Gα13 negatively controls osteoclastogenesis through inhibition of the Akt-GSK3β-NFATc1 signalling pathway

Many positive signalling pathways of osteoclastogenesis have been characterized, but negative signalling pathways are less well studied. Here we show by microarray and RNAi that guanine nucleotide-binding protein subunit α13 (Gα13) is a negative regulator of osteoclastogenesis. Osteoclast-lineage-specific Gna13 conditional knockout mice have a severe osteoporosis phenotype. Gna13-deficiency triggers a drastic increase in both osteoclast number and activity (hyper-activation), mechanistically through decreased RhoA activity and enhanced Akt/GSK3β/NFATc1 signalling. Consistently, Akt inhibition or RhoA activation rescues hyper-activation of Gna13-deficient osteoclasts, and RhoA inhibition mimics the osteoclast hyperactivation resulting from Gna13-deficiency. Notably, Gα13 gain-of-function inhibits Akt activation and osteoclastogenesis, and protects mice from pathological bone loss in disease models. Collectively, we reveal that Gα13 is a master endogenous negative switch for osteoclastogenesis through regulation of the RhoA/Akt/GSK3β/NFATc1 signalling pathway, and that manipulating Gα13 activity might be a therapeutic strategy for bone diseases.

O steoclasts are the principal, if not exclusive, boneresorbing cells [1][2][3] . A balance between bone formation by osteoblasts and bone resorption by osteoclasts is critical to maintain normal bone density and mineral homeostasis [1][2][3] . Osteoclastogenesis is initiated by signals transmitted by the receptor activator of nuclear factor-kB (RANK) ligand (RANKL) and macrophage colony-stimulating factor (M-CSF) [1][2][3][4] . RANKL and M-CSF, binding to their receptors, RANK and cFms, respectively, on the surface of osteoclast precursors, activates many key transcription factors including nuclear factor of activated T-cells, c1 (NFATc1) and CCAAT/enhancer-binding protein alpha (C/EBPa) [1][2][3][4] . Our previous work on cathepsin K 5,6 , Atp6i (refs 7,8), C/EBPa 9 and RGS10 (ref. 10) has contributed to the discovery of positive regulatory machinery controlling osteoclast differentiation and function. Notably, excessive osteoclast activity is responsible for various bone diseases including osteoporosis, rheumatoid arthritis and Paget's disease [1][2][3] . Thus, we anticipate that hyperactivation of osteoclasts should be antagonized by intrinsic negative regulators, which have received limited attention, to maintain normal bone homeostasis. A decrease in the expression of negative osteoclast regulators might trigger the bone loss associated with many pathological bone disorders.
Guanine nucleotide-binding protein subunit a13 (Ga13; encoded by Gna13) belongs to the G12 subfamily of the G protein superfamily. Ga13 regulates cell cytoskeleton organization by regulating the RhoGEF-RhoGTPase signalling pathway 11,12 . Here we utilize a genome-wide screening strategy and characterize Ga13 as an intrinsic negative regulator of osteoclast formation and activity. Absence of Ga13 favours osteoclast formation and enlarges osteoclast size, leading to an osteopenia phenotype in mice. On the other hand, constitutively active Ga13 blocks the formation of multinucleated osteoclasts and their bone resorptive activity. We reveal that Akt-mediated signalling, regulated by RhoA, is critical for the function of Ga13 in osteoclasts.

Results
Gna13 silencing promotes osteoclastogenesis. Given the limited attention attributed to the investigation of negative regulators of osteoclastogenesis, we performed a genome-wide screening to identify factors that negatively regulate osteoclast formation ( Fig. 1). Towards this end, we compared gene expression in human blood monocyte-derived osteoclasts with that in their precursors, and found that Gna13 was not only drastically induced by RANKL, but also its expression pattern was comparable to various osteoclast genes (Fig. 1a), and much higher than the genes encoding several other G proteins (Fig. 1b). Moreover, gene expression analysis confirmed that Gna13 was highly expressed in osteoclasts and osteoclast-like cells derived from MOCP-5, an osteoclast precursor cell line that was generated in our lab 13 , as compared with several tissues, such as heart, kidney, lung and intestine (Fig. 1c). Furthermore, using murine bone marrow monocytes (BMMs), which are widely used as primary osteoclast precursors, we demonstrated that whereas Ga13 expression was only mildly induced by M-CSF (B2-fold at both mRNA and protein level), its expression was strongly induced by combined stimulation with M-CSF and RANKL (B6-fold at mRNA level and 8-fold at protein level) (Fig. 1d,e). These results indicated that Ga13 might have important roles in osteoclasts. Hence, we then silenced Ga13 expression by lentiviral-mediated expression of short hairpin RNA against Gna13 in BMMs. The knockdown efficiency (B90%) was confirmed by immunobloting as compared with control cells (Fig. 1f). Interestingly, our data showed that Ga13 silencing strongly increased osteoclast formation and also drastically enhanced osteoclast size (Fig. 1g,h). These results suggested that Ga13 might be a negative regulator of bone resorption.
Depletion of Gna13 causes osteopenia in mice. To further investigate the role of Ga13 in osteoclast formation and activity, we generated knock-out mice through specific deletion of Gna13 in the osteoclast lineage (Fig. 2). Mice bearing loxP sites encompassing the Gna13 exon2 (Gna13 f/f mice) 14 were crossed with those expressing Cre recombinase driven by the lysozyme M promoter (LysM-Cre mice) or the Cathepsin K promoter (Ctsk-Cre mice). The offspring were intercrossed to get Gna13 f/f LysM-Cre mice or Gna13 f/f Ctsk-Cre mice and wild-type (WT) mice. Mouse genotypes were confirmed by PCR ( Supplementary  Fig. 1a-c). The deletion of Ga13 in Gna13 f/f LysM-Cre BMMs and Gna13 f/f Ctsk-Cre pre-osteoclasts was confirmed by quantitative PCR and immunoblotting (Fig. 2a,b). While LysM-Cre deletes Gna13 in the osteoclast precursors (Fig. 2a,b), Ctsk-Cre works at a late stage of osteoclast differentiation ( Supplementary Fig. 1d).
Two-month-old WT and Gna13 f/f LysM-Cre or Gna13 f/f Ctsk-Cre mice have no obvious gross morphological changes. As assessed by X-ray analysis, skeletal mass was decreased in distal femurs of both male and female Gna13 f/f LysM-Cre mice as compared with those of WT littermates (Fig. 2c). Quantitative microtomography (m-CT) analysis showed that Gna13 f/f LysM-Cre mice exhibited B30% reduction in bone volume/tissue volume (BV/TV) and trabecular number (Tb.N) as well as B40% increase in trabecular space (Tb.Sp), B50% decreased in bone mineral density (BMD) as compared with WT littermates (Fig. 2d,e). Tartrate-resistant acid phosphatase (TRAP) activity was markedly increased in the primary spongiosa of 2-month old Gna13 f/f LysM-Cre mouse femur as compared with that of WT littermates ( Fig. 2f-h). Histomorphometric parameters of distal femurs showed that, in the trabecular bone area, Gna13 f/f LysM-Cre mice had B3-fold increased in osteoclast numbers per bone surface (N.Oc/BS) as compared with WT littermates (Fig. 2g). Interestingly, osteoclast number was even more drastically increased in the cortical bone area in the mutant mice ( Fig. 2f,g). Similarly, X-ray and m-CT analyses showed that skeletal mass was decreased in distal femurs of 2-month-old Gna13 f/f Ctsk-Cre mice (Fig. 2i,j, Supplementary Fig. 2c). TRAP staining showed B6-fold increase in osteoclast numbers in femoral sections of Gna13 f/f Ctsk-Cre mice as compared with that of WT controls (Fig. 2k, Supplementary Fig. 1d-f). Histomorphometry based on Trichrome staining showed that trabecular bone number was reduced without much change of osteoblast number and osteoblast surface per bone surface in the Gna13 f/f Ctsk-Cre mice, confirming that the reduced bone density was not caused by decreased bone formation ( Supplementary Fig. 2a,b). In fact, as measured by calcein labelling ( Supplementary Fig. 2d), mineral apposition rate was increased by about 35% in Gna13 f/f Ctsk-Cre mice. Consistently, serum alkaline phosphatase (ALP) was also increased by about 35% in Gna13 f/f Ctsk-Cre mice ( Supplementary Fig. 2e). The accelerated bone formation and remodelling might be caused by robust osteoclast formation. These results show that osteoclast-lineage-specific Gna13 deletion promotes osteoclastic bone resorption leading to lower bone density in vivo.
Gna13-depletion hyper-activates osteoclast. Given our data that Gna13 deficiency in vivo led to an increase in osteoclast development (Fig. 2), we examined the effects of using lower doses of RANKL in osteoclastogenesis under Gna13 deficiency. Unlike WT BMMs, Gna13-deficient cells formed osteoclasts when attended by permissive RANKL dose (one tenth of the optimum 10 ng ml À 1 RANKL) (Fig. 3a) and formed 25% more osteoclasts at 10 ng ml À 1 RANKL (Fig. 3a,b). Further, BMMs deficient in Gna13 generated osteoclast earlier (Bday 3) than WT BMMs (Bday 5) ( Supplementary Fig. 3a,b). Gna13 deficiency not only increased the sensitivity of BMMs to RANKL, but also increased the size of osteoclasts (Fig. 3a,b). More osteoclasts with 3-8 nuclei or 413 nuclei were observed in the Gna13-deficient group (increased by 50% and 100%, respectively) (Fig. 3b). Consistently, gene expression analysis revealed that the expression of the osteoclast genes encoding NFATc1, Ctsk, Acp5, Atp6v0d2 and dc-stamp was higher in Gna13 f/f LysM-Cre, but not in WT, BMMs cultured with M-CSF and RANKL for 3 days (Fig. 3c). Importantly, among the aforementioned genes, dc-stamp and Atp6v0d2 are known to have roles in osteoclast fusion 15,16 , indicating that Ga13 negatively regulates osteoclast formation by also promoting the fusion of mononucleated osteoclasts into multinucleated osteoclasts.
To further analyse the role of Ga13 in osteoclastic bone resorption, we performed in vitro bone resorption assays and analysed bone resorption pits by wheat germ agglutinin (WGA) staining and scanning electron microscope (Fig. 3d). Data showed that the total resorption area by Gna13 f/f LysM-Cre osteoclasts was significantly higher than that of WT osteoclasts (Fig. 3e). Collagen I C-terminal telopeptide (CTX-I), released during bone resorption 17 , were drastically increased in Gna13 f/f LysM-Cre cell culture medium (Fig. 3f). In addition, WGA-fluorescein isothiocyanate (FITC) staining showed that pits resorbed by Gna13 f/f LysM-Cre osteoclasts are also larger (B14-fold) and deeper (B3-fold) than those by WT cells (Fig. 3g,h, Supplementary Fig. 4). Collectively, the results showed that Ga13 negatively regulated osteoclastic bone resorption.
Next, we co-stained the osteoclasts on bone slides with Rhodamine-conjugated-Phalloidin (red), to analyse F-actin ring formation, a critical structure of mature osteoclasts 18 , anti-Ctsk antibody (green) and Hoechst (blue) (Fig.3i,j). Data showed that Gna13-deficient osteoclasts exhibited 1.5-fold increased F-actin ring formation (Fig. 3i,k), and the average size of F-actin rings was also about threefold larger in these cells than control cells (Fig. 3j,l).
Furthermore, immunostaining showed a more polarized location of Ctsk in Gna13 f/f LysM-Cre cells (Fig. 3j), indicating a change in Ctsk transportation and secretion. We thus analysed intracellular and medium Ctsk by western blot (Fig. 3m,n). Data showed a slight increase (B1.5-fold) in intracellular Ctsk level in Gna13 f/f LysM-Cre osteoclasts (Fig. 3m). However, secreted Ctsk concentration in the culture medium by Gna13 f/f LysM-Cre osteoclasts was drastically elevated (B4-fold) (Fig. 3n). These results demonstrate that Gna13-deficient osteoclasts had enlarged actin ring with increased Ctsk secretion, both of which would contribute to enhanced bone resorption besides enhanced osteoclast formation.
Ga13-RhoA brakes Akt-GSK3b-NFATc1 signalling. RANKL and M-CSF orchestrate multiple signalling pathways to promote osteoclast formation 1-3 . Expression of RANKL receptor (RANK) and M-CSF receptor (cFms) were similar in WT and Gna13-deficient cells ( Supplementary Fig. 3j,k). To analyse signal transduction in response to RANKL or M-CSF stimulation (including Akt, p38, JNK and Erk activation), BMMs were starved for 5 h and then stimulated with RANKL or M-CSF for 5-60 min ( Fig. 4a,b). Akt phosphorylation was enhanced in Gna13 f/f LysM-Cre cells by either RANKL or M-CSF as compared with WT BMMs (Fig. 4a,b). Nonetheless, activation of the p38, JNK and Erk signalling pathways were similar between WT and Gna13 f/f LysM-Cre cells ( Supplementary Fig. 5).
Given the drastic increase in Akt activation, we then applied the Akt inhibitor MK2206 2HCl to attenuate Akt activity, which inhibited osteoclastogenesis in a dose-dependent manner (Fig. 4c-e). Notably, Gna13-deficient cells were more resistant to the Akt inhibitory effect on osteoclastogenesis. In the presence of 0.1B0.2 mM MK2206 2HCl (Fig. 4d,e), Gna13-deficient cells formed osteoclasts at normal level comparatively to WT cells. The results indicate that Gna13 deficiency promotes osteoclastogenesis by increasing Akt activity.
It has been established that the PI3K/Akt-GSK3b-NFATc1 signalling cascade is critical for osteoclast differentiation 20,21 . Specifically, Akt can inhibit GSK3b, which can promote NFATc1 translocation from the nucleus into the cytoplasm and thereby abrogate osteoclast differentiation. Thus, we characterized GSK3b phosphorylation level and NFATc1 nuclei translocation in WT and Gna13 f/f LysM-Cre cells. We found that GSK3a/b phosphorylation was hyper-activated in Gna13 f/f LysM-Cre monocyte/macrophage in response to RANKL and M-CSF (Fig. 4o). NFATc1 expression increased during osteoclastogenesis in WT and Gna13 f/f LysM-Cre cells, and more quickly in Gna13 f/f LysM-Cre cells (Fig. 4p). Consistently, while similar amounts of NFATc1 were observed in the cytoplasm, much more NFATc1 was observed in the Gna13 f/f LysM-Cre nucleus than WT (Fig. 4q). The results demonstrated that Ga13 mediates Akt-GSK3b-NFATc1 signalling to promote osteoclast gene expression.
Furthermore, we infected RAW264.7 cells with the same retrovirus we used in BMMs in Fig. 5m,o. Similarly, Ga13CA overexpression inhibited osteoclast-like cell formation ( Fig. 5o left panels; 5p), which could be rescued by AktCA overexpression (Fig. 5o right panels; 5p). Akt activation was greatly decreased in RAW264.7 cells overexpressing Ga13CA (Fig. 5q,r), while the activation of p38 and Erk remains unchanged ( Supplementary  Fig. 7b). The overexpression of Ga13CA and AktCA was confirmed by western blot (Fig. 5s,t). The results indicated that Ga13CA also inhibited osteoclast-like cells formation from the RAW264.7 cells by inhibiting Akt phosphorylation.
Ga13CA protects mice from pathological bone loss. To investigate Ga13 clinical implication potential, we overexpressed Ga13CA locally in mice using adeno-associated virus (AAV) gene expression to investigate its effects in protecting against bone degradation. AAV vectors have advantages compared with many other viral and non-viral based gene delivery platforms in both function and safety for correcting genetic-based diseases 23 . To test the use of AAV-Ga13CA, we utilized three disease models of bone loss: the human tumour-necrosis factora (hTNFa)-transgene expressing mouse model of autoimmune arthritis 24 (named TNFa-RA) (Fig. 6), ovariectomized (OVX) animal model of osteoporosis, (Fig. 7a-d) and calvaria-adjacent lipopolysaccharide (LPS)-injection mouse model of osteolysis (Fig. 7e,f) 25 . Successful infection of AAV in vivo was monitored by yellow fluorescent protein (YFP) expression ( Supplementary  Figs 8 and 9a). Successful overexpression of G13CA in vivo was confirmed by immunofluorescence staining ( Supplementary  Fig. 9b).
TNFa-RA mice were accompanied by excessive osteoclast formation and bone destruction (Fig. 6). Local administration of AAV-Ga13CA to ankles greatly reduced bone loss, as assessed by X-ray in the TNFa-RA mice as compared with those injected with AAV-YFP (Fig. 6c,d). Not surprisingly, osteoclast number was dramatically decreased in TNFa-RA mice with local injection of AAV-Ga13CA to ankles (Fig. 6g,h). Unexpectedly, local administration of AAV-Ga13CA also relieved ankle swelling (Fig. 6a,b), cartilage destruction ( Fig. 6e lower panels, Fig. 6f) and inflammation cell infiltration (Fig. 6e upper and middle panels,  Fig. 6f) in TNFa-RA mice. Hence, reduced bone loss and decreased osteoclast number might be a combined effect of Ga13CA and inhibited inflammation.
OVX-induced bone loss and osteoclast number increased due to estrogen depletion ( Fig. 7a-d, sham þ PBS group compared with OVX þ YFP group). OVX mice were subjected to calvariaadjacent subcutaneous injection of AAV-YFP or AAV-Ga13CA. As assessed by X-ray and u-CT, application of AAV-Ga13CA reduced OVX-induced bone loss (Fig. 7a,b). As assessed by TRAP staining of whole calvarial and calvarial sections, AAV-Ga13CA dramatically reduced OVX-induced osteoclast number increase (Fig. 7c,d) as compared with YFP group.
In the LPS-injection model, both WT and Gna13 f/f LysM-Cre mice were subjected to calvaria-adjacent subcutaneous injection of LPS or PBS control (Fig. 7e,f). Compared with local injection of PBS, LPS induced a dramatic increase in osteoclast number and bone resorption (Fig. 7e,f). The absence of Ga13 further promoted the induction of osteoclastogenesis and bone loss in both PBS and LPS treated mice (Fig. 7e,f). Local administration of AAV-Ga13CA to calvaria had a marked therapeutic effect on osteoclast formation and bone destruction by LPS, compared with local injection of AAV-YFP control (Fig. 7e,f). These results confirm that it is possible to overexpress Ga13CA in vivo to target pathologic bone loss.
Full images of western blots are presented in Supplementary  Fig. 10.

Discussion
We proposed a model that Ga13-RhoA antagonizes osteoclast formation and activity by attenuating the Akt-GSK3b-NFATc1   signalling axis (Fig. 8). Briefly, as a downstream of RANK and c-Fms, Akt phosphorylates and inactivates GSK3b, which has a role in NFATc1 nuclei exportation, so as to promote osteoclast differentiation; Meanwhile, Ga13 activates RhoA which in turn inhibits Akt phosphorylation and activity, so as to antagonize the over-activation of osteoclast (Fig. 8). The critical role of PI3K-Akt activity in osteoclast differentiation and activation is welldocumented 20,21,[26][27][28][29] . Although several endogenous factors that oppose osteoclast differentiation were described, most are downregulated under osteoclastogenic condition. In contrast, Ga13 is induced by the combined stimulation of RANKL and M-CSF, and in turn inhibit osteoclast differentiation so as to avoid over-activated osteoclast and excessive bone loss. Hence, our data reveal that Ga13 is an intrinsic control mechanism during osteoclastogenesis. In addition, we comprehensively study and report a signalling cascade (Ga13-RhoA-Akt-GSK3b-NFATc1) that negatively regulates osteoclastogenesis and osteoclast function. Most importantly, our results from both loss-of-function and gain-of-function strategies proved that Ga13 is a key negative regulator and main switch in osteoclastogenesis. It was reported that RhoA activity is essential for podosome formation and osteoclastic bone resorption 30 . On the other hand, it was also reported that increased RhoA activity destabilizes the sealing zone in osteoclasts, and inhibition of RhoA stabilizes the sealing zone correlating with acetylation and stabilization of the microtubule network in these cells [31][32][33][34] . Our study showed that inhibition of RhoA activity through Ga13 knockout, RhoA inhibitor and RNA interference favours osteoclastogenesis, while RhoA activator inhibits osteoclastogenesis. Hence, these studies suggest that RhoA is the downstream effector of Ga13 to negatively regulate osteoclastogenesis. Our finding is underscored by genetic evidences that have suggested an important role for the RhoGEF-RhoGTPase (for example, Arhgef3-RhoA) pathway in osteoporosis [35][36][37][38] . Genome-wide linkage studies have identified chromosome region 3p14-p21, in which RHOA gene was located, as a quantitative trait locus for BMD 35,36 . One RHOA singlenucleotide polymorphism (SNP) rs17595772, and one ARHGEF3 SNP were reported to be significantly associated with decreased BMD) in postmenopausal women 37,38 . However, the mechanisms underlying the correlation between RhoGEF-RhoGTPase pathway and BMD remain to be explored.
Current anti-resorptive agents (for example, bisphosphonates and denosumab) are effective but far from ideal. The major problem of bisphosphonates is their permanent deleterious affect on normal bone remodelling. Agents preventing bone resorption by inhibiting late differentiation of osteoclasts without affecting normal bone remodelling are more desired, which are unlikely to interfere with the coupling of osteoblastic bone formation to osteoclastic bone resorption, critical for maintaining normal bone homeostasis 39 . Here we showed that overexpression of Ga13CA markely inhibits osteoclast formation and bone resorption both in vitro and in vivo. Importantly, although multi-nuclear cell formation was blocked, Ga13CA over-expressing monocytes were TRAP þ stained (Fig. 5a), indicating that Ga13CA regulated osteoclast differentiation at a late stage and could be a good target for bone loss without affecting normal bone remodelling. Interestingly, Ga13 might have dual functions in inflammatory disease models (TNFa-RA and LPS-injection models), by concurrently inhibiting osteoclast differentiation and inflammation so as to achieve the ultimate outcome of bone loss protection. Hence, activating Ga13, an endogenous osteoclast inhibitor, may serve as a novel therapeutic approach targeting bone loss in various bone pathological states.
LysM-Cre transgenic mice were purchased from the Jackson laboratory. Ctsk-Cre knock-in mice were provided by Dr S. Kato (University of Tokyo, Tokyo, Japan) 40 .
GeneChip analysis. Human peripheral blood mononuclear cells were cultured for 7 days with human RANKL and human M-CSF. GeneChip data were analysed using Affymetrics scanner and accompanying gene expression software.
Radiographic analysis.. For X-ray analysis, excised 2-month-old mouse femurs were scanned using a high-resolution soft X Histology. Histology was performed as previously described 40,41 . Briefly, the femurs of 2-month-old mice were fixed with 70% ethanol followed by plastic embedding and Golden trichrome staining, or with 10% neutral buffered formalin followed by decalcification in 10% EDTA for 2 weeks, paraffin embedding and TRAP staining using Phosphatase, Leukocyte Acid (TRAP) kit (Sigma-Aldrich, 387A-1KT). Osteoclastic and osteoblastic perimeters were measured and analysed using Osteomeasure in a blinded fashion.
In-vivo calcein labelling. Calcein labelling was previously described 41 . Briefly, 2-month-old mice were intraperitoneally injected with 20 mg kg À 1 of calcein in a 2% sodium bicarbonate solution, 8 days and 2 days before killing of mice. Calvarias were fixed in 4% PFA, soaked in 30% glucose in PB, embedded in OCT and frozen sectioned. Mineral apposition rate is the distance between the midpoints of the two labels divided by the time between the midpoints of the interval.
Serum ALP assay. Two-month old mouse serum was collected after 6-hour fasting, and the serum ALP activity was detected and quantified using Alkaline Phosphatase Assay Kit (Colorimetric) (ab83369) according to the manufacturer's instructions.
In vitro osteoclast differentiation assay. Mature osteoclasts were generated as described 10 . Briefly, isolated BMMs from C57BL/6 mice were cultured in MEM a (pH 6.9) containing 10% FBS, 10 ng ml À 1 recombinant RANKL and 10 ng ml À 1 recombinant M-CSF for 5 days. Recombinant murine M-CSF and RANKL were obtained from R&D systems, Inc. Cell culture medium was obtained from Gibco, Life Technologies Corporation. Mature osteoclasts were characterized by staining for TRAP activity using a commercial kit (Sigma-Aldrich, 387A-1KT) and TRAP þ MNCs were enumerated per well in a 24-well plate.
In vitro bone resorption assay. For bone resorption assay, osteoclasts were seeded on bovine bone slides. Concentration of bovine cross-linked C-telopeptide of type I collagen (CTX-1) in the medium was measured using CrossLaps for Culture ELISA (CTX-I) kit (Immunodiagnostic Systems Limited) following the manufacturer's instructions. Bone resorption pits were sonicated in PBS, stained with 2 ug ml À 1 WGA-lectin (Sigma-Aldrich, L-3892) for two hours and then using DAB peroxidase (horseradish peroxidase) Substrate Kit (Vector laboratories, SK-4100) (ref. 42). To analyse bone resorption pits depth and volume, bone slides were sonicated in PBS, stained with 2 ug ml À 1 FITC-WGA (Sigma-Aldrich, L-4895) staining for 2 h and visualized by confocal microscopy 43 .
RhoA activation assay. Pre-osteoclasts (derived from BMMs with 2-day treatment of RANKL and M-CSF) were starved for 4 h and stimulated by RANKL and M-CSF for different times. Cell lysates was harvested and RhoA-GTP activity was analysed using commercial kits (Rho Activation Assay Biochem Kit, Cytoskeleton, Inc.) according to the manufacturer's instructions.
Constructs and virus production and infection. TRC lentiviral mouse Gna13 targeted short hairpin RNA (Clone ID: TRCN0000098146, target sequence 5 0 -CGCGATCAACACAGAGAACAT-3 0 ), TRC Lentiviral pLKO.1 Empty Vector and mouse Gna13 full-length cDNA were purchased from Open Biosystems, Thermo Fisher Scientific Inc. Lentiviral pLB vector and packaging plasmids (pCMV-VSV-G and pCMV-Dr8.92) were purchased from Addgene. pHR-EF-IRES-Bla (referred to as pHR in the following text) was a gift from Dr Ling Tian in Dr W. Timothy Garvey's lab and used for lentivirus-mediated overexpression 22 . Ga13 constitutively active form (Ga13 Q226L, referred to as Ga13CA in the text) full-length cDNA was purchased from Missouri cDNA Resource Center. Akt constitutively active form cDNA was purchased from addgene. Ga13CA was ligated into pHR-EF-IRES-Bla at the BamHI and XhoI sites and substituted the green fluorescent protein fragments, to construct pHR.Ga13CA vector. Retrovirus vector pMXs-ires-puro was a gift from Dr Xu Feng. A 3xFLAG sequence was inserted into pMXs-ires-puro to construct pMXs-3xFLAG-ires-puro vector. Ga13CA was ligated into pMXs-3xFLAG-ires-puro at the EcoRI and NotI sites to construct pMXs-3xFLAG-G13CA-ires-puro vector (referred to as pMXs-Ga13CA). AktCA was ligated into pMXs-ires-puro vector to construct pMXs-AktCA vector. For lentivirus production, lentivirus vectors (with 10% pLB vector or 10% pHR vector) and packaging plasmids were co-transfected into HEK-293T cells using a calcium phosphate co-precipitation method. The lentiviral supernatant was harvested 48-72 h post-transfection. For retrovirus production, pMXs vectors was transfected into 293GPG cells (gift from Dr Xu Feng) using a calcium phosphate co-precipitation method and retrovirus supernatant were harvested between 48-96 h (ref. 46). Ga13CA was ligated into pAAV-CMV vector from Invitrogen at EcoRI and SalI sites to produce pAAV-CMV-G13CA vector (referred to as pAAV-G13CA). Lentivirus and retrovirus titers were determined by transfecting HEK293T cells with serial dilutions of virus supernatant. AAV was produced and tittered according to the manufacturer's instructions (Life Technology, Inc). BMMs or RAW264.7 cells were transduced with virus supernatant in the presence of 8 mg ml À 1 polybrene (Sigma) for 24 h before induced with RANKL.
OVX-induced bone destruction and AAV-G13CA treatment. OVX or sham was performed on two-month-old female mice. One week later and two weeks later those mice were administered with a local calvarial injection of 30 ul AAV (titer ¼ 10 9-10 ml À 1 ) expressing YFP or G13CA. Mice were harvested 5 weeks after OVX operation and fixed in 4% PFA. Calvaria bone were analysed by X-ray, u-CT and whole-mount TRAP staining. Calvarial were also decalcified for 3 days, immersed in 30% sucrose overnight and then submitted to frozen section and TRAP staining and immunofluorescent staining.
LPS-induced bone destruction and AAV-G13CA treatment. Seven-week-old female mice were administered with a local calvarial injection of AAV expressing YFP or G13CA. Six days later those mice were administered with a local calvarial injection of LPS (sigma) at 25 mg kg À 1 body weight or PBS and analysed after 5 days following a previously described model 25 . For analysis, calvarial with skin tissue were fixed in 4% PFA for 6 h at 4°C, decalcified for 3 days, immersed in 30% sucrose overnight and then submitted to frozen section and TRAP staining.
TNFa-induced rheumatoid arthritis and AAV-Ga13CA treatment. 13-week old WT and human-TNFa-transgenic (hTNFtg) mice was injected with AAV (titer ¼ 10 9-10 ml À 1 ) expressing YFP or G13CA over ankles. The injection was performed four times every week successively. Photographic and radiographic analysis were performed before and 35 days after the first injection. Samples were harvested 35 days after the first injection, and Safranin O (SO) stain, H&E stain and TRAP stain were performed. Hind paw volume was quantified by water displacement method 47 . Bone destruction in X-ray was quantified by Larsen method in RA lesion area 48 . Cartilage destruction was assessed using safranin O staining and was measured as OARSI grade in ankle joint area 49 .
Statistics. Data represent mean±s.d. Analysis was performed using GraphPad Prism, version 5. Statistical significance was assessed using a two-tailed Student's t-test or ANOVA analysis, as indicated in the figure legend, considering a P valuer0.05 as significant.
Data availability. The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files.