Low kindlin-3 levels in osteoclasts of kindlin-3 hypomorphic mice result in osteopetrosis due to leaky sealing zones

Integrins are a family of ubiquitously expressed adhesion receptors that anchor cells within their ambient extracellular matrix. This task requires the binding of talin and kindlin proteins to the integrin cytoplasmic domain, which then triggers the formation of an active integrin conformation capable of stably binding to the extracellular ligand. In addition, both the talin and kindlin protein families initiate the formation of adhesion complexes that are strongly connected with the actin cytoskeleton and serve as signaling centers of the cell (Calderwood et al., 2013; Moser et al., 2009b).

Integrins are also key adhesion receptors for hematopoietic cells. Many of these cells owe their characteristic properties to the rapid activation of their integrins. For instance, platelets and neutrophils circulate in the blood system in a non-adhesive state but swiftly shift their integrins towards an active state to form platelet aggregates during hemostasis or to enable leukocyte adhesion and extravasation to combat infection. We have previously shown that these processes are very sensitive to a genetic reduction in kindlin-3 (also known as Fermt3) expression (Klapproth et al., 2015). While platelets and neutrophils rely on fast integrin activation, which is also achieved by stoichiometric levels of talin-1 and kindlin-3, other hematopoietic cells, such as macrophages and osteoclasts, remain in tissues for long periods and consequently might be less dependent on rapid integrin activation.

In this study, we therefore focused on how a strong reduction of kindlin-3 expression impacts osteoclast function. Osteoclasts are large multinucleated cells that form by fusion of mononucleated hematopoietic precursors of the myeloid lineage. Their main task is to degrade bone matrix. Therefore, osteoclasts polarize and form large circular adhesion structures, so called sealing zones (Takito et al., 2018). These integrin-mediated adhesion structures form tight contacts with the bone surface and encase the ruffled border, a specialized membrane compartment, which secretes protons and proteinases into the subjacent resorption lacunae to degrade both the mineral and organic bone components. Sealing zones consist of strongly interconnected punctate-like adhesion structures, also known as podosomes. Each podosome comprises a dense actin core surrounded by a ring of integrins and integrin-associated adaptor proteins that connect the adhesion complex with the actin cytoskeleton and thereby provide strong matrix adhesion (Collin et al., 2006; Linder and Aepfelbacher, 2003; Pfaff and Jurdic, 2001; van den Dries et al., 2019).

We have previously shown that, in the absence of kindlin-3, macrophages and osteoclasts can only form immature podosomes consisting of a smaller actin core than normal podosomes and lacking a discrete ring of integrins. The absence of integrin-mediated signaling leads to defects in cytoskeletal organization and cell spreading. Consequently, osteoclasts from kindlin-3-null mice fail to form sealing zones and exhibit strongly reduced bone resorption, which results in a severe osteopetrosis (Schmidt et al., 2011). Osteoblasts do not contribute to the increased bone mass of kindlin-3-knockout mice (Schmidt et al., 2011), because, like osteocytes and chondrocytes, they differentiate from mesenchymal progenitors that only express kindlin-2 (also known as Fermt2) and not kindlin-3. Consistent with this, recent studies in kindlin-2 mutant mice have reported defects in skeletal development, bone homeostasis and bone formation (Cao et al., 2020; Fu et al., 2020; Wu et al., 2015). Deletion of kindlin-2 in osteocytes and osteoblasts results in a striking osteopenia and mechanical defects in weight-bearing long bones, since impaired focal adhesion formation and defective cytoskeletal organization results in defective mechanotransduction (Cao et al., 2020; Fu et al., 2020; Qin et al., 2021).

In the current study, we investigated osteoclast function at very low kindlin-3 levels that still allow the formation of podosomes (Klapproth et al., 2019). We show that kindlin-3 hypomorphic osteoclasts are able to adhere to the bone matrix through formation of sealing zones in vitro and in vivo. However, these structures are leaky, leading to impaired bone resorption as they are unable to maintain high concentrations of proteinases, such as cathepsin K, in the resorption lacunae. Consequently, kindlin-3 hypomorphic mice develop mild osteopetrosis.

Kindlin-3-deficient mice die shortly after birth, as they suffer from severe bleeding and leukocyte adhesion deficiency (Moser et al., 2009a, 2008). In contrast, kindlin-3 hypomorphic mice, which carry a neomycin resistance cassette within an intron of the Fermt3 gene and therefore shows strongly reduced gene expression, are viable with a normal life expectancy, and are protected from spontaneous bleeding and infections (Klapproth et al., 2015). In order to study osteoclasts from kindlin-3 hypomorphic mice, we first determined their kindlin-3 expression levels. For this purpose, we differentiated bone marrow cells into osteoclast-like cells by adding macrophage colony-stimulating factor (M-CSF; also known as Csf1) and RANKL (also known as Tnfsf11) to the culture medium and determined their kindlin-3 levels by western blot analyses and densitometric measurements. In contrast, the kindlin-3-null cells used in this study were differentiated from fetal liver because it was not possible to isolate sufficient bone marrow cells from kindlin-3-knockout mice due to their early lethality and severe osteopetrosis. Similar to our previous results in platelets and neutrophils (Klapproth et al., 2015), the presence of a neomycin resistance cassette within one Fermt3 allele (+/n) defined a hypomorphic allele, and reduced kindlin-3 protein expression to ∼50%. The presence of two hypomorphic Fermt3 alleles (n/n) or one hypomorphic and one Fermt3-null allele (n/−) lowered kindlin-3 expression in osteoclast-like cells to 10% and 5%, compared to wild-type cells, respectively (Fig. 1A,B) (Klapproth et al., 2015). We then determined the relative protein levels of kindlin-3, talin-1 and β-integrin subunits in wild-type macrophages and osteoclast-like cells compared to platelets, neutrophils and fibroblasts by mass spectrometry (Fig. S1). This relative quantification showed that macrophages and osteoclasts have a slightly higher ratio of talin-1 to kindlin-3 than platelets, neutrophils and fibroblasts (note that fibroblasts express kindlin-2; Fig. S1). Furthermore, the measurements confirmed that fibroblasts predominantly express β1 integrins, neutrophils β2 integrins and platelets β3 integrins, while osteoclast-like cells form all three integrin classes (Fig. S1). Taken together, these data suggest that kindlin-3 hypomorphic osteoclasts from n/− mice express more than 50 times less kindlin-3 than talin-1.

It is important to note that although kindlin-3-deficient osteoclast-like cells showed reduced adhesion and failed to spread into pancake-shaped polykaryons, their fusion and differentiation into multinucleated osteoclasts is hardly affected by the absence of kindlin-3 (Schmidt et al., 2011). To test the impact of low kindlin-3 levels on osteoclast adhesion and spreading in vitro, we generated osteoclast-like cells from the bone marrow of wild-type and kindlin-3 hypomorphic (n/n and n/−) mice using either low (20 ng/ml) or high (100 ng/ml) concentrations of M-CSF to trigger their differentiation and fusion into large polykaryons. Wild-type osteoclast-like cells started to reorganize their cytoskeleton and relocate their nuclei to the cell periphery within 4 days in culture. In contrast, kindlin-3 hypomorphic cells exhibited strongly delayed spreading but managed to form large, tartrate resistant acid phosphatase (TRAP)-positive, multinucleated and spread osteoclast-like cells 1 to 2 days later, depending on whether 10 or 5% of kindlin-3 protein was expressed (Fig. 1C). Notably, the delay in spreading could be only partially rescued by supplementing the medium with high concentrations of M-CSF. Of note, this concentration was sufficient to rescue spreading of integrin β3-deficient osteoclasts (Fig. 1D; Faccio et al., 2003). As signals originating from integrin adhesion sites control actin polymerization and cytoskeletal organization, we tested whether reduced adhesion signaling in kindlin-3 hypomorphic osteoclasts is responsible for the delay in cell spreading. Therefore, we generated pre-osteoclasts expressing different levels of kindlin-3 (+/+, n/n and n/−) by culturing bone marrow cells for only 2.5 days in M-CSF- and RANKL-containing medium, and plated them for 20 min on osteopontin, which is the major ligand of the principal osteoclast integrin αVβ3. With decreasing amounts of kindlin-3, we found a stepwise reduction in phosphorylation of Pyk2 and paxillin, two central signaling and adaptor proteins within the adhesion complex (Fig. 1E). In parallel, we tested pre-osteoclast adhesion on osteopontin, which also showed reduced cell adhesion at low kindlin-3 expression levels. Of note, the adhesion was significantly higher compared to kindlin-3-deficient pre-osteoclasts (Fig. 1F). Thus, we conclude that reduced integrin-mediated adhesion and signaling are responsible for the spreading defect of kindlin-3 hypomorphic osteoclasts.

To investigate whether other integrin classes are also functionally impaired at low kindlin-3 levels, we performed experiments with macrophages, which are related to and capable of differentiating into osteoclasts (Udagawa et al., 1990). Adhesion of hypomorphic (n/n and n/−) and kindlin-3-null macrophages to the β1 and αv integrin ligand fibronectin (Fig. S2A), the β2 integrin ligand ICAM-1 (Fig. S2B) and the α4β1 integrin ligand VCAM-1 (Fig. S2C) was reduced in a kindlin-3-dose-dependent manner. Adhesion to fibronectin and VCAM-1 was fully rescued in the presence of the exogenous integrin-activating agent Mn²⁺, suggesting that a reduced number of active-ligand bound integrins account for the adhesion defect. In line with this, we detected reduced levels of active β1 integrins on the surface of kindlin-3 hypomorphic macrophages by flow cytometry using the activation-epitope specific antibody 9EG7 (Fig. S2D; Lenter et al., 1993). As a consequence of reduced integrin activity, hypomorphic macrophages showed diminished spreading on glass, cell-culture plastic and fibronectin-coated plastic (Fig. S2E–G). Importantly, the described defects in integrin function were not caused by altered integrin surface expression as shown by flow cytometry (Fig. S2H). Moreover, we tested the phagocytic activity of macrophages as a function of kindin-3 expression in vivo. In this model, recruitment of macrophages into the peritoneal cavity was induced by lipopolysaccharide (LPS) injection and phagocytic uptake of fluorescent beads was measured 6 h later by flow cytometry. While more macrophages, identified by F4/80, CD115 and CD11b-positive FACS staining, were collected from hypomorphic mice, uptake of fluorescent beads was significantly reduced (Fig. S2I,J).

Taken together, low kindlin-3 expression in osteoclasts and macrophages results in reduced numbers of active integrins and impairs integrin-mediated adhesion and signaling.

Macrophages and pre-osteoclasts arrange their podosomes into clusters, whereas mature, multi-nucleated osteoclasts arrange them into podosome belts at the cell periphery or more centrally located rings, also called sealing zones, depending on whether they are cultured on glass or on mineralized matrix, respectively (Georgess et al., 2014; Jurdic et al., 2006). Several studies have shown that kindlin-3 colocalizes with integrins and other plaque proteins to the podosomal ring structure in podosomes of macrophages and pre-osteoclasts (Klapproth et al., 2019; Schmidt et al., 2011). Accordingly, we found kindlin-3 in podosomal belts and sealing zones of mature osteoclast-like cells (Fig. S3A,B). Importantly, while the formation of mature podosomes requires kindlin-3, expression of only 5% of kindlin-3 is already sufficient to recruit and assemble integrins and adaptor proteins in rings around the podosome actin core (Klapproth et al., 2019; Schmidt et al., 2011). Nevertheless, a strong reduction in kindlin-3 expression often resulted in the fragmentation of the podosome clusters into several smaller clusters and the actin cores were usually embedded in a more intense actin cloud (Fig. 2A). Moreover, the distribution of plaque proteins such as vinculin was more diffuse within the podosome clusters of hypomorphic cells (Fig. 2A–C). These abnormalities are also indicated by fluorescence intensity profiles through actin cores of podosome clusters of wild-type and hypomorphic cells. Here, control cells show a clear alternation of actin and vinculin, which becomes progressively lost at reduced kindlin-3 expression levels (Fig. 2B,C).

Next, we investigated whether these changes in podosome organization and cluster formation interferes with their rearrangement into podosome belts. Remarkably, kindlin-3 hypomorphic osteoclast-like cells were able to organize their podosomes in well-defined belts, that were, however, significantly thinner compared to those of wild-type cells (Fig. 2D,E). At higher magnification, distinct actin cores and an almost fused actin core ring are visible at the inner rim of wild-type podosome belts (Fig. 2F). The individual as well as the tightly associated actin cores are clearly devoid of the plaque proteins vinculin and paxillin, which colocalize with a less intense actin cloud. In strong contrast, individual actin cores can hardly be seen in podosome belts of kindlin-3 hypomorphic osteoclast-like cells and are embedded in an intense actin cloud, which colocalizes with a diffuse vinculin and paxillin signal (Fig. 2F). To gain further insights into the actin organization of the adhesion belts, we performed high-resolution scanning electron microscopy on ventral membrane preparations of wild-type and kindlin-3 hypomorphic osteoclast-like cells. In line with the immunofluorescence imaging results, actin cores of n/− cells were smaller than those of wild-type cells but were embedded in a denser mesh of actin filaments compared to controls, indicating that low kindlin-3 levels are not sufficient to organize the actin cytoskeleton within the adhesion belt (Fig. 2G; Fig. S3C,D).

We then tested the ability of kindlin-3 hypomorphic osteoclasts to form sealing zones. To this end osteoclast-like cells were cultured on a calcified, bone-like matrix. Despite their low kindlin-3 expression, kindlin-3 hypomorphic osteoclast-like cells also assembled sealing zones that seemed indistinguishable from those of control cells (Fig. 3A). In order to measure the resorptive capacity of kindlin-3 hypomorphic osteoclast-like cells in vitro, we cultured them on a calcified matrix and determined the resorbed area. Whereas osteoclast-like cells differentiated from kindlin-3 deficient fetal liver cells hardly resorb any matrix due to their impaired adhesion and spreading (Schmidt et al., 2011), kindlin-3 hypomorphic osteoclast-like cells also exhibited a progressive reduction in their resorptive activity, which is concordant with the decrease in kindlin-3 levels (Fig. 3B,C).

Sealing zones tightly attach and seal the ventral plasma membrane of the osteoclast to the underlying bone and encase a special membrane compartment, called the ruffled border. Within this compartment, locally secreted protons and proteinases degrade the mineral and organic components of the bone matrix. One of these proteinases is cathepsin K, which is secreted in large amounts at the ruffled border and accumulates in the resorption lacunae to cleave key bone matrix proteins such as collagen I (Saftig et al., 2000). We analyzed accumulation of cathepsin K at the ventral cell membrane by confocal microscopy, where it was enriched in sealing zone encircled areas of control osteoclast-like cells. In strong contrast, cathepsin K was not found to be accumulated at the ventral cell membrane of kindlin-3 hypomorphic osteoclast-like cells (Fig. 3D). Given that cathepsin K activity in cell lysates of wild-type and hypomorphic osteoclast-like cells was similar (Fig. 3E), we assume that the adhesion structures of hypomorphic cells are not tight enough to prevent leakage of the secreted proteinases from the resorption area.

To investigate whether the reduced resorptive activity of kindlin-3 hypomorphic osteoclasts causes an osteopetrotic phenotype in mice, we analyzed tibiae of control and kindlin-3 hypomorphic mice by means of histology and peripheral quantitative computer tomography (pQCT). Hematoxylin and Eosin (H&E) staining of paraffin sections and Masson–Goldner staining of sections of PMMA-embedded bones, which stains the bone matrix, suggested increased trabecular bone in 4.5-month-old kindlin-3 hypomorphic mice (Fig. 4A,B). An increase in trabecular bone mineral density (BMD), as measured by pQCT of bones from 6-month-old mice confirmed these findings (Fig. 4C). Heterozygous mice, which express ∼50% of wild-type kindlin-3 levels, appeared normal. Notably, kindlin-3 hypomorphic mice remained osteopetrotic throughout life, as the difference in trabecular BMD between control and hypomorphic mice was still present at the age of 2 years (Fig. 4D).

Next, we determined the number and location of osteoclasts on bone sections. TRAP staining and histomorphometric analysis of bones of 6-month-old mice revealed more than twofold and threefold elevated osteoclast numbers per bone surface in n/n and n/− mice, respectively (Fig. 4E,F). The eroded surface, which represents the bone surface area covered by osteoclasts, was also increased (Fig. 4G). Interestingly, both the absolute number of TRAP-positive osteoclasts normalized to the bone surface area, as well as the fraction of bone surface covered by osteoclasts was significantly increased in kindlin-3 hypomorphic mice.

To further corroborate the hypothesis that the resorption defect results from leaky sealing zones, we performed transmission electron microscopy (TEM) on ultrasections of the tibia of 7-day-old wild-type and n/− mice. These studies revealed that kindlin-3 hypomorphic osteoclasts, like osteoclasts from control mice, attach to the bone surface, are polarized showing extensively folded membrane structures of the ruffled border and are able to generate resorption lacunae (Fig. 4H). Sealing zones from wild-type osteoclasts, however, were clearly separated from the ruffled border and the cytoplasm (Fig. 4H, indicated by white arrowheads) and appeared as homogeneous patches that were devoid of cell organelles and lie smoothly on the bone surface (Fig. 4H; Fig. S4A, indicated by black arrows). In stark contrast, in kindlin-3 hypomorphic osteoclasts, no clear separation between the ruffled border and the sealing zone was apparent, in particular the folded ventral plasma membrane of the ruffled border extended into the sealing zone thereby preventing close contact with the bone surface (Fig. 4H; Fig. S4A, indicated by white arrows). Furthermore, the sealing zones of kindlin-3-hypomorphic osteoclasts often contained vesicles and were not clearly separated from the rest of the cytoplasm. In line with the reduced resorptive activity, we found intact collagen fibers in the resorption lacunae and between membrane folds of the ruffled border of kindlin-3 hypomorphic osteoclasts pointing towards insufficient resorptive activity (Fig. 4I), which might also be caused by structural defects of the ruffled border. While the ruffled border in control cells comprises an array of lamellar pillars, it appeared as a sponge-like structure in hypomorphic osteoclasts (Fig. 4H; Fig. S4B). In cross sections, the ruffled border of hypomorphic osteoclasts appeared as a laterally fused network rather than a cluster of individual lamella (Fig. S4C). Strikingly, adjacent to the ruffled border, hypomorphic cells usually contained many large cytoplasmic vesicles (Fig. 4H; Fig. S4B). Thus, these ultrastructural analyses indicate that reduced bone resorption in kindlin-3 hypomorphic mice is primarily due to structural defects in the sealing zone and the ruffled border of their osteoclasts.

In the present study, we wanted to investigate the effect of low kindlin-3 expression on hematopoietic cells, whose integrin-mediated functions are not immediately dependent on rapid integrin activation and ligand binding. Therefore, we chose also osteoclasts as a model for the following reasons. Firstly, their fusion and differentiation take place in the bone marrow and they do not require integrin-dependent homing processes to reach their destination. Secondly, their fusion into polykaryons and differentiation are independent of kindlin-3-mediated integrin activity (Schmidt et al., 2011). Finally, in contrast to platelets and neutrophils, osteoclasts form defined highly complex integrin-mediated adhesion structures.

Our studies confirmed a number of findings already obtained on kindlin-3 hypomorphic platelets and neutrophils. A stepwise reduction of kindlin-3 expression in osteoclasts, as well as in macrophages, results in a progressive reduction in cell adhesion and cell spreading, which is due to a reduced number of activated integrins and reduced integrin-mediated signaling. In this context, it is necessary to mention that reduced adhesive properties are most likely the reason for the increased number of macrophages flushed from the peritoneal cavity. Indeed, kindlin-3 hypomorphic macrophages exhibit clearly reduced phagocytic activity. It is remarkable that even 5% of kindlin-3 and thus a shift in the ratio of kindlin-3 to talin-1 of at least 1 to 50 is sufficient to allow attachment of osteoclasts to the bone surface and induce cell spreading and polarization, which is not possible in the absence of kindlin-3. In other words, even minimal levels of kindlin-3 are sufficient to trigger integrin-dependent cell functions.

A key difference between osteoclasts that do not produce kindlin-3 and those that produce only 5% kindlin-3 is that the latter cells are able to adhere to the bone matrix and form resorption structures. In stark contrast, kindlin-3-knockout osteoclasts float within the bone marrow cavity and, despite their greatly increased numbers, only a few are detected at the bone surface. Consequently, the eroded surface, which is the bone surface area that is covered by osteoclasts, is strongly decreased in kindlin-3 deficient mice (Schmidt et al., 2011), while it is increased in kindlin-3 hypomorphic mice, because weak or even kindlin-3-independent adhesion may be possible at high ligand density or when only weak shear forces act on the cells (Moretti et al., 2018, 2013). By contrast, histological studies of bone sections revealed that kindlin-3 hypomorphic osteoclasts bind efficiently to the bone surface, although they showed a strong adhesion defect in in vitro adhesion assays. This may be mainly due to the fact that these cells are only exposed to very low shear forces within the bone marrow cavity, in contrast to circulating neutrophils or platelets, which have to adhere to the vascular endothelium.

Kindlin-3-knockout animals already show massive osteopetrosis at birth. It is therefore all the more surprising that animals that have only 5% of the usual level of kindlin-3 only develop a mild osteopetrosis that does not worsen over time. The rather mild bone phenotype is probably owing to three reasons. First, in addition to their ability to adhere, kindlin-3 hypomorphic mice compensate for the impaired osteoclast function partly by increasing their numbers compared to control animals, which is certainly due to their reduced ability to resorb bone and thereby release calcium ions. As we have already shown for kindlin-3-knockout mice, impaired bone resorption leads to decreased blood Ca²⁺ level, which triggers increased release of parathyroid hormone (PTH) by the parathyroid gland. PTH in turn induces the release of RANKL from osteoblasts, strongly promoting osteoclast differentiation (Schmidt et al., 2011). Second, the strongly elevated osteoclast numbers in kindlin-3 hypomorphic and knockout mice suggests that differentiation and fusion of osteoclasts in vivo can occur independently of integrin-mediated adhesion, especially at high precursor density. Third, kindlin-3 hypomorphic osteoclasts can at least to a certain extent degrade the bone matrix. The latter is evidenced not only by in vitro resorption assays but also by TEM of ultrathin bone sections showing polarized kindlin-3 hypomorphic osteoclasts that have formed a ruffled border framed by a sealing zone on top of resorption bays and the presence of degraded material in intracellular vesicles.

Although only 5% kindlin-3 is sufficient for formation of adhesion belts or sealing zones, our cell biological studies revealed deficits in these adhesion structures, which explains the adhesion defect. Most obvious are the formation of thinner podosome belts by kindlin-3 hypomorphic osteoclast-like cells cultured on glass, in which the podosomes are embedded in a diffuse actin cloud and in which the actin cores are less clearly separated from the surrounding ring of integrins and adaptor proteins. This indicates that kindlin-3 is not only critically involved in integrin activation but also in the assembly and organization of the adhesion structures. Consistent with this, we have recently shown that kindlin-3 regulates the turnover of podosomes by recruiting leupaxin to the adhesion site that controls paxillin phosphorylation via the phosphatase PTP-PEST (also known as Ptpn12) (Klapproth et al., 2019). Furthermore, direct interactions of kindlins with actin, paxillin and ILK have been demonstrated, indicating a central role of kindlins in the organization of the cytoskeleton (Bledzka et al., 2016; Bottcher et al., 2017; Gao et al., 2017; Huet-Calderwood et al., 2014; Theodosiou et al., 2016). The diffuse actin network, as well as the reduced diameter of actin cores of podosomes, also seen in electron microscopy studies, clearly support this. Moreover, our ultrastructural analyses revealed striking differences in the ruffled border, suggesting that the generation of a resorptive milieu by the ruffled border is additionally impaired in kindlin-3 hypomorphic osteoclasts due to the described defects in integrin signaling and cytoskeletal organization.

How kindlin-3 regulates the actin cytoskeleton within this context, remains to be further elucidated. The reduced number of active integrins and the impaired adhesion structure assembly in kindlin-3 hypomorphic cells results in a weaker linkage between the extracellular matrix and the actin cytoskeleton, which affects its maturation. Beside the indirect connection of kindlin-3 to actin, for example, via ILK or paxillin (Bottcher et al., 2017; Fukuda et al., 2014; Gao et al., 2017; Huet-Calderwood et al., 2014; Theodosiou et al., 2016), regulation of several signaling molecules involved in actin cytoskeletal organization, such as Pyk2, Syk, Vav1 and RACK1, depend on the presence of kindlin-3 (Feng et al., 2012; Klapproth et al., 2019; Schmidt et al., 2011; Xue et al., 2013).

In sum, the kindlin-3-knockout and hypomorphic mouse models allowed us to unravel different aspects of the role of kindlin-3 in regulating osteoclast-mediated bone resorption. Whereas studies on kindlin-3-knockout animals showed that osteoclast differentiation and fusion, as well as osteoblast function, are not significantly altered in the absence of kindlin-3, spreading and formation of complex integrin-mediated adhesion structures, such as adhesion belts or sealing zones, are completely abolished. Thus, these cells hardly adhere to the bone surface but predominantly float in the bone marrow cavity, which leads to to massive osteopetrosis being already present at birth (Schmidt et al., 2011). In contrast, minimal amounts of kindlin-3 are sufficient to attach osteoclasts to the bone surface and induce cell spreading and polarization. Importantly, in our study using kindlin-3 hypomorphic mice, we demonstrated that kindlin-3 is also a key organizer of the integrin adhesion complex. Moreover, whereas the cause of osteopetrosis is usually either defective maturation of osteoclasts or impaired generation of an acidic milieu and secretion of proteinases, our study shows that structural problems in the organization of integrin adhesion structures of osteoclasts can also result in osteopetrosis.

Kindlin-3-deficient (−/−) and kindlin-3 hypomorphic (+/n; n/n; n/−) mice were described previously (Klapproth et al., 2015; Moser et al., 2008). Mouse experiments were performed with the approval of the District Government of Bavaria, Germany.

Recombinant murine mouse-colony stimulating factor (M-CSF) and receptor activator of NF-κB ligand (RANKL) were obtained from Peprotech (Hamburg, Germany).

The following antibodies were used for immunofluorescent labeling of cells: mouse anti-vinculin antibody (Sigma-Aldrich, Deisenhofen, Germany); mouse anti-cathepsin K (Calbiochem, Darmstadt, Germany); rabbit anti-paxillin (Abcam, Cambridge, UK); mouse anti-talin (Sigma-Aldrich). Phalloidin dyes and secondary antibodies were obtained from Invitrogen (Darmstadt, Germany).

The following antibodies were used for flow cytometry: hamster IgG anti-integrin β1-PE, isotype control hamster IgG-PE (both from Biolegend, San Diego, USA); rat IgG2a anti-integrin β2, rat IgG2a anti-integrin α4, rat IgG2a anti-integrin α5, rat IgG2a anti-integrin αL, rat IgG1 anti-integrin αV, isotype control rat IgG1, isotype control rat IgG2a, rat anti-integrin β1 clone 9EG7 (all from BD Biosciences); hamster IgG anti-integrin β3, rat IgG2b anti-integrin αM, isotype control rat IgG2b (all from eBioscience, Frankfurt am Main, Germany).

The following antibodies were used for western blotting: mouse anti–glyceraldehyde 3-phosphate dehydrogenase (anti-GAPDH; Merck, Darmstadt, Germany); rabbit anti-Pyk2, rabbit anti-pPyk2 Tyr402, (both from Cell Signaling Technology, Leiden, Netherlands); rabbit anti-paxillin (Abcam); rabbit anti-paxillin Y31; rabbit anti-paxillin Y118 (both from Thermo Fisher Scientific, Darmstadt, Germany); rabbit anti-kindlin-3 antibody is made in house (Ussar et al., 2006). HRP-labeled secondary antibodies were purchased from Jackson ImmunoResearch Laboratories (Suffolk, UK). Catalog numbers and antibody dilutions are provided in Table S1.

Bone marrow-derived macrophages (BMDMs) were generated by treating bone marrow cells with R10 medium [RPMI 1640 containing 10% fetal calf serum (FCS), 25 mM HEPES, 100 U/ml penicillin, 100 µg/ml streptomycin, 2 mM L-Glu, 50 µM β-mercaptoethanol] supplemented with M-CSF for 6 days. Kindlin-3^−/− macrophages were differentiated from fetal liver cells accordingly.

Osteoclast-like cells were also differentiated from bone marrow or fetal liver cells. The cells were kept in α-MEM supplemented with 20 ng/ml M-CSF overnight and nonadherent cells were collected after 24 h. Leukocytes were isolated from the interface after centrifugation at 1000 g for 20 min in leukocyte separation medium (Laboratoires Eurobio, Les Ulis, France), then washed with α-MEM medium and seeded at a concentration of 2000 cells/mm² in osteoclast differentiation medium [α-MEM containing 10% FCS, 100 U/ml penicillin, 100 µg/ml streptomycin and supplemented with 60 ng/ml M-CSF (unless stated otherwise) and 40 ng/ml RANKL]. Cells were cultured at 37°C in 5% CO₂ for 5–8 days, and medium was changed every second day. Staining for TRAP activity was performed after 4–6 days in culture. Pictures were taken with a microscope (Axioskop; Carl Zeiss, Inc., Oberkochen, Germany) equipped with a DC500 camera (Leica), using a 20× NA 0.50 objective and IM50 software. Pictures were edited with Photoshop (Adobe).

For pre-osteoclast generation, 5×10⁶ interface cells were plated on a 10-cm dish and cultured in osteoclast differentiation medium for 3 day. Cells were washed with PBS and lifted with 0.1% Trypsin-EDTA.

Cell culture medium and supplements were purchased from Gibco, Invitrogen, Thermo Fisher Scientific, unless otherwise stated.

BMDM and (pre-)osteoclasts were immunofluorescently labeled as previously described (Schmidt et al., 2011). Cells were plated on 12-mm coverslips or osteologic slides (BD Biosciences) and fixed with 4% paraformaldehyde (PFA) in PBS for 10 min. Permeabilization and blocking was performed with 0.1% Triton X-100 and 1.5% bovine serum albumin (BSA) in PBS for 1 h. Cells were then incubated with the primary antibody in 1% BSA for 3 h at room temperature or at 4°C overnight. F-actin was visualized with Alexa Fluor 488–phalloidin or Alexa Fluor 647–phalloidin.

Cells were imaged at room temperature with a Leica TCS SP5 X confocal microscope (Leica, Bonn, Germany) using 63×NA 1.40 oil objective lenses and Leica Confocal Software (LAS AF). Single channels were imaged sequentially. All pictures were processed with Photoshop (Adobe Systems, San José, California, USA).

The podosome belt width was obtained by measuring the thickness of the vinculin ring of at least 20 belts from osteoclast-like cells that were differentiated in three or more independent preparations.

Long bones of 4.5-month-old mice were dissected, freed from soft tissue, decalcified in 10% EDTA in PBS for 2 weeks and embedded in paraffin. Paraffin sections were stained with H&E and for TRAP activity using a commercial kit (Sigma-Aldrich). For Masson–Goldner staining bones were fixed in 70% ethanol and embedded in polymethylmethacrylate. 3-µm sections of the distal femur were deplasticized and stained for Masson-Goldner with hematoxylin (Gill II; Carl Roth, Karlsruhe, Germany), acid Fuchsin-Ponceau Xylidine, and phosphomolybdic acid-Orange G for visualizing cells, and osteoid and Light Green for visualizing mineralized matrix.

Pictures were taken by bright field microscopy with a microscope (Axioskop; Carl Zeiss, Inc.) equipped with a DC500 camera (Leica), 10× NA 0.3 and 20× NA 0.50 objectives, and IM50 software. Pictures were edited with Photoshop (Adobe).

pQCT of the distal femur was performed with XCT Research SA⁺ (StraTec Medizintechnik). Total and trabecular bone mineral density (BMD) was measured by QCT of the distal femur (XCT Research SA-StraTec) at 7.5% of the bone length below the growth plate of the femur using peel-mode 2.

Histomorphometric analysis was performed according to the standards set forth by the American Society for Bone and Mineral Research as described previously (Bentmann et al., 2010). The following parameters were examined: bone surface, osteoclast number and osteoclast surface. Measurements were performed by a person who was blind to the genotype of the mice.

For TEM, freshly excised tibia from mice killed at 7 days old, were resin-embedded, without decalcification, according to standard procedures [immersion-fixation in cacodylate-buffered (100 mM, pH 7) 4% formaldehyde and 1% glutaraldehyde, post-fixation in 1% osmium tetroxide, en-bloc staining with 1% uranylacetate, graded dehydration with ethanol and embedding in epoxide (glycidether, NMA, DDSA, Serva, Heidelberg, Germany)]. Ultrathin sections at nominal thickness of 60 nm and contrast-stained with lead-citrate and uranylacetate were observed in a Zeiss EM 910 at 100 kV (Carl Zeiss). Micrographs were taken on image-plates scanned at 15 µm resolution (Ditabis, Pforzheim, Germany).

Primary osteoclast-like cells were differentiated on glass coverslips and ‘unroofed’ as described previously (Heuser, 2000; Luxenburg et al., 2007; Schmidt et al., 2011). Briefly, cells were treated with hypotonic buffer (20% PHEM buffer; 6 mM PIPES, 5 mM HEPES, 0.4 mM Mg₂SO₄, and 2 mM EGTA, pH 6.9) and opened by sonication for 2 s. The resulting ventral membrane preparations were fixed with 4% formaldehyde in 0.1 M phosphate buffer (pH 7.4, 1 h at room temperature) and subjected to indirect immunogold labeling essentially as described previously (Luxenburg et al., 2007). Antibodies were mouse anti-actin (Chemicon MAB1501R; 1:500) and goat anti-mouse-IgG coupled to 5 nm colloidal gold, followed by silver enhancement (Nanoprobes HQ silver). Samples were briefly postfixed with 0.5% (w/v) aqueous OsO₄, dehydrated, critical-point dried and sputter-coated with 2 nm platinum. Samples were viewed with a field emission scanning electron microscope (DSM 982-Gemini; Carl Zeiss, Inc.) equipped with secondary electron (SE) and backscattered secondary electron (BSE) detectors at 10 kV. The fine structure of podosomes was documented by using a conventional SE detector; low magnification overviews of ventral plasma membranes required high-contrast imaging with an in-lens detector. BSE imaging was used to map immunogold-silver distribution. Optionally, signals of the SE and BSE detector were mixed by using the signal mixer of the microscope. Pseudocolouring was performed with algorithms in Adobe Photoshop CS6.

Adhesion assays were performed as described previously (Chavakis et al., 2008; Schmidt et al., 2011). Non-cell-culture treated 96-well plates (Falcon, BD Biosciences) were coated with 4 μg/ml recombinant human intercellular adhesion molecule-1 (ICAM-1) and recombinant human vascular cell adhesion molecule-1 each (VCAM-1), 2 µg/ml osteopontin (all from R&D Systems) or 5 µg/ml bovine plasma fibronectin (Merck-Millipore) in coating buffer (20 mM Tris-HCl pH 9.0, 150 mM NaCl, 2 mM MgCl₂) overnight at 4°C. The wells were blocked for 1 h at room temperature with 1.5% BSA in PBS.

5×10⁴ macrophages or pre-osteoclasts in RPMI 1640, 0.1% FCS and 25 mM HEPES pH 7.4 (Gibco, Thermo Fisher Scientific) with or without 1 mM MnCl₂ were seeded on indicated ligands for 30 min at 37°C, washed twice with PBS and fixed with 4% PFA in PBS for 15 min. Each experiment was performed in quadruplets. After fixation, adherent macrophages were stained with 5 mg/ml Crystal Violet in 2% ethanol for 30 min, washed and the remaining dye was dissolved in 2% SDS and the staining intensity was measured in an ELISA plate reader at 595 nm. The number of adherent pre-osteoclasts was determined by acquisition of three phase contrast pictures from each well and counting adherent cells.

To visualize cell spreading, cells were seeded for 2 hours on different surfaces, fixed with 4% PFA in PBS for 15 min, imaged with an Axiovert 40C microscope (Zeiss) equipped with a 20× NA 0.3 objective (Zeiss).

Bone marrow derived pre-osteoclasts were trypsinized and kept in serum-free α-MEM in suspension for 1.5 h at 37°C. Then cells were either kept in suspension or plated on a surface coated with 2 µg/ml osteopontin for 20 min. Cells were carefully washed with ice cold PBS and lysed in Tris-buffered saline (TBS), containing 1% Triton X-100, protease inhibitors (Roche, Penzberg, Germany) and phosphatase inhibitor cocktails (Sigma-Aldrich); 40 µg of lysates were subjected to 10% SDS-PAGE and subsequent western blot analyses.

Osteoclast-like cells differentiated in vitro from bone marrow cells were plated in osteoclast differentiation medium on Biocoat™ Osteologic™ slides coated with bone biomaterial (BD Biosciences) for 1 week. To quantify matrix resorption, cells were removed from the slides by treating them with 1% Triton X-100 and mechanical agitation. Pictures were obtained with a microscope (Axioskop; 20× NA 0.50 objective), and the resorbed area was quantified using MetaMorph software (Molecular Devices, LLC., San Jose, CA, USA).

Functional cathepsin K concentration was measured in protein lysates of cultured osteoclast-like cells using a Cathepsin K Activity Assay kit (BioVision, Inc., Milpitas, USA). The assay was performed according to the manufacturer's instructions using 50 µg protein lysate per measurement.

Cells were incubated with Fc receptor-blocking antibody (Merck-Millipore) in FACS buffer (PBS with 1% FCS and 2 mM Na₂EDTA) and then stained with appropriate fluorophore-conjugated monoclonal antibodies for 30 min on ice. Analysis was performed with a FACSCalibur, a FACSAria or a FACSCanto (BD Biosciences) and data were analyzed using FlowJo software.

BMDMs were treated with Fc receptor-blocking antibody for 10 min. The cells were kept either unstimulated or activated with 2 mM Mn²⁺ and incubated with Alexa Fluor 647-conjugated anti-CD29 antibody (clone 9EG7, BD Biosciences) for 30 min at 37°C. Antibody labeling was performed using an Alexa Fluor 647 antibody labeling kit from Invitrogen according to the manufacturer's protocol. Cells were analyzed using a FACSAria.

Peritoneal inflammation was induced by intraperitoneal stimulation with 1 µg TNF-α (R&D Systems) in 350 µl PBS. After 4 h, 200 µl of pHrodo Green E coli BioParticles Conjugate (2 mg/ml; Thermo Fisher Scientific) was administered by intraperitoneal (i.p.) injection. Mice were killed 2 h later and their peritoneal cavity was washed with 10 ml of ice-cold PBS. The total number of leukocytes in the peritoneal lavage fluid was measured using the ProCyte Hematology analyzer (IDEXX, Westbrook, Maine, USA). Subsequently, FACS stainings were performed using antibodies directed against CD45 (APC/Cy7, Biolegend), CD11b (PerCP-Cy5.5, eBioscience/Thermo Fisher), Gr-1 (PE, eBioscience), CD115 (APC, Biolegend), F4/80 (eFluor 450, eBioscience). After erythrocyte lysis (1:10, BD FACS Lysing solution, BD Biosciences), samples were washed and analyzed by flow cytometry.

Samples for mass spectrometry (MS) analysis were prepared by in-solution digestion. Tissues were lysed in 4% SDS in PBS, heated to 95°C and sonicated. 20 µg of protein were precipitated overnight in 4 volumes (v/v) of ice-cold acetone. Protein pellets were extracted by centrifugation at 13,000 g for 10 min and dissolved in urea buffer (6 M Urea, 2 M Thiourea in 10 mM HEPES, pH 8.0). Urea-containing samples were reduced by applying DTT at a final concentration of 5 mM for 1 h at room temperature, alkylated with a final concentration of 40 mM 2-iodoacetamide (IAA) for 30 min in the dark at room temperature, digested with 1 µl LysC for 2–3 h at room temperature, diluted with 50 mM ammonium bicarbonate to a urea concentration of 2 M, incubated with 1 µl 0.5 mg/ml trypsin overnight at room temperature, acidified to 1% formic acid and purified using Stop and Go extraction tips (StageTips) (Rappsilber et al., 2003).

Proteome samples were analyzed using a liquid chromatography tandem mass spectrometry on a Q-Exactive™ HF-X Hybrid Quadrupole-Orbitrap™ (Thermo Fisher Scientific). Chromatographic peptide separation was achieved on PoroShell 120 packed 50 cm analytical columns coupled to an EASY-nLC 1000 HPLC system and a binary buffer system consisting of buffer A (0.1% formic acid) and buffer B (80% acetonitrile and 0.1% formic acid). Samples derived from in-solution digestion were measured over a 90 min gradient, raising the content of buffer B from 4% to 23% over 65 min, from 23% to 55% over 13 min and from 55% to 95% over 3 min. The column was washed with 95% buffer B for 9 min. Full MS spectra (350–1650 m/z) were recorded at a resolution (R) of 60,000, maximum injection time (max. IT) of 20 ms and automatic gain control (AGC) target of 3 × 10⁶. The 22 most abundant ion peptides in each full MS scan were selected for HCD fragmentation at nominal collisional energy (NCE) of 28. MS2 spectra were recorded at R=15,000, a maximum IT of 22 ms, and an AGC target of 1 × 10⁵.

The generated MS raw data were analyzed using MaxQuant analysis software and the implemented Andromeda software (1.5.3.8) (Cox and Mann, 2008; Cox et al., 2011). Peptides and proteins were identified using the mouse UniProt database with common contaminants. All parameters in MaxQuant were set to the default values. Trypsin was selected as the digestion enzyme, and a maximum of two missed cleavages was allowed.

Methionine oxidation and N-terminal acetylation were set as variable modifications, and carbamidomethylation of cysteines was chosen as a fixed modification. The label-free quantification (LFQ) algorithm was used to quantify the measured peptides and the ‘match between runs’ option was enabled to quantify peptides with a missing MS₂ spectrum. iBAQ quantification was selected.

Subsequent statistical analysis was performed using Perseus (1.5.5.3) software. Potential contaminants and reverse peptides were excluded, and values were log₂ transformed. iBAQ values of selected proteins were normalized to the median iBAQ values of kindlin-3.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD026711.

Data are presented as means±95%c.i. (in vitro data) or means±s.d. (ex vivo data). Statistical significance was analyzed by applying the following tests: macrophage adhesion, mixed-effects analysis followed by Sidak's multiple comparison test; 9EG7 binding, paired one-way ANOVA followed by Tukey's multiple comparison test; peritonitis, ordinary one-way ANOVA followed by Tukey's multiple comparison test; osteopontin adhesion, ordinary one-way ANOVA followed by Tukey's multiple comparison test; width of podosome belt, ordinary one-way ANOVA followed by Tukey's multiple comparison test; resorption, ordinary one-way ANOVA followed by Tukey's multiple comparison test; bone mineral density and histomorphometry, ordinary one-way ANOVA followed by Tukey's multiple comparison test or two-tailed unpaired Student's t-test.

We thank the members of the Research Department Molecular Medicine at the Max Planck Institute of Biochemistry for their input and fruitful discussions.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259040