Immune complexes regulate bone metabolism through FcRγ signalling

Autoantibody production and immune complex (IC) formation are frequently observed in autoimmune diseases associated with bone loss. However, it has been poorly understood whether ICs regulate bone metabolism directly. Here we show that the level of osteoclastogenesis is determined by the strength of FcRγ signalling, which is dependent on the relative expression of positive and negative FcγRs (FcγRI/III/IV and IIB, respectively) as well as the availability of their ligands, ICs. Under physiological conditions, unexpectedly, FcγRIII inhibits osteoclastogenesis by depriving other osteoclastogenic Ig-like receptors of FcRγ. Fcgr2b−/− mice lose bone upon the onset of a hypergammaglobulinemia or the administration of IgG1 ICs, which act mainly through FcγRIII. The IgG2 IC activates osteoclastogenesis by binding to FcγRI and FcγRIV, which is induced under inflammatory conditions. These results demonstrate a link between the adaptive immunity and bone, suggesting a regulatory role for ICs in bone resorption in general, and not only in inflammatory diseases. Bone and the immune system are functionally intertwined. This study shows that osteoclastogenesis is modulated by the intensity of Fcγ receptor signalling, which is shaped by the balance between the positive and negative Fcγ receptors expressed on osteoclasts and the availability of their ligands, immune complexes.

T he immune and bone systems share numerous regulatory factors, including cytokines, receptors and signalling molecules. Therefore, as a pathology in one of these systems may very well have an impact on the other, understanding the interplay between the immune and skeletal systems is crucially important 1 . Bone homeostasis depends on a dynamic balance of bone formation and resorption, which are mediated by osteoblasts and osteoclasts, respectively 2,3 . Tipping the balance in favour of osteoclasts leads to diseases characterized by a low bone mass, including osteoporosis, whereas impaired osteoclastic bone resorption results in diseases with a high bone mass, including osteopetrosis. Osteoclasts originate from bone marrow-derived monocyte/macrophage lineage cells (BMMs) and their differentiation is mediated by many of the same regulators utilized in the immune system. Receptor activator of NF-kB ligand (RANKL) and macrophage colony-stimulating factor (M-CSF) are key cytokines in osteoclastogenesis 3,4 .
RANKL activates the differentiation process by inducing the master transcription factor for osteoclastogenesis, nuclear factor of activated T cells c1 (NFATc1), via the tumour necrosis factor receptor-associated factor 6 and c-Fos pathways 5 . The induction of NFATc1 is also dependent on the calcium signal activated by the co-stimulatory signals for RANK. The immunoreceptor tyrosine-based activation motif (ITAM)-bearing adaptor proteins, Fc receptor common g subunit (FcRg, which is encoded by Fcer1g) and DNAX-activating protein 12 kDa (DAP12, which is encoded by Tyrobp), play a crucial role in the transduction of the co-stimulatory signals for RANK 6,7 . FcRg associates with the immunoglobulin (Ig)-like receptors, such as osteoclast-associated receptor (OSCAR) and paired Ig-like receptor-A (PIR-A), while DAP12 associates with triggering receptor expressed in myeloid cell-2 and signal-regulatory protein b1 (ref. 6). The surface expression of these Ig-like receptors is known to be dependent on the association with their adaptors FcRg and DAP12 (ref. 6).
Phosphorylation of ITAM recruits the protein tyrosine kinases including Syk, which activate the phosphorylation of PLCg in the signalling complex containing Tec family members activated by RANK 8 . Activation of PLCg stimulates calcium oscillations, in turn leading to the robust induction of NFATc1 (refs 1,5). The essential role of ITAM-mediated signalling in osteoclastogenesis has been demonstrated by the severe osteopetrotic phenotype that manifests in Fcer1g À / À Tyrobp À / À mice 6,7 . It is evident that FcRg plays a crucial role in osteoclastogenesis because of the much more severe osteopetrosis that develops in Fcer1g À / À Tyrobp À / À mice than in Tyrobp À / À mice 6,7 ; however, Fcer1g À / À mice exhibit a normal bone phenotype, making the specific role of FcRg in bone homeostasis enigmatic. FcRg acts as the common subunit of the activating FcgRs expressed in essentially all innate immune cells, including monocyte/macrophage lineage cells 9,10 ; however, the function of FcgRs in the regulation of osteoclastogenesis is only poorly understood.
It is well documented that enhanced bone resorption is associated with the activation of the immune system observed in autoimmune or inflammatory diseases, such as rheumatoid arthritis (RA) 11 , systemic lupus erythematosus (SLE) 12 and inflammatory bowel disease 13 . The role of FcgRs has been explored in arthritis models [14][15][16][17][18] ; however, it has been difficult to observe their direct effects on bone metabolism due to their central contribution to the onset of autoimmune disease and inflammation. Thus, the direct regulation of bone homeostasis by IgG immune complexes (ICs) has not been established.
In mice, three classes of activating FcgRs including FcgRI, FcgRIII and FcgRIV, which associate with FcRg, and one immunoreceptor tyrosine-based inhibitory motif-bearing inhibitory receptor, FcgRIIB, have been characterized 9,10,19 . These receptors bind to four subclasses of mouse IgG (IgG1, IgG2a, IgG2b and IgG3) with different affinities. High-affinity FcgRI (or FcgRIA in human) binds particular IgG subclasses (IgG2a in mice and IgG1, IgG3 and IgG4 in humans) in both monomeric and IC forms, whereas the other FcgRs have a markedly lower affinity for various IgG subclasses and are only activated by binding IgG ICs. The inhibitory FcgRIIB is co-expressed with FcgRs in several types of immune cells, including mast cells, neutrophils and macrophages, such that the balance of their expression determines the threshold of the activation of FcRg signalling in response to the ICs. For example, IgG1 binds FcgRIII and FcgRIIB with an activating-to-inhibitory (A/I) ratio of 0.1 (ref. 20), whereas IgG2a and IgG2b bind FcgRIV and FcgRIIB with an A/I ratio of 70 and 7, respectively 20 . Here we addressed whether and how IgG ICs regulate bone homeostasis by binding FcgRs on osteoclast precursor cells.
Here we report that osteoclastogenesis is determined by the strength of FcRg-mediated ITAM signalling. Under physiological conditions, the activation of FcgRIII-mediated FcRg signalling is counterbalanced by the inhibitory receptor FcgRIIB and the signalling through other FcRg-associating Ig-like receptors is inhibited by the sequestration of FcRg by FcgRIII. Under pathological conditions such as autoimmune diseases associated with hypergammaglobulinaemia, IgG ICs induce osteoclastogenesis by acting on the highly expressed positive FcgRs without any attenuating effect by the negative receptor. This study provides clear evidence for direct regulation of bone homeostasis by IgG ICs, and provides insights into the pathogenesis of osteoporosis associated with autoimmune diseases. Thus, FcgRs place bone metabolism under the control of the IgGs, with physiological and pathological consequences constituting a vital link between the bone and adaptive immune system.

Results
The osteoporotic phenotype of mice deficient in FccRIII. Among four mouse FcgRs, we identified FcgRIIB and FcgRIII as the ones predominantly expressed in BMMs at the mRNA level ( Supplementary Fig. 1a), and also at the protein level (Fig. 1a). To investigate whether FcgRIII activates osteoclast differentiation, we analysed the bone phenotype of mice deficient in FcgRIII. Unexpectedly, microcomputed tomography (mCT) and dualenergy X-ray absorptiometry indicated that Fcgr3 À / À mice possessed an osteoporotic phenotype ( Fig.1b and Supplementary  Fig. 1b,c). Bone morphometric analysis revealed an increase in osteoclast number and bone resorption (Fig. 1c,d). There were no substantial differences in the number or function of osteoblasts ( Supplementary Fig. 1d), showing that it is the increased osteoclastic bone resorption activity that is responsible for the low bone mass in Fcgr3 À / À mice.
In vitro osteoclast differentiation was evaluated by counting the multinucleated cells positive for the osteoclast marker tartrateresistant acid phosphatase (TRAP) after stimulation of BMMs with RANKL in the presence of M-CSF. Osteoclast formation was increased in Fcgr3 À / À cells (Fig. 1e), while the cell proliferation rate was unchanged ( Supplementary Fig. 1e). Enhanced osteoclastogenesis was also observed in a co-culture of Fcgr3 À / À BMMs and osteoblasts, although Fcgr3 À / À osteoblasts exhibited a normal ability to support osteoclastogenesis ( Supplementary  Fig. 1f), suggesting that FcgRIII suppresses osteoclast differentiation by a cell-autonomous mechanism.
FccRIII inhibits ITAM signalling by sequestering FcRc. How does the FcRg-associating receptor FcgRIII suppress osteoclastogenesis? We examined the activation of signalling pathways downstream of ITAM in the absence of FcgRIII. Although FcgRIII has been thought to activate the ITAM signal through FcRg, we observed hyperactivation of PLCg2 and calcium oscillations in Fcgr3 À / À osteoclast precursor cells (Fig. 2a, left and middle). Consistent with this, the expression of NFATc1 was significantly elevated in Fcgr3 À / À cells during osteoclastogenesis (Fig. 2a, right). In contrast, retroviral expression of FcgRIII in wild-type, but not in Fcer1g À / À cells, suppressed the differentiation into osteoclasts (Fig. 2b) and the Fcgr3 deficiency rescued the defect in ITAM signalling in Tyrobp À / À cells ( Supplementary Fig. 3). These results suggest that FcgRIII negatively regulates the FcRg-associated co-stimulatory signal under physiological conditions.
FcgRIII is highly expressed in osteoclast precursor cells, but is downregulated during the course of osteoclast differentiation ( Fig. 2c and Supplementary Fig. 1a). When we sorted the FcgRIII high and FcgRIII low populations, FcgRIII low , but not FcgRIII high cells, differentiated into osteoclasts efficiently (Fig. 2d), suggesting that the downregulation of FcgRIII expression is an important step in the early phase of osteoclast differentiation. Since FcgRIII expressed on BMMs constantly associated with FcRg regardless of the presence or absence of IgG1 (Fig. 2e), we hypothesized that FcgRIII may function as an 'inhibitory receptor' by depriving other activating receptors of the FcRg subunit. Consistent with this hypothesis, the overexpression of FcRg rescued the osteoclastogenesis of FcgRIII high cells (Fig. 2f). Besides activating FcgRs, FcRg associates with co-stimulatory receptors, such as OSCAR and PIR-A in osteoclast precursor cells. Flow cytometric analysis revealed that OSCAR high cells are found mainly in the FcgRIII low population, despite a comparable cellular protein level of OSCAR in the FcgRIII high and FcgRIII low populations (Fig. 2g). The association with FcRg, and the cell surface expression of OSCAR and PIR-A, were detected in Fcgr3 À / À cells much more potently than Fcgr3 þ / þ cells (Fig. 2h,i). Collectively, FcgRIII expression modulates the level of FcRg available for the other receptors, thus effectively inhibiting the surface expression of these receptors. Since we did not observe a significant increase in FcgRI or FcgRIV expression in Fcgr3 À / À cells and the ligands for these receptors IgGs in Fcgr3 À / À mice ( Supplementary  Fig. 2), we can exclude the possibility that the higher expression of FcgRI and FcgRIV contributes to the increased osteoclastogenesis in Fcgr3 À / À cells.
Fcgr2b À / À mice develop an osteoporotic phenotype. The question thus arises as to how FcgRIII regulates osteoclastogenesis in the presence of its ligands. In immune cells, FcgRIIImediated FcRg signalling is antagonized by FcgRIIB, which also binds to IgG1 (ref. 20). This led us to investigate the role of FcgR signalling in osteoclastogenesis Fcgr2b À / À mice. Fcgr2b À / À mice exhibited an osteoporotic phenotype due to an increase in the osteoclast number and eroded surface without any abnormality in osteoblastic bone formation at the age of 12 weeks (Fig. 3a,b and Supplementary Fig. 4a-c). However, there was no significant difference in osteoclast formation between the wild-type and Fcgr2b À / À cells cultured in conventional culture medium containing 10% fetal bovine serum (FBS; Fig. 3c, upper, and Supplementary Fig. 4d). Since the affinity of IgGs to FcgRs differs among species and the amount of IgGs contained in FBS is low, we replaced the FBS with serum isolated from wild-type mice. Interestingly, osteoclast formation was much more efficient in Fcgr2b À / À cells than in wild-type cells when cultured in the medium supplemented with the mouse serum (Fig. 3c, lower), presumably because of the presence of mouse IgGs in the serum.
To demonstrate the relevance of IgGs in this osteoclast formation assay, we depleted IgGs from an aliquot of mouse   serum. The enhanced osteoclastogenesis in Fcgr2b À / À cells was not detected in the culture with the IgG-depleted mouse serum (Fig. 3d, upper), indicating that IgG is responsible for the osteoclastogenic effect. Fcgr2b À / À mice contain an elevated level of autoantibodies in the serum due to the hyperactivation of various immune cells including B and plasma cells 21,22 . As expected, osteoclast formation in Fcgr2b À / À cells was further increased when cultured in the medium supplemented with the serum isolated from Fcgr2b À / À mice, whereas in wild-type cells osteoclast formation was not enhanced (Fig. 3d, lower). Consistent with this, the osteoporotic phenotype of Fcgr2b À / À mice became increasingly evident with age, in accordance with the increase in IgG production ( Supplementary Fig. 4e-h).
To confirm that the osteoporotic phenotype in Fcgr2b À / À mice is caused by an increase in the activating FcgR-mediated signals, Fcgr2b À / À mice were crossed with Fcer1g À / À mice, in which no activating FcgRs are functionally expressed on the cell surface 23 . Indeed, the low bone mass phenotype observed in Fcgr2b À / À Fcer1g þ / À mice was not seen in Fcgr2b À / À Fcer1g À / À double knockout (DKO) mice, and osteoclast number and bone resorption in DKO mice was much lower than that in Fcgr2b À / À Fcer1g þ / À mice ( Fig. 3e and   Supplementary Fig. 5a). There was no significant difference in the bone mass and bone histomorphometric parameters between the control Fcgr2b þ / À Fcer1g þ / À and DKO mice ( Fig. 3e and Supplementary Fig. 5a,b). The augmented osteoclastogenesis in Fcgr2b À / À Fcer1g þ / À cells stimulated by the mouse serum was rescued in DKO cells (Fig. 3f), suggesting that the enhanced osteoclastogenesis in Fcgr2b À / À mice was dependent on FcRg and associating FcgRs. These results indicate that FcgRIIB inhibits osteoclastogenesis by counteracting the FcRg-associated activating FcgRs in the presence of its ligand IgGs.
ICs increased osteoclast formation in Fcgr2b À / À mice. Normal serum contains only a low amount of ICs; however, the level of the autoantibody-containing IC increases in numerous autoimmune diseases. Does the osteoclastogenic effect of the serum  **

BMMs
BMMs Figure 3 | Osteoporotic phenotype in accordance with autoimmune disease in Fcgr2b À / À mice. (a) mCT of the femur of 12-week-old female Fcgr2b þ / À and Fcgr2b À / À mice (see Fig. 1b legend for the details). Representative data of 10 mice are shown. Bone volume and degree of trabecular separation were determined with the mCT analysis (right, n ¼ 10). (b) Bone histomorphometric analysis of the tibiae of 12-week-old female Fcgr2b þ / À and Fcgr2b À / À mice and the parameters for osteoclastic bone resorption, as determined by the bone morphometric analysis (n ¼ 10). (c) Osteoclast differentiation in Fcgr2b þ / À and Fcgr2b À / À cells cultured in 10% fetal bovine serum (upper) or 5% mouse serum isolated from WT mice (lower). Representative data (left) and quantification (n ¼ 3) are shown. (d) Effect of IgGs in mouse serum isolated from WT mice (upper) or Fcgr2b À / À mice (lower) on osteoclast differentiation in Fcgr2b þ / þ and Fcgr2b À / À cells. (e) The mCT analysis of the femur of 12-week-old female Fcgr2b þ / À Fcer1g þ / À (Control), Fcgr2b À / À Fcer1g þ / À and Fcgr2b À / À Fcer1g À / À (DKO) mice. Representative data of nine mice are shown. Bone volume was determined with the mCT analysis (right, n ¼ 9). (f) Osteoclast differentiation in Fcgr2b þ / À Fcer1g þ / À , Fcgr2b À / À Fcer1g þ / À and Fcgr2b À / À Fcer1g À / À cells cultured in WT mouse serum. All quantification experiments were performed using triplicate culture wells. All data are representative of more than three independent experiments and are shown as the mean±s.e.m. Statistical analyses were performed using unpaired two-tailed Student's t-test (*Po0.05; **Po0.01; n.s., not significant). NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7637 ARTICLE from Fcgr2b À / À mice depend on IgG ICs? Size-exclusion chromatography revealed that the serum isolated from Fcgr2b À / À mice contained a higher amount of IgGs (corresponding to the same molecular weight as the mouse IgG used as a reference, or higher, as shown in the dotted rectangle in Fig. 4a) than wild-type mice. The IgG-containing fractions in the serum from the Fcgr2b À / À mice were separated into three peak fractions and confirmed to contain IgGs (Fig. 4b). Among these fractions, one of them (F14) corresponds to the molecular weight of the fraction containing monomeric IgG and two fractions (F4 and F6) contained a higher molecular weight, which had previously been shown to contain ICs 24 . We tested the effect of these fractions on osteoclast differentiation. The IC fractions (F4 and F6), but not the monomeric IgG fraction (F14), dramatically increased osteoclast differentiation in Fcgr2b À / À cells (Fig. 4c). This is consistent with the finding that IgG ICs, but not monomeric IgGs, activate immune cells by crosslinking FcgRs 9,10 . These results suggest that ICs play a functional role in osteoporosis in Fcgr2b À / À mice.
To further define the role of the pathological ICs in osteoclastogenesis, we generated ICs by mixing the serum isolated from wild-type mice and an F(ab 0 ) 2 segment that recognizes and crosslinks mouse IgG by binding to its light chain. Osteoclast formation was markedly enhanced in Fcgr2b À / À , but not in wild-type cells treated with the mouse serum containing the F(ab 0 ) 2 segment, and this enhancement was abrogated by depletion of IgGs from the serum (Fig. 4d). Furthermore, we tested the effect of another type of soluble IC, composed of monoclonal mouse IgG1 specific for trinitrophenyl (TNP). This IC increased osteoclast formation in Fcgr2b À / À , but not in Fcgr2b À / À Fcer1g À / À cells (Fig. 4e). Thus, the ICs were able to increase osteoclast differentiation through the activating FcgRs when the inhibitory effect was relatively compromised (that is, there was a downregulation of FcgRIIB or a stimulation by IgG subclasses with a higher affinity to the activating receptors).
Since nonsialylated IgGs have a higher affinity to activating FcgRs and IgG sialylation is decreased in several autoimmune diseases 25,26 , we examined the sialylation status of the IgGs in Fcgr2b À / À mice. We found that the sialylation ratio of the IgGs in the Fcgr2b À / À mice was significantly less than that in the control mice ( Supplementary Fig. 6a). Compared with normal mouse IgGs, osteoclastogenesis was more effectively promoted by the same concentration of IgGs purified from the Fcgr2b À / À mice ( Supplementary Fig. 6b). De-sialylation of the IgGs in the control mice, but not in the Fcgr2b À / À mice, had a stimulatory effect on osteoclastogenesis ( Supplementary  Fig. 6b,c) These results suggest that the de-sialylation of IgGs contributes to enhanced osteoclastogenesis under inflammatory conditions.

The engagement of FccRs by ICs stimulates osteoclastogenesis.
IgG1 is the most abundant IgG subclass in the serum under physiological conditions, while the production level of other subclasses, such as IgG2a and IgG2b, increases during the course of the immune response 9 . In mice, the activity of IgG1 is dependent on FcgRIII, whereas either FcgRIV alone, or a combination of FcgRIV with either FcgRI or FcgRIII, is crucial for the activity of the IgG2a and IgG2b subclasses 9,10 . We analysed the effect of three mouse IgG subclasses and their responsible FcgRs on osteoclast differentiation. We cultured BMMs derived from wild-type, Fcgr2b À / À , Fcgr3 À / À , Fcgr2b À / À Fcer1g À / À and Fcer1g À / À mice on plate-bound monoclonal IgGs, which have an ability to crosslink FcgRs. IgG1 crosslinking increased osteoclast formation in Fcgr2b À / À , but not in wild-type cells, through the activation of ITAM signalling (Fig. 5a, left and   Fig. 5b). This suggests that the positive effect of FcgRIII was inhibited by FcgRIIB in wild-type cells, since IgG1 binds FcgRIII with an extremely low A/I ratio 20 . Since the Fcgr2b gene is in close proximity to the Fcgr3 gene on the same chromosome, it is extremely difficult to generate Fcgr2b À / À Fcgr3 À / À mice, but Fcgr2b À / À Fcer1g À / À can be used to prove the involvement of the activating FcgRs, including FcgRIII. The IgG1-mediated increase in osteoclastogenesis was abrogated in Fcgr2b À / À Fcer1g À / À cells (Fig. 5a, left). Consistent with this, the enhancement of osteoclastogenesis in IgG1-stimulated Fcgr2b À / À cells was abrogated by the knockdown of the FcgRIII expression using short hairpin RNA (shRNA; Fig. 5c and Supplementary  Fig. 7), indicating that IgG1 indeed acts on FcgRIII. In contrast, the crosslinking by IgG2a and IgG2b resulted in increased osteoclast formation, even in wild-type cells, and this increase was also observed in Fcgr3 À / À , but not in Fcgr2b À / À Fcer1g À / À and Fcer1g À / À cells (Fig. 5a, middle and right). When the expression of FcgRI or FcgRIV was knocked down by shRNA, the stimulatory effect of IgG2a crosslinking on osteoclastogenesis was markedly suppressed (Fig. 5d and Supplementary Fig. 7). These results suggest that the ICs containing either IgG2a or IgG2b function through FcgRI and FcgRIV, to which IgG2a and IgG2b bind with a higher affinity than the inhibitory FcgRIIB 20 .
In vivo evidence that ICs directly regulate bone resorption. To test whether the ICs directly induce bone loss in vivo, we investigated the effect of a local administration of the IgG ICs to the calvarial bone. We observed a marked increase in osteoclast number, eroded surface and trabecular bone loss when IgG2a ICs were injected into wild-type, Fcgr2b À / À and Fcgr3 À / À mice, but not Fcer1g À / À mice ( Fig. 6a and Supplementary Fig. 8a). The bone loss was observed only in the Fcgr2b À / À mice when they were injected with IgG1 ICs (Fig. 6a and Supplementary  Fig. 8a). The infiltration of inflammatory cells such as lymphocytes and neutrophils, as observed in the lipopolysaccharideinduced bone destruction sites, were not detected in this model ( Supplementary Fig. 8b), suggesting that the ICs directly induced local bone loss through the activating FcgRs without any immune response activity.
When we intravenously injected the IgG1 ICs into 3-week-old Fcgr2b À / À mice (before the onset of autoimmune symptoms), we observed an increase in osteoclast number and a generalized bone loss without any discernible signs of inflammation or increase in the serum cytokine levels (Fig. 6b,c and Supplementary Fig. 8c-e), suggesting that the circulating ICs directly promoted the osteoclastogenesis that resulted in the systemic bone loss.

Altered FccR expression in arthritis.
To explore the contribution of the IC-induced bone loss to arthritis, we examined the effect of serum IgGs on osteoclast precursor cells derived from mice with collagen-induced arthritis (CIA). The serum isolated from CIA mice induced osteoclastogenesis more effectively than the control mouse serum (Fig. 6d). These effects were abrogated by the depletion of IgGs from the serum (Fig. 6d), indicating that the high IgG concentration in CIA serum is responsible for the efficient osteoclastogenesis. On the other hand, BMMs derived from CIA mice underwent differentiation into osteoclasts more efficiently than control BMMs in the presence of the CIA serum, which was accompanied by a higher activation of ITAM signalling (Fig. 6d,e). To investigate the mechanism by which CIA BMMs were more sensitive to IgGs, we examined the expression level of FcgRs in CIA BMMs. The expression of FcgRIIB was decreased, while that of FcgRIII and FcgRIV was increased in CIA BMMs (Fig. 6f). Consistent with this, the IgG1 IC markedly (a) Fractionation of mouse serum isolated from Fcgr2b þ / þ and Fcgr2b À / À mice using size-exclusion chromatography. Purified mouse IgG was used as a reference. Fractions with a molecular weight corresponding to IgG and higher (in the dotted rectangle) were used for the analysis in Fig. 2b. Representative data of three independent experiments are shown. (b) ELISA analysis of IgG in the peak fractions (F4, 6, 14). (c) Effect of IgG-containing fractions on osteoclast formation in Fcgr2b À / À cells. (d) Effect of IgG IC generated by mixing the WT mouse serum and a F(ab') 2 segment on osteoclast differentiation of Fcgr2b À / À cells or WT cells. (e) Effect of soluble IC composed of TNP-BSA together with a-TNP IgG1 (IgG1 IC) on osteoclast differentiation in WT, Fcgr2b À / À and Fcgr2b À / À Fcer1g À / À cells. All quantification experiments were performed using triplicate culture wells. All data are representative of more than three independent experiments and are shown as the mean±s.e.m. Statistical analyses were performed using unpaired two-tailed Student's t-test (*Po0.05; **Po0.01; ***Po0.001; n.s., not significant). NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7637 ARTICLE induced osteoclast formation in the BMMs derived from CIA but not control mice (Fig. 6g). The IgG2a IC also enhanced osteoclastogenesis of the BMMs derived from CIA mice much more efficiently than control mice (Fig. 6g). When the expression of FcgRIII and FcgRIV was knocked down by shRNA, the stimulatory effects of the IgG1 and IgG2a ICs were abrogated, respectively (Fig. 6h,i). These results suggest that the downregulation of inhibitory FcgRIIB and the upregulation of activating FcgRIII and FcgRIV underlie the hyper-responsiveness of the CIA BMMs to IgG ICs.

Discussion
IgG antibodies are important for protecting the host from microbial infections by the activation of the FcgRs and immune cells. IgGs and IgG ICs are also involved in the pathogenesis of several autoimmune diseases by both excessive and prolonged activation of immune reactions leading to tissue injury. Recently, the investigation of the mechanisms of intravenous Ig has brought into focus the crucial role of the sialylated IgGs in antiinflammatory effector functions [25][26][27] . Thus, the diverse biological and pathological properties of IgGs in the immune system have been established, the elucidation of which has contributed to advances in therapeutic applications of the IgGs, such as vaccination, intravenous Ig and antibody treatments. However, the direct effect of IgGs and ICs in other biological systems has been scarcely reported. FcgRs recognize IgGs and IgG ICs so as to transmit intracellular signals through FcRg, which activates ITAM signalling, required for co-stimulatory signals in osteoclast differentiation. Since the FcgR system consists of  multiple positive and negative receptors, the effect of IgGs varies among the different subtypes and the effect of monomeric and polymeric IgG also differs, it has been difficult to elucidate the molecular basis of IgG regulation of osteoclastogenesis despite the wealth of in vitro and in vivo observations of the involvement of FcgRs in bone metabolism 6,7,14,[6][7][8][9][10][11][12][13][14][15][16][17][18]28,29 . This study affords a detailed picture of the regulation of osteoclastogenesis by the IgG-FcgR system, in which both ligands and receptors are differentially regulated under physiological and pathological conditions, providing clear evidence for the IC-mediated regulation of bone metabolism.
FcgRIII has been extensively investigated as a major receptor of IgG ICs and a crucial partner of FcRg in host defense. One would thus expect that FcgRIII would function as the major positive receptor in terms of osteoclastogenesis among the FcgRs, considering the fact that FcRg-mediated ITAM signalling constitutes the crucial positive signal for osteoclastogenesis. In contrast with this expectation, our finding of an osteoporotic phenotype and enhanced osteoclastogenesis in Fcgr3 À / À mice reveals that FcgRIII functions as an inhibitory receptor in osteoclastogenesis under a physiological setting. At a very low concentration of IgG ICs, ITAM signalling was suppressed in Fcgr3 À / À mice, because FcgRIII sequestrates FcRg from Ig-like receptors such as OSCAR and PIR-A, resulting in the reduced surface expression of these Ig-like receptors (Fig. 7a). This competition between ITAM-associating receptors for surface expression is analogous to the increased expression of FcgRIII in mast cells lacking FceRI during the development of anaphylactic reactions due to the increased availability of FcR b and/or g chains 30 . Consistent with the notion that FcgRIII functions as an inhibitory receptor in osteoclastogenesis, FcgRIII expression decreases as osteoclastogenesis proceeds and efficient osteoclast formation was observed in FcgRIII low osteoclast precursor cells. This is an interesting case of a competitive expression mechanism in which downregulation of FcgRIII is crucial for the physiological process of osteoclastogenesis (Fig. 7a).
Fcgr2b À / À mice also exhibited an osteoporotic phenotype because of enhanced osteoclastogenesis; however, this increased osteoclastogenic potential was not due to a cell-autonomous effect of Fcgr2b À / À osteoclast precursor cells, but, rather, was observed only when Fcgr2b À / À cells were cultured in the mouse serum. Osteoclastogenesis in Fcgr2b À / À cells was increased further when cultured with serum from Fcgr2b À / À mice, which contains a high level of IgG ICs. This led us to hypothesize that serum IgGs would stimulate osteoclastogenesis when FcgRIIB is downregulated. Indeed, depletion of IgGs abrogated the ability of the mouse serum to increase the osteoclastogenesis in Fcgr2b À / À cells. Size-exclusion chromatography indicated that polymerized IgGs, rather than monomeric IgGs, were responsible for this osteoclastogenic effect. We also confirmed in vitro the positive effect of ICs from Fcgr2b À / À mice on osteoclastogenesis on Fcgr2b À / À cells and the systemic administration of IgG ICs caused bone loss in Fcgr2b À / À mice without any involvement of an inflammatory reaction. Enhanced osteoclastogenesis is not observed in Fcgr2b À / À Fcer1g À / À DKO mice, thereby showing that enhanced FcRg signalling is causative of the phenotype. Thus, ICs acts on the positive FcgRs expressed in osteoclast precursor cells in order to stimulate osteoclastogenesis.
IgG subclasses are produced in a manner depending on the host conditions and display distinct functional activity as a result of binding to FcgRs with distinct affinities. The effects of each subclass on the immune reactions have been extensively studied 9,31 ; however, it has remained essentially enigmatic in the context of bone cell regulation. Our in vitro experiments demonstrated that the IgG1 ICs stimulate osteoclastogenesis only in Fcgr2b À / À cells and that such stimulatory effect is abolished by the knockdown of FcgRIII expression, suggesting that IgG1 IC-induced osteoclastogenesis is mainly mediated by FcgRIII and the downregulation of FcgRIIB is required.
On the other hand, IgG2a and IgG2b ICs enhance osteoclastogenesis even in wild-type cells and the effect of the IgG2a ICs is eliminated by the knockdown of FcgRI or FcgRIV expression, suggesting that FcgRI and FcgRIV also transduce osteogenic signals despite their low expression level when stimulated by excessive IgG2 ICs. The potent osteoclastogenic ability of IgG2 ICs may be explained by a high A/I ratio (FcgRI and IV/FcgRIIB) of IgG2 than that (FcgRIII/FcgRIIB) of IgG1 (ref. 20). Consistent with these results, the local injection of IgG1 ICs into the calvaria results in bone destruction only in Fcgr2b À / À mice, while the injection of IgG2a ICs does so even in wild-type mice.
In summary, we have demonstrated in mice that FcgRIIB and FcgRIII are predominantly expressed in osteoclast precursor cells under physiological conditions. In normal mouse serum, the total amount of IgG ICs is extremely low, and most of the IgGs present  is IgG1, which binds to FcgRIII and FcgRIIB with an extremely low A/I ratio 20 . Therefore, the IgG1-induced ITAM activation through FcgRIII is counterbalanced by FcgRIIB. Under pathological conditions, such as autoimmune inflammation, the total amount of IgG1 ICs increases and the ratio between FcgRIIB and FcgRIII is tilted in favour of FcgRIII (Fig. 6f), possibly by inflammatory mediators such as C5a and IFN-g (ref. 32). In addition, the IgGs were less sialylated in cases of inflammation ( Supplementary Fig. 6a). As a result, the binding of the IgG1 ICs to FcgRIII powerfully stimulates the ITAM signal in the absence of the inhibition by FcgRIIB. Activation of the autoimmune reaction is also associated with the production of IgG2 ICs and increased expression of FcgRIV ( Fig. 6) 17 , further facilitating the ITAM signal. Thus, the strength of the ITAM signal is dependent on both the availability of the IgG ICs and the expression patterns of the activating FcgRs and inhibitory FcgRIIB (Fig. 7b).
The It will be important to take into consideration the difference in FcgR systems between human and mice in understanding and applying to human diseases.
Consistent with our findings, clinical observations have indicated that the incidence and severity of erosive bone damage in RA correlate with the titres of anti-IgG antibodies (rheumatoid factor) and anticitrullinated peptide antibody (ACPA) 34 . Interestingly, ACPA was also shown to activate osteoclastogenesis independently of FcgRs 29 . Monocytes from patients with RA and SLE displayed an increased expression of activating receptor FcgRIIIA/B 35,36 . Patients with RA carrying the FcgRIIIA158V allele, which has a higher affinity for the human IgG1 than another allele (FcgRIIIA158F), were associated with a more severe bone erosion 37 . In addition, serum IgGs from patients with a variety of different autoimmune diseases were less sialylated 25,38 .
Since genetic ablation of Fcer1g alone does not cause any obvious abnormality in the bone phenotype, the contribution of FcRg-mediated ITAM signalling to osteoclast formation has been a mystery. FcRg-associated receptors in osteoclasts can be divided into two types: Ig-like receptors, such as OSCAR and PIR-A, and FcgRs, such as FcgRI, FcgRIII and FcgRIV. Under physiological conditions, FcRg-associated FcgRs are virtually nonfunctioning, as described above, the bone phenotype of Fcer1g À / À mice would be expected to be caused by an impairment of Ig-like receptor-mediated signals rather than FcgR-mediated signals. However, the loss of FcRg signalling mediated by Ig-like receptors can be compensated for by the enhancement of DAP12-mediated ITAM signalling (Supplementary Fig. 9). In fact, introducing a combined deficiency of DAP12 into Fcer1g À / À mice results in a much more severe form of osteopetrosis than Tyrobp À / À mice, showing that the lack of FcRg is compensated for by a DAP12dependent mechanism. It is likely that DAP12-mediated signalling is upregulated because of a lifelong impairment of FcRg in Fcer1g À / À mice. In contrast, osteoclastogenesis is impaired in FcgRIII high cells owing to the low availability of FcRg because upregulation of DAP12 signalling may not efficiently occur in wild-type mice in which the FcRg-mediated signals are still intact. It is also interesting to note that PLCg phosphorylation is activated without RANKL stimulation under certain conditions in which FcgR signalling is enhanced (Figs 2a,5b and 6e and Supplementary Fig. 3c,g).
IC-induced osteoclastogenesis may contribute to both osteoporosis and the local bone erosion that occurs in various autoimmune diseases including RA and SLE, chronic inflammatory diseases 39,40 , multiple myeloma 41 and monoclonal gammopathy of undetermined significance 42 , which are all characterized by a high IgG production. The direct effect of the ICs on osteoclasts may also explain, at least in part, the bone-protective effect of the B-cell-depleting anti-CD20 antibody rituximab in RA patients 43,44 . The observation that ACPA-producing plasma cells accumulate in the inflamed synovium in RA suggested a correlation between bone erosion and the locally produced IgGs 45 . In contrast, the circulating ICs may make a significant contribution to the systemic bone loss commonly observed in various diseases associated with hypergammaglobulinaemia. It will be important to evaluate the level of IgG ICs in the serum as well as inflamed synovium in RA 46 . This study thus has shed light on the direct regulatory role of ICs in bone metabolism and may provide a molecular basis for future therapeutic approaches to inflammatory bone disease.

Methods
Mice and analysis of the bone phenotype. Mice were kept under specific pathogen-free conditions, and all animal experiments were performed with the approval of the Institutional Review Board at the University of Tokyo. Generation of Fcgr2b À / À , Fcgr3 À / À , Fcer1g À / À and Tyrobp À / À mice have been described elsewhere 23,[47][48][49] . All mice were backcrossed with C57BL/6 mice more than 10 times. Fcgr2b À / À Fcer1g À / À mice were purchased from Taconic and backcrossed more than eight times with C57BL/6 mice. Twelve-week-old sex-matched mice and their littermate controls were used for the analysis of the bone phenotype unless otherwise mentioned. Fcgr3 À / À mice were previously reported to have no major bone defect, possibly because this was based on a comparison among various knockout strains without littermate controls 17 . Fcgr2b À / À mice were analysed at the age of 12 weeks, unless otherwise described, when they had an increase in IgG production, but had not developed any obvious autoimmune symptoms such as glomerulonephritis, oedema and weight loss 21 . Femurs and tibiae were subjected to mCT and histomorphometric analyses, respectively. mCT scanning was performed using a ScanXmate-A100S Scanner (Comscantechno). Three-dimensional microstructural image data were reconstructed and structural indices were calculated using the TRI/3D-BON software (RATOC). The bone mineral was calculated using the TRI/3D-BON-BMD-PNTM software (RATOC). For histological analyses, tibiae were dehydrated and embedded in glycol methacrylate. Longitudinal sections, 3-mm thick, were cut on a Microtome and stained with Toluidine blue or TRAP. Static parameters of bone formation and resorption were measured in a defined area between 0.3 and 1.2 mm from the growth plate by using an OsteoMeasure bone histomorphometry system (Osteometrics). For static parameters we measured the osteoblast surface per bone surface, the number of osteoclasts per bone perimeter, the osteoclast surface per bone surface and the eroded surface per bone surface. For dynamic histomorphometry, mineral apposition rate and mineralized surface per bone surface were measured under ultraviolet light and used to calculate bone formation rate with a surface referent.
Osteoclast differentiation in vitro. Nonadherent bone marrow cells were cultured in a-MEM (Invitrogen) with 10% FBS (Sigma) containing 10 ng ml À 1 M-CSF (R&D Systems) for 2 days to obtain BMMs. Subsequently, the BMMs were cultured in the presence of 10 ng ml À 1 M-CSF and 25 ng ml À 1 RANKL (Pepro Tech) with 10% FBS or 5% mouse serum for 3 days. RANKL and M-CSF were used at these concentrations throughout the paper unless otherwise described. Co-culture of BMMs and osteoblasts was performed in the presence of 10 nM 1a,25dihydroxyvitamin D3 (Sigma) with 10% FBS or 5% mouse serum. The differentiation into osteoclasts was evaluated by counting TRAP-positive multinucleated (more than three nuclei) cells. Osteoclastogenesis of the BMMs from CIA mice (3 weeks after the secondary immunization) was compared with those from ageand sex-matched nonimmunized DBA/1J mice.
Treatment with plate-bound IgGs and IgG ICs. Mouse monoclonal IgG subclassspecific antibodies were purchased from BD Biosciences. Immunoglobulins were immobilized on the culture plates in PBS at 20 mg ml À 1 at 4°C for 12 h, and then the plates were washed extensively to remove all unbound IgGs before cells were seeded. Soluble ICs were prepared by mixing aliquots of mouse serum or IgGdepleted mouse serum with a goat F(ab 0 ) 2 a-mouse k (Southern Biotechnology) at 30 mg ml À 1 for 12 h. Another type of soluble ICs were prepared by incubating TNP-BSA (Wako) with a-TNP IgG1 antibody (BD Biosciences) at a molar ratio of 1:20 in PBS overnight at 4°C. TNP-BSA alone was used as control. The ICs were added to BMMs at 300 ng ml À 1 together with RANKL.
IC-induced bone loss. Eight-week-old female mice were administered with a local calvarial injection of soluble IC (TNP-BSA with a-TNP IgG antibody) containing each IgG subclass at 100 mg or TNP-BSA alone. The IC-induced bone destruction was observed within 2 days of administration. After 3 days, calvarial tissues were embedded and frozen in 5% carboxy-methylcellulose sodium, and serial sections were stained for TRAP (with haematoxylin). Parameters such as the osteoclast number, eroded surface and the length of the inflammatory cell layer were determined. For the effect of ICs on systemic bone loss, 3-week-old female Fcgr2b À / À mice were given weekly intravenous injections of 300 ml of TNP-BSA alone or together with a-TNP IgG1 for 3 weeks. Three days after the last injection, bone analysis was performed.
Induction of CIA. Eight-week-old male DBA/1J mice were immunized with an emulsion that consisted of 50 ml of chicken type II collagen (Sigma-Aldrich, 4 mg ml À 1 ) and 50 ml of adjuvant given intradermally into the base of the tail at two sites. For immunization, we used complete Freund's adjuvant (Difco Laboratories). Three weeks after the primary immunization, mice were challenged with the collagen/incomplete Freund's adjuvant emulsion (Difco Laboratories). We judged the development of arthritis in the joint using the following criteria: 0, no joint swelling; 1, swelling of one finger joint; 2, mild swelling of the wrist or ankle; 3, severe swelling of the wrist or ankle. The scores for all the fingers of the forepaws and hind paws, wrists and ankles were totalled for each mouse (with a maximum possible score of 12 for each). Three weeks after the secondary immunization, serum and bone marrow cells were isolated from the mice with a score of more than 9.
Statistical analysis. Statistical analysis was performed using the unpaired twotailed Student's t-test (*Po0.05; **Po0.01; ***Po0.001; n.s., not significant; n.d., not detected in all figures). All data are expressed as the mean±s.e.m. Results are representative examples of more than four independent experiments. We estimated the sample size considering the variation and the mean of the samples. We tried to reach the conclusion using as small a size of samples as possible. We usually excluded samples if we observed any abnormality in terms of size, weight or apparent disease symptoms in mice before performing experiments. However, we did not exclude animals here, as we did not observe any abnormality in the present study. Neither randomization nor blinding was carried out in this study. Statistical tests are justified as appropriate for every figure, and the data meet the assumptions of the tests.