A Glanzmann’s mutation in β3 integrin specifically impairs osteoclast function

Introduction

The osteoclast is a polykaryon of monocyte/macrophage lineage (1, 2). It differs from other members of this family by its capacity to resorb bone, an event which necessitates contact between the osteoclast and bone matrix. Once this proximity is achieved, bone-derived signals induce the osteoclast to undergo dramatic polarization eventuating in formation, at its interface with matrix, of a unique ruffled membrane which is the cell’s resorptive organelle. Thus, the means by which the osteoclast recognizes bone and transmits matrix-derived intracellular signals is critical to the cell’s capacity to resorb the skeleton.

Integrins are heterodimeric transmembrane proteins consisting of α and β subunits, which not only mediate cell-cell and cell-matrix interaction, but also act as signaling receptors (3). The integrin αvβ3 is expressed by osteoclasts, and blocking studies establish that binding of this complex to bone is essential to the resorptive process (4, 5). Consistent with this posture, β3^–/– mice become progressively osteosclerotic with age, a phenomenon due to dysfunctional osteoclasts, which fail to adequately polarize and develop abnormal ruffled membranes (6, 7). In culture, these cells do not adequately organize their cytoskeleton, and thus fail to spread normally. When placed on whale dentin, cultured β3^–/– osteoclasts only superficially excavate the surface and fail to generate normal resorptive lacunae (7).

Despite the critical role played by αvβ3 in skeletal resorption, the molecular mechanisms by which the integrin regulates osteoclast function are incompletely understood. For example, the occupied integrin activates c-src (8, 9), a molecule central to the osteoclast’s capacity to organize its cytoskeleton and resorb bone (10, 11), but the components of the integrin mediating this event are unknown.

Using a retroviral strategy, we demonstrate full rescue of β3^–/– osteoclast function with a full-length β3 cDNA. This observation permitted us to ask if the β3 cytoplasmic domain, known to transmit intracellular signals in many cell types, is essential for osteoclast function. We find that, in contrast to the complete rescue achieved by full-length β3, deletion of the integrin subunit’s cytoplasmic domain renders it completely ineffective in β3^–/– osteoclasts. To identify the components of the β3 cytoplasmic domain regulating osteoclast function, we generated a series of point mutants known, in other circumstances, to mediate β3 integrin signaling. Of the six mutants, only the S⁷⁵²P substitution, which also characterizes a form of the human bleeding disorder Glanzmann’s thrombasthenia (12), fails to rescue the spreading and resorptive capacity of β3^–/– osteoclasts or activate c-src upon ligation.

Methods

Results

In vitro, osteoclasts generated from β3^–/– BMMs develop an abnormal cytoskeleton manifested by failure to spread on culture dishes. Furthermore, mutant osteoclasts maintained on whale dentin slices excavate shallow resorption lacunae (7). Thus, we asked if expressing human β3 integrin (hβ3) rescues the abnormal phenotype of β3^–/– murine osteoclasts. We chose the human integrin because (a) high affinity mAb’s specific for hβ3 are available, and (b) the cytoplasmic domains of human and murine β3 are identical, making it very likely that hβ3 would be effective in murine cells.

Because authentic osteoclast precursors, namely primary BMMs, are difficult to efficiently transfect, we used a retroviral approach (13). To this end, we cloned the full-length hβ3 cDNA into the ΔU3 retroviral construct. To determine if integrin-transmitted intracellular signals are essential to osteoclastic bone resorption, we also constructed a retrovirus vector encoding hβ3 lacking its cytoplasmic tail (hβ3Δc) (Figure 1a).

β3-deficient BMMs were transduced with ΔU3-hβ3 or ΔU3-hβ3Δc, and surface expression of each was analyzed by flow cytometry with an antibody specific for the extracellular domain of hβ3. ΔU3-β-gal served as negative control. Retroviral transduction with either hβ3 construct yields equivalent surface expression of the WT and mutant integrin (Figure 1b). Following confirmation of hβ3 and hβ3Δc expression, transduced macrophages were cocultured with ST2 stromal cells in osteoclastogenic conditions. Nontransduced WT and β3^–/– BMMs served as controls. Eight days later, the cultures were analyzed for osteoclast expression of the β3 external domain and stained for TRAP activity, a marker of osteoclast differentiation (Figure 2a).

WT cultures consist of very large (150–400 μm diameter), well-spread TRAP-expressing, multinucleated cells. ST2 stromal cells either overlie or are pushed aside by these osteoclasts. β3^–/– cultures also contain numerous TRAP-expressing multinucleated cells, a manifestation of the fact that αvβ3 is not essential for osteoclastogenesis (7). These osteoclasts, however, spread poorly and therefore appear much smaller. Occasional larger osteoclasts are present but they also fail to spread well or push ST2 cells aside. Expression of full-length hβ3 completely restores the spreading capacity of the mutant osteoclasts. In contrast, expression of the truncated hβ3Δc yields a culture indistinguishable from that containing nontransduced β3^–/– cells. Immunofluorescent staining of mature osteoclasts derived from hβ3- and hβ3Δc-transduced precursors confirms that retroviral-driven expression of these cDNAs persists and localizes with fibrillar actin as BMMs undergo osteoclast differentiation (Figure 2b).

We next turned to the functional implications of the morphological rescue of β3-deficient osteoclasts. To this end, we generated osteoclasts on whale dentin slices, and after 8 days assessed resorption lacunae formation by scanning electron microscopy. β3^+/+ osteoclasts form well-demarcated deep resorption pits, while those excavated by cells lacking the integrin are substantially fewer in number, shallow, and poorly defined (Figures 2c and 5b). Reflecting their recovered spreading capacity, the resorptive activity of β3^–/– cells transduced with full-length hβ3 integrin is indistinguishable from their WT counterparts. In contrast, deletion of its intracellular tail abrogates the capacity of the integrin to restore the resorptive activity of β3-deficient cells.

Having established that the cytoplasmic domain of the β3 integrin is essential for osteoclast function, we turned to the individual amino acids mediating this event. On the basis of known regulatory capacity of αv-associated β integrin subunits in other systems (17–22), we mutated specific residues in the β3 cytoplasmic domain (Figure 3a). The mutants were cloned into ΔU3 retroviral construct, which was used to generate retrovirus for transduction of β3^–/– osteoclast precursors.

Once again, osteoclasts generated from uninfected β3^–/– BMMs fail to spread (Figure 3, b and c). In contrast, infection of β3^–/– osteoclast precursors with hβ3, or any mutant, save one, restores the spreading capacity of their osteoclast progeny, establishing that these altered amino acids are not essential for organization of the osteoclast cytoskeleton. hβ3(S⁷⁵²P) is the only mutant which obviates rescue of β3^–/– osteoclasts’ capacity to spread. Flow cytometric analysis of transduced BMMs demonstrates that the failure of hβ3(S⁷⁵²P) to rescue spreading is not due to diminished surface expression of this mutant in osteoclast precursors (Figure 4a). Furthermore, immunofluorescent staining of mature osteoclasts demonstrates persistent expression of hβ3(S⁷⁵²P) and two other representative mutant β3 cDNAs (Figure 4b).

Mirroring spreading, the shallow, poorly-defined pits characteristic of β3^–/– osteoclasts are completely normalized by hβ3 and representative, nondisruptive mutants, hβ3(D⁷²³A) and hβ3(Y⁷⁴⁷F/Y⁷⁵⁹F) (Figure 5a). The lacunae formed by hβ3(S⁷⁵²P), in contrast, appear morphologically similar to those produced by β3^–/– osteoclasts. Quantitative analysis reveals that the number of resorptive pits formed and the percent of dentin surface excavated by β3^–/– osteoclasts transduced with hβ3 or hβ3(D⁷²³A) are indistinguishable from those generated by WT cells (Figure 5b). Alternatively, β3^–/– cells bearing hβ3Δc, or the hβ3(S⁷⁵²P) mutant, mirror nontransduced β3^–/– osteoclasts for these same indices of resorptive activity. Pit depth is also completely rescued in hβ3 and hβ3(D⁷²³A) cultures. Interestingly, while hβ3Δc and especially hβ3(S⁷⁵²P) transductants generate no more pits than do virgin β3^–/– osteoclasts, they partially normalize pit depth (P < 0.01 compared with both WT and β3^–/– osteoclasts).

We next turned to c-src activation, an intracellular signal mediated by αvβ3, in osteoclasts and asked if the event required the β3 cytoplasmic domain. Early osteoclasts were generated from β3^–/– BMMs transduced with hβ3, hβ3Δc, or the hβ3(S⁷⁵²P) and hβ3(Y⁷⁴⁷F/Y⁷⁵⁹F) mutants. The transductants were lifted with EDTA and kept in suspension or plated on the αvβ3 ligand vitronectin. After 1 hour, phosphorylation of c-src at the activation-specific Y⁴¹⁶ site was determined by immunoblot (Figure 6). Adhesion to vitronectin activates c-src in osteoclasts generated from β3^–/– marrow macrophages transduced with intact hβ3 or hβ3(Y⁷⁴⁷F/Y⁷⁵⁹F). In contrast, no such activation occurs in osteoclasts bearing hβ3Δc and hβ3(S⁷⁵²P), mirroring the functional effects of these mutations.

Discussion

Osteoclastic bone resorption is initiated by matrix recognition, and formation, at the cell-bone interface, of an isolated, acidified microenvironment, which is the site of skeletal degradation (1). This physical intimacy between the cell and bone indicates that attachment molecules on the osteoclast are pivotal to skeletal remodeling. This posture is buttressed by experiments performed, in vitro and in vivo, demonstrating that αvβ3 blockade blunts the osteoclast’s ability to resorb bone (4–6). The clinical relevance of this observation is underscored by the capacity of soluble organic mimetics of the αvβ3 ligand to prevent experimental, postmenopausal osteoporosis (5).

With these experiments in mind, and the wish to determine the role of the αvβ3 integrin in skeletal development, we generated β3-deficient mice (7). Because the platelet integrin αIIbβ3 is not expressed in these animals, they serve as a model of the human bleeding dyscrasia Glanzmann’s thrombasthenia (23). Reflecting osteoclast dysfunction, β3^–/– mice are hypocalcemic and develop bone sclerosis as they age (7). While the bone phenotype in patients with Glanzmann’s thrombasthenia is unknown, a reasonable possibility holds that they too may have increased bone mass. A likely clinical consequence of this phenomenon would be protection against pathological bone loss such as that attending cessation of ovarian function.

The fact that β3^–/– osteoclasts fail to normally organize their cytoskeleton, in vitro and in vivo, represents compelling evidence that the integrin transmits matrix-derived signals essential to the resorptive process. Given that the majority of known signaling events mediated by αvβ3 depend upon the β3 cytoplasmic domain, we asked if such was the case regarding the osteoclast. To address this issue, we first expressed full-length hβ3 in β3^–/– osteoclasts. This undertaking was complicated by the fact that primary macrophages, which are osteoclast precursors, cannot be transfected with high efficiency by traditional methods. Thus, we utilized a retroviral strategy (13) which permits effective expression of the transgene in virtually all osteoclast precursors. These transduced cells, when placed in osteoclastogenic conditions, differentiate into osteoclasts indistinguishable from WT, both in their capacity to spread and, most importantly, to resorb bone. In contrast, β3^–/– osteoclasts are unaltered by the β3 integrin transgene lacking the cytoplasmic domain. Thus, signal transduction mediated by the β3 cytoplasmic tail is critical for integrin function in the osteoclast.

Previous studies have demonstrated that binding of RGD-containing peptides to the integrin triggers various intracellular signals, including changes in intracellular calcium (24–26), and activation of c-src (8), PYK2 (9), p130cas (27), and PI3-kinase (28). Despite these observations, the structural components of the β3 cytoplasmic tail mediating osteoclast function have not been elucidated.

In an attempt to identify the amino acid residues in the β3 cytoplasmic tail critical to osteoclastic bone resorption, we generated single and double amino acid mutants based upon the demonstration, in other cells, that they alter β integrin function. D⁷²³ is implicated in forming a salt bridge with the α integrin subunit, thereby stabilizing an activated conformation (20). P⁷⁴⁵, Y⁷⁴⁷, and Y⁷⁵⁹ are located in two NPXY/NXXY motifs which are conserved among most β integrin subunits and mediate many aspects of integrin function (17–19). Mutation of integrin β1A-P⁷⁸¹, which corresponds to β3-P⁷⁴⁵, dampens expression and inactivates β1A in the mouse embryonic stem cell line GD25 (22).

Of the six point mutants known to impact β integrin function, only S⁷⁵²P fails to rescue β3^–/– osteoclasts. This mutation has also been documented in Glanzmann’s thrombasthenia (29), in which platelets fail to aggregate. Thus, the residue regulating human platelet function also regulates the cytoskeletal organization and bone resorptive activity of osteoclasts. Specifically, hβ3(S⁷⁵²P) fails to rescue both the impaired capacity of β3^–/– osteoclasts to spread, and the frequency with which they form resorptive lacunae. Similar to platelets (30), the effect of the mutation is specific for proline, as the more conservative mutation, S⁷⁵²A, is as effective as WT hβ3 in rescuing osteoclast spreading (data not shown). This result suggests that local secondary structure, and not phosphorylation of S⁷⁵², likely mediates its central role in osteoclast function.

It is of interest that, like S⁷⁵², Y⁷⁴⁷ and Y⁷⁵⁹ are, in combination, also essential for platelet function (31). Both tyrosines, when phosphorylated, bind the signaling molecules SHC and GRB2, as well as the cytoskeletal protein myosin, in platelets (32, 33). Unlike S⁷⁵², however, these combined mutations fail to impact osteoclast function. Thus, the osteoclast and platelet may share some β3-mediated signaling pathways, while others appear cell-specific.

c-src is a tyrosine kinase essential for osteoclast function (10, 11, 34). The mechanism by which this proto-oncogene activates the osteoclast is complex, but clearly involves both the kinase and protein docking domains (35, 36). We find that αvβ3-dependent activation of c-src requires the cytoplasmic domain of β3, and is abrogated by the S⁷⁵²P, but not the Y⁷⁴⁷F/Y⁷⁵⁹F, mutation. Thus, the ability of β3 to activate c-src correlates with its capacity to stimulate osteoclast function.

While the S⁷⁵²P mutation affects both osteoclast and platelet function, the downstream signaling mechanisms appear to be distinct. Deletion of c-src has not been reported to cause platelet dysfunction (34), in contrast to osteoclasts, and therefore is unlikely to represent the relevant β3-mediated signaling pathway in platelets.

Integrin signaling is bidirectional, with inside-out signals affecting ligand binding affinity, and outside-in signals determining intracellular events occurring upon integrin-ligand interaction (37). In the context of overexpression in Chinese hamster ovary (CHO) cells, Chen et al. (29) showed that the S⁷⁵²P mutation disrupts binding of the ligand-induced binding site (LIBS) antibody PAC1, suggesting a defect in inside-out signaling. In contrast, we find that ς osteoclasts expressing hβ3(S⁷⁵²P) bind AP5, another β3-specific LIBS antibody (R. Faccio, unpublished observations). Given that AP5 recognizes the activated form of αvβ3, this observation suggests that the β3(S⁷⁵²P) mutation, in osteoclasts, arrests outside-in signaling and thus, ligand-induced c-src activation.