Paxillin Contracts the Osteoclast Cytoskeleton

Abstract

Osteoclastic bone resorption depends upon the cell’s ability to organize its cytoskeleton via the αvβ3 integrin and osteoclastogenic cytokines. Since paxillin associates with αvβ3, we asked if it participates in skeletal degradation. Unlike deletion of other αvβ3-associated cytoskeleton-regulating molecules, which impairs the cell’s ability to spread, paxillin-deficient (Pax^−/−) osteoclasts, generated from embryonic stem cells, “superspread” in response to RANK ligand (RANKL) and form large, albeit dynamically atypical, actin bands. Despite their increased size, Pax^−/− osteoclasts resorb bone poorly, excavating pits approximately 1/3 normal depth. Ligand-occupied αvβ3 or RANKL promotes paxillin serine and tyrosine phosphorylation, the latter via c-Src. The abnormal Pax^−/− phenotype is rescued by WT paxillin but not that lacking its LD4 domain. In keeping with the appearance of mutant osteoclasts, WT paxillin, overexpressed in WT cells, contracts the cytoskeleton. Most importantly, the abnormal phenotype of Pax^−/− osteoclasts likely represents failed RANKL-mediated delivery of myosin IIA to the actin cytoskeleton via the paxillin LD4 domain but is independent of tyrosine phosphorylation. Thus, in response to RANKL, paxillin associates with myosin IIA to contract the osteoclast cytoskeleton thereby promoting its bone-degrading capacity.

Introduction

The skeleton-degrading capacity of the osteoclast depends upon organization of its distinctive cytoskeleton to polarize machinery to the cell-bone interface. This process involves generation of two unique resorptive organelles, the ruffled border and actin ring. The ruffled border is a complex structure present only in bone-resorbing polykaryons. It is formed by insertion of lysosome-derived vesicles into the bone-apposed plasma membrane under the aegis of synaptotagmin VII and autophagy proteins ^(1,2). The ruffled border contains an electrogenic H+ATPase and charge-coupled Cl⁻ channel, which, in concert, transport HCl into the resorptive space. The acidic milieu demineralizes bone, exposing its organic matrix, which, in turn, is degraded by cathepsin K, again delivered by vesicle insertion into the ruffled border.

Because bone resorption involves an unusually low, local pH and abundant cathepsin K at this site, the process requires intimacy of the cell and bone surface to form a microenvironment isolated from the general extracellular space. This isolation is accomplished by the actin ring or sealing zone, formed in proximity to bone, which encompasses the ruffled border.

While peripheral “belts” of actin may form in glass- or plastic-adherent cells, the actin ring appears exclusively in mineralized matrix-residing osteoclasts and is the product of podosome migration ^(3–5). Podosomes are punctuate, matrix adhesive structures, which, while analogous to focal adhesions, are more labile. They contain a central actin core and peripheral “cloud” of cytoskeletal-regulating proteins ^(4,6). In non-adherent cells of osteoclast lineage, podosomes are arranged in clusters distributed diffusely throughout the cytoplasm. Upon adherence to mineralized substrate, they organize into actin rings to initiate the resorptive process.

The fact that osteoclast-bone contact is required for the polykaryon to polarize its resorptive machinery indicates the event is mediated by extracellular, matrix-derived intracellular signals ⁽⁷⁾. In fact, αvβ3, the principal integrin expressed by osteoclasts, is central to this polarization process. In the context of macrophage-lineage cells, abundant αvβ3 is unique to the osteoclast and is expressed only when precursors are exposed to receptor activator of nuclear factor κB ligand (RANKL) ⁽⁸⁾. αvβ3 is therefore a marker of commitment of macrophages to the osteoclast phenotype. The integrin assumes its high affinity conformation under the aegis of macrophage colony-stimulating factor (M-CSF) and when in contact with bone, activates a canonical signaling pathway required for osteoclast cytoskeleton polarization and efficient resorption ^(9,10). This cytoskeleton-organizing signaling sequence involves c-Src and Syk phosphorylation and recruitment of the co-stimulatory ITAM protein, Dap12 ⁽¹¹⁾. The relatively osteoclast-specific guanine nucleotide exchange factor, Vav3, is phosphorylated by Syk leading to transit of cytoskeleton-organizing Rac to its active, GTP-associated state.

An important confirmation of the biological relevance of this pathway is the fact that deletion of any member yields a relatively similar osteoclast phenotype in that the cells fail to spread or effectively resorb bone and have a “crenated” appearance. Also indicative of cytoskeleton dysfunction, absence of many of these signaling molecules obviates normal actin ring and ruffled border formation ^(11,12).

In addition to activating signaling molecules, such as tyrosine kinases and guanine nucleotide exchange factors, αvβ3 in osteoclasts, associates with adaptor proteins such as talin (unpublished observation) and kindlin3 ⁽¹³⁾. Deletion of these adaptor proteins, like other components of the canonical signaling pathway, yields a non-spread phenotype.

Paxillin is a key cytoskeleton-organizing adaptor protein ⁽¹⁴⁾. Its association with αvβ3, in osteoclast podosomes, suggests its absence would yield the non-spread appearance common to deletion of any component of the integrin-induced canonical signaling pathway ⁽¹⁵⁾. This hypothesis is supported by the fact that paxillin regulates Rac, a Rho GTP-ase, whose activation, by M-CSF in the osteoclast, requires αvβ3 and whose absence eventuates in resorptive inefficiency and severe osteopetrosis ^(14,16).

Exploration of the role of paxillin in osteoclast function is compromised by the fact that mice deleted of the protein die in early embryogenesis. To circumvent this difficulty we generated Pax^−/− and Pax^+/− osteoclasts from murine embryonic stem cells (ESCs). Paxillin haploinsufficient osteoclasts are indistinguishable from those derived from normal bone marrow macrophages (BMMs). Homozygous-deleted cells, however, exhibit an unexpected phenotype as in contrast to the “crenated” appearance of cells lacking other αvβ3-associated proteins, they “superspread” in response to RANKL and have large, atypical actin bands. Despite their size, Pax^−/− osteoclasts fail to effectively resorb bone. Most importantly, the unexpected phenotype of these Pax^−/− polykaryons likely reflects absence of phosphotyrosine-independent RANKL-induced myosin IIA association with the paxillin LD4 domain.

Materials and Methods

Reagents and Plasmids

Recombinant murine M-CSF and murine glytathione-S-transferase-RANKL were expressed as described ⁽¹⁷⁾. Anti-paxillin antibody was obtained from BD Transduction Laboratory (San Diego, CA), anti-phospho-Y31 paxillin from ECM Bioscience (Versailles, KY) and anti-phospho-serine 126 paxillin from Biosource (Carlsbad, CA). Anti-β3 integrin subunit, anti-phosphorylated ERK, anti-phosphorylated AKT, anti-ERK, anti-AKT, anti-phosphorylated IkBα, anti-IkBα and anti-myosin IIA antibodies were obtained from Cell Signaling Technology (Danvers, MA). Anti-NFATc1 and anti-cFms antibodies were from Santa Cruz (Santa Cruz, CA); Anti-β-actin antibody, Anti-α-tubulin antibody, peroxidase-labeled wheat germ agglutinin and TRITC conjugated wheat germ agglutinin were purchased from Sigma (St. Louis, MO). Anti-cathepsin K antibody was from Millipore (Billerica, MA). SU6656 was purchased from Calbiochem (San Diego, CA).

Full-length human paxillin construct was kindly provided from Dr. Kenneth Jacobson (University of North Carolina) and was subcloned into BamH1 and Xho1 site of pMX retroviral vector or EcoR1 and BamH1 site of lentiviral vector. We used standard molecular biological methods to construct cDNAs coding ΔLD4 paxillin. Lentiviral constructs, lentiviral packaging plasmids(pHR’8.2deltaR) and lentiviral envelop plasmid (pCMV-VSV-G) were obtained from Dr. Yunfeng Feng (Department of Internal medicine, Washington University School of Medicine, St. Louis, Missouri, USA).

Osteoclast Differentiation of ESCs

Pax+/− ES Cell clone #43 and Pax^−/− ES Cell clone #17 ⁽¹⁸⁾ were used for osteoclast differentiation. As previously described ^(19,20), undifferentiated ESCs were maintained on feeder cells on a gelatin coated dish in DMEM (GIBCO) containing 15% FBS, 1.5×10⁻⁴ MTG, 1.5% leukemia inhibitor factor (LIF), 100U/ml streptomycin and penicillin. For differentiation into osteoclasts, ESCs were trypsinized and added to T25 flasks (4×10⁵ cells/flask) in IMEM (GIBCO) containing 15% FBS, 1.5×10⁻⁴ MTG, 1.5% leukemia inhibitor factor (LIF), 100U/ml streptomycin and penicillin, for two days to delete feeder cells. The purified ESCs were trypsinized and plated for embryoid body (EB) formation by 5 days of liquid differentiation. Single cells, derived from trypsinized EBs, were suspended in methylcellulose replating medium containing α-MEM with 10% FBS and 100ng/ml M-CSF for a three-dimensional culture for 5 days. Macrophage-colonies were collected by digesting the methylcellulose using cellulose, and the cells were cultured in α-MEM with 10% FBS and 100ng/ml M-CSF. For osteoclast differentiation, the ESC-derived macrophages were cultured in α-MEM with 10% FBS plus 30ng/ml M-CSF and increasing amounts of RANKL.

Bone marrow macrophage isolation and osteoclast culture

Primary BMMs were prepared as described ⁽¹⁷⁾. Briefly, marrow was extracted from femora and tibia of 6–8 wk-old mice with α-MEM and cultured in α-MEM containing 10% FBS, 100 IU/ml penicillin and streptomycin with 100ng/ml MCSF in bacterial dishes. For osteoclast culture, a total of 5×10³ cells were cultured in 200 µl α-MEM containing 10% FBS and various amounts of RANKL and 30ng/ml M-CSF in 96-well tissue culture plates for 5 days. Cells were fixed and stained for TRAP activity after 5d in culture using a commercial kit (387-A; Sigma-Aldich).

Actin ring staining and bone resorption assay

Osteoclasts were generated on bone slices from bone marrow- or ESC- derived macrophages by exposure to 100ng/ml RANKL and 30ng/ml MCSF. For actin ring staining, the cells were fixed in 4% paraformaldehyde and permeablized in 0.1% Triton X-100, rinsed in PBS, and immunostained with AlexaFluor 488-phalloidin (Invitrogen). For bone resorption assay, osteoclasts were removed and resorption pits were visualized by incubation of the specimen with 20 µg/ml peroxidase-conjugated wheat germ agglutinin as described previously and stained with 3,3’ -diaminobenzidine (Sigma).

Western blotting and immunoprecipitation

Cultured cells were washed twice with ice-cold PBS and lysed on RIPA buffer containing 20mM Tris, pH7.5, 150mM NaCl, 1mM EDTA, 1mM EGTA, 1% Triton X-100, 2.5mM sodium pyrophosphate, 1mM β-glycerophosphate, 1mM Na₃VO₄, 1mM NaF, and 1×protease inhibitor mixture (Roche). After incubation on ice for 10 min, the cell lysates were clarified by centrifugation at 15,000 rpm for 10 min. 40µg of total lysate was subjected to 10% SDS-PAGE and transferred onto PVDF membrane. The filters were blocked in 0.1% casein (USB)/PBS and 0.1% Tween 20 for 1h and incubated with primary antibodies at 4°C overnight followed by probing with secondary antibodies coupled with fluorescent (Invitrogen). The proteins were visualized using Odyssey (LI-COR, Lincoln, NE).

Virus production and macrophage infection

Lentiviral vectors were combined with the mixture of packaging plasmid (pHR’8.2 delta R) and the envelope (pCMV-VSV-G) plasmid, and were transfected into 293T cells by calciumphospho-precipitation. Virus-containing medium were collected 48 hr after transfection, ESC-derived macrophages were incubated with the virus-containing medium for 24 h in the presence of 100 ng/ml M-CSF and 10 µg/ml protamine (Sigma-Aldrich). Cells were then selected at 2µg/ml puromycin for 3 days. WT or mutant paxillin in retrovirus vectors were transiently transfected into Plat-E packaging cells by calciumphospho-precipitation. Virus was collected 48 hr after transfection. BMMs were infected with virus for 24 hr in the presence of 100 ng/ml M-CSF and 4 µg/ml polybrene (Sigma). Cells were selected in the presence of M-CSF and 1 µg/ml blasticidin (Calbiochem) for 3 days prior to use as osteoclast precursors.

Dynamic imaging of osteoclasts on bone

Pax^+/− or Pax^−/− macrophages derived from ESCs, transduced with GFP-Actin were maintained using standard culture conditions (37°C and 5% CO₂, 95% air atmosphere) with RANKL and M-CSF, for 8 days, in bone powder coated Bioptechs (non-liquid perfused) Delta T culture system, consisting of a heated, indium-tin-oxide-coated glass dish attached to a calibrated Bioptechs micro-perfusion peristaltic pump. All cultures were observed with the 20× objective (NA, 0.4) of an inverted automated wide-field epifluorescence DIC microscope (Leica DMIRE2, Leica Microsystems, Wetzlar, Germany). An objective lens heater was used to improve temperature homogeneity. Images (608×512 pixels spatial and 12-bit intensity resolution) were recorded with a cooled Retiga 1300 camera (Qimaging, Burnaby, BC, Canada) every 2 minutes in 2×2 binned acquisition mode, using 100–300 msecond exposures. Dynamic images were composed using ImageJ.

Statistics

Statistical significance was determined using Student’s T-test. Data are represented as mean ±SD.

Results

Osteoclastogenic cytokines and integrin activation induce paxillin phosphorylation

Paxillin is a constitutively expressed (not shown) podosomal protein, which co-localizes with αvβ3 in osteoclasts. Because integrin occupancy, in these cells, activates other cytoskeleton-regulating proteins, we asked if the same obtains regarding paxillin. Thus, cytokine-starved, pre-fusion osteoclasts were lifted and maintained in suspension or plated on the αvβ3 ligand, vitronectin, for 30 min. As determined by immunoblot with phospospecific antibody, integrin occupancy phosphorylates the critical paxillin residues, Y31 and S126, in the bone resorptive cell (Fig. 1A) ⁽¹⁴⁾.

Fig. 1. Osteoclastogenic cytokines and integrin activation induce paxillin phosphorylation.

(A–C) WT bone marrow macrophages (BMMs) were cultured with RANKL and M-CSF for 3 days in tissue culture dishes to generate pre-fusion osteoclasts. A) Cells were maintained in suspension (S) or plated on vitronectin (A) for 30 minutes. B) M-CSF or C) RANKL were added with time. D) WT osteoclasts, pretreated either with SU6656 (2µM) or DMSO for 20 minutes, were exposed to RANKL (100 ng/ml) or PBS for 30 minutes. E) WT or c-Src^−/− osteoclasts were treated with RANKL or PBS for 30 minutes. (A–E) Paxillin immunoprecipitates were immunoblotted for phosphorylated paxillin^Y31 (p-Pax^Y31) and total paxillin (Pax). Total cell lysates (TCL) were immunoblotted for phosphorylated paxillin^S126 (p-Pax^S126) and total paxillin. F) WT osteoclasts were treated with RANKL with time. Paxillin immunoprecipitates were immunoblotted for phosphorylated c-Src^Y416 (p-Src) and paxillin. G) Osteoclasts derived from Pax^+/− and Pax^−/− ESC-derived macrophage-like cells were treated with RANKL. β3 integrin in paxillin immunoprecipitates was determined by immunoblot. H) WT and Syk^−/− pre-fusion osteoclasts were maintained in suspension (S) or plated on vitronectin (A) for 30 minutes. (upper panel) Paxillin immunoprecipitates were immunoblotted for p-Pax^Y31 and paxillin. (lower panel) p-Pax^S126, paxillin and Syk content, in lysates, was determined by immunoblot.

In addition to its osteoclastogenic properties, RANKL and M-CSF promote activity of differentiated osteoclasts ^(21,22). To determine if they impact paxillin, cytokine-starved pre-fusion osteoclasts were exposed to either cytokine with time. Again, immunoblot with phosphospecific antibody demonstrates that RANKL or M-CSF induces phosphorylation of the functional paxillin tyrosine and serine residues (Fig. 1B, C).

Integrin- and cytokine-activated c-Src is key to organize the osteoclast cytoskeleton and we therefore asked if the tyrosine kinase mediates RANKL-stimulated paxillin phosphorylation in the resorptive cell. As shown in Fig. 1D, and E, the Src family kinase inhibitor, SU6656 or genetic deletion of c-Src, arrests RANKL-induced Pax^Y31 phosphorylation while not impacting that of Pax^S126. Furthermore, paxillin and c-Src associate upon RANKL exposure (Fig. 1F). The αvβ3 integrin is central to organization of the osteoclast cytoskeleton, and its association with paxillin, in the bone resorbing cell, is stimulated by RANKL (Fig. 1G). Moreover, deletion of Syk, which associates with αvβ3 and is activated by c-Src, also obviates integrin-induced Pax^Y31 phosphorylation ⁽¹¹⁾ (Fig. 1H). These data suggest the Pax^−/− osteoclasts might have cytoskeleton defects reflecting arrest of the canonical signaling pathway emanating from αvβ3.

Pax^−/− osteoclast spreading is hypersensitive to RANKL

Given paxillin is an integrin-associated protein, a reasonable hypothesis holds that it regulates the osteoclast cytoskeleton and thus its absence should yield a robust phenotype. Exploration of this issue is challenged, however, by the early embryonic lethality of Pax^−/− mice. To circumvent this difficulty, embryonic bodies differentiated from Pax^−/− and Pax^+/− ESCs were cultured with M-CSF until macrophage-like cells appeared which were then exposed to M-CSF and RANKL. As seen in Fig. 2A, Pax+/− ESCs yield characteristic osteoclasts in a RANKL dose-dependent manner.

Fig. 2. Pax^−/− osteoclast spreading is hypersensitive to RANKL.

A) Pax^+/− and Pax^−/− ESC -derived macrophage-like cells were cultured with M-CSF and increasing amounts of RANKL for 4 days after which they were stained for TRAP activity. Arrows indicate spread osteoclasts. B) Wells containing TRAP-stained cells detailed in A. C) Number of spread osteoclasts/field in A. D) Pax^+/− and Pax^−/− ESC-derived macrophage-like cells were cultured with RANKL and M-CSF with time. Osteoclast differentiation markers were determined by immunoblot. E) and F) ESC-derived Pax^+/− and Pax^−/− macrophage-like cells, cultured with M-CSF for 3 days, were serum and cytokine starved overnight. They were then exposed to either E) RANKL (100 ng/ml) or F) M-CSF (100 ng/ml) with time. Signaling molecules were identified by immunoblot. (*p<0.05, **p<0.01; Scale bar: 200 µm.)

Osteoclasts lacking signaling molecules, which organize their cytoskeleton, typically exhibit a common phenotype characterized by failure to spread and a “crenated” façade ^(11,13). Given its critical role in cytoskeletal organization, in other cells, we were surprised, therefore, that Pax^−/− osteoclasts effectively spread and, in fact, formed more and larger tartrate resistant acid phosphatase (TRAP)-expressing polykaryons, than their heterozygous counterparts, particularly at low concentrations of RANKL (Fig. 2A,B,C). A possible explanation for the larger and more abundant osteoclasts in the absence of paxillin is accelerated differentiation. To test this hypothesis, Pax^−/− and Pax^+/− macrophages were cultured in low or high dose RANKL for up to 4 days and markers of osteoclast maturation, including the β3 integrin subunit, NFATc1, and c-Src assessed daily, by immunoblot. There were no differences in expression of these proteins by Pax^−/− or Pax^+/− cells (Fig. 2D). Moreover, activation of osteoclastogenic signaling molecules by RANKL or M-CSF was unaltered absent paxillin (Fig. 2E,F). Therefore, the hypersensitivity of Pax^−/− osteoclasts to RANKL as regards a “superspread” façade is not a manifestation of accelerated differentiation.

Pax^−/− osteoclasts fail to effectively resorb bone

To determine the role of paxillin in the bone-resorptive capacity of osteoclasts, Pax^−/− and Pax^+/− ESC-derived macrophages were maintained on cortical bone in the presence of M-CSF and RANKL. After 5 days, the cells were removed, and resorption pits stained with peroxidase-conjugated wheat germ agglutinin. Whereas Pax^+/− osteoclasts generate well-demarcated lacunae, the surface area of those formed by Pax^−/− polykaryons is greater but the pits have indistinct margins (Fig. 3A) and are only 1/3 as deep as control, as measured by confocal microscopy (Fig. 3B). To determine the genesis of the distinct features of lacunae formed by mutant and control osteoclasts, we prepared plastic embedded sections of cells residing on bone (Fig. 3C). Mirroring the planar appearance of wheat germ agglutinin-stained pits, high magnification of thin sagittal sections demonstrates Pax^−/− osteoclasts are superspread and generate shallow excavations with indistinct borders. Thus, paxillin mediates osteoclast function but not formation.

Fig. 3. Pax^−/− osteoclasts fail to effectively resorb bone.

Pax^+/− and Pax^−/− ESC-derived macrophage-like cells were cultured with RANKL and M-CSF on bone slices for 5 days. A and B) Osteoclasts were removed and the bone slices stained with peroxidase-labeled (A) or fluorescent-labeled (B) wheat germ agglutinin to visualize resorption lacunae. B) Depth of resorption lacunae as measured by confocal microscopy. C) High resolution sagittal images of plastic-embedded ESC-derived Pax^+/− and Pax^−/− osteoclasts cultured on bone. D) Pax^+/− and Pax^−/− ESC-derived macrophage-like cells were cultured on bone for 5 days in the presence of M-CSF and increasing amounts of RANKL. Fibrillar actin was visualized by immunofluorescence following FITc-phalloidin staining. E) Histomorphometric analysis of ring area/cell size. F) Pax^+/− and Pax^−/− ESC-derived macrophage-like cells, cultured with RANKL and M-CSF on glass for 5 days, were stained with FITC-phalloidin and α-Tubulin. (* p<0.05, **p<0.01, ***p<0.001 Scale bars: 50 µm (A, B, C) and 100 µm (D and F).

As resorptive efficiency often reflects cytoskeletal organization, we next examined podosome belts and actin rings. To this end, Pax^−/− and Pax^+/− osteoclasts, generated on bone slices with M-CSF and RANKL, were stained with FITC-phalloidin to visualize fibrillar actin (Fig. 3D, E). Similar to wild type (WT) cells, numerous small actin rings form within individual Pax^+/− osteoclasts ⁽²³⁾. With increasing RANKL, the actin ring circumference increases, thereby augmenting the cytoplasmic area encompassed by these structures. At all concentrations of RANKL, however, Pax^−/− osteoclasts contain a large, single circular actin-rich structure, which encompasses virtually the entire cell. In addition, the rate of actin ring remodeling on bone is retarded in paxillin-deficient osteoclasts (Movie). In contrast to abnormal actin rings generated on bone, actin belts and tubulin localization are not affected in paxillin-deficient osteoclasts cultured on glass (Fig. 3F), thus, paxillin-mediated actin distribution requires signals emanating from the osteoclast’s natural substrate, bone.

To confirm that the dysfunction of Pax^−/− osteoclasts reflects absence of the cytoskeletal protein, ESC-derived macrophages were lentivirally transduced with WT paxillin or empty vector and cultured on bone in the presence of RANKL and M-CSF (Fig. 4A). Paxillin, but not empty vector, completely normalizes the actin rings and resorptive pits of Pax^−/− osteoclasts (Fig. 4B,C,D).

Fig. 4. Rescue of cytoskeleton and resorptive function of ESC-derived Pax^−/− osteoclasts.

ESC-derived Pax^−/− macrophage-like cells were lentivirally transduced with WT paxillin or empty vector and cultured on bone for 5 days in the presence of M-CSF and increasing amounts of RANKL. A) Paxillin expression confirmed by immunoblot. B) Visualization of fibrillar actin by immunofluorescence following FITc-phalloidin staining. C) Osteoclasts were removed and the bone slices stained with peroxidase-labeled wheat germ agglutinin to visualize resorption lacunae. D) Depth of resorption pits illustrated in C. (***p<0.001; Scale bars: 100 µm (B); 50 µm (C)).

Excess paxillin inhibits osteoclast spreading

The above data establish that osteoclasts lacking paxillin “superspread” raising the possibility that an abundance of the cytoskeletal protein may constrict the cell. To determine if such is the case, we retrovirally expressed paxillin in WT BMMs and exposed them to M-CSF and increasing amounts of RANKL. In fact, WT cells containing excess paxillin are less spread than their vector-transduced counterparts resulting in reduced TRAP staining in wells bearing the over-expressing osteoclasts (Fig. 5A,B,C). The paucity of TRAP staining does not reflect arrested differentiation as appearance of osteoclastogenic markers is unaltered by excess paxillin (Fig. 5D). Confirming that its over-expression alters the osteoclast cytoskeleton in a manner reciprocal to that of its absence, actin ring area of paxillin-transduced WT cells mirrors reduced cell size (Fig. 5E,F).

Fig. 5. Excess paxillin constricts the osteoclast cytoskeleton.

A) WT BMMs, retrovirally transduced with empty vector or WT paxillin, were cultured with M-CSF and increasing amounts of RANKL for 5 days after which the cells were stained for TRAP activity. B) Wells containing TRAP-stained cells detailed in A. C) Number of spread osteoclasts/field illustrated in A. D) WT BMMs retrovirally transduced with empty vector or WT paxillin were cultured with RANKL and M-CSF for 1–5 days or M-CSF, alone, for 5 days (Mϕ). Osteoclast differentiation markers were determined by immunoblot. E) WT BMMs, retrovirally transduced with empty vector or WT paxillin, were cultured with M-CSF and RANKL for 6 days after which actin rings were visualized by FITc-phalloidin staining. F) Histomorphometric analysis of actin ring area/cell size. (*p<0.05, ***P<0.001; Scale bars: 200 µm (A) and 100 µm (E)).

Bone resorption requires paxillin LD4 domain but not tyrosine phosphorylation

Paxillin’s LD domains are its most extensively characterized motifs. LD4, in particular, associates with a variety of proteins, which regulate integrin-mediated actin organization ⁽²⁴⁾. To explore this issue in the context of osteoclasts, Pax^−/− macrophages, transduced with vector or WT paxillin or that lacking LD4 (ΔLD4), were cultured, on bone for 5 days, in the presence of M-CSF and RANKL. Confirming the essential nature of the domain, WT paxillin rescues the capacity of Pax^−/− osteoclasts to form normal actin rings and effectively resorb bone, whereas ΔLD4 does not (Fig. 6). Similarly, the fact that RANKL, M-CSF and integrin occupancy promotes paxillin tyrosine phosphorylation (Fig. 1) suggests those residues, which regulate many of the adaptor protein’s biological properties, also mediate its capacity to organize the osteoclast cytoskeleton. To determine if such is the case, we generated osteoclasts from Pax^−/− macrophages transduced with Pax^2Y-F wherein tyrosines 31 and 118 are mutated to phenylalanine. In contrast to the failure of ΔLD4 to rescue the Pax^−/− osteoclast phenotype, Pax2^Y-F induces actin ring and resorption lacunae formation indistinguishable from the WT construct (Fig. 6).

Fig. 6. Optimal bone resorption requires paxillin LD4 domain but is independent of tyrosine phosphorylation.

Pax^−/− macrophages, lentivirally transduced with WT paxillin, Pax^ΔLD4, Pax^2Y-F, or empty vector. The cells were differentiated into osteoclasts on bone by 5 days exposure to M-CSF and RANKL. A) Visualization of fibrillar actin by FITC-phalloidin staining. B) Quantification of ring area. C) Visualization of resorption lacunae stained with peroxidase-conjugated wheat germ agglutinin. D) Depth of resorption pits illustrated in C. (*** p<0.001 vs vector;** p<0.01 vs vector; Scale bars: 100 µm (A) and 50 µm (C))

Paxillin/myosin IIA association regulates the osteoclast cytoskeleton

Osteoclasts lacking αvβ3 or its effector molecules, such as c-Src or Syk, fail to spread and have a phenotype distinctly different than those lacking paxillin. Interestingly, however, non-muscle myosin IIA associates with osteoclast adhesion structures and its siRNA knockdown yields a “superspread” appearance similar to those lacking paxillin suggesting a mechanistic association between the two molecules ⁽²⁵⁾. We therefore exposed cytokine-starved, bone-residing osteoclasts to RANKL, with time, and determined myosin IIA content in paxillin immunoprecipitates. In fact, RANKL progressively induces paxillin/myosin IIA association, which is confirmed by immunostaining (Fig. 7A, B). To further determine if myosin IIA/paxillin recognition regulates the osteoclast cytoskeleton we asked if the essential PaxLD4 domain, mediates the association. To this end, we transduced Pax^−/− macrophages with HA-tagged WT paxillin or that lacking its LD4 domain (Pax^ΔLD4). As seen in Fig. 7C, absence of paxillin LD4 precludes its immunoprecipitation with myosin IIA. Thus, organization of the osteoclast cytoskeleton, by paxillin, is independent of Pax^Y31 and Pax^Y118 phosphorylation but requires myosin IIA.

Fig. 7. Myosin IIA mediates paxillin organization of the osteoclast cytoskeleton.

A) Cytokine-starved bone-residing WT osteoclasts were exposed to RANKL, with time. Paxillin immunoprecipitates were immunoblotted for myosin IIA and paxillin. B) WT macrophages were cultured with RANKL and M-CSF, on glass, for 5 days. Paxillin and myosin IIA localization in osteoclasts was determined by microscopy. Yellow staining in merged image represents co-localization of the two immunoreactive proteins. C) Cytokine-starved Pax^−/− pre-fusion osteoclasts, transduced with HA tagged WT paxillin or Pax^ΔLD4, were exposed to RANKL with time. HA (paxillin) immunoprecipitates were immunoblotted for myosin IIA and HA. (Scale bar: 50 µm).

Discussion

Paxillin is a 559 amino acid, ubiquitous scaffolding protein, that binds actin and recruits cytoskeletal and signaling molecules to an integrin-associated complex ⁽²⁶⁾. It is constitutively expressed by osteoclasts, present in their actin rings and localizes to podosomes where it directly recognizes the cytoplasmic domain of the β3 integrin subunit ⁽¹⁵⁾. Paxillin is essential for cytoskeletal organization and motility of virtually all cell types ⁽²⁷⁾.

Because of the early embryonic lethality of the Pax^−/− mouse ⁽²⁸⁾, the precise role of paxillin in osteoclasts is unresolved. To address this issue we developed techniques for generating large numbers of authentic osteoclasts from ESCs lacking one or both paxillin alleles. ESC-derived heterozygous osteoclasts are indistinguishable from WT polykaryons obtained by culturing BMMs in M-CSF and RANKL. Their numbers increase parri passu with RANKL abundance and expression of osteoclast differentiation markers appear as expected.

Absence of the αvβ3 does not impact osteoclast differentiation and the same holds regarding other integrins expressed by the resorptive cell ^(12,13). These matrix-binding heterodimers are central to organizing the osteoclast cytoskeleton and their absence, alone and in combination, typically yields non-spread TRAP-expressing polykaryons. This phenotype is mirrored by absence of a variety of podosomal signaling and adaptor molecules in proximity to the integrin. As paxillin is an important integrin-associated protein, and in other cells its absence or inhibition retards spreading ^(18,29,30), we predicted that its deletion in osteoclasts would yield the same “crenated” phenotype. We were surprised, therefore, that Pax^−/− osteoclasts actually superspread and form larger actin bands on bone, at lower concentrations of RANKL, than do their WT counterparts. The fact that WT cells, in which paxillin is overexpressed, are small, further substantiates that the protein restricts osteoclast size.

The osteoclast cyclically spreads and constricts during the resorptive cycle, with the area of bone-degrading cells smaller than those which are inactive ⁽³⁾. Our data indicate paxillin participates in physiological constriction of the osteoclast cytoskeleton, an essential feature of skeletal degradation.

The bone resorptive capacity of large osteoclasts, such as appear in Paget’s disease ⁽³¹⁾ and SHIP-1 deficiency ⁽³²⁾, is characteristically robust but this is not the case regarding those lacking paxillin. While the surface area of individual resorptive lacunae generated by Pax^−/− osteoclasts is large, the pits are morphologically abnormal as they lack well-defined margins. Most importantly, the depth of Pax^−/− osteoclast-excavated pits is substantially reduced. This observation establishes expression of osteoclast resorptive capacity as a function of lacuna area may in some circumstances, be misleading and the precise appearance of these pits provides important information.

Osteoclast cytoskeletal organization is substrate-dependent and therefore differs in cells residing on physiological (i.e. mineralized) and non-mineralized surfaces ⁽³⁾. These differences are reflected by podosome distribution, among the most sensitive morphological features of osteoclast cytoskeletal function. Whereas osteoclasts cultured on glass distribute their podosomes in a large peripheral belt, those residing on bone form numerous smaller actin rings, which separate the osteoclast resorptive microenvironment from the general extracellular space. In fact, individual osteoclasts simultaneously form multiple resorptive sites and thus each contains numerous actin rings which increase in abundance with progressive RANKL concentrations ^(5,23). The development of techniques to kinetically visualize osteoclasts on bone established that these physiologically-relevant sealing zones remodel asynchronously within a given cell ⁽³³⁾. In contrast, osteoclasts residing on non-mineralized substrates ultimately organize podosomes into a single peripheral actin belt distinctly different in composition than actin rings ⁽³⁾. While Pax^−/− osteoclasts, on bone, organize fibrillar actin, the appearance of these structures mirrors actin belts more than sealing zones. Attesting to dysfunction, the remodeling rate of the “belts” of Pax^−/− cells is substantially slower than control, consistent with paxillin’s regulation of podosomal disassembly ⁽¹⁴⁾. Thus, the inability of Pax^−/− osteoclasts to generate well-defined, deep resorptive lacunae may reflect abnormal podosome organization.

In addition to their osteoclastogenic properties, M-CSF and RANKL stimulate bone resorption by mature osteoclasts ^(21,22). Hence, organization of the actin cytoskeleton reflects both integrin activation and signals emanating from cytokine receptors. In fact, ligand occupancy of αvβ3 and c-Fms stimulate similar cytoskeleton- regulating molecules, including c-Src, and M-CSF activates the integrin in the bone resorptive cell.

RANKL promotes paxillin association with αvβ3 and induces phosphorylation of key serine and tyrosine residues, the latter via c-Src whose adaptor and kinase functions regulate the osteoclast cytoskeleton ^(6,34). c-Src constitutively associates with αvβ3 ⁽¹¹⁾. Upon integrin occupancy, it tyrosine phosphorylates Syk, thereby stimulating a cytoskeleton-organizing complex of which Rac is a distal effector. Paxillin also participates in activating Rac in other cells, a process requiring c-Src-mediated paxillin^Y31 phosphorylation ⁽¹⁴⁾. The association of both paxillin and RANK, with c-Src, supports the concept of a paxillin-containing signaling complex emanating from the cytokine receptor.

Given the unusual appearance of Pax^−/− osteoclasts, relative to those lacking αvβ3-activated signaling molecules, suggested the means by which the adaptor protein organizes the cell’s cytoskeleton may not involve the integrin. This conclusion was buttressed by the fact that RANKL-stimulated Rac activation which, in the osteoclast, requires the integrin ⁽³⁵⁾, is not impaired in Pax^−/− cells.

Paxillin is a multidomain, focal adhesion adaptor involved in integrin and growth factor signaling ⁽³⁶⁾. Its LD4 motif binds a protein which regulates cell spreading and motility ⁽²⁴⁾. The appearance of Pax^−/− osteoclasts suggested that in this cell, this paxillin-associated, cytoskeletal organizing protein may be myosin IIA. This hypothesis was buttressed by the fact that, similar to paxillin, active myosin IIA causes cells to contract and reduces their size ⁽³⁷⁾ as does paxillin overexpression in WT osteoclasts. Most importantly, like absence of paxillin, myosin IIA deficiency promotes spreading of osteoclasts and increases their size due to accelerated fusion, without impacting differentiation ⁽²⁵⁾. Despite the increased dimensions of myosin IIA-deficient cells, they undergo cytoplasmic fragmentation and like Pax^−/− osteoclasts, are dysfunctional ⁽³⁷⁾. In keeping with a key role for myosin IIA in the osteoclast cytoskeleton-organizing capacity of paxillin, the two proteins co-localize in podosome belts and their association is induced by RANKL. Supporting the relevance of paxillin’s recognition of myosin IIA under the influence of RANKL, it does not occur in the absence of the paxillin LD4 domain, which is essential for osteoclast cytoskeletal organization. Thus, paxillin is a component of osteoclast adhesive structures and its activation likely involves RANK-induced signals, which promote association with myosin IIA to constrict the actin cytoskeleton to its bone-resorptive conformation.

Tyrosine phosphorylation of paxillin is central to much of its biological activity such as focal adhesion disassembly in glass residing cells ⁽³⁸⁾. Furthermore, tyrosine phosphatase inhibition in transformed cells, cultured on glass, expands their belts of invadapodia, structures analogous to podosomes ⁽²⁹⁾. On the other hand, mutation of its two most biologically relevant tyrosines, namely Y31 and Y118 did not impair the cytoskeleton-organizing capacity of paxillin, in osteoclasts, or the cell’s bone degrading facility. While initially unexpected, these observations are consistent with the fact that dominant/negative suppression of Pax^Y31 and Pax^Y118 phosphorylation does not alter NBT-II cell adhesion or spreading ⁽³⁹⁾. Most importantly, Pax^2Y-F, while inhibiting the contractile response of smooth muscle to acetylcholine, does not dampen myosin light chain phosphorylation ⁽⁴⁰⁾. These data indicate that delivery of myosin IIA to the cytoskeleton, in a phosphotyrosine-independent manner, is the dominant mechanism whereby paxillin regulates osteoclastic bone resorption.

Supplementary Material

Supp Movie S1. Online Supplemental Material.

Movie. Movie of paxillin^−/− and paxillin^+/− osteoclast cytoskeleton organization. Pax^−/− (right panel) and Pax^+/− (left panel) GFP-actin-transduced macrophages, cultured in a indium- tin-oxide-coated glass dish containing a thin layer of pulverized bovine bone, were differentiated into osteoclasts by 9 days exposure to RANKL and M-CSF. The plates were attached to a calibrated Bioptechs micro-perfusion peristaltic pump plated. Cultures were observed with the 20× objective using 100−300 msecond exposures. Actin ring remodeling is substantially more rapid in Pax^+/− than Pax^−/− osteoclasts. Scale bar: 100 µm.