Osteoblastic molecular scaffold Gab1 is required for maintaining bone homeostasis

Dysregulation of osteoblast and osteoclast activities and consequent loss of bone homeostasis are associated with a range of pathological diseases including osteoporosis (Karsenty and Wagner, 2002). Previous studies have shown that insulin and insulin-like growth factor 1 (IGF-1) have anabolic effects on osteoblasts through serine-threonine Akt kinase (AKT) activation to maintain bone mass and turnover (Fujita et al., 2004; Fulzele et al., 2007; Scheid and Woodgett, 2001; Thomas et al., 1996). Human beings and animals with diabetes frequently exhibit impaired bone mineralization and defective bone strength (Thrailkill et al., 2005). However, the molecular mechanisms underlying the osteoblastic responses to IGF-1 and insulin during osteoblast development and function have not been fully investigated.

The growth factor receptor-bound protein 2 (Grb-2)-associated binder (Gab) proteins are scaffolding adapter molecules, consisting of Gab1, Gab2 and Gab3, which are engaged in signal relay from cytokine and growth factor receptors (Liu and Rohrschneider, 2002). On stimulation, Gab proteins undergo rapid tyrosine phosphorylation, creating multiple docking sites to recruit and activate Src homology-2 (SH2) domain-containing proteins such as Shp2 (Src homology 2 containing tyrosine phosphatase) and phosphoinositide 3-kinase (PI3K) (Holgado-Madruga et al., 1996; Takahashi-Tezuka et al., 1998). For Gab1 and Gab2, the Gab-SHP2 interaction is considered to be an essential component for extracellular signal-regulated kinase (ERK) activation, whereas the association between Gab and the p85 subunit of PI3K is important for mediating the PI3K-AKT signaling pathway (Gu and Neel, 2003). Reports on Gabs knockout mice have clearly demonstrated the crucial roles of the Gab proteins during development and in maintenance of tissue homeostasis (Bard-Chapeau et al., 2005; Bentires-Alj et al., 2006; Cai et al., 2002; Gu et al., 2001; Itoh et al., 2000; Nakaoka et al., 2007; Nishida et al., 2002; Sachs et al., 2000; Wada et al., 2005). Gab1 is ubiquitously expressed within various tissue cell types, and plays a crucial role in transmitting key signals that control a diverse set of biological responses (Bard-Chapeau et al., 2005; Cai et al., 2002; Holgado-Madruga et al., 1997; Ingham et al., 2001; Itoh et al., 2000; Sachs et al., 2000; Yamasaki et al., 2003). Disruption of the Gab1 gene in mice results in embryonic lethality caused by impaired development of the heart, placenta and skin (Itoh et al., 2000). Gab2 knockout mice exhibit impaired allergic responses and osteopetrosis caused by defective osteoclastogenesis (Gu et al., 2001; Wada et al., 2005). Although crucial roles of Gab1 involved in the signal transduction of many growth factors have been proposed, the function of Gab1 in postnatal bone metabolism remains poorly understood.

In this report, we show that Gab1 is required for osteoblast activity and bone homeostasis. Osteoblast-specific disruption of Gab1 led to low trabecular bone mass with decreased bone formation and impaired bone mechanical properties. Furthermore, Gab1 deficiency is associated with declined mineralization and increased susceptibility to apoptosis in osteoblasts. We also show that activation of AKT and ERK in response to IGF-1 and insulin is attenuated in Gab1-deficient osteoblasts.

To elucidate the physiological role of Gab1 in osteoblast function and bone homeostasis, we generated an osteoblast-specific Gab1 knockout mouse line by breeding the Gab1^flox/flox mouse (Wada et al., 2005) with the Osteocalcin-Cre transgenic mouse (OC-Cre)(Tan et al., 2007). The Gab1^flox/flox;OC-Cre mice (Gab1 cKO) were born normally and were fertile. Cre-mediated deletion of exon 3 of Gab1 in different tissues isolated from a Gab1^flox/flox;OC-Cre mouse revealed that an efficient excision occurred exclusively in calvaria, femurs and spine, all of which were bone tissues containing osteoblasts (supplementary material Fig. S1A,B).

X-ray and three-dimensional microcomputed tomography (micro-CT) scanning analyses of vertebrae and femora revealed that male Gab1 mutant mice exhibited lower trabecular bone mass than their wild-type littermates at 6.5 months of age (Fig. 1A-C). Cortical bone thickness, the most important parameter for biomechanical strength, was reduced from 268.7±54.9 μm in controls to 199.9±32.5 μm in Gab1 mutant mice at 6.5 months of age (Fig. 1D). Bone mineral density (BMD) of femurs from male mutant mice was about 20% lower than that of controls at 2 months and 6.5 months of age (Fig. 1E). Consistently, BMD of female mutant femurs exhibited a significant decrease as compared with sex-matched controls at 2 months and 15 months (Fig. 1F). Micro-CT analysis in trabeculae from female femurs also showed decreased bone mass at 2 months of age (supplementary material Fig. S2A-H). Detailed histological analysis of tibias revealed decreased bone mass in mutants at the age of 2 months (Fig. 1G). We also generated another osteoblast-specific Gab1 knockout mouse by breeding the Gab1^flox/flox mouse with the Collagen1a1-Cre transgenic mouse (Col1a1-Cre) (Zha et al., 2008). The Gab1^flox/flox;Col1a1-Cre mice exhibited similar phenotypes as Gab1^flox/flox;OC-Cre mice (supplementary material Fig. S3A-F). The results showed that targeted disruption of Gab1 in osteoblasts resulted in osteopenia in mutant mice.

To determine the cellular basis underlying the bone mass loss in Gab1 mutants, we performed bone histomorphometric analysis of the proximal tibias of male mice at 2 months of age. von Kossa staining showed a significantly decreased bone volume in the mutant mice (Fig. 2A-D). The ratio of trabecular bone volume to tissue volume (BV/TV) in tibias of Gab1-deficient mice was 33% lower than that of controls (Fig. 2G). The decreased bone volumes were correlated with decreased trabecular number (TbN) and increased trabecular separation (Tb.Sp) in mutants compared with controls (Fig. 2H,I). In addition, trabecular thickness was decreased in mutant mice (Fig. 2J). Number of osteoblasts per tissue area (N.Ob/T.Ar) and number of osteoblasts per bone perimeter (N.Ob/B.Pm) were significantly decreased in Gab1 mutants (Fig. 2K,L). Double calcein labeling revealed a dramatically declined bone formation rate (BFR) in 2-month-old mutant mice as compared with controls (Fig. 2E,F). The ratio of bone formation rate to bone surface (BFR/BS) and the mineral appositional rate (MAR) were also markedly reduced in mutants (Fig. 2M,N). These results suggested that loss of osteoblastic Gab1 resulted in impaired osteoblast function.

Decreased osteoblast activity might result in the reduction of osteoclast function. As expected, we found that Gab1-deficient mice had less numbers of tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts on tibia sections at 2 months of age (Fig. 3A-D). Bone histomorphometric analyses revealed that all bone resorption parameters, the percentage of bone surface covered by mature osteoclasts (Oc.S/BS), the number of mature osteoclasts in tissue area (N.Oc/T.Ar) and the number of mature osteoclasts (N.Oc/B.Pm) were significantly decreased in Gab1 mutant mice (Fig. 3E-G). Real-time PCR analysis showed that expression of genes encoding TRAP, cathepsin K (CathK) and matrix metalloproteinase 9 (MMP-9), lysosomal enzymes essential for osteoclastic bone resorption, were markedly reduced in femoral bones of Gab1 mutant mice (Fig. 3H-J). Co-culture of calvarial osteoblasts and a common osteoclast progenitor population derived from the wild-type spleen showed that TRAP-positive cells were significantly reduced, when osteoblasts were derived from Gab1 mutant mice as compared with those from controls (Fig. 3K,L). Furthermore, expression of the receptor activator of NF-κB ligand (RANKL) was markedly downregulated in Gab1 mutant osteoblasts, whereas the expression of osteoprotegerin (OPG) was not changed (Fig. 3M,N). Quantitative analysis of mouse serum RANKL concentrations also showed a reduction in the Gab1 mutant mice as compared with controls (Fig. 3O). These results indicated that the osteoclastogenic activity of the Gab1 mutant osteoblasts was significantly decreased.

The mechanical properties of bone are determined by the amount and structure of bone. To investigate the role of Gab1 in the macromechanical properties of bone, we performed a three-point bending test whereby right femoral mid-diaphyses were deflected to failure by a loaded bar from above (Fig. 4A). Femurs from Gab1-deficienct mice exhibited altered mechanical properties, indicated by the load-deflection diagram (Fig. 4B). Young's elastic modulus in mutant mice exhibited a 41% decrease relative to that of wild-type mice (Fig. 4C). Femurs from 2-month-old Gab1 mutant mice showed a 42% decrease in bone strength (maximum load) as compared with wild-type controls (Fig. 4D), consistent with the decreased cortical bone area. In addition, stiffness and energy-to-failure were reduced significantly (by 49% and 50%, respectively) relative to wild-type controls (Fig. 4E,F). Ultimate stress (the stress causing bone fracture) of Gab1-deficient bone decreased by nearly 50% compared with controls (Fig. 4G). These results demonstrated that loss of osteoblastic Gab1 eventually affected the intrinsic bone mechanical properties.

To explore the cellular mechanism underlying the osteopenia phenotype of Gab1 mutant mice, primary osteoblast proliferation and function were measured in in vitro culture systems. The Gab1 mutant calvarial osteoblasts isolated from Gab1^flox/flox;Col1a1-Cre mice showed a decreased percentage of Ki-67-positive osteoblasts as compared with that of controls (35.6±7.87% in wild type versus 21.1±1.8% in mutants) (Fig. 5A,B). Early osteoblast differentiation was evaluated by alkaline phosphotase (ALP) staining. In order to avoid the effect of reduced proliferation, we analyzed osteoblast differentiation at high density in cultures. ALP staining and relative ALP activity in cell extracts were unchanged in mutants compared with controls at all stages examined (Fig. 5C,D). Furthermore, the amount of ALP in serum revealed no obvious difference between Gab1 mutant mice and control mice (Fig. 5E). These data suggested that Gab1 was not necessary for early osteoblast differentiation. However, when late differentiation of osteoblast was assayed by Alizarin Red staining and von Kossa staining, the deposition of mineralized extracellular matrix (bone nodules) was reduced in Gab1 mutants at day 18 in culture (Fig. 5F). Next, in order to identify candidate genes that might be responsible for the impaired osteoblast maturation, the expression of osteoblast marker genes was analyzed by real-time PCR in cultures at day 7. Expression levels of transcripts for collagen type 1a (Col1a), osteocalcin (OC) and bone sialoprotein (Bsp) were significantly reduced, whereas mRNA expression of ALP was not changed (Fig. 5G). In 2-month-old total bone extracts, the reduction of Col1a, OC and Bsp were also confirmed by real-time PCR (data not shown), which was in agreement with reduced bone mineralization in vivo. We also performed in situ hybridization on bone sections from 2-month-old mice and immunofluorescence staining on primary osteoblasts to confirm that the expression of Col1a was indeed decreased in Gab1 mutants (supplementary material Fig. S4A-D). In addition, we studied whether silencing of Gab1 in a preosteoblastic cell line could mimic the defective mineralization phenotype of osteoblasts derived from Gab1 mutant mice. Western blot suggested that Gab1 was knocked down efficiently in MC3T3-E1 cells (Fig. 5H). We found that interference of Gab1 reduced cell proliferation (data not shown), decreased bone nodule formation but did not influence early differentiation (Fig. 5I). These results imply that Gab1 is essential for late differentiation and mineralization of osteoblasts.

To investigate the role of Gab1 in mediating survival signals to osteoblasts, osteoblastic apoptosis was detected using the TUNEL (terminal dUTP nick-end labeling) assay and Annexin-V flow cytometry analysis. The rate of apoptotic cells was significantly increased in cultured primary Gab1 mutant osteoblasts relative to control cells, as revealed by the TUNEL assay (Fig. 6A,B). In addition, the percentage of Annexin-V-positive osteoblasts was significantly increased under serum-free induction (Fig. 6C,D). To confirm the anti-apoptotic role of Gab1, we performed the TUNEL assay on Gab1-silenced MC3T3E1-S14 preosteoblastic cells. Consistently, Gab1-silenced preosteoblasts also showed enhanced susceptibility to serum-deprivation-induced cell apoptosis (Fig. 6E,F). These results suggested that Gab1 deficiency in osteoblasts enhanced their susceptibility to apoptosis.

It has been shown that Gab1 serves as a scaffolding protein downstream of several receptor protein-tyrosine kinases, including the insulin and epidermal growth factor (EGF) receptors. On stimulation of receptors, Gab proteins recruit the p85 subunit of PI3K and Shp2, resulting in activation of both AKT and ERK, respectively (Cunnick et al., 2000; Ingham et al., 2001; Koyama et al., 2008). So, we next examined whether AKT and ERK signal transduction cascades were affected by Gab1 deficiency. We found that phosphorylation of AKT and ERK significantly decreased in primary mutant osteoblasts isolated from calvaria (Fig. 7A). Phosphorylation of AKT and ERK stimulated by IGF-1 and insulin in Gab1-deficient cells was in all cases lower than that of controls (Fig. 7B,C). By contrast, the levels of activation of AKT and ERK induced by basic fibroblast growth factor (bFGF) and EGF were not significantly different between Gab1-deficient and control osteoblasts during the time course of the experiment (Fig. 7D,E). These data indicate that Gab1 is involved in mediating IGF-1- and insulin-stimulated activation of AKT and ERK in osteoblasts.

Our results firmly establish that Gab1 has a function in bone metabolism that is distinct from that of Gab2, although there is a high degree of homology and structural similarity between Gab1 and Gab2 (Nishida and Hirano, 2003). Gab2 has been shown to couple RANK to downstream signaling required for osteoclastogenesis, and has a negative regulatory role for osteoblast differentiation (Itoh et al., 2007; Wada et al., 2005). We have shown for the first time that disruption of Gab1 in mature osteoblasts resulted in a low-bone-turnover osteopenia phenotype at 2 months of age, demonstrating an essential function of Gab1 in osteoblasts. Altered skeleton caused by Gab1 deletion leads to poor mechanical properties. It is conceivable that the osteopenic phenotype of osteoblast-specific Gab1 mutant mice is largely due to the impaired activation of AKT and ERK after Gab1 ablation. Supportively, Akt1^−/− and Akt2^−/− mice exhibit skeletal abnormalities with decreased bone mass (Kawamura et al., 2007), whereas activation of AKT by Pten deletion in osteoblasts leads to the osteopetrosis phenotype (Liu et al., 2007). In addition, ERK-MAPK activation has been demonstrated to stimulate osteoblast differentiation and skeletal development through a pathway involving RUNX2 (Ge et al., 2007).

Gab1 appears to positively regulate osteoblast maturation. Osteoblastic differentiation undergoes several stages, including proliferation, extracellular matrix deposition, matrix maturation, and mineralization (Stein and Lian, 1993). Surprisingly, Gab1 has little effect on early osteoblast differentiation, as suggested by comparable early marker ALP-positive osteoblasts in both control and Gab1 mutant mice. Gab1 mutants suffered from a decreased bone formation rate and a defect in late differentiation and bone mineralization. These results were confirmed by a decreased number of mineralized nodules in cultured Gab1 mutant calvarial osteoblasts and osteoblastic cell line MC3T3-E1. Consistently, expression of OC and Bsp, late markers of osteoblast differentiation, were significantly reduced in both Gab1 mutant bones and primary osteoblasts. In addition, synthesis of extracellular matrix Col1a was markedly decreased. Impaired late differentiation of osteoblast and insufficient expression of the bone matrix components might account for the low-bone-mass phenotype and the impaired mechanical properties of bone from Gab1 mutant mice.

Gab1 is essential for regulating osteoblast survival. Gab1 has been shown to play an important positive role in cell survival. Mutant Gab1 lacking PI3K-binding sites induces apoptosis and overexpression of Gab1 in PC12 cells alone is sufficient to protect cells from apoptosis (Holgado-Madruga et al., 1997; Korhonen et al., 1999). The Shp2 interaction with Gab1 has been well studied, and deletion of Shp2 in trophoblast stem cells leads to increased apoptosis (Yang et al., 2006). Inhibition of PI3K and AKT and of MAPK and ERK activities strongly increased osteoblast apoptosis under serum-free conditions (Liang et al., 2008). In this study, we showed that both primary Gab1-deficient osteoblasts and Gab1 knockdown pre-osteoblastic cells exhibited increased apoptosis, providing evidence to demonstrate that Gab1 played a protective role in osteoblast apoptosis.

More importantly, disruption of adaptor Gab1 in osteoblasts resulted in decreased expression of RANKL, which in turn affected osteoclast differentiation. The OPG-RANKL system has long been known as a key regulator that couples the function of osteoblasts and osteoclasts (Kearns et al., 2008). Recently, it has been shown that the Wnt signaling pathway in osteoblasts determines osteoclastgenesis through regulating expression of OPG (Boyce et al., 2005; Glass et al., 2005; Holmen et al., 2005), whereas Hedgehog signaling in osteoblasts promotes osteoclast formation by upregulating parathyroid hormone-related protein (PTHrP) and RANKL expression (Mak et al., 2008). We, and others, have also revealed that TGF-β signaling plays a crucial role in the regulation of the RANKL-OPG axis (Hofbauer et al., 1998; Murakami et al., 1998; Okamoto et al., 2006; Tan et al., 2007). Here, we have demonstrated that, in mature osteoblasts, signals mediated by Gab1 are required for osteoclastgenesis through positive regulation of RANKL expression.

Our results suggest that adaptor Gab1 might participate in mediating IGF-1- and insulin-triggered AKT and ERK activation in osteoblasts. IGF-1 and insulin have been shown to play important roles in the anabolic regulation of bone metabolism through effects on collagen synthesis (Kream et al., 1985; Rosen and Luben, 1983; Thomas et al., 1996) and ALP production (Canalis, 1983). Mice with osteoblast-specific deletion of IGF-1 receptor exhibited osteopenia phenotype with normal size and weight (Zhang et al., 2002). Insulin receptor substrate-1 (Irs1)-deficient mice show low turnover osteopenia, with the impairment of osteoblast proliferation and differentiation, and support of osteoclastogenesis (Ogata et al., 2000). Our Gab1 mutant phenotypes are similar to those of mice deficient in IGF-1 receptor and IRS-1, described above. Overexpression of Gab1 in Irs1-deficient fibroblasts can partially restore biological effects mediated by IGF-1 signaling (Winnay et al., 2000). In the current study, we found that the phosphorylation levels of AKT and ERK stimulated by IGF-1 or insulin, but not by bFGF or EGF, were significantly diminished in Gab1-deficient osteoblasts as compared with controls, suggesting that decreased bone mass in Gab1 mutant mice is possibly due to the reduction in sustained activation of AKT and ERK of osteoblasts in response to IGF-1 and insulin.

In summary, we have provided first genetic evidence demonstrating that osteoblastic Gab1 is essential for maintaining normal postnatal bone homeostasis and bone mechanical property. We show that selective deletion of Gab1 in osteoblasts results in decreased proliferation, defective mineralization and increased apoptosis. Moreover, differentiation of osteoclasts is also decreased due to impaired capacity of Gab1 mutant osteoblasts in supporting osteoclastogenesis. Mechanistically, we show that the activation of AKT and ERK is significantly attenuated in Gab1-deficient osteoblasts in response to IGF-1 and insulin.

Mice lacking Gab1 in ostoblasts were generated by breeding homozygous carrying the floxed Gab1 alleles (Gab1^flox/flox) (Bard-Chapeau et al., 2005) with OC-Cre (Tan et al., 2007) or Col1a1-Cre (Zha et al., 2008) transgenic mice. Gab1^flox/flox;OC-Cre or Gab1^flox/flox;Col1a1-Cre mice were obtained by breeding the Gab1^flox/flox mice with Gab1^flox/flox;OC-Cre or Gab1^flox/flox;Col1a1-Cre mice. The Gab1^flox/flox littermates were used as controls. Animals were handled in accordance with institutional guidelines. Genotyping for Gab1 locus and Cre transgene were detected by PCR analysis on genomic DNA extracted from mouse tails using primers described previously (Bard-Chapeau et al., 2005; Tan et al., 2007). Primers used to detect the specific deletion of exon 3 of Gab1 in the bone by PCR were described previously (Bard-Chapeau et al., 2005).

Radiographic analyses were carried out using a soft X-ray system (Contour Plus). BMD of the left femur was measured by dual energy X-ray absorptiometry with a Piximus Mouse Densitometer (GE Lunar Medical System). The femurs were scanned and reconstructed with 8 μm isotropic voxel size on a microcomputed tomography analysis (micro-CT) system (eXplore Locus SP, GE Medical Systems). The reconstructed 3D imagines of femurs were analyzed using Microviewer (GE Medical Systems). A fixed threshold (1600) was used to separate the bone and marrow phases. Specifically, trabecular bone architecture was analyzed at the distal femoral metaphysis, whereas cortical bone morphology was evaluated at the femoral midshaft.

Tissues were fixed in 10% formalin overnight, decalcified in 15% EDTA-PBS and embedded in paraffin. Sections 6 μm thick were cut and stained with hematoxylineosin and TRAP according to standard methods. For in vivo fluorescent labeling, 2-month-old animals were injected with calcein (20 mg/kg body weight) intraperitoneally at days 7 and 2 before tissue collection. After dehydration, the undecalcified tibiae were embedded in methylmethacrylate, and 5-μm sections were cut and stained with von Kossa stain or toluidine blue. Histomorphometric analyses were performed using OsteoMeasure (OsteoMetrics, Decatur, GA) bone analysis software. The regions of interest for trabecular bone data collection were measured in an area 1.5 mm in length, from 0.3 mm below the growth plate of the proximal tibias. All histomorphometric parameters are reported in accordance with standard criteria (Parfitt et al., 1987).

To test for differences in bone failure properties between Gab1 mutant and control mice, femur from 2-month-old mice were rehydrated at room temperature in phosphate buffered saline (PBS) and femoral biomechanical properties assessed by three-point bending. Strength tests were performed at the right femur midshaft with a displacement rate of 6 mm/minute (span length, 8 mm) using a mechanical testing machine (model 5865; Instron, Norwood, MA). Whole-bone mechanical properties, including maximum load, stiffness, energy-to-failure and Young's elastic modulus, were determined using load-deflection diagrams.

Primary osteoblasts were isolated from calvariae of 3-day-old neonatal Gab1^flox/flox;Col1a1-Cre mice by serial digestion using 0.1% collagenase I (Gibco) as previously described (Tan et al., 2007). Cells were grown in α-MEM containing 10% FBS until confluent. Cells were then re-plated for differentiation at 8×10⁴ cells per well, or for proliferation at 2×10⁴ cells per well in a 24-well plate. Osteoblasts were grown on the glasses in a 24-well plate and the Ki-67 (Abcam) assay used to detect proliferation. For differentiation, medium was supplemented with 50 mg/ml L-ascorbic acid and 10 mM β-glycerophosphate at day 2 of culture. Staining of cultures for ALP activity was conducted using Sigma Kit 86R. Bone nodules were identified morphologically by staining with Alizarin Red S (Sigma-Aldrich) and von Kossa stain. Osteoblast-osteoclast co-culture experiments were performed as described (Jochum et al., 2000). Briefly, calvarial osteoblasts (5×10⁴ per 24-well culture dish) were co-cultured with nonadherent monocyte/macrophage progenitor cells derived from wild-type spleen (5×10⁵ per 24-well culture dish) in medium supplemented with 10 nM 1,25-dihydroxy vitamin D3. TRAP staining was performed after 8 days of culture.

To determine the osteoblast apoptosis, primary osteoblasts and MC3T3-E1 cell line were grown on glasses in a 24-well plate and then the culture medium changed to a serum-free medium for 24 hours. Apoptotic osteoblasts were then further identified by an in situ cell-death detection kit (Promega) according to the manufacturer's instructions. FITC-labeled Annexin-V staining was used to confirm the presence of apoptosis. Osteoblasts were plated on 100-mm plates (1×10⁶ per plate) and cultured for 1 day, and then cultured in serum-free medium for 24 hour before harvesting for Annexin-V—FITC staining. After staining with Annexin-V—FITC, osteoblasts were analyzed by FACS and 20,000 cells were counted.

To inhibit the expression of Gab1 in MC3T3-E1 pre-osteoblasts, three oligoribonucleotides specifically targeting mouse Gab1 were synthesized by Invitrogen. The coding sequences were: RNAi-G1, 5′-CCTAACAGAACCCTCTTTG-3′; RNAi-G2, 5′-ATGATGTATGACTGCCCAC-3′; and RNAi-G3, 5′-AGCCACATCCAACTCATGA-3′. The targeted oligoribonucleotides were introduced into pSUPERretro Vector, and retroviral vector pSUPERretro-Gab1 was transfected into PLAT-E cells with FuGene6 (Roche, Basel, Switzerland) to be packaged. The infection of MC3T3-E1 by retrovirus was described previously (Sun et al., 2008).

Total RNA was isolated from mouse bones and osteoblasts using the TRizol reagent (Invitrogen) according to the manufacturer's instructions. Total RNA (5 μg) was reverse transcribed to cDNA with the use of the first-strand cDNA synthesis kit (Invitrogen). Quantitative PCR was performed to measure the relative mRNA levels using the LightCycler system (Roche) with SYBR Green. The primers have been described previously (Tan et al., 2007).

Primary osteoblasts were plated on 60-mm plates (5×10⁵ per plate) and cultured for 2 days. Culture medium was changed to α-MEM without serum for 6 hours and then stimulated with insulin (100 nM), IGF-1 (10 nM), EGF (100 ng/ml) or bFGF (1 nM). Total cell proteins were extracted by sonication in RIPA buffer containing protease inhibitors (Roche). After measurement of the protein content according to the BCA protein assay kit (Thermo Scientific, Rockford, IL), protein samples were separated on SDS-PAGE gels and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore). Primary antibodies used included: Gab1, phosphorylated ERK, phosphorylated AKT, total ERK, total AKT (Cell Signaling), β-actin (Sigma) and tubulin (Santa Cruz Biotechnology).

In situ hybridization was performed as previously described (Yang et al., 2008). For immunofluorescence, primary osteoblasts cultured for 7 days were fixed in 4% PFA for 15 minutes, washed in PBS containing 0.5% Triton X-100, blocked for 1 hour at room temperature using goat serum, and incubated with collegen antibody (1:500; Abcam) for 2 hours at room temperature. After washing, FITC-conjugated goat anti-rabbit antibody was used to detect the primary antibody. For nuclear staining, osteoblasts were treated with DAPI (Sigma).

Sera were collected from 2-month-old female mice and RANKL levels determined using the Quantikine mouse RANKL ELISA kit (R&D Systems).

All results were expressed as mean ± standard deviation. All statistical analyses were performed using the SPSS software package. In the Student's t-test, P<0.05 was accepted as significant.

We thank Bin Zhao for undecalcified sections, Xiwen Xiong and Chuwen Lin for BMD detections and Jun Wang for helpful instruction in microCT assays. This work was supported by Chinese National Key Program on Basic Research (2005CB522506; 2006CB943501; 2006BAI23B01-3), Key Project for Drug Discovery and Development in China (2009ZX09501-027), Key Project for Infectious Diseases in China (2008ZX10002-016), National Natural Science Foundation of China (30671030, 30871396, 90607004) and E-Institutes of Shanghai Municipal Education Commission (E03003).