Transglutaminase 2 regulates osteoclast differentiation via a Blimp1-dependent pathway

Transglutaminase 2 (TG2) performs multiple reactions, including transamidation, and also plays a role in signal transduction as a GTP-binding protein. In this study, we reveal that TG2 controls osteoclast differentiation and bone homeostasis in mice. Osteoclasts specifically expressed the TG2 isoform among eight TG family members. Suppression in TG2 expression with siRNA led to increased osteoclast formation from primary mouse precursor cells in response to receptor activator of nuclear factor kappaB ligand (RANKL). This osteoclastogenic effect of TG2 knockdown was associated with enhanced induction of c-Fos and NFATc1 by RANKL. Moreover, TG2 knockdown up-regulated B lymphocyte-induced maturation protein 1 (Blimp1), which represses anti-osteoclastogenic genes, in a manner dependent on the NF-κB signaling pathway. To the contrary, TG2 overexpression inhibited osteoclast formation and the expression of osteoclastogenic genes. Consistent with these in vitro results, TG2 knockout mice exhibited lower trabecular bone mass and increased number of osteoclasts compared with wild-type mice. Taken together, our results provide strong evidence that TG2 plays an important role in bone metabolism by suppressing excessive osteoclastogenesis via the regulation of the NF-κB-Blimp1 signaling pathway.

(IRF-8) and v-maf musculoaponeurotic fibrosarcoma oncogene family protein B (MafB) 14 . As IRF-8 and MafB have been shown to reduce the expression and function of NFATc1 15,16 , Blimp1 might ensure the maintenance of NFATc1 activity during osteoclastogenesis by repressing IRF-8 and MafB.
The transglutaminase (TG) family consists of eight distinct members: factor XIII A (FXIIIA), TG1 (or keratinocyte TG), TG2 (or tissue TG), TG3 (or epidermal TG), TG4 (or prostate TG), TG5 (or TG X), TG6 (or TG Y), and TG7 (or TG Z) 17 . TG family enzymes catalyze posttranslational modifications of various substrates via transamidation, esterification, and hydrolysis reactions in a Ca 2+ -dependent manner. These TG-mediated reactions have an impact on diverse cellular responses, including proliferation 18 , differentiation 19 , death 20 , and migration 21 . Consequently, TGs modulate multiple biological processes, including development, tissue remodeling, inflammation, and wound healing [22][23][24] . The TG2 isoform is distinguished from other members of TG family by several unique characteristics, such as a ubiquitous expression pattern, GTP-binding and -hydrolysis, and cell-matrix interaction regulation 25 . The functions of TG2 depend on its subcellular localization and presence of its regulators. Ca 2+ and GTP are the most important regulators of TG2 and act as switches between the two distinct functional entities of TG2, transglutaminase and GTP hydrolase, via allosteric modulation 23,26,27 .
Some studies have reported association of TG family members, especially TG2 and FXIIIA, with bone development and matrix mineralization 28 . TG2 and FXIIIA were shown to be expressed in hypertrophic chondrocytes and osteoblasts 28, 29 . In preosteoblastic MC3T3-E1 cells, a high level of FXIIIA activity was detected during osteoblastic differentiation and inhibition of TG activity suppressed mineralization 30 . These and other observations have implicated TG2 and FXIIIA in matrix crosslinking and assembly, the essential processes of bone mineralization 28 . However, the relationship between TG2 and osteoclasts has not been investigated. To gain a better understanding on the role of TG2 in the regulation of bone homeostasis, we examined whether TG2 is involved in the differentiation of osteoclasts using loss-and gain-of-function approaches.

TG2 is involved in the regulation of RANKL-induced osteoclast differentiation.
We first examined the expression profile of TG family members in osteoclast precursors (BMMs) and pre-fusion osteoclasts (pOCs; BMMs treated with RANKL and M-CSF for two days). Among the TG family members, TG2 mRNA was detected at a high level in both BMMs and pOCs (Fig. 1a). FXIIIA mRNA was only weakly expressed, while the other TG isoforms were not detected. This expression pattern of TG genes was maintained during differentiation. As TG2 was the only TG members expressed at a significant level, we next investigated the role of TG2 in osteoclastogenesis by evaluating the effect of TG2 reduction using a siRNA-mediated gene knockdown system. Two different oligonucleotides, targeting different regions of TG2 mRNA, were tested. A greater than 50% reduction in TG2 expression was achieved in BMMs transfected with TG2 siRNA #1 or #2 compared to BMMs transfected with control siRNA (Fig. 1b and Supplementary Figure 1a). To test the effect of TG2 knockdown on osteoclast formation, the siRNA-transfected BMMs were incubated with RANKL and M-CSF for three days and then stained for TRAP activity. As shown in Fig. 1c and Supplementary Figure 1b, the generation of osteoclasts (TRAP-positive MNCs) was facilitated by TG2 knockdown. The surface area per osteoclast and the number of osteoclasts generated were increased in TG2-knockdown samples ( Fig. 1d and Supplementary Figure 1c), indicating more number of large osteoclasts were formed. For the rest of the study TG2 siRNA #1 (designated as TG2 siRNA) was used. Similar results were obtained with the macrophage cell line RAW264.7, which has been used as an osteoclast precursor cell in many studies (Supplementary Figure 2). Moreover, TG2 knockdown in BMMs committed to osteoclastic lineage by priming with RANKL for one or two days also accelerated mature osteoclast formation (Supplementary Figure 3). As the c-Fos and NFATc1 transcription factors are essential for osteoclastogenesis and are induced and activated by RANKL 31 , we next investigated whether TG2 knockdown altered RANKL induction of c-Fos and NFATc1. The mRNA levels of c-Fos and NFATc1 were significantly higher in TG2-knockdown cells than in control cells (Fig. 1e). In addition, the protein levels of the c-Fos and NFATc1 were increased more upon RANKL stimulation in TG2-knockdown cells (Fig. 1f). Consistently, a significantly higher level of NFATc1 transcriptional activity was observed in pOCs generated from TG2-knockdown cells compared with the control cells (Fig. 1g). Finally, when knockdown of c-Fos or NFATc1 was simultaneously carried out with TG2 knockdown, the enhancement of osteoclastogenesis by TG2 down-regulation was abrogated (Fig. 1h). These results suggest that, among TG family members, TG2 has a specific role in regulating osteoclast differentiation.
Actin ring formation and bone resorption activity are increased by TG2 knockdown. As TG2 appears to be involved in the regulation of osteoclast differentiation, we next examined whether the activity of osteoclast could be modulated by TG2. The sealing zone is a ring-shaped cytoskeletal structure composed of highly polymerized actin, which is crucial for tight adhesion to the bone surface and the bone-resorbing activity of osteoclasts 12,32 . Therefore, we investigated the effect of TG2 knockdown on actin ring formation and bone resorption. The number of actin ring-positive osteoclasts generated from TG2-knockdown BMMs was higher than that from the control siRNA-transfected BMMs ( Fig. 2a and b). The sealing zone area per osteoclast was also significantly increased by TG2 knockdown (Fig. 2b). As TG2 knockdown increased osteoclast size and sealing zone area, we next examined the effects of TG2 deficiency on the expression levels of genes associated with osteoclast fusion and bone resorption activity, such as DC-STAMP, TRAP and ATP6v0d2. As shown in Fig. 2c, the mRNA levels of DC-STAMP, TRAP and ATP6v0d2 were higher in TG2-knockdown cells. Next, bone resorption activity was assessed using dentine slices. Consistent with the increase in the number of osteoclasts generated, TG2 knockdown caused an increase in the total area of the resorption pits formed (Fig. 2d and e). Moreover, both the resorption area per osteoclast and the depth of the pits were increased by TG2 knockdown (Fig. 2e). These data suggest that TG2 knockdown enhances actin ring formation and the resorption activity of osteoclasts, in addition to the differentiation of osteoclasts. Blimp1 is up-regulated by TG2 knockdown via the NF-κB signaling pathway. A previous study showed that TG2 had a suppressive effect on the expression of Blimp1 transcription repressor in lymphocytes 33 . Therefore, we explored the possibility that Blimp1 is involved in TG2-mediated regulation of osteoclastogenesis. We first examined whether Blimp1 expression was affected by TG2 knockdown in osteoclastogenic culture conditions. The mRNA level of Blimp1 was increased by RANKL treatment, and this induction of Blimp1 was significantly greater in TG2-knockdown cells than in control cells (Fig. 3a). TG2 knockdown also enhanced the RANKL induction of Blimp1 at the protein level (Fig. 3b). Next, we investigated the effect of TG2 knockdown on intranuclear level of Blimp1. Analyses of cytoplasmic and nuclear fractions revealed an increase in nuclear Blimp1 upon TG2 knockdown (Fig. 3c). Blimp1 protein was not readily detected in the cytoplasmic fraction. TG2 knockdown also increased the amount of NFATc1 protein in the nuclear fraction upon RANKL stimulation (Fig. 3c).
Because Blimp1 has been reported to have a p65 binding site in its promoter region and has been shown to be up-regulated by activation of the NF-κB signaling pathway in B cells 34 , we examined whether the increase in Blimp1 level resulting from TG2 knockdown was due to activation of the NF-κB pathway. To this end, we compared RANKL stimulation of NF-κB in control and TG2-knockdown cells. RANKL stimulation induced phosphorylation of IκB within 5 min, and the level of RANKL-induced IκB phosphorylation was further enhanced by TG2 knockdown at 5 and 30 min (Fig. 3d upper). Consistently, the total IκB level was lower in TG2-knockdown cells. Moreover, phosphorylation of p65 was increased by TG2 knockdown (Fig. 3d lower). We also examined whether p65 transcriptional activity was increased by TG2 knockdown. TG2 knockdown enhanced the activity of p65 15 min after RANKL stimulation (Fig. 3e). In contrast, the RANKL stimulation of MAPK pathways was not affected by TG2 knockdown (Supplementary Figure 4). Since multiple putative binding sites of p65 were predicted to be present on the promoter region of Blimp1, we performed the chromatin immunoprecipitation assay to assess legitimate binding of p65 regulated by TG2 knockdown. Indeed, p65 bound to predicted sites 2 and 3 of Blimp1 promoter and the enrichment by TG2 knockdown was observed at the site 2 (Fig. 3f). In addition, the NF-κB inhibitor BAY 11-7085 attenuated the increase in p65 phosphorylation by TG2 knockdown (Fig. 3g). BAY 11-7085 also significantly mitigated the effect of TG2 knockdown on the induction of Blimp1 and NFATc1 (Fig. 3h). Interestingly, the expression level of c-Fos was not affected by BAY 11-7085 treatment (Fig. 3h). These data suggest that TG2 knockdown up-regulates expression of Blimp1 and NFATc1 through the NF-κB pathway during osteoclast differentiation. We next examined the effect of TG2 knockdown on the expression of IRF-8 and MafB, transcription repressors suggested to be down-regulated by Blimp1 and suppress NFATc1 expression [14][15][16] . As shown in Fig. 3i, TG2 knockdown reduced the expression of MafB and IRF8 in RANKL-treated BMMs (Fig. 3i). Therefore, the regulation of NFATc1 by TG2 may involve the Blimp1-IRF8/MafB pathway in osteoclasts.

Blimp1 mediates the effect of TG2 knockdown on osteoclastogenesis.
To further clarify the role of the TG2-NF-κB-Blimp1 axis in RANKL-induced osteoclastogenesis, we investigated the effect of TG2 and Blimp1 double-knockdown on osteoclast formation. Significant reduction in both TG2 and Blimp1 expression upon dual transfection of siRNAs was verified by real-time PCR (Fig. 4a). As shown earlier in Fig. 1c, TG2 knockdown significantly enhanced the formation of TRAP-positive MNCs by RANKL. This enhancement of osteoclast formation by TG2 knockdown was clearly attenuated by the addition of Blimp1 knockdown (Fig. 4b). Both the number of TRAP-positive MNCs and the surface area of osteoclasts were decreased by knockdown of Blimp1 (Fig. 4c). We also examined whether knockdown of TG2 and Blimp1 affected the expression of c-Fos and NFATc1. As previously shown in Figs 1 and 3, TG2 knockdown increased the expression level of NFATc1. Notably, Blimp1 knockdown prominently suppressed the up-regulation of NFATc1 resulting from TG2 knockdown (Fig. 4d). In contrast, the expression level of c-Fos was only slightly affected by Blimp1 knockdown (Fig. 4d). These results indicate that Blimp1 mediates the increases of osteoclast formation and NFATc1 induction by TG2 knockdown.

TG2 overexpression suppresses RANKL-induced osteoclast differentiation. We next investigated
the role of TG2 in osteoclastogenesis by evaluating the effects of TG2 overexpression in RAW264.7. Using a lentiviral system, a TG2-overexpressing RAW264.7 subline was established. The mRNA and protein levels of TG2 were much higher in these TG2-transduced cells (Fig. 5a). When these cells were induced to differentiate with RANKL, osteoclast formation was less active than in control cells (Fig. 5b). Both the total number of TRAP-positive MNCs and the surface area per osteoclast were significantly lowered by TG2 overexpression (Fig. 5c). Consistently, TG2 overexpression reduced the expression of Blimp1, c-Fos and NFATc1 at both mRNA ( Fig. 5d) and protein levels (Fig. 5e). These results support that TG2 plays a negative role in osteoclast differentiation via the Blimp1-NFATc1 axis.

TG2 knockout (KO) mice have decreased bone mass.
To obtain in vivo evidence to support a role of TG2 in the regulation of osteoclastogenesis and bone metabolism, we analyzed the radiological and histomorphometric parameters of TG2 KO mice bones. Three-dimensional micro-CT analyses of femurs from 13-week-old female mice revealed that the trabecular bone volume was significantly lower in TG2 KO mice compared with wild-type (WT) mice (Fig. 6a). Similarly, sagittal and transaxial images of femurs showed reduction in trabecular bone mass in TG2 KO mice (Fig. 6b). The trabecular bone volume per tissue volume (BV/TV) was 52% lower indicating the fold-change from control are shown (left). BMMs transfected with indicated combination of siRNAs were cultured with M-CSF plus RANKL for four days. Cells were stained for TRAP activity (right upper). Scale bar, 100 μm. TRAP-positive MNCs (≥3 nuclei) were counted as osteoclasts (right lower). **P < 0.005 versus control siRNA or between indicated groups. Full length gels and Western blots are presented in Supplementary Figure 9. Actin ring formation and bone resorption are enhanced by TG2 knockdown. (a,b) BMMs were transfected with control or TG2 siRNA and cultured with M-CSF (30 ng/ml) and RANKL (100 ng/ml) for three days. Cells were stained with rhodamine-phalloidin conjugate and DAPI to visualize F-actin and nuclei, respectively. The number of actin rings was counted, and the sealing zone (SZ) area per osteoclast was measured. **P < 0.005 versus control siRNA. Scale bar, 100 μm. (c) After transfection with control or TG2 siRNA, BMMs were incubated with M-CSF (30 ng/ml) and RANKL (100 ng/ml) for the indicated number of days. The mRNA levels of DC-STAMP, TRAP and ATP6v0d2 were measured by real-time PCR. (d,e) BMMs were transfected with control or TG2 siRNA, and a dentine resorption assay was performed. Black areas indicate resorbed lacunae on the dentin slices. The total resorption area, resorption area per osteoclast, and depth per pit were calculated from four randomly selected images. CTL, control. *P < 0.05; **P < 0.005 between indicated groups. Scale bar, 100 μm. The protein levels of TG2, Blimp1, c-Fos, and NFATc1 were analyzed by Western blotting. C, control siRNA; T, TG2 siRNA. (c) Cytoplasmic and nuclear fractions were obtained from cells cultured for two days with M-CSF and RANKL. Western blotting was performed to detect Blimp1 and NFATc1. PARP (full length) and α-tubulin were used as loading controls for cytoplasmic and nuclear extracts, respectively. CTL, control. (d) Cells cultured for two days after siRNA transfection were deprived of serum and factors for 5 h. Cells were then re-stimulated with RANKL (500 ng/ml) for the indicated times. Whole cell lysates were subjected to Western blot analysis with anti-phospho-IκB, -IκB and anti-phospho-p65 antibodies. (e) Nuclear fractions were prepared from cells stimulated as in (d). A p65 in TG2 KO mice compared with WT mice (Fig. 6c). The trabecular thickness (Tb.Th) and trabecular number (Tb.N) were also decreased in TG2 KO mice (Fig. 6c). Consistently, TG2 KO mice had a significantly higher level of trabecular separation (Tb.sp) compared with WT mice (Fig. 6c). To determine whether the decreased bone mass in TG2 KO mice was related to increased osteoclastogenesis, we analyzed bone tissue sections stained for TRAP activity or with hematoxylin and eosin (H&E) after demineralization (Fig. 6d). The number of osteoclasts per bone perimeter (N.Oc/B.Pm) and the level of osteoclast surface per bone surface (Oc.S/BS) were higher in TG2 KO versus WT mice (Fig. 6e). The eroded surface per bone surface (ES/BS) was also higher in TG2 KO mice (Fig. 6e). These results suggest that TG2 deficiency increased the generation of osteoclasts competent to perform the resorption function. As previous reports have implicated a positive role of some TG isoforms in osteoblast differentiation 28, 30 , we also analyzed whether the low bone mass phenotype of TG2 KO mice was associated with reduced in vivo osteoblastogenesis. The number of osteoblasts per bone perimeter (N.Ob/B.Pm) was not significantly different between WT and TG2 KO mice (Fig. 6e). Taken together, these observations indicate that TG2 affects the generation of osteoclasts, but not osteoblasts, in mice.
Osteoclastogenic potential of TG2 KO BMMs is higher than that of WT BMMs. We next investigated whether the reduced bone mass and increased osteoclastogenesis in TG2 KO mice were due to a cell autonomous difference in osteoclast precursors by comparing the osteoclastogenic potential of BMMs from TG2 KO and WT mice. The generation of osteoclasts from TG2 KO BMMs was about 4-fold higher than that from WT BMMs in the in vitro differentiation cultures (Fig. 7a). Both the number of TRAP-positive MNCs and the surface area per osteoclast were higher in TG2 KO cell cultures (Fig. 7b). The number of actin ring-positive cells and the sealing zone area per osteoclast were also higher in TG2 KO cells (Supplementary Figure 5). The cell proliferation rate was not different between WT and KO BMMs (Supplementary Figure 6). Consistent with the result of the siRNA experiments shown in Fig. 3, an enhancement of Blimp1, c-Fos, and NFATc1 was observed in TG2 KO cells ( Fig. 7c and d). We further examined the activation of NFATc1 by evaluating the intranuclear localization of NFATc1 in osteoclasts. Treatment with RANKL increased nuclear detection of NFATc1, and the percentage of cells with nuclear NFATc1 was higher in TG2 KO osteoclasts ( Fig. 7e and Supplementary Figure 7). Conversely, the expression of IRF8 and MafB was lower in TG2 KO cells (Fig. 7f). Collectively, these observations indicate that TG2 KO potentiates osteoclastogenesis by enhancing Blimp1-dependent NFATc1 activation in response to RANKL stimulation.

TG2 deficiency increases the activation of NF-κB signaling.
To further validate the target signaling pathways affected by TG2 deficiency, we evaluated the activation of the NF-κB pathway in WT and TG2 KO BMMs. RANKL-induced IκB phosphorylation was higher in TG2 KO BMMs than in WT BMMs 5 min after stimulation (Fig. 8a). The level of total IκB was lower in TG2 KO BMMs 30 min after RANKL stimulation (Fig. 8a). In addition, nuclear p65 level increased upon RANKL stimulation at 15 and 30 min, and this increase was greater in TG2 KO BMMs than in WT BMMs (Fig. 8b). Nuclear translocation of p65 was also assessed by confocal microscopy. As shown in Fig. 8c, TG2 KO BMMs had a higher percentage of positive cell for nuclear p65 compared to WT BMMs. This observation indicates that p65 translocation to the nucleus in response to RANKL was enhanced by TG2 deficiency. The MAPK activation by RANKL was not affected by the absence of TG2 (Supplementary Figure 8). These results from TG2 KO BMMs, together with the data from TG2 knockdown cells, support that the NF-κB signaling pathway is crucial for TG2-mediated regulation of osteoclast differentiation.

Discussion
Previous studies have suggested that TG activity is involved in the regulation of cell adhesion, matrix deposition, and differentiation of osteoblasts 28-30 . However, the relationship between TGs and osteoclast regulation has not been investigated. In this study, we explored the functional role of the TG2 isoform in osteoclast differentiation. Our in vitro and in vivo experiments revealed that TG2 negatively regulates osteoclastogenesis via a NF-κB-and Blimp1-dependent pathway.
Among the eight TG family members, TG2 was selectively expressed in osteoclasts (Fig. 1a). The reduction in TG2 level by TG2-targeted siRNA enhanced RANKL-dependent osteoclast differentiation in vitro and increased induction of c-Fos, NFATc1, and Blimp1 by RANKL (Figs 1 and 3). This negative association between TG2 level and osteoclastogenesis was confirmed using TG2 KO mice that displayed increased numbers of osteoclasts on the bone surface and less trabecular bone mass than WT mice (Fig. 6). Conversely, TG2 overexpression in RAW264.7 transcription activity assay was performed as described in Materials and methods. *P < 0.05 between indicated groups. (f) Seven predicted NF-κB binding sites on the promoter of Blimp1 are shown (upper). Chromatin immunoprecipitation assay with control IgG or p65 antibody was performed with RAW264.7 cells transfected with control or TG2 siRNA. PCR was performed using primer sets to detect the seven predicted sites (lower). C, control siRNA; T, TG2 siRNA. (g) BMMs transfected with control or TG2 siRNA were cultured with or without BAY 11-7085 (5 μM) for two days in the presence of M-CSF (30 ng/ml) and RANKL (100 ng/ml). Cells were then serum-starved for 5 h and re-stimulated with RANKL (500 ng/ml) for 5 min. Cell lysates were prepared and subjected to Western blotting. C, control siRNA; T, TG2 siRNA. (h) siRNA-transfected BMMs were cultured with or without BAY 11-7085 (5 μM) for two days in the presence of M-CSF (30 ng/ml) and RANKL (100 ng/ml). Western blotting was performed to assess the levels of the indicated proteins. C, control siRNA; T, TG2 siRNA. (i) siRNA-transfected BMMs were cultured in the presence of M-CSF (30 ng/ml) and RANKL (100 ng/ml). The mRNA levels of Irf8 and MafB were assessed by real-time PCR. **P < 0.005 versus control siRNA. C, control siRNA; T, TG2 siRNA. Full length Western blots are presented in Supplementary Figure 9.
cells suppressed osteoclastogenesis, further supporting the inhibitory role of TG2 in osteoclast differentiation (Fig. 5). The inhibitory function of TG2 despite apparent up-regulation by RANKL at early stages ( Fig. 3a and b) suggests that TG2 forms a negative feedback loop to fine-tune osteoclastogenesis, preventing excessive osteoclast differentiation and activation. This type of negative feedback regulation in osteoclastogenesis has been reported from our and other groups 35,36 .
In this study, we found that Blimp1 is an important mediator of TG2 function in the regulation of osteoclastogenesis. Blimp1 was first identified as a transcriptional repressor of the interferon beta promoter and has been characterized to play crucial roles in immune cell development, including the terminal differentiation of B-cells to plasma cells 13 . More recently, Blimp1 was also shown to be involved in the regulation of osteoclastogenesis through down-regulation of anti-osteoclastogenic transcription factors like IRF8 and MafB, which inhibit RANKL induction of NFATc1 14 . In addition, inhibition of Bcl6, which suppresses the expression of NFATc1, DC-STAMP, and cathepsin K, by Blimp1 was suggested to contribute to efficient osteoclast differentiation 37 . The first clue to the relationship between Blimp1 and TG2 was obtained in a study of lymphocytes 33 . In that study, spleen cells from TG2 KO mice showed a higher percentage of plasma cells and elevated level of Blimp1 mRNA in   response to immunization 33 , implying a negative regulation of Blimp1 expression by TG2. However, the potential mechanisms by which TG2 regulates Blimp1 expression have not been elucidated to date.
The regulation of Blimp1 expression by TG2 in RANKL-treated BMMs appeared to occur at the level of transcription (Figs 3a, 5d and 7c). This transcriptional regulation of Blimp1 is consistent with the results of a study using spleen B cells from TG2 KO mice 33 . TG2-mediated cross-linking and the subsequent inactivation of transcription factors, such as Sp1, have been suggested as the mechanism by which TG2 down-regulates gene expression 38 . Another mechanism of TG2-dependent transcriptional inhibition involves AP1; in this mode, TG2 interacts with c-Jun and consequently reduces dimmer formation between c-Jun and c-Fos 39 . We cannot exclude the possibility of an analogous means of direct inhibition by TG2 of the binding and/or activation of transcription factors that target the Blimp1 promoter. However, our Western blot results did not show the presence of TG2 protein in nuclear fractions (Fig. 3c), suggesting that other mechanisms are more likely to be responsible for the increase in Blimp1 caused by TG2 deficiency. Indeed, in our study, RANKL-induced phosphorylation of IκB was potentiated by TG2 knockdown (Fig. 3d), and an NF-κB inhibitor abrogated Blimp1 up-regulation by TG2 knockdown (Fig. 3h). Therefore, it is possible that TG2 inhibits a kinase that phosphorylates IκB, perhaps IKK, or stimulates a phosphatase that targets phosphorylated IκB in the cytosol. In this regard, the function of TG2 as a  . (a,b) BMMs obtained from WT and TG2 KO mice were cultured with M-CSF (30 ng/ml) and RANKL (100 ng/ml) for three days. Cells were then stained for TRAP. Scale bar, 100 μm. TRAP-positive MNCs were counted, and the surface area per osteoclast was measured. *P < 0.05; **P < 0.005 versus WT cells. (c,d) BMMs were cultured with M-CSF and RANKL for the indicated number of days. The mRNA levels of Blimp1 were determined by real-time PCR. *P < 0.05; **P < 0.005 versus WT cells. Cell lysates were analyzed by Western blotting with the indicated antibodies. (e) BMMs from WT and TG2 KO mice were treated with M-CSF and RANKL for two days. Cells were stained for NFATc1 and laminB. Anti-laminB staining marked the nuclear envelope. Cells with nuclear NFATc1 staining were counted. Scale bar, 20 μm. *P < 0.05 versus WT cells. (f) BMMs were cultured with M-CSF and RANKL for one day. The mRNA levels of Irf8 and MafB were determined by real-time PCR. *P < 0.05; **P < 0.005 versus WT cells. Full length Western blots are presented in Supplementary Figure 10.
GTPase signaling molecule 40 might be of particular interest. Further investigations will be required to reveal the precise mechanism accounting for this novel TG2 action on the NF-κB signaling pathway.
c-Fos also appears to be involved in the regulation of osteoclastogenesis by TG2. In TG2-deficient BMMs, the extent of c-Fos induction by RANKL was stronger relative to that of control cells (Figs 1, 3, 4 and 7). Conversely, TG2 overexpression led to weaker c-Fos induction by RANKL (Fig. 5). This effect of TG2 on c-Fos expression was not affected by NF-κB inhibition (Fig. 3h). In addition, the effect of Blimp1 knockdown on c-Fos expression was minimal (Fig. 4d). Therefore, the regulation of c-Fos by TG2 is likely to be independent of the NF-κB-Blimp1 axis. As c-Fos is a component of the AP1 transcription factor and its expression is positively auto-regulated, the Whole cell lysates were analyzed by Western blot using anti-phospho-IκB and -IκB antibodies. (b) Nuclear and cytosolic fractions prepared from WT and TG2 KO BMMs were stimulated with RANKL and subjected to Western blot analysis. PARP (full length) and α-tubulin were used as loading controls for nuclear and cytoplasmic extracts, respectively. The ratios of p65 to PARP were determined by densitometry. (c) WT and TG2 KO BMMs were serum-starved for 5 h and stimulated with or without RANKL (500 ng/ml) for 15 min. Cells were stained for p65 and laminB. Anti-laminB staining marked the nuclear envelope. Cells with nuclear p65 staining were counted. Scale bar, 20 μm. *P < 0.05 versus WT cells. (d) Schematic illustration of a proposed mechanism through which TG2 regulates osteoclastogenesis. A reduction in TG2 enhances the activation of NF-κB p65, which increases the expression of Blimp1. Blimp1 together with c-Fos augments the expression and nuclear translocation of NFATc1 to enhance osteoclastogenesis induced by RANKL. Full length Western blots are presented in Supplementary Figure 10.
suppressive effect of TG2 on c-Fos induction might be attributed to the aforementioned mechanism of interaction of TG2 with c-Jun and the resulting reduction in AP1 dimer formation. Although the identification of the direct targets of TG2 required for regulation of c-Fos and NF-κB was beyond the scope of this study, we clearly demonstrated that two separate pathways involving NF-κB and Blimp1 on one axis and c-Fos on the other axis mediated the effect of TG2 on osteoclastogenesis (Fig. 8d).
In our study, TG2 KO mice showed increased numbers of osteoclasts on the bone surfaces compared with WT mice. However, the number of osteoblasts was not significantly different between TG2 KO and WT mice (Fig. 6e). Therefore, the lower bone mass of TG2 KO mice is likely to be due to the higher bone-resorbing activity of osteoclasts rather than to lower bone-forming activity by osteoblasts. This observation might appear to be inconsistent with previous reports where TG enzymatic activities were implicated in matrix mineralization and osteoblastic differentiation 29,30 . However, clear evidence supporting the contribution of TGs, specifically the TG2 isoform, to bone formation or osteoblast differentiation has not been provided to date. Indeed, TG2 KO mice did not show any evident defect in bone development upon macro-and microscopic examination 41 . The normal skeletal development of TG2 KO mice has been suggested to be due to compensatory overexpression of FXIIIA and TGF-β 42 . While it remains to be determined whether FXIIIA compensates for the lack of TG2 activity during osteoblastogenesis in TG2 KO mice, TG2 deficiency was not compensated for osteoclastogenesis in our study (Figs 6 and 7). The lack of expression of other TG isoforms, with only fairly weak expression of FXIIIA, during osteoclastogenesis (Fig. 1a) might explain the overt effect of TG2 deficiency on osteoclasts but not on osteoblasts.
Our present study revealed for the first time a regulatory role of TG2 in osteoclast differentiation. We further demonstrated that this TG2 function was mediated by the NF-κB-Blimp1 and c-Fos pathways that converge on NFATc1. Further investigation of the direct target molecules of TG2 and the mode of TG2 action on the targets will be required to comprehensively understand the role of TG2 associated with the regulation of osteoclastogenesis and bone metabolism.

Materials and Methods
Reagents. Re combinant RANKL and M-CSF were purchased from Peprotech EC Ltd. (London, UK).
Protein G Agarose for 1 h and incubated with p65 antibody (1:200) or control rabbit IgG overnight at 4 °C with rotation. Next day, the antibody/antigen/DNA complexes were collected by Protein G agarose incubation for 1 h. The complexes were eluted with 1% SDS and 0.1 M NaHCO 3 and reverse crosslinked with NaCl (overnight incubation at 65 °C). DNA was purified using spin columns and PCR was performed (32 cycles) using primers covering putative binding sites of p65 on the Blimp1 promoter. All primer sets are listed in Supplementary Table 1. Preparation of cytoplasmic and nuclear proteins. Cytoplasmic and nuclear proteins were prepared from BMMs using the NE-PER Nuclear and Cytoplasmic Extraction Reagents Kit according to the manufacturer's instructions. In brief, cultured cells were washed with PBS and treated with the cell lysis buffer provided in the kit. After centrifugation, the supernatant was transferred to a new tube as cytoplasmic extract. The pellet was washed twice and lysed with the nuclear lysis reagent provided in the kit. After centrifugation, the supernatant was used as the nuclear protein extract. Protein concentrations were determined using a detergent-compatible colorimetric assay kit (Bio-Rad Laboratories, Hercules, CA, USA).

RT-PCR and real-time PCR.
After isolation of total RNA using TRIZOL (Invitrogen), reverse transcription was performed using 2 μg of total RNA and Superscript II reverse transcriptase (Invitrogen) according to the manufacturer's protocol. PCR was performed using 26 cycles of denaturation at 95 °C for 15 sec and annealing/ extension at 60 °C for 60 sec. PCR products were separated in 1.5% agarose gels and stained with ethidium bromide. The HPRT housekeeping gene was used as an input control. Real-time PCR was performed using the ABI 7500 real-time system with the KAPA SYBR FAST qPCR kit (Kapa Biosystems, Wilmington, MA, USA). The detector was programmed with the following PCR conditions: 40 cycles of 3 sec denaturation at 95 °C and 33 sec amplification at 60 °C. All reactions were run in triplicate. The mRNA levels of target genes were normalized to the mRNA level of the HPRT gene, and values indicating the fold-changes from the controls are shown. Relative differences in mRNA levels were evaluated using the comparative cycle threshold method. All primer sets for RT-PCR and real-time PCR are listed in Supplementary Table 2.
Gene cloning and lentiviral transduction. To generate TG2-overexpressing stable RAW264.7 cells, the 2061 bp cDNA (GenBank accession number NM_009373.3) fragment of TG2 was obtained by RT-PCR of mRNA from BMMs. The PCR reaction steps included an initial denaturation at 95 °C for 5 min, followed by reaction cycles of denaturation at 95 °C for 30 sec, annealing at 56 °C for 30 sec, and extension at 72 °C for 4 min. The following primers were used: TG2, 5′-GGACTAGTGCCACCATGGCAGAGGAGCTGCTC-3′ (sense) and 5′-TCCCCGCGGGGCCGGGCCGATGATAAC-3′ (antisense). The pLenti6 lentiviral vector and the TG2 PCR product were cleaved with SpeI and SacII restriction enzymes and subjected to 1% agarose gel electrophoresis. Each DNA fragment was purified from the gel, and the TG2 fragment was ligated to the pLenti6 vector DNA. To generate lentiviral particl es, 293FT cells were transfected with pLenti6-LacZ (control) or pLenti6-TG2 using PolyFect Transfection Reagent (Qiagen, Leipzig, Germany). Culture supernatant containing viral particles was harvested after 48 h and passed through a 0.45 μm syringe filter. RAW264.7 cells were infected with viral supernatant mixed with polybrene (6 μg/ml) for 18 h. After an overnight incubation, infected RAW264.7 cells were selected with blasticidin (1:2000) for 15 days. Overexpression of TG2 was determined by measuring mRNA and protein levels.
Western blotting. BMMs and RAW264.7 cells were lysed with a lysis buffer containing 120 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% NP40, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na 3 VO 4 , 1 mM NaF, and protease inhibitors (Roche, Mannheim, Germany). The protein concentrations of cell lysates were determined using the DC protein assay kit (Bio-Rad), and equal amounts of protein were loaded onto 10% or 12% SDS-polyacrylamide gels. After transfer onto nitrocellulose membranes (Amersham Pharmacia, Uppsala, Sweden) and blocking with 5% nonfat skim milk, primary antibodies were added, and membranes were incubated overnight at 4 °C. After washing, the membranes were incubated with HRP-conjugated secondary antibodies in 2% skim milk for 1 h. Immunoreactivity was detected with ECL reagents. Comparable loading was verified by reprobing the same membranes with anti-β-actin antibody.
TRAP staining and osteoclast measurement. To generate osteoclasts, BMMs and RAW 264.7 cells were plated in 48-well tissue culture plates at densities of 4 × 10 4 cells/well and 1 × 10 4 cells/well, respectively. Cells were cultured in the presence or absence of RANKL (100 ng/ml) for 3 or 4 days. BMM cultures were supplemented with M-CSF (30 ng/ml). The culture medium was changed at day 2. Under these conditions, TRAP-positive mononuclear cells were detected at day 2, and TRAP-positive MNCs (≥ 3 nuclei) were observed at days 3 and 4. The cells were fixed with 3.7% formaldehyde for 30 min at room temperature. After permeabilization with 0.1% Triton X-100 for 5 min, TRAP staining was performed in the dark for 30 min using a TRAP-staining kit following the manufacturer's instructions. After staining, cells were washed with distilled water and observed under a light microscope. The total number of TRAP-positive MNCs and the surface area per osteoclast were measured using Osteomeasure software (OsteoMetrics, Inc., Decatur, GA, USA).
Resorption assay. Bone resorption assays were performed as previously described 44 . Briefly, BMMs seeded on dentine discs (Osteosite Dentine Discs, Immunodiagnostic Systems Inc., Boldon, UK) were stimulated with RANKL for seven days. After removing cells by treating the dentine discs with 5% sodium hypochlorite for 10 min and wiping with cotton swab, the discs were photographed. The areas and depths of the resorption pits were measured with a Carl Zeiss LSM 5-PASCAL laser-scanning microscope (Carl Zeiss Microimaging GmbH, Goettingen, Germany).
Three-dimensional micro-computed tomography analysis. Femurs of 13-week-old female TG2 KO mice and WT littermate mice (6 per genotype) were fixed in 4% (vol/vol) paraformaldehyde overnight and analyzed by micro-computed tomography using the SkyScan 1072 system (SkyScan, Kontich, Belgium). Trabecular bone parameters were measured in 1-mm thick trabecular regions, starting 1 mm below the proximal edge of the growth plate in the distal end of the femur. A total of 350-400 tomographic slices were acquired and three-dimensional analyses were performed with CT-volume software (ver 2.1.1.0; Skyscan).
Histomorphometric analysis. Bone histomorphometric analyses were performed using paraffin-embedded sections. In brief, femurs were fixed in 4% paraformaldehyde, decalcified in 12% EDTA for 4 weeks, and embedded in paraffin. The paraffin blocks were serially sectioned at a thickness of 6 μm and stained with TRAP/hematoxylin or hematoxylin/eosin according to standard procedures. Trabecular bone parameters were analyzed using Osteomeasure software.

Statistical analysis.
All experiments, except the in vivo KO mouse analysis, were performed three or more times. Data are presented as mean ± SD. Statistical significance was determined using Student's t test for in vitro and in vivo studies. P values less than 0.05 were considered to be statistically significant.