Localization and possible role of two different alpha v beta 3 integrin conformations in resting and resorbing osteoclasts

Integrins are membrane receptors that mediate the interactions between cells and the extracellular matrix. Cellular attachment to immobilized ligands commonly results in cytoskeletal reorganization and clustering of integrins at discrete adhesive sites known as focal contacts or, in osteoclasts (OCs) and cells of monocytic lineage, podosomes(Marchisio et al., 1984). In these specialized structures, an array of submembranous proteins, ranging from structural molecules to regulatory enzymes, form a multimolecular complex linking the actin microfilament network with integrins and, hence, with the extracellular matrix (Miyamoto et al.,1995). The formation of focal contacts or podosomes triggers signaling cascades that, in concert with growth factor-activated transduction pathways, regulate cell proliferation, differentiation, survival and migration(Clark and Brugge, 1995;Schwartz et al., 1995). Integrins adhere to various extracellular matrix proteins, and their affinity for these ligands can be altered by conformational changes in their ligand-binding sites in response to cellular signals(Hughes et al., 1997;LaFlamme et al., 1994). Evidence is accumulating that these `basal' and `activated' forms,characterized by different substrate affinities, are functionally distinct. In M21 melanoma cells, the activation of α_vβ₃integrin results in increased numbers of adherent cells, which bind to the ligand more weakly than the `basal' integrin, facilitating migration(Pelletier et al., 1996). We recently demonstrated that, similar to the α<sub>3</sub>-containing integrins in melanoma cells and platelets(Pelletier et al., 1995),α <sub>v</sub>β<sub>3</sub> also exists in human osteoclast-like cells in both the `basal' and `activated' states(Faccio et al., 1998).

The functional cycle of the OC consists of migration towards bone, followed by adherence to the bone surface, where the cell polarizes and initiates the resorptive process. Motile OCs are non-polarized cells, characterized by the presence at their leading edge of membrane protrusions, called lamellipodia,and, in a row behind the leading edge, by podosome complexes, comprising filaments of actin associated with several actin-binding proteins. The passage from a motile to a resorbing cell involves cytoskeletal reorganization and matrix attachment. Following attachment to the bone surface, podosomes assemble in an actin-rich region of the membrane, forming a tight sealing zone, enclosing the resorptive microenvironment. Following insertion of secretory vesicles, a highly convoluted ruffled membrane, the ruffled border,forms facing the bone surface where resorption will take place, while the basosolateral membrane is involved in membrane trafficking(Fig. 4A).

Integrins are believed to play a role in OC activity by mediating matrix adhesion and regulating the cytoskeletal organization required for both cell migration and formation of the sealing zone and ruffled border. OCs express abundant α_vβ₃ integrin(Zambonin-Zallone et al.,1989), which recognizes bone matrix proteins(Davies et al., 1989;Duong et al., 2000;Lakkakorpi et al., 1991;Nesbitt et al., 1993). Addition of RGD-containing peptides or disintegrins (such as echistatin or kistrin), which bind to α_vβ₃, arrest OC adhesion to bone or cause retraction of attached osteoclasts(King et al., 1994;Nakamura et al., 1996;Nakamura et al., 1999;Sato et al., 1990). Moreover,OCs generated from β₃-deficient mice are dysfunctional in vitro and in vivo (McHugh et al.,2000; Feng et al.,2001).

Recent evidence shows that activation of theα _vβ₃ integrin induces the[Ca²⁺]_I-dependent phosphorylation of Pyk2, a non-receptor tyrosine kinase involved in formation of the sealing zone. Activated Pyk2 forms a complex with c-Src and c-Cbl, and these molecules are all involved in outside-in activation of α_vβ₃signaling, possibly leading to podosome assembly(Sanjay et al., 2001). Nakamura et al. recently showed that Src^-/- prefusion OCs adhered to, but failed to spread on, vitronectin (Vn)-coated surfaces and thatα _vβ₃-integrin-mediated signaling was inhibited in these cells. Moreover, the same authors showed that the spreading defect of Src^-/- pOCs could be rescued by M-CSF treatment, which resulted in recruitment of the α_vβ₃ integrin to adhesion contacts. However, the authors did not directly assess whether or not M-CSF treatment leads to changes in integrin affinity for ligand.

Cross-talk between integrin-mediated adhesion and growth factors has been described in many recent studies; however, the underlying mechanisms remain incompletely understood. PDGF induces the association of its phosphorylated receptor with integrin α_vβ₃ in fibroblasts plated on vitronectin, which correlates with enhanced PDGF-induced cell proliferation (Schneller et al.,1997). Basic fibroblast growth factor (bFGF) enhances cell migration of vascular endothelial cells, leading to concentration of`activated' α_vβ₃ integrin to polarized lamellipodia (Kiosses et al.,2001). Moreover, in transformed cells HGF promotes cell migration by increasing α_vβ₃ integrin affinity(Trusolino et al., 2000).

Recently the crystal structure of α_vβ₃integrin has been solved (Xiong et al.,2001; Beglova et al.,2002), showing that ligand binding itself alters the conformation of this integrin, exposing neo-epitopes, known as ligand-induced binding sites(LIBS). LIBS can also be exposed by intracellular signaling events, a process known as inside-out signaling. Anti-LIBS antibodies can be used as tools to identify integrins in the `activated' conformation or to convert an integrin from the `basal' to the `activated' state. AP5 is an anti-LIBS Ab that recognizes the β<sub>3</sub> N-terminus and is regulated by cation-binding at a site distinct from the LIBS(Honda et al., 1995). At normal extracellular calcium levels, AP5 binds toα <sub>v</sub>β<sub>3</sub> only when the integrin is in the`activated' conformation. However, in low calcium concentrations AP5 can bind to all forms of the integrin leading to its activation. Once AP5 is bound at low calcium concentrations, α_vβ₃ remains in the `activated' state even when the calcium levels are increased. It has also been possible to generate monoclonal antibodies that specifically recognize the `activated' conformation of an integrin at the ligand-binding site. WOW1 is one such antibody, in which the heavy chain hypervariable region 3 of PAC1 Fab was replaced with a single α_v-integrin-binding domain,from the multivalent adenovirus penton base, to engineer an antibody that binds to only activated α_vβ₃(Pampori et al., 1999). Thus,binding of WOW1 can be used as another marker of integrin activation. In contrast to AP5 and WOW1, AP3 is a monoclonal Ab, which binds to theα _vβ₃ integrin regardless of its conformational state. In this paper we defined the fraction of integrin that is recognized by AP5 and/or WOW-1 as the `activated' conformation and the fraction of integrin that binds to AP3 and/or 1A2, but not to AP5 or WOW-1, as the `basal' or `non-activated' conformation.

This study focuses on the functional implications ofα _vβ₃ conformation in human OCs. We utilized different sources of human osteoclasts, giant cell tumors of bone-derived cell line, bone-marrow-derived osteoclasts and osteoclastoma cells, all of which have been shown to express osteoclastogenic markers. We demonstrate distinct localization of the `basal' and `activated' integrin in resting and resorbing OCs. We find that during bone resorption `activated'α <sub>v</sub>β<sub>3</sub> is mainly localized in the ruffled border, whereas the `basal' conformation is responsible for the adhesive properties of the sealing zone. Moreover we demonstrate that HGF and M-CSF,two growth factors that regulate osteoclastogenesis and osteoclast survival(Fuller et al., 1993;Grano et al., 1996), modulate the activation state of α_vβ₃. Growth factor treatment alters the affinity of α_vβ₃ toward its ligands, promoting OC migration to theα _vβ₃ ligand osteopontin and enhancing bone resorption. These data offer a potential explanation for the diverse responses of OC to the same ligand.

Several osteoclast sources were utilized for this research. Primary human osteoclasts were used for immunofluorescence because the antibodies used to detect β₃ in all conformations (AP3 and 1A2) and in the high affinity conformation (AP5 and WOW1) recognize only human β₃integrin. Murine osteoclasts and/or the human osteoclastlike cell line GCT 23 were used for bone resorption, migration and flow cytometry experiments.

Human osteoclasts were obtained as described previously by mechanical disaggregation of small surgery specimens from osteoclastomas(Grano et al., 1996) or from human bone marrow cultures. Briefly, human bone marrow samples obtained by orthopedic surgical procedures were subjected to Ficoll-Hypaque (Ficoll, Sigma Chemical Co.; Hypaque 76, Nycomed, Princeton, NJ) gradient purification, and cells at the gradient interface were collected and cultured overnight in the presence of 20 ng/ml of recombinant human M-CSF (R&D Systems Inc.,Minneapolis, MN). Non-adherent cells were plated onto coverslips or dentine slices in the presence of alpha-MEM supplemented with 20% Fetal Calf Serum(FCS, Gibco Limited, Uxbridge, UK), and Vitamin D3 10^-9 M at 37°C in a water saturated atmosphere with 5% CO₂. After 3 weeks multinucleated TRAP-positive human osteoclasts appeared.

An osteoclast-like cell line, GCT 23, cloned from a human giant cell tumor of bone has been extensively characterized for its osteoclast phenotype(Grano et al., 1994). Cells used for the experiments were all within nine and 14 passages. All the cells were maintained in Iscove's medium (IMEM) supplemented with 10% fetal calf serum, 100 IU/ml penicillin, 100 μg/ml streptomycin, 2.5 μg/ml amphotericin and 50 IU mycostatin at 37°C, in a water saturated atmosphere with 5% CO₂ and fed by medium replacement every 2-3 days.

Macrophages were isolated from bone marrow of four- to eight-week-old mice,cultured overnight in α-MEM containing 10% heat-inactivated FBS and subjected to Ficoll-Hypaque (Ficoll, Sigma Chemical Co.; Hypaque 76, Nycomed)gradient purification. Cells at the gradient interface were collected and cultured in the presence of 10 ng/ml recombinant M-CSF (R&D Systems Inc.)and 100 ng/ml RANKL for 3-4 days.

Recombinant osteopontin (OPN) was the kind gift of K. O. Johanson(SmithKline Beecham, King of Prussia, PA). Laminin from human placenta (LN-2)was from Sigma Chemical Co (St Louis, MO). All the antibodies were protein-A purified IgGs from ascites fluid. LM609 is a blocking antiα _vβ₃ antibodyCheresh, 1991) and was kindly supplied by David Cheresh (The Scripps Research Institute, La Jolla, CA);WOW1, a monoclonal antibody which specifically binds the activated conformation of α_vβ₃, was generously provided by S. Shattil (The Scripps Research Institute, La Jolla, CA); AP3, a monoclonal antibody which recognizes all forms of β₃, and AP5,a monoclonal antibody against the activated conformation ofβ ₃, have been extensively characterized(Honda et al., 1995); 1A2, an mAb against human β₃ was a gift of S. Blystone, Department of Cell and Developmental Biology, State University of New York (SUNY), Upstate Medical University at Syracuse, NY.

Monoclonal antibodies against the α_v integrin subunit were purchased from Telios Pharmaceuticals, Inc (San Diego, CA). Human recombinant macrophage colony stimulating factor (M-CSF) was from Genetics Institute(Cambridge, MA). Affinity-purified HGF was a kind gift of P. M. Comoglio(IRCC, University of Torino, Italy).

Freshly isolated human osteoclasts were plated onto bone slices,hydroxyapatite-coated coverslips (Osteologic, Millenium Biologix Inc.,Kingston, Ontario, Canada) or glass coverslips in α-MEM supplemented with 20% FHS (fetal horse serum) at 37°C in a humidified atmosphere containing 5% CO₂. After 24 hours, cells were incubated with HGF or M-CSF in IMEM supplemented with 0.5% BSA for 30 minutes at 37°C. Control cells were incubated with α-MEM + 0.5% BSA alone. After incubation,cells were washed three times with PBS and fixed in 3% paraformaldehyde, 2%sucrose in PBS pH 7.6 for 10 minutes at room temperature. After rinsing in PBS, cells were made permeable to antibodies by soaking coverslips for 3 minutes at 0°C in Hepes-Triton X-100 buffer (20 mmol/l Hepes pH 7.4, 300 mmol/l sucrose, 50 mmol/l NaCl, 3 mmol/l MgCl₂ and 0.5% Triton X-100).

For indirect immunofluorescence, the primary antibody was layered on fixed and permeabilized cells and incubated in a humidified chamber for 45 minutes at 37°C. The distribution of AP5 and WOW1 were also evaluated in live cells using Abs incubated for 45 minutes at 4°C to prevent receptor internalization. Cells were then fixed and permeabilized as described previously. After rinsing in PBS (pH 7.6), coverslips were incubated with the appropriate Cyanine-3 conjugated secondary Ab (Cy-3, Chemicon, CA) and with 10μg/ml fluorescein-labeled phalloidin (F-PHD, Sigma) for 45 minutes at 37°C. Stained coverslips were than mounted in 20% Mowiol 4-88(Calbiochem-Novabiochem Corporation, La Jolla, CA). Observations were performed by epifluorescence in a Zeiss axioplan microscope.

Osteoclasts plated onto bone slices were viewed with a confocal microscope in the M.I.A. laboratory at DIBIT San Raffaele in Milan, Italy.

GCT23 cells were removed from growth plates by a brief treatment with Trypsin/EDTA (Sigma) and washed three times in a calcium-free buffer based on Hanks Balanced Salt Solution (HBSS). Pretreated cells were incubated with HGF or M-CSF in IMEM supplemented with 0.5% BSA for 30 minutes at 37°C;control cells were incubated with IMEM alone. After incubation with or without the growth factors, cells were washed twice with HBSS buffer supplemented with calcium and magnesium and incubated with the mAb AP5 (50 μg/ml) against the activated β₃ integrin subunit in high calcium buffer for 45 minutes on ice. A positive control was performed by incubating cells with AP5 in HBSS calcium-free buffer. Cells were then rinsed and treated with FITC-conjugated goat-anti-mouse serum (Sigma) for 30 minutes on ice. Finally,cells were suspended in 0.5 ml of HBSS buffer in the presence of propidium iodide (500 ng/ml, Sigma) added immediately prior to analysis on a Becton-Dickinson FACScan. Appropriate forward- and side-scatter gates were set for these cells, and only those cells excluding propidium iodide were included in the analysis of FITC fluorescence.

Murine osteoclasts from bone marrow macrophages were generated on whale dentin slices or on BD BioCoat^TM Osteologic^TM Bone Cell Culture System (Millenium Biologix) as described above. After 3 days in culture in the presence of 25 ng/ml M-CSF and 100 ng/ml RANKL, the culture media was changed and HGF (50 ng/ml), M-CSF (100 ng/ml) or vehicle alone were added for the next 48 hours. Cells on dentine slices were fixed in 4% PFA for 10 minutes at room temperature and TRAP stained. Resorptive pits were analyzed by light microscopy and pictures were taken with a 16× objective.

At the same time point, osteoclasts plated onto dentine slices were removed by a brief treatment with domestic bleach, and resorptive areas were analyzed by light microscopy.

Haptotactic migration assays were performed in transwell plates with an 8μm pore size (Costar, Cambridge, MA) as described previously(Pelletier et al., 1996). Briefly, the lower side of the membrane was coated with OPN or LN at the concentration of 10 μg/ml for 2 hours at room temperature. Both sides of the membrane were blocked with 2% heat-denatured BSA for 1 hour. Cells were preincubated as described above, and 10⁴ cells, diluted in IMEM supplemented with 0.5% FCS, were added per well. Following incubation at the indicated times at 37°C, cells were stained with ADiff-Quick@ (Dade Diagnostics, Aquada, PR). The cells attached to the top surface of the membrane were removed with cotton swabs. Cells that migrated to the lower side were viewed at 300× magnification, and the number of cells per field was counted. Results represent the averages from 15 fields±s.e. of a representative experiment. For inhibitory experiments, cells, after pre-incubation with AP5, HGF or M-CSF, were incubated for an additional 30 minutes at 37°C with the function-blocking mAb LM609 or the PI3-kinase inhibitor Wortmannin (Sigma).

Statistical analysis was performed by unpaired Student's t-test. The results are expressed as average±s.e.

We have demonstrated that α_vβ₃ exists in human osteoclast-like cells in two conformational states(Faccio et al., 1998), with the`activated' state, recognized by the monoclonal Ab AP5, stimulating cell adhesion and migration toward α<sub>v</sub>β<sub>3</sub> ligands. In transformed cells, growth factors promote cell migration by increasing integrin affinity (Trusolino et al.,2000). Thus, our first exercise was to determine whether cytokines or growth factors could regulate α<sub>v</sub>β<sub>3</sub>integrin conformations in human osteoclast-like cell line GCT 23. We turned to HGF and M-CSF because of their impact on OC activity and differentiation(Felix et al., 1994;Fuller et al., 1993;Grano et al., 1996). We used human osteoclast-like cell line and not primary OCs in order to obtain a large number of cells. Flow cytometric analysis of human osteoclast-like cells with AP5 was used to examine the effects of HGF and M-CSF onα <sub>v</sub>β<sub>3</sub> conformations. Since in the presence of low calcium concentrations AP5 induces the activation of allα <sub>v</sub>β<sub>3</sub>(Faccio et al., 1998), we used this condition (AP5 —Ca) to detect total β<sub>3</sub>(Fig. 1A). By contrast, in high calcium, buffer AP5 recognizes only `activated'α _vβ₃, and we used this condition (AP5+Ca) to determine the degree of α_vβ₃ activation in unstimulated cells (Fig. 1A). Pre-incubation of human osteoclast-like cells with HGF (5 ng/ml)(Fig. 1B) or M-CSF (0.5 nM)(Fig. 1C) for 30 minutes at 37°C, prompts increased binding of AP5 in high calcium buffer, indicating that both HGF and M-CSF can induce the `activated'α _vβ₃ conformation. Because phosphatidyl inositol 3-kinase (PI 3-kinase) has been implicated in integrin activation in leukocytes and platelets (Shimizu et al.,1995; Kovacsovics et al.,1995; Zell et al.,1996; Zhang et al.,1996), we analyzed its role in growth-factor-mediatedα _vβ₃ activation in OCs. Addition of wortmannin, the PI 3-kinase inhibitor, to the pre-incubation medium prevents the integrin-activating effect of HGF or M-CSF (data not shown), indicating that PI 3-kinase activity is required forα _vβ₃ activation.

Having established the presence of theα _vβ₃ integrin in different conformational states in human osteoclast-like cells GCT 23, we analyzed their expression in human osteoclasts. Because FACS analysis detects the presence of the activatedα _vβ₃ but not its localization, we turned to immunofluorescence as a method to detect the distribution of the two conformations of α_vβ₃. We used the mAb AP3,which binds to epitopes within residues 287-489(Honda et al., 1995) and recognizes β₃ in both conformational states. In non-polarized primary OCs plated on glass, AP3 highlightsα _vβ₃ on the OC surface in the podosomes(Fig. 2A,B) and in the motile parts of the plasma membrane, called lamellipodia(Fig. 2A,B arrow). At higher magnification (Fig. 2A,insert), AP3 demonstrates that α_vβ₃ is organized in rosette-like structures that surround a core of actin filaments in the podosomes (Fig. 2B,insert). We have also seen this particular organization of theβ ₃ integrin around the podosomal actin core in avian OCs(Marchisio et al., 1984). Similar results were obtained using the mAb 1A2, which binds to an external domain of the β₃ subunit (not shown), and using an anti-α_v mAb, which exhibited a complete overlap with theβ ₃ distribution (data not shown). By contrast, the `activated'conformation of the β<sub>3</sub> chain, which is recognized by AP5, was primarily detected at the edge of the cell, where it colocalized with a meshwork of actin filaments in lamellipodia and membrane ruffles(Fig. 2C,D, insert) but was mostly absent from the podosomes. To confirm that the different distribution of the `activated' receptor was not caused by an artifact dependent on fixation procedures or to non-specific binding of AP5, cells were stained with a second Ab that recognizes `activated' α<sub>v</sub>β<sub>3</sub>,WOW1, before and after fixation. In Fig. 3, WOW1 staining of human osteoclasts shows a complete concordance with the AP5 distribution. In both live and fixed cells, the `activated'receptor colocalized at the cell edge with the cortical actin (A-B and D-E arrows), and, in contrast to the `basal' receptor, was absent from the podosomes (C,F, asterisk). On the basis of these findings, we performed all the following immunofluorescence staining experiments on fixed cells to avoid possible modifications, such as receptor internalization, during the Ab incubation.

As shown in Fig. 4A,resorbing OCs are polarized cells that adhere to the bone through the sealing zone (SZ) and form a specific membrane domain, the ruffled border (RB), which is involved in the dissolution of mineralized bone matrix. Having shown that the two different conformational states of the integrin coexist in non-resorbing OCs plated on coverslips, we examined the distribution of the two conformational states of α_vβ₃ during bone resorption. Human OCs were plated onto bone slices, stained with AP3 or AP5 mAbs and observed by confocal microscopy. In the cross-section from the sealing zone, total α_vβ₃, as visualized with AP3 (Fig. 4B), appeared as a circle at the cell periphery (SZ) that perfectly colocalized with the actin ring (Fig. 4C), suggesting that the receptor participates in recognition of bone matrix. The cross-section shown in panels D and E revealed the α_vβ₃distribution in the ruffled border (Fig. 4D) where actin filaments were also detected(Fig. 4E). The same cell has been analyzed through an optical yz plane (F-G). In this cross section, the overlap of α_vβ₃ and actin in the sealing zone(SZ), as well as their localization in the ruffled border (RB), was evident. By contrast, the receptor in the `activated' conformation, detected by AP5,only partially colocalized in a punctuate pattern with actin in the sealing zone (Fig. 5A,B), a distribution confirmed by analysis of the same cell through the yz plane(Fig. 5E,F in SZ). The`activated' integrin was strongly detected in the ruffled border(Fig. 5C,D;Fig. 5E,F in RB) with a similar distribution to the AP3 staining. On the basis of these observations we can conclude that the `activated' form is present mainly at the edge of the ruffled border, whereas the `basal' conformation is also present in the sealing zone. These findings suggest that the two conformational states ofα _vβ₃ are differently distributed, indicating that they may exert distinct effects on OC activity.

We next determined whether both `basal' and `activated' conformation ofα _vβ₃ could be modulated by growth factor treatment. Human OCs plated on coverslips were treated with HGF, M-CSF or vehicle alone and immunostained with the mAb AP3, which recognizes bothα _vβ₃ conformational states. In untreated cells, α_vβ₃ was organized in rosette-like structures (Fig. 6A-C), as described previously (Fig. 2A,B). By contrast, pre-incubation with HGF or M-CSF for 30 minutes at 37°C alters the distribution ofα _vβ₃, shifting the integrin away from podosomes to the protrusive edges of the plasma membrane(Fig. 6D-F and 6G-J). The distribution of the `activated' α_vβ₃ detected by AP5 and WOW-1 mirrors immunostaining with AP3 (not shown), indicating that the majority of the integrin has been activated by the growth factors.

Flow-cytometric analysis revealed that PI 3-kinase is required for growth-factor-mediated α_vβ₃ activation in osteoclast-like cells (not shown). In mature OCs, PI 3-kinase associate withα _vβ₃(Hruska et al., 1995;Lakkakorpi et al., 1997). In our system, adding the PI 3-kinase inhibitor wortmannin before growth factor stimulation blunted the α_vβ₃ redistribution(Fig. 6K-M). The receptor was still organized in podosomes and failed to localize at the cell edge. These results suggest that PI 3-kinase is important for the activation and redistribution of α_vβ₃ under growth factor stimulation. Thus, HGF and M-CSF mobilizeα _vβ₃ from the cytoskeleton, inducing its relocalization in a PI-3-kinase-dependent manner.

Given the changes in α_vβ₃ conformation in non-resorbing cells in response to HGF and M-CSF, we next turned to the effects of these cytokines on resorbing OCs. Human OCs were plated onto hydroxyapatite-coated coverslips (BD BioCoat™ Osteologic™), which efficiently substitute for bone slices. Although this inorganic substrate does not itself contain the organic matrix components found in bone, OC resorptive activity does depend on the presence of α_vβ₃integrin (R.F. and D. Novack, unpublished). The necessary organic components may be secreted by the osteoclastic cells and/or contained in the serum-supplemented media. Mature human osteoclasts were treated with growth factors or vehicle and immunostained with AP5. As we saw in OCs on bone(Fig. 5), high affinityα _vβ₃ is confined mainly to the ruffled border and is distributed with a punctuate pattern in the sealing zone in the absence of exogenous growth factors (Fig. 7A,C). Treatment with HGF (Fig. 7D,F) or M-CSF (not shown) for 30 minutes modified the integrin conformation, including the fraction in the sealing zone that appeared brightly stained by AP5 (Fig. 7D). However, the `activated' integrin only partially colocalized(yellow) with the actin ring, as demonstrated by the merge betweenα <sub>v</sub>β<sub>3</sub> staining in red and actin in green(Fig. 7F). The `activated'integrin after growth factor treatment was also detected in small protrusions at the cell periphery external to the actin ring. Longer incubation (8 hours)yielded dramatic morphological changes in resorbing OCs. The cells started to migrate and the integrin was eventually found in more organized motile areas of the plasma membrane, such as lamellipodia and fillopodia (not shown).

To assess the functional implications of HGF- and M-CSF-induced alterations in α_vβ₃ activation, we determined the effect of the growth factors on osteoclastic bone resorption. Murine bone marrow macrophages were plated onto whale dentine slices or hydroxyapatite-coated slides and cultured in the presence of 100 ng/ml RANKL and 25 ng/ml M-CSF. After 3 days, the presence of mature OCs was confirmed by TRAP staining of parallel cultures. At this time, the culture media was replaced with a fresh one containing 50 ng/ml of HGF or 100 ng/ml M-CSF. After an additional 48 hours, cells on dentine slices were fixed and TRAP stained, and the bone resorptive lacunae were analyzed by light microscopy(Fig. 8A-C). Compared with untreated cells (A), HGF (B) and M-CSF (C) treatment increased the pit number 2.6-and 2.3-fold, respectively. To evaluate the resorbed area,hydroxyapatite-coated slides were treated with bleach to remove cells, and the cleared area was quantified. Unstimulated OCs resorbed 30% of the total area(Fig. 8D), whereas HGF (E) and M-CSF (F) increased the resorptive area to 65% and 58%, respectively.

Normal resorptive function of OCs depends in part upon their ability to migrate over the bone surface to initiate new sites of bone resorption. Having shown an increased bone resorption in OCs treated with HGF and M-CSF, we determined whether it was correlated with an increased motility. Haptotactic migration of human osteoclast-like cells was measured in a transwell assay in which the bottom of the membrane had been coated with OPN (10 μg/ml). Cells, which had migrated through the membrane to the OPN-coated surface, were counted. Pretreatment with either growth factor increased human osteoclast-like cell migration to levels comparable to those pre-treated with AP5 in low calcium, which we had previously established promotes motility(Faccio et al., 1998)(Fig. 9A). This increase in cell migration was inhibited by the α_vβ₃blocking mAb LM609 (Fig. 9A). In contrast to OPN, human osteoclast-like cell migration toward laminin (LN),a substrate not recognized by α_vβ₃, was not affected by HGF or M-CSF pretreatment nor inhibited by LM609(Fig. 9B), indicating a specific role for the two growth factors inα _vβ₃-dependent migration.

Having established that the PI-3-kinase inhibitor wortmannin blocks growth-factor-mediated redistribution of α_vβ₃(Fig. 6K-M), we next examined whether this treatment would also affect growth-factor-mediated OC migration. After pre-incubation with HGF, M-CSF or AP5 in the presence of low calcium, as a control for integrin activation, wortmannin (200 nM) was added to the medium and cells were allowed to migrate toward OPN or LN. As seen inFig. 9A, wortmannin substantially reduced cell migration towards OPN to levels comparable with anti-α_vβ₃ mAb-mediated inhibition. The inhibitor was completely ineffective in the altering migration toward LN(Fig. 9B). This result and the previous morphologic observations suggest that PI 3-kinase activity is required for growth-factor-induced cell migration via recruitment of the`activated' α_vβ₃ to the motility related sites.

The activity of integrins in several cell types is tightly controlled, and adhesion and/or motility can depend upon a shift from a `basal' to a`activated' state of the receptor. The leukocyte-specific integrin LFA-1 is expressed at the cell surface in an inactive state but can rapidly be converted into an active state, which leads to enhanced ICAM-1 binding during transmigration across the vasculature (vanKooyk et al., 2000). Similarly, theα _IIbβ₃ integrin is expressed in a `basal'state in circulating platelets and is converted into an `activated' state enabling it to bind to fibrinogen during clot formation. Previously we showed that α_vβ₃ exists in different functional states in osteoclast-like cells (Faccio et al., 1998). The `activated' conformation increases affinity to the ligand, leading to a higher efficiency in migration toward osteopontin. Two distinct mechanisms have been put forward for the integrin activation. The first is that there is an increase in the intrinsic affinity of the integrin to its ligand owing to conformational change, and the second is that the avidity of the receptor increases as a result of receptor clustering at the cell surface. Changes in integrin affinity (interaction of a single integrin with its ligand) can be detected by antibodies that identify activation-dependent neo-epitopes.

In the current study we utilized two different antibodies, WOW-1 and AP5,to detect the `activated' α<sub>v</sub>β<sub>3</sub> integrin. WOW-1, a monovalent Ab, binds only to unoccupied, activated integrins and is insensitive to changes in integrin clustering. AP5, in contrast, binds to the N-terminal domain of the β<sub>3</sub> subunit stabilizing and/or inducing its active conformation. Certain cations (e.g. Ca<sup>2+</sup>) can compete with AP5 in binding to the receptor when in the inactive conformation. The advantage of using AP5 is that it can be used to detect the fraction of integrin that is `activated' (at high calcium) and the fraction of`activatable' integrin (at low calcium).

The crystal structure of the α_vβ₃ integrin has been solved recently by two different groups(Beglova et al., 2002;Xiong et al., 2001). The C-termini of the α_v and β₃ extracellular domains are very close together in the α_vβ₃structure (Xiong et al.,2001), mimicking the close association in the juxtamembrane region that maintains integrin in the inactive state. In this genuflected conformation, the activation epitopes are also masked. Integrin activation has been correlated with flexibility at the `genu'. Therefore an extended form,which is commonly seen in electron micrographs of integrins, including integrin simultaneously bound to ligand and activating antibody, exposes a much broader surface, including several activating epitopes. AP5, a monoclonal activating antibody used in the current study, binds to the first six amino acids in the N-terminal end of the β₃ subunit, which are localized in the PSI (plexin-semaphorin integrin) domain(Honda et al., 1995). Unfortunately the structure of this terminal region, localized in the genu motif, has not been solved. It is possible that the binding of AP5 is completely masked in the genuflected conformation (inactive) but is exposed in the extended configuration (active), as demonstrated for other activating antibodies (Beglova et al.,2002).

In the current paper we have identified the localization of two differentα _vβ₃ conformations in resting and resorbing human osteoclasts derived from several sources (bone marrow culture,osteoclastoma and a giant cell tumor of bone-derived cell line). We have utilized two antibodies to detect the `activated'α <sub>v</sub>β<sub>3</sub> conformation, AP5 and WOW-1, which recognize different epitopes localized in the β<sub>3</sub> subunit, and both have shown same pattern of distribution in both live and fixed cells. Both antibodies specifically bind to the fraction of integrin organized at the cell edge, where it colocalizes with filaments of cortical actin. By contrast,the antibodies to total α<sub>v</sub>β<sub>3</sub>, AP3 and 1A2 identify an additional pool of `basal' integrin in the podosomes and in the sealing zone, where it is associated with the actin ring. It is not clear why the `basal' and not `activated' integrin is found in the sealing zone where there is very tight membrane apposition to the bone surface, as observed in electron micrographs of resorbing OC. It is possible that the extended conformation of activated integrin, seen in the crystal structure, is too large to fit in this space. The concentration of integrin in the sealing zone is great, so even with `non-activated' integrin, the avidity is high at this site.

Even in the absence of growth factors, the integrin exists in both `basal'and `activated' forms. Growth factor treatment, which promotes cell motility,induces both redistribution and activation of theα _vβ₃ integrin. The integrin moves from the podosomes, in which the receptor is found in a `basal' conformation, to the new forming leading edges, where the receptor is in an `activated' state. These changes in the integrin are accompanied by functional changes in the OC(increased migration and resorption) and suggest that the two pools of integrin serve different functions.

Evidence that integrins and growth factor receptors share common signaling pathways (Clark and Brugge,1995), suggests that signals induced by growth factors could be responsible for integrin activation(Trusolino et al., 2000;Trusolino et al., 1998). We studied the effects of HGF and M-CSF because of their ability to influence osteoclast differentiation, survival and activity. Flow cytometric analysis confirms the ability of the two growth factors to induce integrin activation as measured by binding of the mAb AP5 in the presence of divalent cations.

Following growth factor exposure, cytoskeletal rearrangements occur,allowing lateral movement of the integrin and leading toα _vβ₃ translocation to the motile regions of the plasma membrane. In resorbing OCs, growth factor treatment causes dynamic reorganization of the `activated' integrin into the marginal area of the cell around the clear zone in fillopodia-like structures. Consequently, this integrin reorganization supports the translocation of OCs to new sites of the bone surface to be resorbed and explains the increased bone resorption found in the presence of HGF and M-CSF (Fig. 8). Migration assays confirm the pivotal role of the activated receptor during this process. HGF or M-CSF pre-incubation induces aα <sub>v</sub>β<sub>3</sub>-dependent and-specific increase in osteoclast motility toward osteopontin but has no effect on cell migration toward LN. Nakamura et al. show that M-CSF treatment causes recruitment of theα <sub>v</sub>β<sub>3</sub> integrin to adhesion contacts and induces stable association with PLC-γ, PI 3-kinase and Pyk2(Nakamura et al., 2001). In that system the authors did not address the question of whether the`activated' α_vβ₃ is mediating the recruitment of this molecular complex. In our study we found that the growth factor treatment induces activation of most α_vβ₃expressed in OCs and that PI 3-kinase is required for this effect.

PI 3-kinase has been implicated in integrin activation in leukocytes and platelets (Shimizu et al.,1995; Kovacsovics et al.,1995; Zell et al.,1996; Zhang et al.,1996). In avian OCs, α_vβ₃ is associated with the signaling PI 3-kinase, c-Src and FAK(Hruska et al., 1995). In that system, interaction of α_vβ₃ with osteopontin increases PI 3-kinase activity and association with Triton-insoluble gelsolin(Chellaiah and Hruska, 1996). In generated murine osteoclasts, PI 3-kinase translocates to the cytoskeleton upon osteoclast attachment to the bone surface(Lakkakorpi et al., 1997). We also found that the functional consequences ofα _vβ₃ activation are PI 3-kinase dependent. First, pre-incubation with wortmannin inhibits cell migration towards theα _vβ₃ ligand OPN but not towards LN. Second,inhibition of PI 3-kinase activity is accompanied by the inability of HGF/M-CSF to recruit α_vβ₃ integrin to motility related sites. Phosphatidylinositides generated by PI 3-kinase can activate GTP-GDP exchangers for the small GTPases Rho and Rac, which, in turn,control organization of the actin cytoskeleton, inducing formation of focal adhesion, podosomes, membrane ruffles and lamellipodia(Chellaiah et al., 2000;Clark et al., 1998;Ory et al., 2000;Ridley et al., 1999). Finally,PI 3-kinase mediates M-CSF- and HGF-driven intracellular signaling. Upon HGF stimulation, the HGF receptor c-met exhibits tyrosine kinase activity, including receptor autophosphorylation, and activates a series of signal transducing proteins. Autophosphorylated c-met associates with PI 3-kinase in vivo and in vitro (Graziani et al., 1991; Ponzetto et al.,1993). Husson et al. also showed the formation of a multiprotein complex between c-fms, c-Cbl, PI 3-kinase and Grb2 upon M-CSF stimulation(Husson et al., 1997). In OCs,a functional association between PI 3-kinase and c-Src has been shown during M-CSF-induced spreading (Grey et al.,2000).

In conclusion, our results indicate that the equilibrium between two integrin conformations is modified by growth factors, resulting in altered inside-out signaling and thus modulating the resorptive capability of the OCs. Although it is clear that HGF and M-CSF govern integrin activity, the precise mechanisms by which PI3-K and other signaling molecules interact with growth factors to modulate integrin conformation are still to be determined.

The authors are grateful to S. L. Teitelbaum, F. P. Ross and D. V. Novack for helpful comments. We also thank T. J. Kunicki for providing the monoclonal antibody AP5. This study was supported by grants from AIRC (Italian Foundation for Cancer Research) and from ASI (Italian Space Agency) to A.Z.Z.