficient cathepsin K secretion in osteoclasts 21 . In addition, the deletion of coronin 1A was reported to enhance the recruitment of LC3 to phagosomes in macrophages 26 . Moreover, the small GTPase Rab7 has been identified as an essential component for both autophagosome fusion to lysosomes in variable cells and ruffled border formation in osteoclasts 12,48,49 . These reports collectively suggested that the autophagic pathway and the lysosomal secretion of bone-degrading enzymes in osteoclasts might share common molecular mechanisms, and it is likely that coronin 1A inhibits cathepsin K secretion by suppressing the cellular autophagic machinery. However, notably, knocking out the Map1lc3b gene (encoding LC3-β , a representative member of MAP1LC3/LC3 protein family in mammals 21,50 ) did not affect osteoclast phenotype 21 , perhaps reflecting functional redundancy among Atg8 family members, such as GABARAP and GATE-16 21,50,51 . In accordance with these reports, osteoclastogenesis, cathepsin K-secretion and bone resorption activity of osteoclasts were not significantly changed by siRNA-mediated knockdown of LC3-β ( Supplementary Fig. 9a-c). Future studies are needed to evaluate the significance of LC3 pathway and interplay among the Atg8 family proteins. During autophagosome formation, Atg12 is conjugated to Atg5 irreversibly, and the Atg12-Atg5 conjugate contributes to LC3 lipidation with its E3 ligase-like function [52][53][54] . Atg16L further forms a complex with the Atg12-Atg5 conjugate and facilitates LC3 lipidation [52][53][54] . Therefore, it may be possible that actin filament-bound coronin 1A attenuates the function of Atg12-Atg5 complex and/or Atg16L by yet unknown mechanisms. Coronin 1A was also identified as an F-actin stabilizer against depolymerization 55 . Moreover, F-actin stabilization on phogosomal membrane by coronin 1A is required to block phagosome-lysosome fusion 25,56 . Therefore, it is possible that coronin 1A not only prevents LC3 lipidation and localization via the actin-related mechanism, but also protects plasma membranes (other than ruffled border membranes) from fusing with lysosomal vesicles through stabilizing actin cytoskeleton to prevent aberrant release of lysosomal enzymes. Further studies are required to prove this hypothesis.
It is noteworthy that previous studies have shown that coronin 1A plays a role in calcium mobilization as well as phagosome/endosome fusion to lysosomes 25,27 . Although calcium mobilization is a prerequisite for the calcium-calcineurin-NFATc1 signalling, the modulation of coronin 1A expression did not affect the NFATc1 activation and osteoclast differentiation. This fact suggests that coronin 1A-independent calcium-calcineurin signalling operates in osteoclastogenesis or that the existence of other coronin family proteins simply masks the gain-and loss-of-function phenotype(s) of coronin 1A mutant cells in osteoclastogenesis. In fact, there are seven coronin genes in mammals. Although coronin family members share a common structure, i.e. WD repeats and a carboxy-terminal coiled-coil domain, functional redundancy among coronin family proteins has not been fully characterized. Indeed, although five coronin members, coronin 1A, 1B, 1C, 2A and 7, are expressed in macrophages, only coronin 1A has been identified as a regulator of phagosomal membrane fusion with lysosomes 25,57 . This may explain the strong phenotype in coronin 1A mutant cells concerning osteoclast function but not osteoclastogenesis. To completely elucidate the functional redundancy among coronin family proteins in osteoclast differentiation and function, further studies will be required.
We showed that coronin 1A knockdown or overexpression strongly affected bone resorption. Nevertheless, to date, an abnormal bone phenotype in coronin 1A-deficient mice has never been reported. The previous study showed that autophagic protein Atg5-deficient osteoclasts exhibit a remarkable reduction in osteoclastic bone resorption activity in vitro; but Atg5-deficient mice only showed a minimal increase in bone mass 21 . However, Atg5-deficient mice were protected from ovariectomy-or ageing-induced bone loss 21 , suggesting that a defect in autophagic machinery may cause a strong defect in osteoclasts in vitro; however, for unknown reasons, bone remodeling appears normal in vivo unless osteoclasts are activated by a relevant 'challenge' . Accordingly, bone mass in coronin 1A-deficient mice also might be affected by pathological challenges.
In the present study, we showed that RANKL stimulation downregulated coronin 1A expression (Fig. 1). The coronin 1A promoter region contains a binding site for the Maf family of transcription factors, as detected by using the MATCH program and the TRANScription FACtor (TRANSFAC) database. MafB, one of the Maf family of transcription factors, is expressed in monocytes and macrophages and has been reported to negatively regulates RANKL-induced osteoclastogenesis 58 ; the expression of MafB itself is downregulated by RANKL stimulation 59 . These data may imply the possible relationship between RANKL-induced suppression of MafB expression and the downregulation of coronin 1A during osteoclastogenesis.
Regarding the role of coronin 1A in osteoclast precursor cells, i.e., in monocyte/macrophage-lineage cells, it has been shown that the inflammatory stimulation-induced coronin 1A phosphorylation triggers cellular switching from phagocytosis to micropinocytosis, which enables efficient uptake and degradation of large amounts of infectious materials 60,61 , suggesting that coronin 1A is important for host defense against infection. How about the role of coronin 1A in premature osteoclasts then? Premature osteoclasts actually express a large quantity of cathpsin K in cells, however, they would not considerably secrete it until they reach a later stage of maturation and firmly stick to the bones. Given the high expression of coronin 1A in macrophages and premature osteoclasts, it raises the intriguing possibility that coronin 1A operates mainly in those cells to tightly regulate lysosomal secretion of vital enzymes such as cathepsins. However, further studies are required to address this point.
In conclusion, we have demonstrated that coronin 1A and its binding to actin play an important role, possibly in concert with autophagic machinery, in regulating cathepsin K secretion in osteoclast-lineage cells. The present study thus provides a novel therapeutic target for bone diseases such as osteoporosis and rheumatoid arthritis.

Assesment of osteoclastic bone resorption activity.
BMMs were cultured on the bone biomimetic synthetic surface plate and differentiation to osteoclasts was induced. The cells were removed from the plates, and bone pits generated by osteoclasts were stained and visualized by von Kossa staining. After washing the wells, pits were imaged by microscopy (Evos; Thermo Fisher Scientific). The pits were analysed using MetaMorph software.
Acridine orange staining assay. The acidification activity was measured with Acridine orange staining assay according to previously described methods 40 . Osteoclasts were incubated in α -MEM with 10% FBS containing 5 μ g/mL Acridine orange (Dojindo) for 15 min at 37 °C, then washed three times with PBS. The cells were incubated in fresh medium without Acridine orange for 10 min at 37 °C and observed by confocal microscope (LSM 710 [Zeiss]) with an appropriate filter set (excitation: 450-to 490-nm band-pass filter and emission: 515-to 565-nm band-pass filter or 600-nm long-pass filter). Images were analysed by using MetaMorph software.

LPS-induced bone resorption model. The recombinant adenovirus vector (pAdenoX-ZsGreen1) that
carried mouse coronin 1A was constructed. To generate recombinant adenoviruses, Adeno-X 293 cells were transfected with the PacI-digested adenovirus by using Adeno-X ™ Adenoviral System 3 (Clontech Laboratories) according to the manufacturer's instructions. In total, 1 × 10 7 infection forming units (IFU) of virus diluted in 50 μ L of PBS were injected onto the calvariae four times at one day intervals. Mice were administered with a local calvarial injection of LPS at 1.5 mg/kg body weight three times at two days intervals. Seven days after the first injection, the calvariae were collected and fixed in 4% paraformaldehyde and analysed histologically. The number of TRAP-positive cells per millimetre of calvarial bone surface was counted. The eroded surface per bone surface (ES/BS, %) was analysed by using MetaMorph software.

Micro-computed tomography analysis. Calvarial bones of the mice injected with adenovirus express-
ing LacZ or coronin 1A were analysed using ScanXmate-A100S Scanner (ComScanTechno, Yokohama, Japan) as described previously 62 . Three-dimensional microstructure images were reconstituted and the void volume on the surface of the calvarial bones was calculated by TRI/3D-BON software (RATOC System Engineering, Tokyo, Japan). Statistical analysis. Data were presented as means ± standard error of the mean (SEM) and compared using the Student's t-test or one-way analysis of variance (ANOVA) plus post hoc multiple comparisons using the Dunnett test. A value of P < 0.05 was considered significant.