Subosteoclastic bone resorption is a result of HCl and proteinase secretion through a late endosome-like bone facing membrane domain called ruffled border. As bone matrix is degraded, it enters osteoclasts' transcytotic vesicles for further processing and is then finally exocytosed to the intercellular space. The present study clarifies the spatial relationship between these vesicle fusion and matrix uptake processes at the ruffled border. Our results show the presence of vacuolar H+-ATPase, small GTPase rab7 as well as dense aggregates of F-actin at the peripheral ruffled border, where basolaterally endocytosed transferrin and cathepsin K are delivered. On the contrary, rhodamine-labeled bone matrix enters transcytotic vesicles at the central ruffled border, where the vesicle budding proteins such as clathrin, AP-2 and dynamin II are also localized. We present a model for the mechanism of ruffled border turnover and suggest that, due to its late endosomal characteristics, the ruffled border serves as a valuable model for studying the dynamic organization of other endosomal compartments as well.
Osteoclasts (OCs) are the cells specialized for mineralized bone resorption. They are formed by the maturation and fusion of the monocyte/macrophage lineage cells, and thus eventually become multinucleated. Osteoclastic bone resorption is essential for the normal skeletal growth and bone remodeling. Pathological enhancement and disturbances in their function are associated with a number of bone diseases, including osteoporosis and various forms of osteopetrosis.
Not only in vivo, but also when cultured in vitro on its natural substrate, bone, OC activation for bone resorption leads to a local matrix digestion in the subosteoclastic space (1,2). As a result, so-called resorption lacuna is formed. Intracellular events involved in this process are so far inadequately understood, but they are linked to drastic changes in the organization of the cytoskeleton and vesicle transportation which, in turn, leads to OC polarization (3). In this process the endocytotic pathway from the basolateral membrane (4) and the biosynthetic pathway of lysosomal enzymes from the Golgi (5) are re-oriented towards the bone surface. Vesicles fuse to the plasma membrane facing the bone, which is followed by hydrochloric acid secretion by the vacuolar H+-ATPase (v-H+-ATPase) and protease (e.g. cathepsin K) release (6–10). The plasma membrane becomes convoluted as a result of the expanded membrane area, and due to the nature of the fusing vesicles, it gains several features that resemble those of the vesicular late endosomal membranes (4). Thus, the ruffled border (RB) is formed.
As a result of the hydrochloric acid and protease secretion, bone matrix under the RB is degraded. This process is followed by degraded matrix internalization, transcytosis through the cell and finally, exocytosis to the intercellular space via the functional secretory domain (FSD) at the center of the basolateral membrane (11–13). The RB and the resorption lacuna underneath it are circumscribed by a sealing zone (SZ) which mediates a tight attachment to bone matrix and has been suggested to inhibit free diffusion of molecules like the tight junctions in epithelial cells (14). However, this assumption has been challenged by recent data suggesting that negatively charged molecules with Mr less than 10 000 diffuse freely under the SZ (15). The SZ is characterized by the absence of vitronectin receptor αVβ3, which is present at the RB and throughout the basolateral membrane (16–19). Some contrary data have also been presented suggesting the exact localization of αVβ3 at the sealing zone (18,19). The SZ is also clearly recognized by dense bundles of intracellular F-actin (actin ring) and cytoskeletal proteins linked to it (20–24).
When the OC has finished the resorption process at one site, it either loses the RB and the round actin ring surrounding it and becomes inactive, or it rolls on the bone surface maintaining its resorptive activity (25). In the first case, a simple resorption lacuna is formed, but when rolling on the bone, a single lacuna with a length of hundreds of micrometers can be formed. The striking feature of the latter is that the actin bundles form a kidney-shaped structure (instead of a ring-shaped), where the convex side shows the direction of motion (25). Whereas the role of F-actin at the sealing zone is most likely related to the attachment and/or OC polarization, the role of F-actin at the RB may have an impact on the RB turnover (26).
Recently, several proteins have been identified as regulators in the intracellular vesicle budding, motility, docking and fusion. The directed assembly of clathrin and the AP2 adaptor complexes leads to the formation of clathrin-coated pits, which bud from the plasma membrane and lysosomes (27,28). Some other proteins are involved in the clathrin-mediated endocytosis as well. For example, dynamin is a membrane-associated large-molecular-weight GTPase that has been shown to form rings and spirals around the neck of budding coated pits. Subsequent GTP hydrolysis triggers constriction of the rings, which leads to separation of endocytotic vesicles from the donor membrane (29–31). Furthermore, small GTPases of the rab family are localized at distinct intracellular compartments, where they regulate different steps of the intracellular membrane and traffic along the biosynthetic/secretory and endocytic pathways (32,33). A substantial amount of evidence suggests that rab7 associates with and regulates the trafficking events related to late endosomes (34,35). Mammalian rab7 is involved in the vesicle transport from early to late endosomes (35) and also from late endosomes to lysosomes (36). In OCs it regulates the endocytic vesicle transport to the RB and controls its formation (37).
Despite our rather profound understanding of the bone resorption phenomenon, the mode of function and the membrane turnover of the RB are still unknown. To address these important questions, we studied the spatial organization of the vesicle fusion and matrix uptake processes at the osteoclast RB. Our data strongly suggest that rather than taking place randomly in a punctuate and chaotic manner and occupying the entire RB, vesicle fusion and matrix uptake take place in a highly ordered fashion at the peripheral and central RB, respectively.
Vesicle fusion to RB is associated with the peripheral subdomain rich in F-actin
When osteoclasts are cultured on cortical bovine bone slices and stained for F-actin and β3 integrin, they reveal different morphologies representing different phases of the dynamic resorption cycle (20,38). For the sake of clarity, we focus here on the two most common types of resorbing OCs. In OCs with a single and round resorption pit, F-actin formed a circular and dense ring outlining the resorption lacuna (referred to as OC-1, Figure 1A–D). In these cells, diffuse F-actin staining was also observed right inside the actin ring (and thus, inside the SZ). This F-actin accumulation, as we will show, represents the peripheral RB. Although some labeling of the β3 subunit of vitronectin receptor was observed throughout the RB, it was abundantly present at the peripheral RB where it colocalized with F-actin. It was also detected at the basolateral membrane but not at the SZ. OCs on large, extended resorption pits regularly demonstrated kidney-shaped F-actin rings (referred to as OC-2, Figure 1E–H). In these cells, αvβ3 integrin colocalized with F-actin at the RB ‘leading edge’ right inside the sealing zone. Another β3 subunit-enriched membrane patch was found outside the actin ring.
We have previously shown that a 30-min incubation period of OC cultures in the presence of iron-loaded transferrin leads to its presence at the RB. During a longer follow-up time, transferrin is also observed at the villous FSD at the top of the cell, suggesting its recycling from the RB to this membrane domain (39). To confirm that this pattern of transferrin localization results from endocytosis and subsequent transcytotic delivery towards the RB rather than diffusion under the sealing zone, we studied its sub/intracellular localization after several incubation periods and compared it to that of fluorescein isothiocyanate (FITC)-labeled dextran (MW 10 000). Negatively charged small-molecular-weight dextran has previously been shown to diffuse during a few minutes pulse under the sealing zone to the subosteoclastic lacuna and bind preferentially to its edge (15). In agreement with previous results (15), we observed that FITC-dextran was abundantly present at the resorption lacuna already after a 5-min incubation period (Figure 2A–C). The diffusion rate of transferrin, however, proved different: after 5 min of incubation transferrin (MW ∼ 80 000) was observed in vesicular compartments near the plasma membrane (Figure 2D–F) and after 10 min at circumnuclear re-cycling compartments (Figure 2G–I). There was no signal in the area inside the actin ring (at the level of the RB), which indicates that transferrin is too large a molecule for free diffusion under the sealing zone. After 20 min, however, transferrin was abundantly present at the RB (Figure 2J–L). On the basis of these observations, we conclude that there is no leakage of transferrin under the sealing zone into the subosteoclastic resorption lacuna. Thus, transferrin observed at the RB must be delivered through the cell.
In order to clarify whether transferrin-transporting vesicles fuse to the RB randomly in a chaotic manner or whether there is a preference towards some area of the RB, we studied transferrin localization at the RB further. Compared to the 20-min follow-up, extending the incubation time to 30 min led to an even more intensive labeling of transferrin at the RB. It colocalized with F-actin at the peripheral part of the RB in OC-1 (Figure 3A–C) and at the arch-shaped leading edge of the RB in OC-2 (Figure 3D–F). Also, electron microscopic studies showed accumulation and fusion of transferrin-labeled vesicles at the peripheral RB but not at the central RB (Figure 3J,K). Interestingly, not only translucent vesicles but also multivesicular bodies were often observed to release transferrin-labeled internal vesicles onto the bone surface (Figure 3K,L). However, despite the association of transferrin with microtubules in the cytoplasm, their distribution at the RB was complementary (Figure 3G–I).
Just as the transcytotic vesicular pathway of transferrin from the basolateral membrane was targeted towards the peripheral RB, also v-H+-ATPase and the lysosomal cysteine proteinase responsible for collagen degradation, cathepsin K, were localized at the periphery and at the leading edge of the RB (Figure 4). Similarly, also rab7, the small GTPase regulating the trafficking events from early to late endosomes (35) and also from late endosomes to lysosomes (36) was present only at the peripheral RB but not at the central RB.
These findings imply that vesicles targeted to the RB fuse and release their contents selectively at the periphery or the leading edge of the RB, in the sequel called the vesicle fusion zone.
Degraded matrix uptake occurs at the central RB
The absence of rab7, v-H+-ATPase, cathepsin K and transferrin from the central RB in OC-1 and their complete absence from the β3-enriched membrane patch of OC-2 prompted us to further explore the nature and function of these membrane domains. In resorbing OCs a punctate labeling pattern of clathrin was not only localized at the plasma membrane and juxtanuclearly corresponding to the Golgi (data not shown), but also at the central RB in OC-1 (Figure 5A–C). In OC-2, an intensive labeling of clathrin was observed behind the actin-rich leading edge and at the β3-enriched membrane outside the actin ring (Figure 4D–F). These observations suggest that endocytic vesicles bud from a subdomain distinct from the area to which the RB-targeted secretory vesicles fuse.
To test this hypothesis further, we performed double labeling of phalloidin and dynamin II, which has been shown to play a central role in coated vesicle scission. Like clathrin, dynamin II was clustered at the central RB in OC-1 (Figure 5G–I). In OC-2, it was observed behind the actin-rich leading edge and at the β3-enriched membrane patch outside the actin ring (Figure 5J–L). However, there was no dynamin II at the peripheral RB rich in F-actin. AP-2 and the AP-2 partner, Eps15 (32) (data not shown), showed similar staining patterns to that of clathrin and dynamin II (Figure 5M–R).
Furthermore, when OCs were cultured on tetramethylrhodamine isothiocyanate (TRITC)-coated bone slices, they internalized fluorescent bone matrix into intracellular vesicular compartments. These compartments did not contain transferrin, which indicates that transferrin is delivered to the RB but is not efficiently internalized from lacuna during the 30-min incubation period (Figure 6A–C). The endocytosis of TRITC-labeled bone matrix occurred through the central part of the RB, while transferrin was delivered to the peripheral RB (Figure 6D–G). These findings imply that degraded matrix uptake and membrane internalization take place at the central RB and at the β3-enriched membrane patch outside the actin ring complementarily to the vesicle fusion zone.
Finally, we observed a complementary distribution of α-tubulin and F-actin at the RB (Figure 7A–F). While F-actin was present inside the actin ring at the peripheral RB and at the leading edge of it, microtubules were abundantly present at the central RB as well as at the β3-enriched membrane patch outside the actin ring. To study their specific roles in vesicle delivery to (F-actin) and from (microtubules), experiments with cytochalasin D, nocodazole and the myosin inhibitor 2,3-butanedione monoxime were performed, but their effect on the cytoskeleton and the vesicular transport could not be interpreted due to rapid loss of OC polarization in the presence of these agents.
On the basis of our results, we propose a novel model for RB dynamics, where secretory events and matrix uptake occur at distinct subdomains (Figure 7G–L). The peripheral RB represents the secretory subdomain, vesicle fusion zone (FZ), while the central RB represents the degraded matrix uptake zone (UZ). Thus, these results suggest that the RB membrane is under continuous turnover like, for example, the synaptic membrane in neurons.
At least two different mechanisms of membrane recycling have been proposed to exist in synaptic terminals. The classical model suggests synaptic vesicle membrane retrieval as coated vesicles after fusion to the synaptic membrane and transmitter release (40). The ‘kiss and run’ mechanism has been suggested to result from immediate pinch-off of the synaptic membrane after the content release without fusion to the plasma membrane (41,42). A new model of the recycling machinery in a synaptic bouton of a Drosophila neuromuscular junction was presented recently (43). In each bouton, there are 1–200 active zones, the size of 1 μm and distributed approximately 0.5 μm apart, where exocytosis of the neurotransmitter takes place. In between are the proteins involved in the formation of clathrin-coated vesicles, such as AP-2, dynamin and Dap160, indicating the site of endocytosis (43-45). Furthermore, fluorescent endocytotic probes are internalized preferentially at these sites in nerve–muscle synapses of vertebrates (46,47), although cargo molecules themselves do not determine the distribution of the endocytotic machinery (43,46). This pattern of organization allows the rapid recycling of synaptic vesicle proteins, since they diffuse within seconds to the surrounding recycling machinery after release from the synaptic vesicles (43). Furthermore, it allows each endocytotic site to retrieve spent vesicular membrane locally and stabilize the position of the active zone (47).
In OCs, the delayed internalization of the fused late endocytotic vesicular membrane is the most likely factor leading to the convoluted RB formation. It is also very likely that a portion of secreted proteinases is taken in with degraded bone matrix after secretion. However, since it takes hours for one OC to finish resorption at one site (48), and since the time needed for the proteases to accomplish their function in the resorption lacuna is a magnitude longer than in the case of neurotransmitters, there is no need for a rapid membrane recycling. Our data strongly suggest that the RB is operated through a crawling mechanism: acidic vesicles first fuse and release their contents on nondegraded bone (Figure 7G,J). After minutes and hours, as the mineralized and organic matrices are dissolved, they are internalized, transcytosed and finally exocytosed (11–13). The removal of matrix results in partial penetration of OC into bone. Simultaneously, acidic vesicle flow onto nondegraded bone surface leads to ‘new’ RB formation (Figure 7H,K). Targeted vesicle fusion to one side of the RB results only in directed RB motion and matrix uptake at the concave part of the RB (Figure 7I,L). The molecular mechanisms behind the RB transition from one stage to another remain to be elucidated.
Interestingly, our data suggest that vesicles targeted to the RB may switch from microtubules to microfilaments before fusion at the vesicle fusion zone. Thus, this mechanism would have a similarity to the phenomena where the organelles in neuronal axons reach the nerve terminal in a microfilament-dependent fashion (49–51). Our suggestion is also supported by the recently published data showing that v-H+-ATPase binds directly to F-actin and forms a detergent-insoluble complex in OCs (26,52,53).
The RB shares a number of characteristics similar to late endosomal compartments in other cell types (4). This is also supported by our present findings showing that neither the early endosomal marker, EEA1, nor the recycling endosomal marker, rab11, was detected at the RB (data not shown). Thus, the giant RB membrane (with a diameter of 10–40 μm) provides an excellent experimental model for investigating the dynamic processes in the endocytotic pathway further. The subdomains at the RB presented here may represent a common mechanism applied in the maintenance of endosomal compartments. This is supported by recent data providing evidence for the existence of subdomains in the early endosomal membrane (54). The late endosomal nature of the RB is also suggested by our results showing the association of multivesicular body-like vesicles with the RB. We find the potential physiological role of the internal vesicles released to bone matrix very interesting. If there is one, they might mediate a cytolytic effect similarly to exosomes released from cytotoxic CD8+ T lymphocytes to the target cell surface (55,56). This way OC could mediate its cytolytic effect on osteocytes engulfed in bone (57,58). However, the release of internal vesicles might simply be a consequence of re-organized endocytic pathways towards the RB during bone resorption (3–5).
Our results, together with the previously reported data (37), suggest that rab7 regulates the transportation and docking/fusion of vesicles to the RB as it does in the late endocytic pathway in other cell types (35). Our results also suggest that proteinases and protons are secreted to the resorption lacuna through the same membrane domain. The pathway from the basolateral membrane to the RB and the secretory pathway of proteinases may fuse before reaching the RB. Also, this step may be regulated by rab7 (59). Together, all these data indicate that the late endosomal pathway is involved in the formation and turnover of the RB, which may be a useful target for the intervention of the OC function.
In addition to mediating membrane budding at the plasma membrane and the trans-Golgi network (TGN), clathrin-associated coat protein complexes have recently been shown to assemble onto endosomes and lysosomes (28,60). The location of clathrin, dynamin II, AP-2 and EPS-15 at the central RB suggests that the clathrin/AP-2 complex has a role in the budding of transcytotic vesicles carrying the degraded bone matrix. Furthermore, the identity of the receptors for the bone matrix degradation products still remains unknown. The distribution of αvβ3 integrin, its colocalization with endocytosed bone fragments in vesicular compartments in osteoclasts (data not shown) and its high affinity for denatured type I collagen make αvβ3 a strong candidate for this role (61).
The present study shows that unlike in neuronal synapses, where local membrane recycling is the major pathway for transmitter uptake (41–43,47), these actions take place in a highly ordered manner in osteoclastic RB. Transferrin that is endocytosed from the basolateral membrane is delivered to the peripheral RB, colocalizes with F-actin, v-H+-ATPase, cathepsin K and rab7. Meanwhile, clathrin, AP-2, dynamin II and the endocytosed bone matrix are concentrated at the central RB. Thus, RB has two subdomains – the peripheral vesicle fusion zone and the central matrix uptake zone. However, the molecular mechanisms controlling RB dynamics yet remain to be elucidated.
Materials and Methods
Reagents and antibodies
Materials used for cell culture were purchased from Life Technologies, Inc. (Carlsbad, CA). The rabbit anti-rab7, the monoclonal antibody against the β3 subunit of vitronectin receptor (F11), rabbit anti-B2 and − 70 kDa subunit of v-H+-ATPase antibodies, rabbit anticathepsin K, mouse anti-α-adaptin (AC1-M11) and rabbit anti-Eps 15 were the generous gifts from Drs S.R. Pfeffer (Stanford University), M.A. Horton (University College London), J.P. Mattsson (AstraZeneca), D. Bromme (Mount Sinai School of Medicine), M.S. Robinson (University of Cambridge) and Paul M.P. van Bergen en Henegouwen (Utrecht University), respectively. The goat anti-rab7 antibody was purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) The mouse anti-EEA1 and monoclonal anti-dynamin II antibodies came from Transduction Laboratories (Lexington, KY). The mouse anti-clathrin heavy chain antibody (X22) was purchased from Affinity Bioreagents, Inc. (Golden, CO), the rabbit antihuman transferrin from Zymed Laboratories, Inc. (So. San Francisco, CA), and the mouse anti-α-tubulin from Sigma-Aldrich (St. Louis, MO). Fluorescein isothiocyanate- and tetramethylrhodamine isothiocyanate-labeled phalloidin were purchased from Sigma-Aldrich and fluorescein-conjugated secondary antibodies from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA) FITC-dextran with an average MW of 10 000 came from Molecular Probes, Inc. (Eugene, OR)
Isolation and culture of osteoclasts in vitro
The procedure used to isolate and culture rat osteoclasts was modified from the original method developed by Boyde et al. and Chambers et al. (1,2). Briefly, osteoclasts were mechanically harvested from long bones of rat pups into alpha MEM buffered with 20 mm Hepes and containing 100 IU/ml penicillin, 100 μg/ml streptomycin and 10% heat-inactivated fetal bovine serum, pH 6.9. The cells were then attached to devitalized cortical bovine bone slices (100–150 μm thick). After 60 min, nonadherent cells were washed away and the cells on bone slices were cultured in the same medium at 37 °C and 5% CO2/95% air for 48 h.
Indirect immunofluorescence and confocal laser scanning microscopy
Osteoclasts cultured on bone slices were fixed with 3% paraformaldehyde in PBS for 20 min, permeabilized in 0.1% Triton X-100 in PBS for 4 min on ice. For rab7 and rab11 detection, cells were permeabilized with 0.2% saponin for 5 min prior to fixation. For AP-2 and B2 as well as 70 kDa subunits of v-H+-ATPase staining, cells were fixed and permeabilized with − 20 °C methanol for 5 min followed by − 20 °C acetone for 30 s. The slices were incubated with primary antibodies in PBS/0.2% gelatin for 30 min followed by incubation with TRITC- or FITC-conjugated secondary antibodies in the same solution. Filamentous actin (F-actin) was stained with fluorescein-conjugated phalloidin (0.3 μg/ml). Samples were viewed with a Leica fluorescence microscope and a Leica TCS-SP confocal laser scanning microscope equipped with an Argon-Krypton laser (Leica Microsystems Heidelberg GmbH).
Transferrin and dextran experiments
After 48 h of culture, iron-loaded human transferrin (0.1 mg/ml) was added into the culture medium for 1, 3, 5, 10, 20 and 30 min, followed by a wash in prewarmed PBS for 20 s and fixation with 3% PFA for 20 min Then, cells were permeabilized with 0.1% Triton X-100 in PBS and immunostained with polyclonal antibodies against human transferrin. In some experiments FITC-dextran was added into the culture medium to a final concentration of 100 μm. After the incubation of 1, 3, 5 and 10 min, the cells on bone slices were washed twice with prewarmed PBS for 20 s, fixed with 3% PFA and subsequently incubated in the presence of TRITC-phalloidin.
Fluorescence labeling of the surface of bone slices
Thin bone slices were incubated in 0.1 m bicarbonate buffer (pH 8.3) containing 5-(6)-carboxytetramethylrhodamine succimidyl ester (0.25 mg/ml) (Molecular Probes) for 2 h followed by several washes with PBS.
For conventional electron microscopy, cultures were fixed with 2.5% glutaraldehyde buffered with 0.08 m sodium cacodylate buffer containing 0.05% calcium chloride, pH 7.3, for 1 h.
Immuno-EM: After incubation with iron-loaded human transferrin as described above, osteoclasts were fixed with McLean and Nakane's periodate-lysine-paraformaldehyde fixative for 2.5 h. Labeling with polyclonal antitransferrin antibody and the immunoperoxidase reaction were carried out according to Brown and Farquhar (62). Bone slices were decalcified with 5% EDTA in 0.1 m phosphate buffer, pH 7.2. Cells were then postfixed in 1% OsO4 in 0.1 m cacodylate buffer (pH 7.2), dehydrated in acetone and embedded in Epon LX 112. Ultrathin sections were shortly stained with lead citrate and examined using a Philips 410 LS transmission electron microscope.
P. Härkönen, K. Metsikkö, J. Salo, N. Walsh, L. Patrikainen and Pirkko Huuskonen are acknowledged for helpful discussions and critical reading of the manuscript and J. Vääräniemi for providing the TRITC-labeled bone slices. This study was supported by grants from The Academy of Finland, Sigrid Juselius Foundation and Emil Aaltonen Foundation.