E3 Ubiquitin Ligase Fbw7 Negatively Regulates Osteoblast Differentiation by Targeting Runx2 for Degradation*

Runx2, a master regulator of osteoblast differentiation, is tightly regulated at both transcriptional and post-translational levels. Post-translational modifications such as phosphorylation and ubiquitination have differential effects on Runx2 functions. Here, we show that the reduced expression and functions of Runx2 upon its phosphorylation by GSK3β are mediated by its ubiquitin-mediated degradation through E3 ubiquitin ligase Fbw7α. Fbw7α through its WD domain interacts with Runx2 both in a heterologous (HEK293T cells) system as well as in osteoblasts. GSK3β was also present in the same complex as determined by co-immunoprecipitation. Furthermore, overexpression of either Fbw7α or GSK3β was sufficient to down-regulate endogenous Runx2 expression and function; however, both failed to inhibit endogenous Runx2 when either of them was depleted in osteoblasts. Fbw7α-mediated inhibition of Runx2 expression also led to reduced Runx2 transactivation and osteoblast differentiation. In contrast, inhibition of Fbw7α restored Runx2 levels and promoted osteoblast differentiation. We also observed reciprocal expression levels of Runx2 and Fbw7α in models of bone loss such as lactating (physiological bone loss condition) and ovariectomized (induction of surgical menopause) animals that show reduced Runx2 and enhanced Fbw7α, whereas this was reversed in the estrogen-treated ovariectomized animals. In addition, methylprednisolone (a synthetic glucocorticoid) treatment to neonatal rats showed a temporal decrease in Runx2 with a reciprocal increase in Fbw7 in their calvarium. Taken together, these data demonstrate that Fbw7α negatively regulates osteogenesis by targeting Runx2 for ubiquitin-mediated degradation in a GSK3β-dependent manner and thus provides a plausible explanation for GSK3β-mediated bone loss as described before.

Runx2, a master regulator of osteoblast differentiation, is tightly regulated at both transcriptional and post-translational levels. Post-translational modifications such as phosphorylation and ubiquitination have differential effects on Runx2 functions. Here, we show that the reduced expression and functions of Runx2 upon its phosphorylation by GSK3␤ are mediated by its ubiquitin-mediated degradation through E3 ubiquitin ligase Fbw7␣. Fbw7␣ through its WD domain interacts with Runx2 both in a heterologous (HEK293T cells) system as well as in osteoblasts. GSK3␤ was also present in the same complex as determined by co-immunoprecipitation. Furthermore, overexpression of either Fbw7␣ or GSK3␤ was sufficient to down-regulate endogenous Runx2 expression and function; however, both failed to inhibit endogenous Runx2 when either of them was depleted in osteoblasts. Fbw7␣-mediated inhibition of Runx2 expression also led to reduced Runx2 transactivation and osteoblast differentiation. In contrast, inhibition of Fbw7␣ restored Runx2 levels and promoted osteoblast differentiation. We also observed reciprocal expression levels of Runx2 and Fbw7␣ in models of bone loss such as lactating (physiological bone loss condition) and ovariectomized (induction of surgical menopause) animals that show reduced Runx2 and enhanced Fbw7␣, whereas this was reversed in the estrogen-treated ovariectomized animals. In addition, methylprednisolone (a synthetic glucocorticoid) treatment to neonatal rats showed a temporal decrease in Runx2 with a reciprocal increase in Fbw7 in their calvarium. Taken together, these data demonstrate that Fbw7␣ negatively regulates osteogenesis by targeting Runx2 for ubiquitin-mediated degradation in a GSK3␤-dependent manner and thus provides a plausible explanation for GSK3␤-mediated bone loss as described before.
Runt domain-containing protein Runx2 is a critical regulator of the osteogenic lineage (1). Null mutation of Runx2 gene results in unmineralized skeleton in mice, and its heterozygotic loss results in human cleidocranial dysplasia (CCD) 2 phenotype (2,3), which demonstrates the importance of Runx2 in bone biology. Runx2 induces transcription of osteopontin (OP) and osteocalcin (OC) by binding to osteoblast-specific cis-acting OSE2 elements in the promoter regions of these genes (4,5). Runx2 expression is regulated by vitamin D 3 , estrogen, TGF-␤/ BMP2, and Wnt signaling pathways (6 -10).
In comparison with transcriptional regulation, the posttranslational regulation of Runx2 in osteoblasts is relatively unexplored (11). Recent studies indicate that phosphorylation, acetylation, SUMOylation, and ubiquitination regulate Runx2 functions differently (11). The effects of phosphorylation by different kinases on the Runx2 protein stability and its subsequent function have been reported earlier (12)(13)(14)(15); however, the identification of E3 ubiquitin ligases in such processes has remained elusive.
F-box protein Fbw7 (also known as Fbwx7 and cdc4) is a component of the SCF (Skp-Cullin-F box) ubiquitin ligase complex that recognizes and binds to its substrates through a consensus phospho-binding motif ((T/S)PXX(S/T/D/E)). known as Cdc4-phosphodegron (CPD). The recognition of CPD by SCF Fbw7 is regulated by phosphorylation of the substrate at a threonine/serine residue within this CPD motif (16). This phosphorylated CPD binds with three conserved arginine residues within WD40 repeats in the C terminus of Fbw7 and is subsequently targeted for degradation by SCF Fbw7 . The majority of the known Fbw7 substrates are phosphorylated by GSK3␤ within their CPD including c-Myc (17), MCL-1 (18), , cyclin E (20), and PGC-1␣ (21,22). These studies suggest Fbw7 to be an important protein involved in the development and regulation of cellular differentiation by targeting multiple substrate molecules for degradation.
Recently, SCF FBW7 was shown to inhibit osteogenesis by regulating OASIS expression through ubiquitin-mediated proteasome degradation (23). In osteoblasts, Runx2 levels are regulated coordinately with the cell cycle machinery, apparently by E3 ubiquitin ligases (14,24). Runx2 expression in osteoblasts is also down-regulated upon its phosphorylation by GSK3␤ (25) and JNK1 (15)  some issues and brought them to the attention of the Journal. Subsequently, some issues were also noted in the article by the Journal. Based on the original data provided for Fig. 2C, the human Runx2 construct resolves at the size of -actin. While they grossly differ in size (size of Runx2 is ~60 kDa while that of -actin is 42 kDa), the withdrawing authors claim that the loading control was possibly probed with -tubulin (~55 kDa), which separates at an almost similar size as Runx2 and was erroneously labeled as -actin.
The Runx2 immunoblots from Figs. 2E and 2J were duplicated, and the FLAG Fbw7 immunoblot from Fig. 2J was inappropriately manipulated. Additionally, the actin immunoblots from Figs. 3B and 3C were duplicated, the graphs shown in Figs. 4A and 4B were duplicated, and a non-matching image due to overcropping was used in the Fbw7 panel for the Ovx + E2 group in Fig. 6A, which the withdrawing authors state were inadvertent errors. Given these errors, the withdrawing authors state that the responsible course of action would be to withdraw the article to maintain the high standards and rigor of scientific literature from the withdrawing authors' group as well as the Journal. However, the withdrawing authors state that these errors do not change the underlying scientific findings of the article.
Osteoblast Culture-Mice calvarial osteoblasts were obtained using sequential digestion method as described previously (31). Briefly, calvaria of 1-2-day-old mouse pups (both sexes) were pooled and then cleaned by removing the sutures and adherent mesenchymal tissues. Further, calvaria were subjected to five sequential (10 -15 min) digestions at 37°C in ␣-MEM with 0.1% dispase and 0.1% collagenase P. Cells collected from the second to fifth digestions were plated in T25 flasks (25 cm 2 ) in ␣-MEM containing 10% FBS and 1% penicillin/streptomycin (complete growth medium). Cultures of mouse calvarial osteoblasts upon 80% confluence were trypsinized and seeded as per experiment.
Co-immunoprecipitation and in Vivo Ubiquitination Assay-MC3T3-E1 cells were transfected with Fbw7␣ and its mutants Fbw7␣⌬F and Fbw7␣ WD . 48 h after transfection, whole cell lysates were prepared in radioimmunoprecipitation assay buffer supplemented with protease and phosphatase inhibitor. Co-immunoprecipitation was performed using equal amounts of precleared lysates with anti-FLAG M2 antibody and precalibrated protein A/G-agarose beads (Millipore) at 4°C overnight. On the next day, co-precipitates were immunoblotted with the indicated antibodies. For in-cell ubiquitination assay, 293T cells were transfected with Fbw7␣ and HA-ubiquitin. 48 h after transfection, whole cell lysates were prepared in radioimmunoprecipitation assay buffer supplemented with protease and phosphatase inhibitors and subjected to co-immunoprecipitation with anti-Runx2 antibody. Co-precipitates were immunoblotted with anti-HA and anti-Runx2 antibodies. A Western blotting and cycloheximide half-life experiment was performed as described previously (32,33). Protein samples from femur bone of different experimental groups were isolated as described by Wejheden et al. (34). Immunoblotting was performed as described previously (21,35).
Luciferase Promoter Assay-1 ϫ 10 5 HEK 293T or MC3T3-E1 cells/well were plated 1 day before transfection. On the next day, cells were transfected with p6OSE-luc promoter, siFbw7, and expression plasmids for Runx2, Fbw7, and its mutants (Fbw7␣⌬F and Fbw7 WD ). 24 h after transfection, cells were lysed and assayed for luciferase activity using luciferase assay reagent (Promega, Madison, WI). Data are presented as means of triplicate values obtained from representative experiments.
Immunohistochemistry-Bones of different groups were decalcified in EDTA followed by block preparation and sectioning. After de-waxing of the samples, 1% rat serum in PBS with 0.1% Triton X-100 (Sigma) was used for antigen retrieval for 90 min. Primary antibodies were prepared in 0.5% BSA in Milli-Q water. Primary antibody dilution was used according to the datasheet of the antibody. Runx2 (Abcam) and Fbw7 (Sigma) were used and incubated overnight at 4°C. Slides were washed three times in PBS followed by incubation with fluorophoretagged secondary antibodies for 1 h at room temperature. Alexa Fluor 488 donkey anti-rabbit IgG (A21206, Invitrogen) for Runx2 and Alexa Fluor chicken anti-mouse IgG (A2120, Invitrogen) for Fbw7 were used. Secondary antibodies were also prepared in 0.5% BSA in Milli-Q water. Slides were washed three times in PBS for 10 min each. 1 g/ml DAPI (Sigma) was used as a nuclear stain. Slides were washed with PBS for 5 min and mounted with anti-fade chemical for subsequent confocal microscopy.
Mineralized Nodule Formation, Alkaline Phosphatase Assay, and Alizarin Staining-For mineralization studies, mouse calvarial osteoblasts were seeded in 6-well plates (20,000 cells/ well) in osteoblast growth medium. After 72 h, siControl and siFbw7-transfected cells were then cultured in complete osteoblast growth medium and differentiation induction medium (DIM) containing ␣-MEM with 10 mM ␤-glycerophosphate and 50 g/ml ascorbic acid supplemented with 10% FBS and cultured for 15 days with a change of medium every 2 days. At the end of the experiment, cells were fixed with 4% paraformaldehyde. Alizarin Red-S was used for staining of mineralized nodules followed by extraction of the stain using 10% cetylpyridinium chloride for colorimetric determination of the dye at 550 nm. Similarly, for mineralization studies in MC3T3-E1 cells, 60 -70% confluent cells were plated and then transfected with Fbw7 Negatively Regulates Runx2-dependent Osteogenesis either siFbw7 or siControl and cultured in either osteoblast growth medium or DIM.
Animal Experiments and Procedures-Sprague-Dawley rats (4 months old; 220 Ϯ 20 g) received ovariectomy (OVX) or were sham-operated. One week after surgery, rats were weight-randomized into three groups (n ϭ 6) for 12 weeks: sham-operated (ovary intact)ϩvehicle, OVXϩvehicle, and OVXϩE2 (17␤-estradiol, 5 g/kg/day subcutaneously) (36). At the end of 12 weeks, all groups were killed, and lactating dams 10 days after parturition were also included (n ϭ 6). Femurs were collected for protein isolation and sectioning. All animal care and experimental procedures performed were approved by Institutional Animal Ethical Committee guidelines. Sprague-Dawley rats (three per cage) were housed in temperature-controlled (22-24°C) rooms with maintained fresh air supply, 100% air exhaust, and 60 -70% relative humidity. Rooms had standard diffuse lighting (200 -300 lux) with automatic maintenance of a diurnal 12-h light cycle. The animals were fed standard ad libitum (chow) diet and had free access to reverse osmosis water. The experimental procedures were approved and conducted in accordance with the Institutional Animal Ethics Committee of Council of Scientific and Industrial Research-Central Drug Research Institute (CSIR-CDRI). For ovariectomy and sham surgery, anesthesia was induced with xylazine (3-10 mg/kg) and ketamine (80 -100 mg/kg) mixture (intraperitoneal injection). Euthanasia and disposal of carcass were in compliance with the Institutional Animal Ethics Committee (IAEC) guidelines.
Data Analysis and Statistics-Results are expressed as mean Ϯ S.E. All data were analyzed using GraphPad Prism 5.0 (GraphPad, San Diego, CA). One-way analysis of variance followed by Tukey's multiple comparison test was used to analyze data involving more than two groups.

Fbw7 Negatively Regulates Steady State Levels of Runx2-
Analysis of Runx2 protein sequence revealed the presence of several potential CPD motifs that may be recognized by Fbw7. Four such motifs starting at amino acid positions Ser 14 , Ser 279 , Thr 305 , and Thr 387 matching with the CPDs in other known substrates of Fbw7 are present in Runx2 (Fig. 1a). These CPD sequences in Runx2 are conserved in all isoforms of human Runx2 as well as in mouse and rat, which prompted us to evaluate whether SCF Fbw7 targets Runx2 for degradation. To this end, whole cell extracts (WCEs) of mouse preosteoblast cell line MC3T3-E1 transfected with three isoforms of Fbw7 (␣, ␤, and ␥ individually) were immunoblotted with anti-Runx2 and anti-FLAG-M2 antibodies, and the results showed that all three Fbw7 isoforms substantially down-regulated endogenous Runx2 protein expression (Fig. 1b). Because Fbw7␣ is the dominant isoform of Fbw7, we next co-transfected Runx2 with increasing amounts of Fbw7␣ in HEK293T cells (henceforth 293T). Immunoblot with anti-Runx2 and anti-FLAG antibodies again showed a persistent down-regulation of Runx2 with increasing amounts of Fbw7␣ (Fig. 1c). This finding suggested that Fbw7␣ (hereby referred to as Fbw7) negatively regulated the steady state levels of endogenous as well as overexpressed Runx2.
We next determined whether Fbw7␣ also regulated steady state levels of Runx2 in osteoblasts. To address this, we transfected mouse preosteoblastic MC3T3-E1 cells with increasing amounts of Fbw7 in the presence or absence of proteasome inhibitors MG132 and lactacystin (LCN). Similar to 293T cells, increasing amounts of Fbw7 transfection in osteoblasts reduced endogenous Runx2 protein, whereas proteasome inhibitors abolished the inhibitory effect of Fbw7 on Runx2 protein (Fig.  1d). To further validate whether Runx2 suppression is Fbw7␣specific, we overexpressed Runx2 with Fbw7␣ and its nonfunctional deletion mutants Fbw7␣⌬F (which lacks the F box domain that binds directly to the SKP1 component of SCF ubiquitin ligase) and dnFbw7␣ WD (which contains only a stretch of WD40 repeats, the protein interaction domains in Fbw7 that bind to the substrate) in 293T cells. Contrary to wild type Fbw7␣, the mutant constructs of Fbw7 instead enhanced Runx2 protein level, thus confirming that Runx2 suppression was indeed mediated by Fbw7 (Fig. 1e). Similar results were observed in MC3T3-E1 cells where transfection with wild type Fbw7 reduced endogenous Runx2 protein level, whereas the non-functional mutants instead stabilized it (Fig. 1f). To ascertain that Fbw7 regulated Runx2 protein stability, MC3T3-E1 cells were treated with cycloheximide to inhibit new protein synthesis, and the rate of degradation of endogenous Runx2 along with Runx2⌬N⌬C (region containing 104 -346 amino acids of mouse isoform 5 that still has two conserved CPDs) in the absence or presence of Fbw7 was assessed. Immunoblot with anti-Runx antibody followed by anti-FLAG showed rapid degradation of both endogenous Runx2 as well as Runx2⌬N⌬C when co-transfected with Fbw7 (Fig. 1g), suggesting that Runx2 down-regulation by Fbw7 was indeed due to loss of Runx2 protein stability. As an E3 ubiquitin ligase, Fbw7 is expected to target its substrate for ubiquitin-mediated proteasome degradation; we thus studied whether Fbw7 ubiquitinated Runx2. To address this, we performed in-cell ubiquitination of endogenous Runx2 by transiently overexpressing His-ubiquitin either alone or together with FLAG-Fbw7 or siFbw7 as indicated. Endogenous Runx2 co-immunoprecipitated with anti-Runx2 antibody and subsequently immunoblotted with anti-His, anti-Runx2, and anti-Fbw7 antibodies, respectively, by repeated stripping and probing the same membrane showed increased polyubiquitinated bands of Runx2 in Fbw7-overexpressed condition, whereas Fbw7 knockdown mitigated Runx2 polyubiquitination, suggesting that Fbw7 indeed promoted ubiquitinmediated proteasome degradation of Runx2 (Fig. 1h). Taken together, these data demonstrated that Fbw7 mitigated steady state levels of Runx2 by ubiquitin-mediated proteasome degradation.
Fbw7 through Its WD Domain Physically Associates with Runx2 in Its Runx2⌬N⌬C Region-Because Fbw7 is the substrate-binding moiety of the SCF Fbw7 ubiquitin ligase complex, we therefore determined whether Fbw7 interacted with Runx2. To address this, we used wild type Runx2 (mouse isoform 5 and human isoform 2) and its deletion mutants schematically represented in Fig. 2a. Although Runx2⌬N⌬C lacks N and C termini, it still contains two of the potential CPDs starting at Ser 279 and Thr 305 (Fig. 1a). Interestingly, Ser 279 and Ser 283 that are present in one of the CPDs starting at Ser 279 have been shown Fbw7 Negatively Regulates Runx2-dependent Osteogenesis DECEMBER 25, 2015 • VOLUME 290 • NUMBER 52 to be phosphorylated by GSK3␤, rendering Runx2 transcriptionally inactive (25). We therefore determined whether Fbw7 also targets this Runx2 mutant by co-transfecting it in 293T cells with Fbw7 and its mutants. Immunoblot with Runx2 and FLAG antibodies showed that wild type Fbw7 but not its mutants down-regulated both Runx2 and Runx2⌬N⌬C, suggesting that CPDs present in Runx2⌬N⌬C were important (Fig.  2b). However, the data did not rule out the possibility that the CPDs present in the N or C terminus of Runx2 were not involved in Fbw7-mediated targeting of Runx2. Co-transfec-  FIGURE 2. Fbw7 through its WD domain physically associates with Runx2 in its Runx2⌬N⌬C region. a, Runx2 and its deletion mutants containing different regions with exact amino acids are schematically represented for both mouse Runx2 (mRunx2, isoform 5, 1-528 amino acids) and human Runx2 (hRunx2, isoform 2, 1-507 amino acids). b, Lysates of 293T cells transfected with 0.5 g of Runx2 or Runx2⌬N⌬C either alone or together with Fbw7␣ and its deletion mutants (Fbw7␣⌬F, Fbw7␣ WD ) were immunoblotted (IB) with anti-FLAG-M2 and anti-Runx2 antibodies. c, lysates of 293T cells transfected with 0.5 g of human Runx2 or its deletion mutants (all contained two of the potential CPDs present in Runx2⌬N⌬C) either alone or together with Fbw7␣ were immunoblotted with anti-FLAG-M2 and anti-Runx2 antibodies. d, Fbw7 co-immunoprecipitates (IP) using anti-Fbw7 antibody from lysates of MC3T3-E1 cells were immunoblotted with anti-Fbw7 and anti-Runx2 antibodies. Cells were treated with MG132 (10 M) 6 h before lysate preparation. e, Fbw7 co-immunoprecipitates using anti-FLAG-M2 antibody from lysates of MC3T3-E1 cells transfected with 0.5 g of Fbw7 were immunoblotted with anti-FLAG-M2 and anti-Runx2 antibodies. Cells were treated with MG132 (10 M) 6 h before lysate preparation and lithium chloride (20 mM) 24 h before lysate preparation. f, Fbw7 co-immunoprecipitates from lysates of MC3T3-E1 cells transfected with 50 nM siGSK3␤ were immunoblotted with anti-Fbw7, anti-GSK3␤, and anti-Runx2 antibodies. g, Fbw7 co-immunoprecipitates using anti-FLAG M2 antibody from lysates of 293T cells transfected with 1.0 g of Runx2 or its deletion mutants (Runx2⌬N⌬C, Runx2C, or Runx2N) either alone or together with FLAG-Fbw7␣ were immunoblotted with anti-FLAG-M2 and anti-Runx2 antibodies. h and i, Fbw7 co-immunoprecipitates using anti-Fbw7 antibody from lysates of HEK293T cells transfected with 1.0 g of FLAG-Runx2N or FLAG-Runx2C either alone or together with FLAG-Fbw7 were immunoblotted with anti-FLAG-M2 antibody. IgG-HC: heavy chain; IgG-LC: light chain. j, Fbw7 co-immunoprecipitates using anti-FLAG M2 antibody from lysates of MC3T3-E1 cells transfected with 0.5 g of FLAG-Fbw7␣ or its deletion mutants (Fbw7␣⌬F or Fbw7␣ WD ) were immunoblotted with anti-FLAG-M2 and anti-Runx2 antibodies. Data are representative of minimum three independent experiments. DECEMBER 25, 2015 • VOLUME 290 • NUMBER 52 tion of human Runx2 (human isoform 2) and its deletion mutants (these mutants also contained the two potential CPDs as present in Runx2⌬N⌬C) either alone or together with Fbw7 in 293T cells showed similar results (Fig. 2c). Further, we assessed endogenous interaction between Runx2 and Fbw7 via co-immunoprecipitation of Fbw7 (anti-mouse) from WCEs of MG132-treated or untreated MC3T3-E1 cells as indicated. Immunoblot with anti-Runx2 antibody (anti-rabbit) showed a clear interaction with Fbw7 (Fig. 2d). Furthermore, we also confirmed this endogenous interaction between Runx2 and Fbw7 via co-immunoprecipitation of Fbw7 from WCEs of MC3T3-E1 cells transfected with FLAG-Fbw7 and treated with proteasome inhibitors MG132 and LiCl as indicated. Immunoblot with anti-Runx2 antibody again showed a clear interaction with Fbw7. An intense interaction was observed in conditions where cells were treated with MG132 or LiCl, which further affirmed that Fbw7 physically interacted with Runx2 under physiological conditions and regulated its protein stability in osteoblasts (Fig. 2e). To further validate the role of GSK3␤ in Fbw7-mediated Runx2 degradation, we transiently knocked down GSK3␤ (50 nM) in MC3T3-E1 cells and subsequently assessed interaction between Runx2 and Fbw7. Endogenous Fbw7 co-immunoprecipitated with anti-Fbw7 antibody followed by immunoblotting with anti-Runx2, anti-Fbw7, and anti-GSK3␤ antibodies, respectively, showed greater interaction of Runx2 with Fbw7 in siGSK3␤-transfected condition as compared with cells alone (Fig. 2f), suggesting that depletion of GSK3␤ rescued Runx2, leading to enhanced interaction between Runx2 and Fbw7. To further define the region of Runx2 interacting with Fbw7, we co-transfected Fbw7 with Runx2 and its mutants in 293T cells. Immunoblot with anti-Runx2 followed by anti-FLAG antibody showed clear interaction of Fbw7 with both Runx2 and Runx2⌬N⌬C, although no interaction was seen with either Runx2C (amino acids 1-104) or Runx2N (amino acids 347-528) (Fig. 2g). Furthermore, to rule out any physical interaction of Fbw7 with either Runx2N or Runx2C, we co-transfected HEK293T cells with FLAG-Runx2N or FLAG-Runx2C either alone or together with Fbw7. Fbw7 was co-immunoprecipitated with anti-Fbw7 antibody followed by immunoblotting with anti-FLAG antibody. Notably, no physical interaction was observed in either condition, suggesting that both the N terminus and the C terminus of Runx2 were dispensable for physical interaction between Runx2 and Fbw7 (Fig. 2, h and i). Moreover, these data were in agreement with our finding that Fbw7 targeted Runx2 and Runx2⌬N⌬C for degradation and also confirmed that CPDs present in Runx2⌬N⌬C were important for Fbw7-mediated down-regulation of Runx2. In a bid to determine the specific region of Fbw7 that interacted with Runx2, we co-transfected 293T cells with Runx2 together with Fbw7 and its mutants. Immunoblot with anti-Runx2 antibody followed by anti-FLAG antibody after stripping the same membrane demonstrated that Runx2 interacted with both Fbw7 and its mutants including Fbw7 WD , which contained only WD repeats (Fig. 2j). Taken together, these data demonstrate that Runx2 through its Runx2⌬N⌬C region interacts with WD domain of Fbw7 in osteoblasts.
Fbw7-mediated Degradation of Runx2 Negatively Impacts Its Transactivation Function-As our data so far demonstrated that Fbw7 destabilizes Runx2 by targeting it for degradation, we next asked whether the Fbw7-mediated Runx2 degradation has any effect on Runx2 transactivation potential. To answer this, we performed a luciferase reporter assay using Runx2-responsive p6OSE2-Luc. We observed that co-transfection of Fbw7 with Runx2 in 293T cells strongly inhibited Runx2 transcriptional activity in a dose-dependent manner (Fig. 4a). In contrast, co-transfection of Fbw7 deletion mutants with Runx2 enhanced transactivation by Runx2. The same set of experiments when performed in osteoblastic cells (MC3T3-E1) also showed similar results (Fig. 4b). Furthermore, overexpression of Fbw7 inhibited Runx2-mediated transactivation, whereas knocking down Fbw7 led to a substantial increase in Runx2 transactivation potential as measured by luciferase activity (Fig.  4c). These data suggest that catalytically active Fbw7 negatively modulates Runx2 protein stability and thus inhibits its ability to transactivate the osteogenic target genes.

Fbw7 Knockdown Restores Runx2 Expression and Enhances
Osteoblast Differentiation-Because Fbw7 negatively regulates Runx2 steady state levels by promoting its ubiquitin-mediated proteasomal degradation, we next studied Runx2 expression and its functional consequences during osteoblast differentiation as well as upon Fbw7 knockdown in osteoblasts. Immunoblot analysis of Runx2 and Fbw7 from MC3T3-E1 cells cultured in DIM showed a sharp increase in Runx2 while showing a simultaneous decrease in Fbw7 protein expression between days 3 and 7, suggesting that down-regulation of Fbw7 is required for increased Runx2 expression (Fig. 5a). We next assessed Runx2 expression in MC3T3-E1 cells upon Fbw7 RNAi. Like in Fig. 5a, siRNA-mediated of knockdown of Fbw7 led to a substantial increase in Runx2 protein expression (Fig.  5b) and significantly enhanced nodule formation by MC3T3-E1 cells even without DIM as compared with siControl (Fig. 5c,  upper panel showing representative photographs). Reportedly, PI3K/Akt signaling is known to promote Runx2-dependent osteoblast differentiation by inhibiting GSK3␤ via its Ser 9 phosphorylation (25). We hypothesized that inactivated GSK3␤, in turn, fails to phosphorylate Runx2, thereby preventing its recognition and subsequent Fbw7-mediated proteasomal degradation. Moreover, increased Runx2 levels are known to up-regulate Akt protein levels, thereby regulating Runx2-dependent osteoblast differentiation (42,43). We, therefore, determined the effect of Fbw7 depletion on GSK3␤ and Akt expression in MC3T3-E1 cells cultured in both growth medium and differentiation induction medium upon Fbw7 RNAi. Expectedly, Runx2 expression robustly increased in MC3T3-E1 after siFbw7 transfection under both growth-induced and differentiation-induced conditions (Fig. 5c, lower panel). We also observed enhanced Akt expression as well as increased GSK3␤ phosphorylation (inactive state of GSK3␤) in siFbw7-transfected conditions (Fig. 5c, lower panel), suggesting that Fbw7 inhibition may restore Runx2 expression and its function. Increased Akt expression upon Fbw7 knockdown may apparently be attributed to stabilized Runx2, which is known to activate Akt (25,27,42,43). Like MC3T3-E1, Fbw7 RNAi in mouse primary calvarial osteoblasts showed similar results (Fig. 5d). Fig. 5e shows Fbw7 silencing in primary cultures (Fig. 5e). These data thus suggest that Fbw7 RNAi promotes osteoanabolic molecules (Runx2 and Akt) and antagonizes an anti-osteoblastic molecule (GSK3␤), leading to osteogenesis.

Fbw7 Negatively Regulates Runx2-dependent Osteogenesis
Reciprocal Relationship between Runx2 and Fbw7 in Vivo-Because our data so far indicated that Fbw7 was a negative regulator of osteoblast differentiation, we examined Fbw7 expression under the conditions of bone loss, viz. OVX and lactation (45). Bone sections from sham (ovary-intact) rats showed copious Runx2 but low levels of Fbw7 immunoreactivity (Fig. 6a). This pattern was reversed in the bone sections of lactating and ovariectomized rats, where Runx2 immunoreactivity was reduced, whereas Fbw7 was increased. E2 supplementation to ovariectomized rats resulted in reversal of the patterns of these two reciprocally expressed proteins and restored their expressions to that observed in the sham group (Fig. 6a). Immunoblotting the protein lysates isolated from the same bone samples with Runx2 and Fbw7 antibodies further confirmed reciprocal expression levels of Runx2 and Fbw7 (Fig.  6b). Calvarial tissue from neonatal mice treated with methyl prednisolone (MP, a synthetic glucocorticoid) or vehicle showed that Runx2 protein increased with time in the vehicle group (8 days as compared with 0 days after MP treatment) but not in the MP-treated group. Corresponding Fbw7 immunoblotting showed a reciprocal expression pattern to Runx2, as Fbw7 expression decreased with time in the vehicle group but increased in the MP group (Fig. 6c).

Discussion
Although the regulation of Runx2 at the transcriptional level is well studied, its regulation at the protein levels is poorly understood. In the present study, we have identified the tumor suppressor Fbw7 as a novel regulator of Runx2. From several lines of experiments, it appears that the presence of multiple CPD motifs within Runx2 renders it a substrate for Fbw7. Furthermore, reduced protein stability of Runx2 and Runx2⌬N⌬C in MC3T3-E1 cells co-transfected with Fbw7 and treated with cycloheximide indicated the involvement of Fbw7 ligase activ- . Fbw7-mediated degradation of Runx2 negatively impacts its transactivation capacity. a, 293T cells were transfected with p6OSE2-Luc reporter vector with increasing amount of expression plasmids for Runx2 either alone or together with Fbw7␣ and its deletion mutants (Fbw7␣⌬F or Fbw7␣ WD ). 24 h after transfection, cells were lysed, and luciferase activity was measured. b, MC3T3-E1 cells were transfected with p6OSE2-Luc plasmid and the indicated amount of expression plasmids for Runx2 either alone or together with Fbw7␣ and its deletion mutants as indicated. After 48 h of transfection, cells were lysed, and luciferase activity was measured. c, MC3T3-E1 cells were transfected with p6OSE2-Luc plasmid either with siFbw7 (25 and 50 nM) or with Fbw7␣ (0.1 and 0.2 g). 48 h after transfection, cells were lysed, and luciferase activity was measured. Data are representative of minimum three independent experiments. Results are given as ϮS.E.; *, p Ͻ 0.05, **, p Ͻ 0.001, ***, p Ͻ 0.0001. One-way analysis of variance with Tukey's multiple comparison test was performed using GraphPad Prism Version 5.00.

Fbw7 Negatively Regulates Runx2-dependent Osteogenesis
ity in down-regulation of Runx2 protein expression. As with other known substrates, Fbw7 through its substrate-interacting domain (WD repeats) interacted with Runx2 in its ⌬N⌬C region, which still contains two putative CPD domains (16,19). Our data demonstrated that CPD starting at Ser 279 might be crucial for this interaction. However, further investigation is required to confirm this. Nonetheless, as no interaction between Fbw7 and either of the Runx2 deletion mutants Runx2N or Runx2C was observed (Fig. 2, h and i), it is quite likely that the CPDs present in the ⌬N⌬C region of Runx2 are involved in its interaction and subsequent degradation by Fbw7. Whole cell lysates were prepared after 0, 3, 7, and 14 days followed by immunoblotting with anti-Runx2 and Fbw7␣ antibodies; ␤-actin was probed as a loading control (lower panel). b, lysates of MC3T3-E1 cells transfected with siControl (si-cont) or siFbw7␣ for 24, 48, and 72 h were immunoblotted with anti-Fbw7 and anti-Runx2 antibodies. ␤-Actin was probed as a loading control. c, MC3T3-E1 cells were transfected with either siControl (siC) or siFbw7␣. 72 h after transfection, culture medium was changed every alternate day for 7 days. Cells were fixed and stained with Alizarin Red, and mineralized nodules were photographed (upper panel). MC3T3-E1 cells were transfected with siControl or siFbw7␣. After 3 and 7 days, WCEs prepared were immunoblotted with anti-Fbw7 and anti-Runx2 antibodies. ␤-Actin was probed as a loading control (lower panel). OBM, osteoblast growth medium. d, primary osteoblast cells isolated from BALB/c mice calvaria were transfected with siControl or siFbw7␣. 72 h after transfection, culture medium was changed alternatively for 14 days followed by Alizarin Red staining and imaging under a microscope. 1 ml of cetylpyridinium chloride was added to each well, readings were taken at 595 nm, and the graph was plotted. e, primary osteoblast cells isolated from BALB/c mice calvaria were transfected with siControl or siFbw7␣. 72 h after transfection, and culture medium was changed every alternate days for 14 days followed by WCEs preparation and immunoblotting with anti-Fbw7 and anti-Runx2 antibodies. ␤-Actin was probed as a loading control. f, lysates of MC3T3-E1 cells treated with 100 ng/ml BMP2 and g, dexamethasone (DEX, 0.1 M)-treated lysates were resolved on 10% SDS-PAGE and immunoblotted with anti-Runx2 and anti-Fbw7 antibodies. Data are representative of minimum three independent experiments. DECEMBER 25, 2015 • VOLUME 290 • NUMBER 52

JOURNAL OF BIOLOGICAL CHEMISTRY 30983
As phosphorylation of CPD motifs within the substrates is required for their degradation by Fbw7 (16), we also demonstrate that GSK3␤, a serine/threonine kinase, may phosphorylate Runx2 in its CPD motifs, leading to its degradation by Fbw7 based on two criteria. First, the majority of the known Fbw7 substrates having CPDs similar to Runx2 are phospho-modified by GSK3␤. Second, GSK3␤ controls osteogenesis through regulating Runx2 activity (25). Our finding that GSK3␤ RNAi rescues degradation of Runx2 even in the presence of overexpressed Fbw7 does strengthen the notion that GSK3␤ apparently phosphorylated Runx2 to be recognized and degraded by Fbw7. This finding also gained support from previous studies showing that GSK3␤ inhibited DNA binding and transcriptional activity of Runx2 by phosphorylating Runx2 in GSK3␤ consensus sequence present at Ser 369 -Ser 373 -Ser 377 ( 369 SPPWSYDQS 377 ) in mouse Runx2 isoform 5 (25). Interest-FIGURE 6. Reciprocal relationship between Runx2 and Fbw7 in vivo. a, cross-sections (5 M) were made from decalcified femur epiphysis of various groups as indicated, and immunofluorescence (40ϫ confocal) was performed as indicated under "Materials and Methods" using anti-Fbw7␣ (mouse) and anti-Runx2 (rabbit) primary antibodies and anti-rabbit (Alexa Fluor 488) and anti-mouse (Alexa Fluor 594) secondary antibodies. Sham, ovary intact; OVX, bilateral OVX and maintained for 12 weeks; OVXϩE2, OVX supplemented with 17-␤ estradiol (5 g/kg/day subcutaneously). For the lactation group, rats 10 days after delivery were taken. b, protein extracts isolated from femurs (devoid of bone marrow) from various groups as in panel A were immunoblotted with anti-Runx2 and anti-Fbw7 antibodies. 293T cells transfected with or without Runx2 were used as positive and negative control, respectively, for Runx2 immunoblot ,whereas MC3T3-E1 served as endogenous positive control. c, in 2-day-old BALB/c mice, MP (2 mg/kg/day subcutaneously) was injected for 4 and 8 days, and mice were subsequently sacrificed. Calvariae were lysed, and total protein was resolved on 12% SDS-PAGE followed by immunoblotting with anti-Fbw7 and anti-Runx2 antibodies. ␤-Actin was probed as a loading control. d, based on our study, schematic diagram depicting regulation of Runx2 protein stability by Fbw7 under physiological and bone loss conditions. Data are representative of minimum three independent experiments.

Fbw7 Negatively Regulates Runx2-dependent Osteogenesis
ingly, the corresponding amino acid sequence is also conserved in human Runx2 ( 279 SPPWSYDQS 287 ; isoform 2) and forms one of the four potential CPDs ( 279 SPPWS 283 ) recognized by Fbw7 (25). Furthermore, a recent study showing that inhibition of GSK3␤ leads to attenuation of MP-induced bone loss in rats (46) further supports our finding that active GSK3␤-mediated Runx2 phosphorylation is apparently involved in its degradation by Fbw7, leading to bone loss. Although there are few studies demonstrating phosphorylation of Runx2 affecting its protein stability, none of them addressed the E3 ubiquitin ligase involved in such regulation (15,25). Our data demonstrating Runx2, GSK3␤, and Fbw7 to physically interact with each other in osteoblasts is again in agreement with previous findings that suggest physiological interaction of GSK3␤ with Runx2 is required for attenuating Runx2 functions (25). It is suggested that GSK3␤-dependent phosphorylation of Runx2 at Ser 369 -Ser 373 -Ser 377 located in its negative regulatory region within Runt domain masks its DNA binding ability (25). However, our finding provides direct evidence that Runx2 phosphorylation by GSK3␤ apparently primes it for ubiquitin-mediated degradation by Fbw7, leading to its functional attenuation. It is also evident from luciferase assay that either knockdown of Fbw7 or overexpression of non-functional Fbw7 mutants is sufficient to rescue Runx2 transactivation potential (Fig. 4). In addition, our data also provide a possible explanation for enhanced Runx2 activity via osteoanabolic agents including insulin and insulinlike growth factor-1, leading to activation of PI3K/Akt pathway that stabilizes Runx2 apparently by inactivating GSK3␤ by phosphorylating it at Ser 9 (26,42,43,47).
Here, it is noteworthy to mention that although Wnt signaling is known to a play key role in maintaining bone mass where GSK3␤ and ␤-catenin are crucial signaling molecules, regulation of Runx2 by GSK3␤ in osteoblasts might be independent of Wnt signaling. This is primarily due to the fact that ␤-catenin rather represents a differentiation switch of mesenchymal progenitors for inducing osteoblastic differentiation and suppressing chondrocytic differentiation at an early stage of skeletal development during embryogenesis (48). In contrast, in osteoblasts, ␤-catenin along with its target TCF proteins regulate expression of osteoprotegerin, a major inhibitor of osteoclast differentiation, suggesting osteoanabolic action of ␤-catenin to be due to decrease in osteoclastic bone resorption and not due to increase in bone mass (25,48,49). Thus it appears that enhanced Runx2 expression upon GSK3␤ inhibition is due to inhibition of Fbw7-mediated ubiquitin-proteasome degradation of Runx2 rather than Wnt signaling involving ␤-catenin activation.
Our hypothesis that GSK3␤-mediated phosphorylation of Runx2 primes it for degradation by Fbw7 is further substantiated by the fact that Fbw7 RNAi completely restored Runx2 protein expression in osteoblasts. The reciprocal relationship between Runx2 and Fbw7 was also observed in vivo, such as in ovariectomized and lactating rats. Further, a progressive decline in Runx2 with a corresponding increase in Fbw7 in the calvarium of neonatal rats injected with MP reiterated the reciprocal relationship between the two proteins. Importantly, the OVX-induced changes in the expression of Runx2 and Fbw7 were reversed by E2 supplementation to ovariectomized rats, which suggested hormonal and physiological regulation of the reciprocal relationship between these two proteins. Recent studies identify GSK3␤ as a potent therapeutic target for the treatment of CCD resulting from Runx2 insufficiency because the inhibition of GSK3␤ by genetic or pharmacological means has been shown to significantly rescue CCD (44). In contrast, GSK3␤ insufficiency has been shown to cause enhanced bone mass in adult mouse without any other abnormalities (25). Our finding that GSK3␤ inhibition by either siGSK3␤ or LiCl restored Runx2 protein expression even in the presence of overexpressed Fbw7 also indicated that GSK3␤ could be a potent therapeutic target in osteoporosis and CCD. Further, our finding that down-regulation of endogenous Runx2 protein levels by overexpression of GSK3␤ is substantially inhibited upon Fbw7 knockdown suggests an important interdependence of GSK3␤ and Fbw7 in inhibiting Runx2 functions apparently by promoting the latter's degradation.
Taken together, our study demonstrated that Fbw7 negatively regulates osteogenesis by targeting Runx2 for ubiquitinmediated degradation in a GSK3␤-dependent manner; this regulation is schematically represented in a hypothetical model (Fig. 6d). These findings also provide a plausible explanation for bone loss induced by GSK3␤ activation as reported in several studies (25,46). Although GSK3␤ is considered to be a therapeutic target in CCD for which characterization of small molecule inhibitors is underway, therapeutic targeting of Fbw7 in such patients either alone or together with GSK3␤ may also be a possible treatment option for CCD.
Author Contributions-A. K. T. and N. C. designed the study. Y. K., I. K., G. T., M. P. K., K. K., N. S., and J. K. K. conducted the study. Y. K., I. K., N. S., G. T., K. K., and A. K. T. analyzed the data. A. K. T., N. C., and S. S. interpreted the data. A. K. T., Y. K., I. K., and G. T. drafted the manuscript. A. K. T., N. C., and S. S. approved the final version of the manuscript. A. K. T., Y. K., and G. T. take responsibility for the integrity of data analysis.