Gelsolin Deficiency Blocks Podosome Assembly and Produces Increased Bone Mass and Strength

Osteoclasts are characterized by unique cell adhesion structures found in highly motile cells called podosomes, a property they share with several types of cancer cells and monocyte and/or macrophages. When osteoclasts are cultured on glass surfaces, multiple rows of podosomes are localized in the area corresponding to the clear zone in osteoclasts (Marchisio et al. 1984; Turksen et al. 1988; Zambonin-Zallone et al. 1989; Lakkakorpi and Vaananen 1991; Teti et al. 1991). Signaling molecules associated with podosomes such as pp^60c-src (Boyce et al. 1992; Schwartzberg et al. 1997), phosphatidylinositol 3′-kinase (PI3-K) (Buck and Horwitz 1987; Lakkakorpi et al. 1997; Nakamura et al. 1997), and rhoA (Zhang et al. 1995; Maejima-Ikeda et al. 1997) have been shown to be essential for cytoskeletal organization and bone resorption. Osteoclasts from src^−/− mice have an altered actin distribution with no peripheral podosome arrangement (Schwartzberg et al. 1997), which is similar to pp^60c-src antisense oligonucleotide treatment of cells (Chellaiah et al. 1998). Podosomes consist of an F-actin core surrounded by the actin-binding proteins vinculin, talin, and α-actinin (Marchisio et al. 1984). The presence of gelsolin in osteoclast podosomes is known but its function is not (Zambonin-Zallone et al. 1983, Zambonin-Zallone et al. 1989; Marchisio et al. 1984, Marchisio et al. 1987; Lakkakorpi and Vaananen 1991).

Gelsolin, an actin-binding protein, controls the length of actin filaments in vitro, and cell shape and motility in vivo by a variety of mechanisms (Cunningham et al. 1991). Gelsolin severs assembled actin filaments, caps the fast-growing plus ends, and promotes growth of actin filament by creating nucleation sites (Stossel 1989). Severing and capping are regulated by calcium ions and pH, which synergistically activate gelsolin's binding to actin. Binding of gelsolin to phosphoinositides causes the release of gelsolin from the filament end (uncapping), providing a site for rapid monomer addition (Janmey and Stossel 1987; Janmey et al. 1987).

In avian osteoclasts, we have shown previously that osteopontin (OP) binding to integrin α_vβ₃ in osteoclasts stimulates gelsolin-associated PI3-K, leading to increased levels of gelsolin-associated polyphosphoinositides such as phosphatidylinositol 4,5-bisphosphate, phosphatidylinositol 3,4-bisphosphate, phosphatidylinositol 3,4,5,-trisphosphate, uncapping of actin barbed ends, and actin filament formation. Inhibition of PI3-K blocks OP-stimulated actin filament formation, and cell motility (Chellaiah and Hruska 1996). Recently, we have demonstrated that OP stimulation of gelsolin-associated pp^60c-src is upstream of PI3-K. Antisense oligonucleotides to pp^60c-src blocked OP stimulation of gelsolin-associated PI3-K as well as the downstream effects on the cytoskeleton and bone resorption (Chellaiah et al. 1998).

Gelsolin-deficient (Gsn^−/−) mice, generated by gene targeting methods (Witke et al. 1995), express mild physiological dysfunction during hemostasis, inflammation, and possibly skin remodeling, but gelsolin was not required for oocyte development, spermatogenesis, embryonal development, or longevity. Gsn^−/− dermal fibroblasts have excessive actin stress fibers and migrate more slowly than wild-type (Gsn^+/+) fibroblasts, but have increased contractility in vitro (Witke et al. 1995). The skeleton of Gsn^−/− mice has not been formally analyzed, but tooth eruption and bone development appear normal. Bone cells from Gsn^−/− mice have not been studied.

We show here that elimination of gelsolin resulted in osteoclasts devoid of podosomes and actin rings. Mechanisms of cell attachment substituting for podosomes were expressed, but these failed to support cell motility. However, they were sufficient to enable cell polarization and bone resorption. Gsn^−/− osteoclasts failed to respond to OP with stimulation of motility and bone resorption. Endochondral bone formation was abnormal with delayed resorption of calcified cartilage and thicker metaphyseal trabeculae. Adult Gsn^−/− mice developed thicker cortical and trabecular bone, which was more resistant to fracture, compatible with a mild imbalance between bone resorption and formation.

Rhodamine phalloidin was obtained from Molecular Probes. All the other chemicals, including antigelsolin antibody, were purchased from Sigma Chemical Co.

Mouse osteoclasts were generated in vitro using the mouse bone marrow and stromal cell coculture system essentially as described (Shioi et al. 1994). After coculture for 6–9 d, ST2 cells were removed by treatment with collagenase and dispase (1 mg/ml each in PBS) at 37°C for 20 min as described (Wesolowski et al. 1995). After washing the cultures with PBS, all of the ST2 cells (99%) were removed. The remaining attached cells were rinsed with PBS, and the cultures were kept in cell dissociation buffer for an additional 5 min. Cells were removed from the plates by gentle scraping. Some of the removed cells were replated and stained with either trypan blue or for tartrate-resistant acid phosphatase (TRAP). Cells excluded trypan blue, and they were 99% TRAP-positive. These TRAP-positive cells were used for migration and bone resorption assays as described below.

The mouse osteoclasts generated in cocultures were placed in 6-well clusters containing glass coverslips. Osteoclasts attached to the plates and were treated with OP (25 μg/ml) or vehicle as described previously (Chellaiah and Hruska 1996). Actin staining was carried out as described previously (Chellaiah et al. 1998).

Osteoclasts were made as described above. After coculture for 6–9 d, ST2 cells were removed by treatment with collagenase and dispase (1 mg/ml each in PBS) at 37°C for 20 min as described (Wesolowski et al. 1995). Plates were washed two to three times with serum-free α-MEM medium and the remaining attached cells were kept in serum-free α-MEM medium for 2 h. Cells were treated with PBS or OP for 15 min. For each treatment four to six wells were used. The cells were fixed and rhodamine-phalloidin binding to F-actin was done as described (Chellaiah and Hruska 1996; Chellaiah et al. 1998).

Substrates such as recombinant chicken OP, vitronectin, or fibronectin were diluted in PBS and used for phagokinesis assays, which were performed in 6-well tissue culture dishes. Colloidal gold particles were made essentially as described (Takahashi et al. 1995). Osteoclasts isolated (see above) were seeded at low density (10³ cells/well). Once the cells were attached to the wells, the substrates were added as soluble protein (OP, 25 μg/ml; vitronectin or fibronectin, 10 μg/ml) in α-MEM containing 1% serum and 2% BSA. Some wells were treated with an antibody to OP (50 μg/ml) and OP (25 μg/ml). The cells migrated on this substrate, and phagocytized the gold particle to produce a white track free of the particle in a time-dependent manner. The migrating cells were visible as a black body (Takahashi et al. 1995). Areas were measured after 14-h incubations with the proteins. Cell motility was evaluated by measuring the areas free of gold particles. By using a grided reticule (Boyce Scientific, Inc.) in the eyepiece of a Nikon microscope, areas free of gold particles were measured using a 10× objective. Areas free of gold particles were represented as area moved in mm².

To assay cell migration, Transwell migration chambers (8-μm pore size) were used (Senger et al. 1996). Undersides of the membranes were coated with vitrogen 100 (collagen type 1; 30 mg/ml) at room temperature for 2 h according to the manufacturer's instructions. Osteoclasts isolated as described above were pelleted and resuspended in α-MEM medium containing 1% serum and 2% BSA (50,000 cells/ml). Cells were added to the upper chamber in the above mentioned medium (100 μl) and substrates like OP (25 μg/ml) or vitronectin or fibronectin (10 μg/ml) were added to the lower chamber in α-MEM medium containing 1% serum and 2% BSA (600 μl). Cell migration was allowed to proceed at 37°C in a standard tissue culture incubator for 12–14 h. Cells that migrated to the undersides were stained, visualized, and counted as described (Senger et al. 1996). Data are presented as the number of migrated cells field (mean ± SEM) and all assays were performed in quadruplicate. Statistical significance was calculated as mentioned below in data analysis.

Whale dentine slices (1.5 cm² in size and 0.75 mm in thickness) were cut from previously prepared rectangular sticks on a low-speed saw and were stored in 70% ethanol until required. Before being used in a resorption assay, the slices were sonicated for 3–5 min in distilled water and washed for 2 h in two changes of fresh distilled water. The slices were resterilized in fresh 70% ethanol and placed into wells of a 48-well tissue culture plate. After three washes with α-MEM serum-free media, the slices were placed overnight in a 37°C incubator in the same media. The osteoclast suspension (2 × 10⁴ cells) was added to each well, and after 2 h of adherence the culture media were replaced with α-MEM containing either vehicle (PBS) or 25 μg/ml mouse OP expressed in bacteria. After incubation for 48 h, cells were scraped from the slices and the slices were washed 7–10 times with water. Pits were stained with acid hematoxylin (Sigma Chemical Co.) for 6 min, washed well with water, and counted. The number of pits was determined microscopically. The effect of test substances on bone resorption was determined by scanning the pit areas and pit depth as described below.

Photomicrographs were obtained with the Zeiss LSM 410 scanning laser confocal microscope system, which employs a Zeiss 135 Axiovert inverted microscope fitted with an Omnicron Argon/Krypton laser for confocal excitation. The dentine slices mounted on No. 1 coverslips were viewed with a 40 × 0.6 NA air objective. Images were recorded in the epi-reflection mode using the 488-nm line of the argon laser in a 512 × 512 pixel format. Data collection and processing were accomplished with the Zeiss LSM 410 software package. Photomicrographs were produced from a Sony Mariograph color video printer or stored in a TIFF image format for post collection image processing. An XY digital image of each pit was first obtained. The area of the pit was determined from the free-hand traced perimeter using the LSM software Area Measurement function. The depth profile of the pit was obtained using the LSM Z Scan function to form an XZ profile. The maximum pit depth was determined from this scan using the LSM software Distance Measurement function. All depth measurements were corrected for the presence or absence of immersion oil.

Osteoclasts cultured on whale dentine slices or glass coverslips were immunostained with an antibody (E11; kindly provided by Dr. Stephen Gluck, University of Florida College of Medicine, Gainesville, FL) to 31-kD subunit of vacuolar H+ ATPase (Hemken et al. 1992) or gelsolin antibody as described (Wu et al. 1991; Chellaiah et al. 1996). In brief, cells were fixed with 3% paraformaldehyde and permeabilized with 10 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 1 mM CaCl₂ containing 0.1% Triton X-100 for 1 min. The cells were washed and incubated with E11 antibody or gelsolin antibody (1:100 dilution) for 2 h, washed, and counter stained with Cy2-conjugated goat anti–mouse IgG for 2 h. The cells were washed and mounted on a slide in a mounting solution (Vector Laboratories, Inc.). The cells were viewed on a Zeiss LSM 410 confocal laser scanning microscope and photomicrographs were obtained as described (Gupta et al. 1996). Images were recorded as described below.

After ST2 cells were removed as described above in osteoclast preparation, cells were washed three times with ice-cold PBS and lysates were made from osteoclasts derived from Gsn^−/− and Gsn^+/+ mice as described previously (Chellaiah and Hruska 1996). Protein contents were measured using the Bio-Rad protein assay reagent kit. Equal amounts of lysate proteins were immunoprecipitated with an antibody to gelsolin (Sigma Chemical Co.) or nonimmune serum. Immunoprecipitations and Western analysis were carried out as described (Chellaiah and Hruska 1996).

Equal amounts of protein lysates made from PBS or OP-treated cells were immunoprecipitated with an antibody to gelsolin. The immunecomplexes, adsorbed to protein A–Sepharose pellets, were assayed for pp^60-src or lipid kinase (PI3-K) activity as described previously (Chellaiah and Hruska 1996; Chellaiah et al. 1998).

Images were collected on the Zeiss laser scanning confocal system in the dual imaging mode. One detector was set to image in FITC fluorescence (488 nm laser line excitation, 515–540 nm FITC emission), and the second detector was set to record the backscattered epi-reflection image using the 568-nm laser line. The two images were collected simultaneously. The green FITC fluorescence image was stored in the green frame, and the 568-nm epi-reflection image was stored in the red frame. The composite color image was viewed as green fluorescence plus red epi-reflection.

The distal tibia were fixed in phosphate-buffered 10% formalin, pH 7.4, and decalcified in 14% EDTA for 10–14 d. The bones were washed sequentially with 50, 70, and 90% ethanol and embedded. 5-μm thick longitudinal sections were made and stained with toluidine blue stain or TRAP staining to identify osteoclasts. To estimate mineral opposition rate and bone formation rate, mice were injected with calcein at 2 and 5 d before sacrificing. The distal right femur was kept in 90% ethanol and embedded. Longitudinal sections were made. Static and dynamic histomorphometric measurements were made using a computer and digitizer tablet (Osteomeasure; Osteometrics Inc.) interfaced to a Leitz microscope (Leitz Wetzlar) with a drawing tablet. All measurements were done to the metaphyseal region distal to the growth plate region. To estimate bone formation rate, double labeled and single labeled areas were traced and calculated as described (Jilka et al. 1996; Weinstein et al. 1997). Terminology used is that recommended by the Histomorphometry Nomenclature Committee of the American Society of Bone and Mineral Research (Parfitt et al. 1987).

The left femora from 16 mice was designated for biomechanical testing. Femora were isolated, cleaned of soft tissues, wrapped in gauze that was soaked with PBS, wrapped again in plastic, placed in a sealed vial, and stored at –20°C until testing. Before testing, specimens were thawed to 23°C. Four-point bending tests were conducted using a materials testing machine (8500R; Instron) fitted with an appropriate load transducer (111 N force cell; Lebow 3397). Tests were conducted using a four-point fixture with a 9-mm spacing between the outer (support) points and a 4-mm spacing between the inner (loading) points. The bones were flexed in the anterior-posterior plane (Jepsen et al. 1996). The inner points were displaced at a constant rate of 0.03 mm/s. Torque rotation and force displacement data were collected using a computerized data acquisition system (Labview 5.0; National Instruments).

After testing, force values (F) from the bending tests were converted to bending moment values (M) using the relation M = Fa/2, where a was the distance between the outer and inner points (2.5 mm). Displacement data were normalized by a geometric factor to allow direct comparison between other experiments done using different testing geometries. Displacement values were divided by (3aL − 4a2)/6, where L is the distance between the outer points. From the moment versus normalized displacement curves, four parameters were computed: ultimate moment (Nmm), bending rigidity (Nmm²), displacement at failure (mm/mm²), and energy to failure (Nmm×[mm/mm²]). Failure was defined at the point of ultimate (maximum) moment, whereas bending rigidity was defined as the slope of the linear region of the curve.

All comparisons were made as “% control”, which refers to vehicle-treated cells. The other treatment groups in each experiment were normalized to each control value. Data presented are mean ± SEM of experiments done at different times normalized to intraexperimental control values. For statistical comparisons, analysis of variance (ANOVA) was used with the Bonferroni corrections (Instat for IBM, version 2.0; Graphpad software).

As shown in Fig. 1 (upper panels) and Fig. 2 (+/+ PBS), osteoclasts plated on glass coverslips arranged their actin cytoskeleton in peripheral rows of dot-like structures (podosomes). In some cells, podosomes were further organized into actin ring–like structures found in actively resorbing osteoclasts (Fig. 2, PBS +/+) (Taylor et al. 1989; Kanehisa et al. 1990; Lakkakorpi et al. 1993; for review see Horton et al. 1996). Podosomes were not observed in Gsn^−/− osteoclasts (Fig. 1, lower panels). Instead, F-actin was present in a peripheral web-like structure, which was often bipartite (Fig. 2, lower panels) or present diffusely throughout the cell as seen in the higher magnification (Fig. 1, lower panels). The peripheral row of dot-like structures was never observed in Gsn^−/− osteoclasts.

When osteoclasts from Gsn^+/+ mice were treated with OP, the entire cell subsurface was occupied with podosomes within 15 min (Fig. 2, OP +/+). Conversely, OP treatment had no effect on the cytoskeleton of the Gsn^−/− osteoclasts (Fig. 2, OP −/−). In contrast to the podosomes of osteoclasts derived from Gsn^+/+ mice, the actin cytoskeleton of Gsn^−/− osteoclasts demonstrated either an actin web (Fig. 2, lower panels) or a diffuse distribution throughout the cell (Fig. 1). Discrete dot-like podosome structures were never observed in either type of Gsn^−/− osteoclasts.

Gelsolin is an actin capping protein of podosomes (Zambonin-Zallone et al. 1983, Zambonin-Zallone et al. 1989; Marchisio et al. 1984, Marchisio et al. 1987). Immunostaining of osteoclasts derived from cells of wild-type mice with antigelsolin antibody showed that gelsolin was strongly colocalized with actin in podosomes (Fig. 3 A, Gsn^+/+) and colocalized along the inner edge of actin ring–like structures (Fig. 3 A, Gsn^+/+). Immunostaining of osteoclasts isolated from Gsn^−/− mice documented the absence of gelsolin expression in the Gsn^−/− mice (Fig. 3 A, Gsn^−/−). The few green spots observed in the Gsn^−/− mice are either nonspecific or bleed through from the red channel. In addition, Western analysis of the antigelsolin immunoprecipitates made from the lysates of Gsn^−/− osteoclasts demonstrates the absence of gelsolin (Fig. 3 B, lane 2).

F-actin levels were measured in both Gsn^+/+ and Gsn^−/− mice to detect any increase in F-actin content due to the severing activity by any other protein. Since it has been shown that Cap 32/34 and CapG are expressed in appreciable levels in embryonic cells, and that these proteins may function in capping and uncapping of actin filaments, we wanted to see whether these proteins can increase the F-actin content of the cells. Platelets isolated from Gsn^−/− mice demonstrated diminished capping and severing activities. The resting level actin filaments were similar in osteoclasts derived from Gsn^+/+ and Gsn^−/− mice (Fig. 4). OP treatment produced a threefold increase in F-actin in Gsn^+/+ osteoclasts and produced no change in Gsn^−/− osteoclasts (Fig. 4). The increases in F-actin seen in murine osteoclasts in response to OP is similar to those observed in avian osteoclasts (Chellaiah and Hruska 1996), except that the F-actin change was distributed as podosomes in the mouse and as stress fibers in the avian osteoclasts. Consistent with our previous observation, vinculin, α-actinin, profilin, or CapG had no effect on actin filament formation (Chellaiah and Hruska 1996).

We have shown previously that binding of OP to α_vβ₃ stimulated association signal generating molecules with gelsolin in avian osteoclasts (Chellaiah and Hruska 1996; Chellaiah et al. 1998). To see whether OP stimulation in mouse osteoclasts produced similar activation as seen in avian osteoclasts, in vitro kinase assays were performed on the immune complexes generated by using a gelsolin antibody. In vitro immune complex kinase assays for pp^60c-src have demonstrated that OP induced an increase in phosphorylation of pp^60c-src (Fig. 5A and Fig. B) and increased its activity towards phosphorylating an exogenous substrate, casein (Fig. 5 B). In addition, phosphorylation of two other proteins with the molecular masses of 125 kD (Fig. 5A and Fig. B) and 85 kD (Fig. 5 B) was observed. p85 protein was identified as PI3-K in avian osteoclasts (Chellaiah and Hruska 1996). Coimmunoprecipitation of PI3-K with gelsolin was further confirmed by a PI3-K assay (Fig. 5 C). OP stimulation of PI3-K activity associated with gelsolin is evident from the phosphorylation of the exogenous substrate, phosphatidylinositol 4,5-bisphosphate to phosphatidylinositol 3,4,5-trisphosphate (Fig. 5 C).

Immunolocalization of vinculin was performed in Gsn^+/+ and Gsn^−/− osteoclasts to further define their adhesion structures. Vinculin is involved in assembling focal adhesion plaques and mediating anchorage to the cytoskeleton (Steimle et al. 1999). It has also been shown to have regulatory roles in adhesion, spreading, and motility of cells in culture (Xu et al. 1998; Rodriguez Fernandez et al. 1992, Rodriguez Fernandez et al. 1993). In the Gsn^+/+ osteoclasts (Fig. 6, PBS +/+), vinculin was arrayed peripherally, forming a double band with areas of actin ring structure between the double band. These results are in agreement with prior studies in rat and avian osteoclasts (Taylor et al. 1989; Lakkakorpi et al. 1993). In areas of peripheral podosomes, vinculin was organized in a single intense band at the cell periphery and in the podosome structures. After OP treatment, changes in vinculin distribution (Fig. 6, OP +/+) lagged behind those of actin (Fig. 1), gelsolin (Fig. 3), and the α_vβ₃ integrin (data not shown), which were dramatically redistributed at 15 min. Vinculin was detected in the newly formed podosomes of the OP-treated Gsn^+/+ osteoclasts, probably translocating from an intracellular pool different from the formed adhesion structures. In Gsn^−/− osteoclasts, vinculin was also arrayed peripherally in discrete attachment sites with a unique structure not resembling podosomes or focal adhesions (Fig. 6, middle panels). In Gsn^−/− osteoclasts with a diffuse actin distribution as in Fig. 1, vinculin was also arrayed diffusely with discrete attachment sites scattered throughout the cells (Fig. 6, lower panels).

Since retraction of filopodia is delayed in the growth cones of Gsn^−/− neurons (Lu et al. 1997), and Gsn^−/− leukocytes and fibroblasts are hypomotile in vitro (Witke et al. 1995), we examined the motility of osteoclasts derived from Gsn^−/− bone marrow cells. In phagokinesis assays, the distance migrated by Gsn^+/+ osteoclasts was stimulated threefold by exogenous addition of OP (Fig. 7 A). Whereas basal phagokinesis of Gsn^−/− osteoclasts did not differ from wild-type, OP addition failed to stimulate migration (Fig. 7 A). In haptotaxis assays, the basal motility represents the ability of the cell to sense and respond to an adhesion substrate as opposed to the random movements in the phagokinesis assays. The motility of Gsn^−/− osteoclasts was severely impaired, and again, there was a lack of response to the chemotactic factor, OP (Fig. 7 B). The rate of motility is related to the amount of F-actin levels in the cells. It has been shown that migrant fibroblast cells showed increased actin levels in response to serum levels (Arora and McCulloch 1996). We have shown previously in avian osteoclasts increased F-actin levels in response to OP (Chellaiah and Hruska 1996; Chellaiah et al. 1998). In this study, we have shown that OP increased the F-actin levels and migration of osteoclasts isolated from Gsn^+/+ mice, whereas OP had no effect on osteoclasts from Gsn^−/− mice.

Bone resorption appears to proceed by the intricate coordination of several processes, including attachment, polarized secretion of acid and proteases, and active motility of osteoclasts along the bone surface (Kanehisa and Heersche 1988; Blair et al. 1989; Zambonin-Zallone et al. 1989). Even though osteoclasts isolated from Gsn^−/− mice attach and spread well on glass coverslips and bone, the polarization efficiency was measured by immunostaining the osteoclasts with an antibody to the H⁺ ATPase proton pump responsible for mineral dissolution (Hemken et al. 1992). Osteoclasts derived from Gsn^−/− cells were cultured on dentine slices, and the H⁺ ATPase, which is found in the ruffled border overlying resorption pits, was found to be concentrated at this location (Fig. 8). This indicates that signals for the osteoclasts to polarize are normal. Secondly, we analyzed the number and size of resorption pits produced in the basal conditions (PBS). The number of resorption pits was similar between Gsn^+/+ and Gsn^−/− mouse osteoclasts (Fig. 9 A). Compilation of pit area (XY) and pit depth (XZ) using confocal microscopy scans (Fig. 9 B) on multiple dentine slices with osteoclasts from multiple preparations (Table) revealed competency of the Gsn^−/− cells in basal conditions. However, response of Gsn^−/− osteoclasts to OP treatment was absent, whereas pit complexity, pit area, and pit depth were increased in Gsn^+/+ osteoclasts (Fig. 9 A and Table). Since OP is a chemotactic factor, bone resorption in vivo probably proceeds under the influence of this factor regulating the direction of osteoclast motility. Thus, the lack of response to OP may reflect a diminished capacity for bone resorption in vivo.

Under laboratory conditions, the homozygous mutant mice remained healthy for at least 18 mo and showed normal somatic development and reproductive capacity. They behaved as expected and appeared to hear normally upon startling. Skeletal radiographs revealed no deformity of the long bones and are not shown. There was a detectable increase in cortical thickness of the tibial and femoral diaphysis at age 14 wk (Table), but this was not apparent at 3 or 9 wk of age. Despite this increase in diaphyseal thickness, femoral lengths were not different, and the club-like deformities reported in other osteopetrotic mutations (Marks 1989; Hayman et al. 1996) were absent.

To assess skeletal histomorphometry, mice were injected with calcein to label areas of bone mineralization on two occasions, and histomorphometry of nondecalcified and decalcified bone sections was performed as described in Materials and Methods. Bone sections were stained for cartilage proteoglycans with toluidine blue, and for TRAP to aid in detection of osteoclasts. As shown in Fig. 10, at 14 wk of age the chondrocytes of the epiphyseal growth plates were reduced in number, mildly disorganized, and the cartilage matrix was expanded. The bone trabeculae in the metaphyses below the primary spongiosa of Gsn^−/− mice had retained cartilage proteoglycan detected by the toluidine blue stain. Furthermore, the metaphyseal trabeculae of Gsn^−/− bones (Fig. 10A and Fig. C, arrowheads) were thicker as confirmed by formal histomorphometric measurements (Table). Osteoclast numbers and osteoclast perimeters in the metaphyses tended to be higher in the bones from Gsn^−/− mice, as shown in the sections stained for TRAP (Fig. 10 B). This increase was not statistically significant (Table), but it may have been adaptive for the decreased bone resorption. The intensity of the TRAP stain did not differ between wild-type and mutant osteoclasts. The decrease in bone resorption of mutant mice was demonstrated by changes in eroded perimeters. The perimeters of eroded surfaces were significantly reduced in Gsn^−/− mice (P < 0.05), demonstrating that osteoclast-mediated bone resorption was diminished.

The cancellous bone area (P < 0.005) and trabecular thickness (P < 0.01) were significantly higher in Gsn^−/− bones as compared with those from Gsn^+/+ mice (Table). The impact of gelsolin deficiency was not limited to trabecular bone, as cortical bone assessed in the middiaphyseal femoral midshaft was increased in width (P < 0.05) (Table). The increase in bone mass was age-dependent and not detected until after 9 wk. Age-related skeletal abnormalities of bone thickness and mineralization have been noted previously in mild osteopetrosis (Hayman et al. 1996). The increase in bone mass reported here probably resulted from the defect in bone resorption and the normal rates of bone formation producing an imbalance in skeletal remodeling similar to that reported in other mild osteopetrotic states (Hayman et al. 1996). The osteoblastic parameters of bone modeling revealed no change in osteoblast number or perimeter (Table). There was an increase in mineral appositional rate in Gsn^−/− mice, but the bone formation rates were normal (Table). Increased mineralization of the skeleton has been observed frequently in the osteopetroses, suggesting a link between osteoclast dysfunction and mineralization. The increase in the mineral appositional rate of Gsn^−/− mice may have been due to increased activity of osteoblasts capable of forming bone matrix in an area covered by these cells. Therefore, conclusions regarding the cause of the increase in bone mass require studies of osteoblast function beyond the scope of this report, since development of a mouse endochondral osteoblast model is required.

The imbalance between bone resorption and formation and the increased mineralization reported here may have affected bone strength, and the ability of the skeleton to resist damage (fracture) may be altered. To test this possibility, we performed four-point bending tests of mouse bones to assess their mechanical strength, as described in Materials and Methods. As shown in Fig. 11 and Table, the Gsn^−/− bones required greater energy and greater maximal displacement to produce femoral failure compared with the Gsn^+/+ femurs. Changes in bone stiffness resulting from over mineralization and decreased remodeling seen in some osteopetroses were not observed in the bones from the Gsn^−/− mice.

This report is the first demonstration that podosome assembly is critically gelsolin-dependent. The role of gelsolin in actin filament assembly and stimulus–response coupling associated with cell motility is well known (Stossel et al. 1985; Janmey and Stossel 1987; Janmey et al. 1987; Stossel 1989; Yin 1989; Weeds and Maciver 1993). However, most cells function well without gelsolin, indicating that gelsolin's functions are effectively substituted for in most cells by other actin severing or capping proteins. In these cells, attachment to matrix utilizes focal adhesions, and motility is often accomplished by lamellipodia. On the other hand, we demonstrate here that cells such as the osteoclast, which rely on podosomes for attachment and motility, are rendered hypomotile by the absence of gelsolin, and furthermore, the actin ring structures of the osteoclast that derive from podosome fusion are not formed. As a result, gelsolin deficiency is associated with abnormal actin cytoskeletal architecture, reduced rates of osteoclast motility, which contribute to reduced bone resorption in vivo, and podosome-associated signal transduction is blocked.

Gelsolin is the prototype of a large family of proteins that sever actin filaments, nucleate actin filament assembly, and block the fast exchanging end of actin filaments (Stossel 1989). Therefore, the opportunity for redundancy of function in actin dynamics is high (Yin 1987, Yin 1989; Vandekerckhove and Vancompernolle 1992; Weeds and Maciver 1993; Schafer and Cooper 1995) and the redundancy of actin-regulating proteins is manifested by the Gsn^−/− mouse. Since gelsolin is not expressed during development, cell motility in support of development should be normal, and transgenic mice lacking gelsolin expression exhibit normal motility of embryonic fibroblasts (Witke et al. 1995). However, dermal fibroblasts from adult Gsn^−/− mice have increased actin stress fiber size, decreased actin severing, and hypomotility. In addition, resting platelets isolated from Gsn^−/− mice have an increased F-actin content, and they are slow to remodel their actin cytoskeleton when activated (Barkalow and Hartwig 1995). According to the model of regulated treadmilling for regulation of actin dynamics through the filament-nucleating activity of the Arp 2/3 complex (Machesky and Insall 1999), the findings in gelsolin deficiency are compatible with capping protein serving to complement for gelsolin deficiency with the only outcome being reduction in the rates of motility (Andre et al. 1989; Witke et al. 1995). Thus, our findings that gelsolin is required for podosome assembly is surprising. They establish that the podosome is a unique cell attachment structure, and that in its case, gelsolin capping of actin filaments cannot be substituted. The data also raise the question as to whether podosome assembly utilizes regulated treadmilling and the Arp 2/3 complex for actin filament elongation or whether gelsolin uncapping barbed ends is sufficient for this purpose.

Gsn^−/− osteoclasts were adherent through novel cell–matrix interaction sites, e.g., vinculin-containing focal contacts distinct from focal adhesions, and they were able to polarize their actin cytoskeleton and plasma membranes sufficiently to enable bone resorption. These redundant mechanisms of actin regulation were sufficient for basal rates of bone resorption measured on dentine slices in vitro. However, Gsn^−/− osteoclasts had a major decrease in motility in vitro, demonstrating that the redundant cell–matrix attachment mechanisms were unable to support the rates of assembly and disassembly observed in podosomes that enable high motility rates. This may be the critical component of the osteoclast phenotype in vivo, since the histomorphometric findings demonstrated that bone resorption measured by eroded perimeters was diminished. The findings of hypomotility in vitro and reduced bone resorption in vivo are compatible with the physiologic character of bone resorbing osteoclasts that display high motility rates. Therefore, our data suggest that motility is an important component of bone resorption.

Whereas hypomotility of leukocytes and fibroblasts was a central focus of the reported phenotype of Gsn^−/− mice (Witke et al. 1995), these cells are motile through the process of cellular extensions, filopodia and lamellipodia, establishing new contacts, and pulling the trailing edge of the cell (Lauffenburger and Horwitz 1996). Gelsolin has been shown to be an obligate downstream effector of rac-stimulated fibroblast motility (Azuma et al. 1998), where its severing of cortical actin is required for membrane ruffling, and the loss of gelsolin severing may have produced the decrease in motility. In the studies reported here related to the absence of podosome assembly, the basis for the defect appears related to gelsolin-capping function that must be required in order for the podosome structure to be completed. Several studies have shown a relationship between gelsolin content and cell migration (Chaponnier et al. 1990; Cunningham et al. 1991; Arora and McCulloch 1996; Kwiatkowski 1999). Increasing gelsolin content in cultured fibroblasts by transfection proportionally enhances the rate of cell migration toward a chemoattractant (Cunningham et al. 1991; Arora and McCulloch 1996). Large increases in gelsolin content are also associated with the enhanced motility exhibited during differentiation of myeloid cell lines into macrophage-like cells (Kwiatkowski 1988). However, it has been shown that high gelsolin content is associated with reduced cell migration and increased adhesion in smooth muscle and mammary carcinoma cells (Chaponnier and Gabbiani 1989; Chaponnier et al. 1990). Thus, the capping or severing functions of gelsolin may be variably important depending upon the cell type, but only in the case of podosome assembly is gelsolin critical.

Osteoclasts are unique in that their mechanism of cell motility is podosome based. They do not express focal adhesions, and they use the speed of podosome assembly and disassembly to generate high rates of motility. Gelsolin is present in podosomes of osteoclasts but not in focal adhesions (Marchisio et al. 1987). The greater reliance on podosomes for cell attachment in osteoclasts called immediate attention to podosome absence in Gsn^−/− osteoclasts. Even the actin ring structure of the bone resorbing osteoclast is a coalescence of dense rows of podosomes, and they were absent in Gsn^−/− osteoclasts. Osteoclasts are uniquely motile in that they continue to resorb bone beneath one area of the cell undersurface while another area is reorganizing and moving (Taylor et al. 1989). The result is the production of complex resorption pits with a serpentine character (Taylor et al. 1989). The rapidity of podosome assembly and disassembly enables this feature of coordinating cell function with motility (Kanehisa and Heersche 1988; Kanehisa et al. 1990), another property shared between metastatic cancer cells and osteoclasts. Thus, as reported herein, impaired motility of Gsn^−/− osteoclasts is associated with diminished rates of bone resorption.

Gelsolin plays a key role in integrin-based signaling through podosomes as it is organized in a signaling complex containing PI3-K, pp^60c-src, and other signaling molecules (Chellaiah and Hruska 1996; Chellaiah et al. 1998). As shown here, this complex is expressed in murine wild-type osteoclasts, and the role of this complex is demonstrated by the absence of response to OP in the Gsn^−/− osteoclasts. The receptor for OP signaling is the α_vβ₃ integrin (Chellaiah and Hruska 1996; Chellaiah et al. 1998), also localized to the osteoclast podosome (Zambonin-Zallone et al. 1989). In the Gsn^−/− osteoclasts, α_vβ₃ was expressed in the unique focal contacts, and OP was secreted into the resorption space (Chellaiah, M., unpublished observations). Thus, the major deficiency in the response to OP was the organization of signaling around gelsolin in the podosome for motility. The Gsn^−/− osteoclasts organized their cytoskeleton and polarized sufficiently to resorb bone. Thus, signal transduction through α_vβ₃ and the focal contacts that substitute for podosomes may explain the basal rates of bone resorption in vitro, and osteoclast function sufficient to support development of the skeleton and normal rates of endochondral bone growth in vivo.

Histomorphometric analysis of bones from Gsn^−/− mice documented reduced resorption of calcified cartilage in the epiphyses of the long bones. The epiphyses were mildly disorganized with increased cartilage matrix. The metaphyseal bone trabeculae were thicker than those observed in Gsn^+/+ bones, and they had retained cartilage proteoglycans. Cancellous bone area was increased, and the eroded surface area was decreased. Both trabecular and cortical bone were thicker, and the cause of this appeared to be decreased osteoclast function in the presence of normal osteoblast function. Osteoblast number and surfaces were normal, as were bone formation rates, which tended to be elevated although not significantly. The deficiency in osteoclast function was not sufficient to affect endochondral bone growth, and the shortening and clubbing of the femurs reported in some osteopetrotic states was not observed (Marks 1989; Hayman et al. 1996). The effect of thicker trabeculae and cortices on bone strength was an increase in load required to produce failure, and an increase in the ultimate normalized displacement to produce fracture.

Skeletal diseases characterized by sclerosis resulting from reduced osteoclast-mediated bone resorption are referred to as the osteopetroses (Marks 1989). Osteopetrosis is a heterogeneous group of disorders that range in severity from lethal to the very mild disorder reported here. The lethal disorders result from an absence of marrow cavity formation and the limited development of hematopoiesis. The milder disorders are associated with shortening and deformity of the long bones and axial skeleton, but tooth eruption, hearing, and skull plate development (Hayman et al. 1996). The osteopetrosis reported here is even milder in that endochondral bone lengths were normal. The milder forms of osteopetrosis have been associated with significant effects on skeletal mineralization demonstrating progressive increases in bone density as observed here by the increase in mineral apposition rate (Hayman et al. 1996).

An effect of osteoclast dysfunction on mineralization was suggested in the Acp5 TRAP-deficient homozygous mutant mice. TRAP may dephosphorylate OP in the osteoclast resorption space (Ek-Rylander et al. 1994), and as a result increase mineralization by affecting the role of OP in the bone formation that follows cessation of osteoclast function.

Thus, the findings reported here of thickened calcified cartilage trabeculae, delayed cartilage resorption, and diminished eroded perimeters demonstrate a mild osteopetrosis that was not developmentally important nor sufficient to affect growth, but it was sufficient to increase trabecular and cortical bone thickness with age resulting in increased skeletal strength. The findings reported here resemble the long-term outcome of aminobisphosphonate administration to humans, producing a mild imbalance between bone resorption and bone formation that results in an increase in bone mass (Liberman 1995; Saag et al. 1998). As a result, these studies suggest that the gelsolin-based signaling complex of the osteoclast podosome is an attractive target for pharmacologic regulation of bone resorption.

The authors wish to thank Dr. Stephen Gluck for the E11 antibody to the 32-kD subunit of the vacuolar H+ ATP osteoclast proton pump; Drs. John Freeman and Brian Bennett for assistance with the confocal microscopy; Mrs. Crystal Idleburg for assistance with the bone sectioning and staining; and Kathy Jones and Helen Odle for secretarial assistance.

This work was supported by grant AR41677 (K.A. Hruska and M. Chellaiah) from the National Institutes of Health and the Monsanto/Washington University consortium.