β-Catenin Directly Sequesters Adipocytic and Insulin Sensitizing Activities but Not Osteoblastic Activity of PPARγ2 in Marrow Mesenchymal Stem Cells

Lineage allocation of the marrow mesenchymal stem cells (MSCs) to osteoblasts and adipocytes is dependent on both Wnt signaling and PPARγ2 activity. Activation of PPARγ2, an essential regulator of energy metabolism and insulin sensitivity, stimulates adipocyte and suppresses osteoblast differentiation and bone formation, and correlates with decreased bone mass and increased fracture rate. In contrast, activation of Wnt signaling promotes osteoblast differentiation, augments bone accrual and reduces total body fat. This study examined the cross-talk between PPARγ2 and β-catenin, a key mediator of canonical Wnt signaling, on MSC lineage determination. Rosiglitazone-activated PPARγ2 induced rapid proteolytic degradation of β-catenin, which was prevented by either inhibiting glycogen synthase kinase 3 beta (GSK3β) activity, or blocking pro-adipocytic activity of PPARγ2 using selective antagonist GW9662 or mutation within PPARγ2 protein. Stabilization of β-catenin suppressed PPARγ2 pro-adipocytic but not anti-osteoblastic activity. Moreover, β-catenin stabilization decreased PPARγ2-mediated insulin signaling as measured by insulin receptor and FoxO1 gene expression, and protein levels of phosphorylated Akt (pAkt). Cellular knockdown of β-catenin with siRNA increased expression of adipocyte but did not affect osteoblast gene markers. Interestingly, the expression of Wnt10b was suppressed by anti-osteoblastic, but not by pro-adipocytic activity of PPARγ2. Moreover, β-catenin stabilization in the presence of activated PPARγ2 did not restore Wnt10b expression indicating a dominant role of PPARγ2 in negative regulation of pro-osteoblastic activity of Wnt signaling. In conclusion, β-catenin and PPARγ2 are in cross-talk which results in sequestration of pro-adipocytic and insulin sensitizing activity. The anti-osteoblastic activity of PPARγ2 is independent of this interaction.


Introduction
Regulation of marrow MSC fate toward adipocyte or osteoblast lineage involves multiple mechanisms including modulation of lineage-specific transcription factors [1]. Such modulation may comprise of direct interactions between transcription factors and their co-modulators, which is often coordinated by changes in the activity of signaling pathways. The example of such interaction includes regulation of Wnt signaling and PPARc2 activity.
PPARc nuclear receptor is an essential regulator of energy metabolism and a key transcription factor for adipocyte differentiation [2]. The transcriptional activity of PPARc is controlled by binding of lipophilic ligands to the ligand binding pocket. The natural ligands consist of polyunsaturated fatty acid derivatives and eicosanoids [2]. Synthetic ligands include a class of antidiabetic drugs, thiazolidinediones (TZDs), which bind to PPARc with high affinity, activate its adipogenic activity, and act as insulin sensitizers [2]. PPARc protein is expressed in mice and humans as two different isoforms, PPARc1 and PPARc2, due to alternative promoter usage and alternative splicing [3]. In mice, PPARc2 differs from PPARc1 by the presence of 30 amino acids (28 amino acids in humans) located at the N-teminus of the AF-1 domain. PPARc1 is ubiquitously expressed, whereas PPARc2 expression is restricted to adipocytes, including marrow adipocytes [2,4]. Although both isoforms have overlapping transcriptional activities, PPARc2 seems to be more specific for lipids and carbohydrates metabolism. The most common PPARc polymorphism (Pro12Ala), which is associated with alterations of physiological metabolic status, is located in the unique AF-1 domain of PPARc2 protein [5], and PPARc2 but not PPARc1 can restore adipocytic differentiation in cells previously ablated from both PPARc isoforms [6,7]. The studies of the PPARc role in marrow MSCs differentiation suggest PPARc2 function in commitment to adipocyte lineage, while PPARc1 in control of osteoblast differentiation and production of mineralized matrix [4,8,9].
PPARc2 expression and activity increases in marrow MSCs with aging and upon treatment with TZDs, and it correlates with decreased number of osteoblasts and decreased bone formation, and increased number of adipocytes and accumulation of fat in the bone marrow [10,11]. In contrast, insufficiency in PPARc activity in MSCs leads to increased number of osteoblasts and increased bone mass, and decreased adipocyte number and fat accumulation in the bone marrow [12,13]. In vitro studies suggest a role for PPARc2 isoform in commitment of marrow MSCs to adipocytic lineage at the expense of osteoblastic lineage [4,14]. An analysis of PPARc2 transcriptome of U-33/c2 cells, which represent a model of MSC differentiation under the control of PPARc2, showed that its activation with TZD rosiglitazone (Rosi) leads to simultaneous induction of adipocytic and suppression of osteoblastic gene expression, including suppression of multiple members of Wnt signaling pathway [14]. Although Rosi activates both proadipocytic and anti-osteoblastic properties of PPARc2, these activities can be separated by using ligands of different chemical structures, as we have demonstrated previously [15,16]. Indeed, the possibility to separate different activities of PPARc by manipulating with its phosphorylation status has been recently demonstrated in respect to PPARc anti-diabetic and proadipocytic properties. Insulin-sensitizing activity requires blocking phosphorylation of serine 273 [17], while pro-adipocytic activity requires dephosphorylation of serine 112 within PPARc protein [18,19]. However, the mechanism by which PPARc2 acquire antiosteoblastic activity is not yet elucidated.
Osteoblast differentiation is regulated by a number of osteogenic pathways, including Wnt signaling [20]. Binding of Wnt glycoprotein ligands to LDL-related protein 5/6 (Lrp5/6) and Frizzled (Fzd) co-receptors triggers release of b-catenin from protein degradation complex, its translocation to the nucleus and activation of TCF/LEF transcriptional complex, which facilitates the expression of canonical Wnt-controlled genes regulating cell proliferation and differentiation [21]. The association between naturally occurring mutations in human Lrp5 receptors and high or low bone mass phenotype demonstrates an essential role of Wnt signaling in the regulation of the skeletal homeostasis [20,22,23]. However, the phenotypes of mice with genetically altered components of canonical signaling, including b-catenin and Wnt10b, point toward an interesting phenomenon that different members of the canonical Wnt pathway have distinct effects on the skeleton [24][25][26][27][28][29]. Moreover, they indicate that the same protein may have different functions during MSC differentiation. For example, b-catenin ablation in early mesenchymal progenitors has deleterious effects on skeletal development due to suppressed osteoblast differentiation [25,30], whereas its ablation in lineage committed osteoblasts increases support for osteoclastogenesis without affecting osteoblastic bone formation [24,28].
b-Catenin-mediated Wnt signaling requires interaction with other transcriptional regulators. Besides TCF/LEF complex, bcatenin may interact with a number of transcription factors and nuclear receptors including PPARc [31][32][33]. The interaction between both proteins is facilitated through TCF/LEF binding domain of b-catenin and helices 7 and 8 of PPARc [33,34]. It has been demonstrated that pro-adipocytic activity of PPARc leads to b-catenin dissociation from the complex and its subsequent degradation [33].
Differentiation of marrow MSCs towards osteoblasts relies on functional Wnt10b/b-catenin canonical signaling [35,36], while their differentiation to adipocytes requires PPARc2 [4,12]. Wnt10b/b-catenin signaling suppresses PPARc2 activity and adipogenesis, while PPARc2 suppresses Wnt10b/b-catenin signaling and osteoblastogenesis, suggesting fully reciprocal communication between PPARc2 and canonical Wnt signaling [15,35]. However, selective activation of PPARc2 anti-osteoblastic properties leads to suppression of Wnt10b expression, while selective activation of PPARc2 pro-adipocytic properties does not affect Wnt10b expression [15]. This suggests a possibility that the cross talk between these two regulatory pathways may not be fully reciprocal and may rely in part on a different mechanism, which can be manipulated to advantage one of the MSC phenotypes. Indeed, it has been demonstrated that Wnt signaling, independently of b-catenin activity, may silence PPARc expression by phosphorylation of histone lysine methyltransferase and recruitment the corepressor complex to PPARc promoter region [37,38].
In this study, we demonstrate that PPARc2 pro-adipocytic and pro-insulin signaling activities require degradation of b-catenin protein and that the degradation of b-catenin does not directly affect anti-osteoblastic activity of PPARc2. This activity rather depends on Wnt10b, which is under negative control of PPARc2 and under positive, PPARc2-independent, control of b-catenin.

Plasmids
Luciferase gene reporter constructs, TOP-Flash and FOP-Flash (Millipore, Billerica, MA), were used to measure b-catenin transcriptional activity. The TOP-Flash construct contains three copies of the TCF/LEF sites, whereas FOP-Flash has mutated copies of TCF/LEF sites and is used as a control for measuring nonspecific activation of the reporter. b-Catenin expression plasmid constructed on a pSPORT6 vector backbone was purchased from Invitrogen. pEF-PPARc2 expression construct, a kind gift from Dr. J. Gimble (Pennington Institute, LA), was generated by cloning PPARc2 cDNA into pEF-BOS expression vector downstream of the translation elongation factor 1a (EF-1a) promoter sequence, which permits the levels of ectopically expressed transcript to be within the physiological range ( Figure  S1A and B) [39]. PPARc transcriptional activity was measured using p2AOx construct which has inserted two tandem copies of the rat PPRE sequence specific for acyl-CoA promoter region into pCPSluc vector [40]. Site-directed mutagenesis of the D409 residue of the ligand binding domain (LBD) of PPARc2 was performed using the following primers: forward 59-GCA CTG GAA TTA GAT GCC AGT GAC TTG G-39 and reverse 59-CCA AGT CAC TGG CAT CTA ATT CCA GTG C-39. The accuracy of the introduced mutation was confirmed by DNA sequencing.

Cell Lines and Cell Culture Conditions
Murine marrow-derived U-33 cells represent a clonal cell line spontaneously immortalized in the long term bone marrow culture b-Catenin Regulates PPARc2 Activity [4]. In basal conditions, the U-33 cells display both preosteoblastic phenotype characterized by high alkaline phosphatase enzyme activity and expression of osteoblast-specific gene markers, and preadipocytic phenotype as assessed by levels of expression of adipocyte-specific gene markers [4]. To study the effect of adipocyte-specific transcription factor PPARc2 on marrow MSC differentiation, U-33 cells were stably transfected with either pEF-PPARc2 expression construct (referred to as U-33/c2 cells) or an empty vector pEF-BOS (referred to as U-33/c cells) as described previously [4]. U-33/c2 and U-33/c cells were maintained in MEM-a supplemented with 10% FBS, 1% penicillin/streptomycin solution (Invitrogen), and 0.5 mg/ml G418 for positive selection of transfected cells. The human kidney HEK293 cells were grown in high glucose DMEM supplemented with 10% FBS and 1% penicillin/streptomycin solution. All cultures were grown at 37C in a humidified atmosphere containing 5% CO 2 .

Protein Isolation, Fractionation, and Western Blot Analysis
Cell lysate fractionation was performed as described previously [41]. In brief, cells were washed with PBS and scraped into hypotonic lysis buffer (10 mM Tris-HCl pH 7.5, 140 mM NaCl, 5 mM EDTA, 2 mM dithiothreitol, and protease inhibitors), homogenized, and spun at 1,0006g for 10 min at 4C to pellet down nuclei. The remaining supernatant was centrifuged at 100,0006g for 90 min at 4C to yield the high molecular weight protein fraction containing b-catenin bound to the destruction complex (protein-bound or transcriptionally inactive b-catenin) and the cytosol fraction containing b-catenin released from the complex (protein unbound or transcriptionally active b-catenin). For whole cell lysis, cells were scraped into lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% NP-40, 50 mM NaF, and protease inhibitors) and spun at 7,0006g for 5 min to remove cell debris. For detection of phospho-proteins, cells were scraped into the same lysis buffer and subjected to 5-sec freeze/thaw three times prior to centrifugation at 12,0006g for 5 min. Protein concentration was measured using BCA Protein Assay kit and proteins were separated on 10% SDS-PAGE. For detection of proteins, the following antibodies were used: PPARc (1:167), bcatenin (1:1000), Akt (1:1000), phospho-Akt (1:1000), and b-actin (1:5000). Proteins were visualized using Odyssey Infrared Imaging System (LI-COR Biosciences) after incubation with IR-Dyeconjugated secondary antibodies at dilution of 1:10,000. A relative quantity of proteins was determined by densitometric measure- Immunocytochemistry U-33/c2 cells were briefly washed with PBS buffer and permeabilized by incubating with ice cold methanol for 10 min. After washing, cells were blocked using 5% goat serum in 0.2% Triton-X for 1 h and incubated with mouse anti-b-catenin (1:50) and/or goat anti-PPARc2 (1:50) diluted in 5% goat serum containing 0.1% Triton-X for 1 h at room temperature. To visualize b-catenin and PPARc2 cells were incubated with either goat anti-mouse Alexa Fluor 488 or chicken anti-goat Alexa Fluor 594 respectively for 1 h at room temperature. As a negative control, cells were incubated with Alexa Fluor antibodies without prior incubation with primary antibodies. Finally, the cells were mounted using Prolong Gold anti-fade reagent with DAPI (Invitrogen). Images were taken within 24-48 h after immunostaining.

Quantitative Real-time RT-PCR Analysis
Total RNA was extracted using RNeasy Mini kit. Its purity and concentration were determined using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). After DNase treatment, 0.75mg of RNA was converted to cDNA using the iScript cDNA synthesis kit. The amount of cDNA corresponding to 7.5 ng of RNA was used for each reaction containing Power SYBR Green mix and was processed using StepOne Plus System (Applied Biosystems, Carlsbad, CA). Relative gene expression was determined by the DD-Ct method using 18S RNA levels for normalization. Primers were designed using Primer Express 3.0 software (Applied Biosystems). All primers used in this study are listed in Table S1.
b-Catenin Knockdown using siRNA U-33/c2 cells (1610 5 cells) were transfected with 200 ng of bcatenin siRNA or nonspecific random siRNA as a negative control using Lipofectamine 2000. Seventy two hours after transfection, total RNA and protein were extracted and analyzed for b-catenin knockdown and expression of target genes.
Luciferase Gene Reporter Assay b-Catenin transcriptional activity was measured in U-33/c2 cells. Cells were seeded in a 24-well plate at the 1610 4 cells/cm 2 density and transfected with 0.3 mg of either TOP-Flash or FOP-Flash plasmid. Twenty four hours later, cells were treated with either 1 mM Rosi, 10 mM LiCl, 10 mM GW9662, or a combination of Rosi with either LiCl or GW9662 for 48 h and luciferase activity was measured using Dual Luciferase Reporter Assay System. Transcriptional activites of PPARc2 constructs were measured in Hek293 cells. Cells were seeded in a 24-well plate at the 9610 4 cells/cm 2 density and transfected with 0.2 mg of either pEF-BOS empty vector or non-mutated or mutated pEF-PPARc2 expression plasmids, mixed with 0.2 mg of either pSPORT6 empty vector or pSPORT6 expression plasmid containing wild type b- Analysis of Adipocyte and Osteoblast Differentiation of U-33/c2 cells For adipogenesis assay, U-33/c2 cells were seeded in 6-well plate at the 1610 4 cells/cm 2 density. Twenty four hours later, they were treated with either 1 mM Rosi, 10 mM LiCl, 10 mM GW9662 or a combination of Rosi with either LiCl or GW9662 for 3 days. Lipid accumulation was assessed using Oil Red O staining [15]. For analysis of osteoblast differentiation, the U-33/ c2 cells were seeded in 96-well plate at the 1610 4 cells/cm 2 density. After 24 h of growth, they were treated with either 1 mM Rosi, 10 mM LiCl, or 10 mM GW9662 or a combination of Rosi and LiCl or GW9662 for 3 days. Cells were washed with HEPES and alkaline phosphatase (ALP) activity was measured as previously described [15]. The ALP activity was normalized to cell number measured using Cell Titer 96 AQ ueous Non-Radioactive Cell Proliferation Assay kit.

Statistical Analysis
All experiments were performed in triplicates. Statistical analysis of results was conducted using one-way ANOVA and t-test, as applicable. All data showed represent means and standard deviation of the means (SD). Statistical significance was set to p,0.05.

Rosi-activated PPARc2 Decreases b-catenin Protein Levels
The U-33/c2 cells, and their negative control U-33/c cells, represent a model of marrow MSC differentiation under control of PPARc2 transcription factor [4]. Stable transfection with PPARc2 under the control of elongation factor 1a (EF1a) promoter in U-33/c2 cells produces basal expression of PPARc2 transcript ( Figure S1A) and protein ( Figure S1B). Activation of ectopic PPARc2, but not naturally expressed PPARc1, with TZD Rosi converts U-33/c2 cells to terminally differentiated adipocytes ( Figure S1C), while suppressing osteoblast phenotype of U-33 cells and expression of numerous members of the Wnt signaling pathway, including b-catenin [4;14]. A detailed analysis of bcatenin gene expression as a function of time following Rosi treatment showed that a decrease in the level of b-catenin transcript occurred relatively late, when cells acquired phenotype of fully differentiated adipocytes marked by significant accumulation of fat droplets and expression of lipid metabolism gene markers (Table 1) [14]. The decrease in b-catenin expression was preceded by a decrease in Wnt10b expression, which occurred as early as 6 h after treatment, the time point which marks in U-33/ c2 cells a state of fate determination (Table 1) [14].
Despite the late response of b-catenin gene expression, its protein levels were decreased much earlier after Rosi treatment (Figure 1). In cytoplasm, a majority of b-catenin protein is sequestered between two different forms, either bound to the multiprotein complex which targets it for proteolytic degradation (protein-bound or inactive form) or free of the complex en route to   the nucleus to function as a transcriptional regulator (proteinunbound or active form) [21]. To distinguish between transcriptionally active and inactive forms of cytosolic b-catenin, protein lysates were fractionated as described in Material and Methods to yield protein bound and protein unbound forms of b-catenin, respectively. As shown in Figure 1A, the fraction of proteinunbound b-catenin decreased by 4-fold in U-33/c2 cells after 1 h treatment with Rosi. No decreases in the level of protein-bound bcatenin and in the level of b-catenin transcript, were observed at this time point (Figure 1A and 1B). After 72 h treatment, the protein level of total b-catenin was decreased by 5-fold ( Figure 1C) and was paralleled with a decrease in transcript levels by 2.5 fold ( Figure 1D). No change in b-catenin transcript and protein levels were observed at this time point in control U-33/c cells treated with Rosi ( Figure 1C and 1D). Interestingly, the basal levels of bcatenin protein in untreated U-33/c2 cells were lower as compared to untreated U-33/c cells suggesting that even a sole presence of non-activated PPARc2 isoform has a negative effect on the levels of b-catenin protein. Immunofluorescence analysis of PPARc2 and b-catenin cellular localization showed that in untreated cells both proteins localize in the cytoplasm, where they may physically interact, as demonstrated previously ( Figure 1E) [33].
Presented results indicate that the PPARc2 negative regulation of b-catenin protein levels involves two mechanisms; a rapid proteolytic degradation and a long-term suppression of b-catenin gene expression.

Stabilization of b-catenin with LiCl Protects from PPARc2mediated Degradation
Phosphorylation by glycogen synthase kinase 3b (GSK3b) targets b-catenin for proteosomal degradation. LiCl prevents bcatenin phosphorylation which includes inactivating autophosphorylation of GSK3b [21]. LiCl treatment of U-33/c2 cells counteracted the negative effect of Rosi on b-catenin protein levels (Figure 2A and 2C) without counteracting Rosi negative effect on its transcript levels ( Figure 2B). Moreover, LiCl treatment resulted significant translocation of b-catenin to the nucleus, which still occurred in cells treated simultaneously with LiCl and Rosi ( Figure 2D, Figure S3). This suggests that inhibition of GSK3b activity with LiCl prevents proteolytic degradation of b-catenin and that GSK3b is implicated in b-catenin degradation after Rosi treatment. Consistent with b-catenin translocation to the nucleus, a protective effect of LiCl was also seen at the level of b-catenin transcriptional activity tested in luciferase gene reporter assay using TOP-Flash construct carrying b-catenin responsive TCF/ LEF elements ( Figure 2E). As expected, luciferase activity was significantly decreased by Rosi-activated PPARc2 (by 4-fold), while this activity was increased by over 12-fold in the presence of LiCl. Consistent with b-catenin stabilization and nuclear translocation, simultaneous treatment with Rosi and LiCl not only preserved basal b-catenin activity, but even increased it by 4-fold of that of vehicular control ( Figure 2E).
Stabilization of b-catenin Suppresses PPARc2 Proadipocytic and Insulin Sensitizing Activity, but not Antiosteoblastic Activity In U-33/c2 cells, activation of PPARc2 with Rosi increases expression of adipocyte-specific genes, induces adipocyte formation, and simultaneously decreases the expression of osteoblastspecific gene markers and suppresses osteoblast phenotype [14]. Stabilization of b-catenin with LiCl in the presence of Rosiactivated PPARc2 significantly inhibited adipocyte development, as measured by intracellular lipids accumulation ( Figure 3A), and suppressed the expression of genes positively regulated by this transcription factor including FABP4/aP2 and Cidec ( Figure 3B and 3C). To verify the role of GSK3b in PPARc2 mediated degradation of b-catenin and suppression of adipogenesis, U-33/ c2 cells were treated with a GSK3b-specific reversible competitive ATP inhibitor 6-6-bromoindirubin-39-oxime (BIO). As showed in Figure S2A and S2B, BIO treatment offered partial protection of b-catenin in the presence of Rosi and subsequently inhibited adipogenesis as measured by formation of lipid droplets.
Since PPARc activation with anti-diabetic Rosi increases insulin signaling in adipocytes, we examined the effect of b-catenin stabilization on the expression of gene markers of this pathway. Upon Rosi treatment, both insulin receptor and FoxO1 gene expression increased respectively by 14-and 5-fold, however stabilization of b-catenin in the presence of activated PPARc2 decreased this effect by 2-fold ( Figure 3D and 3E). Furthermore, bcatenin stabilization prevented phosphorylation of Akt, which is a downstream mediator of insulin signaling and indicator of cell sensitivity to insulin ( Figure 3F). These results suggest that stabilization of b-catenin suppresses positive adipocytic and insulin sensitizing PPARc2 activities.
In contrast, b-catenin stabilization did not protect against the PPARc2-mediated suppression of osteoblast phenotype. As shown in Figure 4A, alkaline phosphatase (ALP) enzyme activity was decreased by Rosi and was not restored in the presence of LiCl. Similarly, LiCl did not protect from PPARc2 suppressive effects on the expression of Dlx5, Col1a1 and Wnt10b ( Figure 4B-D). This indicates that the status of b-catenin protein is in relationship to the positive pro-adipocytic and insulin sensitizing PPARc2 activities, but not to the suppressive anti-osteoblastic activity.

Inhibition of PPARc2 Pro-adipocytic Activity Stabilizes bcatenin and Mimics LiCl Effect
In order to assess a contribution of PPARc2 pro-adipocytic activity to b-catenin stability, we inhibited Rosi-induced PPARc2 activity with GW9662 selective antagonist previously shown to block adipogenesis induced by TZD treatment [42]. In U-33/c2 cells, GW9662 inhibited Rosi-induced lipid accumulation and expression of FABP4/aP2 and Cidec ( Figure 5A-C) but it did not affect Rosi-induced suppression of ALP activity and expression of Dlx5, Col1a1 and Wnt10b ( Figure 5D-G). Since the pattern of U-33/c2 cells response to GW9662 was identical to the pattern observed in the presence of LiCl, we analyzed b-catenin protein degradation status. As shown in Figure 5H-J, GW9662 prevented b-catenin protein degradation mediated by Rosi-activated PPARc2 and restored b-catenin localization in the nucleus ( Figure  S3). Consistently, GW9662 restored b-catenin transcriptional activity as measured in TOP-Flash gene reporter construct ( Figure 5K). Similar to LiCl (Figure 2B), treatment with GW9662 alone did not affect b-catenin transcript levels and treatment in combination with Rosi did not prevent Rosi negative effect on b-catenin transcript levels (data not shown).
To test whether PPARc2 anti-osteoblastic activity is dependent on the protein domain conferring the pro-adipocytic activity and b-catenin degradation we introduced previously reported mutation of PPARc1 into PPARc2 protein sequence [33]. It has been shown that substitution in the PPARc1 protein sequence of aspartic acid in the position 379 with alanine abrogates proadipocytic activity, prevents b-catenin binding and proteosomal degradation [33]. We introduced the same mutation in the position of D409 of PPARc2 protein sequence, and verified the stability of D409A mutant in HEK293 cells ( Figure 6A), To avoid interference with endogenous non-mutated PPARc protein, we examined the effect of mutated PPARc2 on b-catenin stabilization and activity in HEK293 cells, which naturally express minimal amounts of both PPARc isoforms and b-catenin (data not shown). HEK293 cells were transiently co-transfected with b-catenin and either non-mutated or mutated PPARc2 expression constructs. As expected, activation with Rosi of non-mutated form of PPARc2 decreased b-catenin protein levels, however activation of D409A mutant did not have an effect on levels of b-catenin protein ( Figure 6B). Consistent with a loss of adipocytic activity, mutation D409A abrogated PPARc2 transcriptional activity as measured using PPRE-controlled luciferase reporter gene assay ( Figure 6C). This result was validated in marrow-derived U-33/c cells. The expression of adipocyte-specific gene marker FABP4/aP2, which is under the control of PPREs, was suppressed in U-33/c cells transfected with D409A construct as compared to non-mutated construct ( Figure 6D). Most importantly, mutation D409A retained the suppressive effect of PPARc2 on the expression of Dlx5, Runx2, and Osterix confirming that the anti-osteoblastic activity of PPARc2 is independent of pro-adipocytic activity and b-catenin degradation ( Figure 6E-G). Moreover, Wnt10b was also downregulated with mutation D409A and in the presence of stabilized b-catenin providing further proof for a PPARc2mediated suppression of Wnt10b independent of b-catenin protein status ( Figure 6H). These results together indicate that PPARc2 pro-adipocytic, but not its anti-osteoblastic acitivity, is responsible for a decrease in b-catenin protein levels and that the antiosteoblastic activity is independent of PPARc2 pro-adipocytic activity and interaction with b-catenin.

siRNAs Silencing of b-catenin Affects Adipocytic but not Osteoblastic Gene Expression
To directly test the role of b-catenin in regulation of PPARc2 pro-adipocytic and anti-osteoblastic activities, we silenced cellular b-catenin using specific siRNA and analyzed alterations in expression of phenotype-specific gene markers. Down-regulation of b-catenin transcript by 70% ( Figure 7B-D), paralleled with a 2fold decrease in b-catenin protein levels ( Figure 7A), significantly enhanced transcript levels for adipogenic gene markers FABP4/ aP2 and Cidec ( Figure 7B). At the same time, transcript levels for osteoblast-specific gene markers Runx2, Dlx5 and Col1a1 were not affected ( Figure 7C), confirming our previous observation that the expression of these genes is not directly controlled by b-catenin. Similarly, expression of Sfrp1 and Wisp1, Wnt signaling components shown previously to regulate osteoblast differentiation [43,44] and being under the control of PPARc2 [14] remained unchanged. However, the expression of Wnt10b was decreased by 2-fold of its basal levels ( Figure 7D). These data support the suppressive effect of b-catenin on adipogenic gene expression and indicate its positive effect on Wnt10b expression. This observation, together with the results presented in Figure 4D and Figure 5G, suggest that Wnt10b is under control of both b-catenin and PPARc2.
Since mutation D409A had been characterized as unable to degrade b-catenin, one would expect that high levels of b-catenin will have a positive effect on expression of Wnt10b. This expectation was supported by the observation that b-catenin silencing decreased Wnt10b expression independently of PPARc2 ( Figure 7D). However and as shown in Figure 6H, mutant D409A suppressed Wnt10b expression. This suggests that PPARc2 negative effect on Wnt10b expression is dominant over b-catenin positive effect, at least in this experimental system.

Discussion
The results presented here demonstrate that PPARc2 activities positively regulating adipocyte-specific and insulin signalingspecific gene expression are sequestered through interaction with b-catenin, whereas PPARc2 anti-osteoblastic activity, which requires suppression of osteoblast-specific transcriptome, is independent of this interaction. We have confirmed that b-catenin degradation is an essential step for a direct activation of PPARc2 pro-adipocytic transcriptional activity mediated through PPRE [33,34] and we have shown that b-catenin degradation is also required for induction of mechanisms increasing insulin sensitivity.
Most importantly, we have demonstrated that the PPARc2 antiosteoblastic activity is regulated by a different mechanism, which does not depend on direct cross-talk with b-catenin but involves negative regulation of Wnt10b expression.
The functional interaction between b-catenin and PPARc2 is two-directional. Stabilization of b-catenin by inactivation of degradation process with either LiCl or BIO GSK3b inhibitor suppresses pro-adipocytic activity of PPARc2, whereas inhibition of pro-adipocytic activity of PPARc2 by either selective antagonist GW9662 or D409A mutation stabilizes b-catenin. At the same time, stabilization of b-catenin in the presence of Rosi does not suppress the PPARc2 anti-osteoblastic activity.
We hypothesize that PPARc2 anti-osteoblastic activity results from negative, and b-catenin independent, regulation of Wnt10b expression, which is an essential activator of pro-osteoblastic canonical Wnt signaling. Indeed, Wnt10b pro-osteoblastic and anti-adipogenic activity has been demonstrated in plethora of in vitro and in vivo studies [26,35,36,45]. Accordingly, overexpression of Wnt10b in MSCs induces osteoblast gene expression and inhibits PPARc2 expression [35], and ectopic expression of Wnt10b in adipocytes produces animals with high bone mass, which are resistant to the bone loss with aging [26]. In contrast, mice deficient in Wnt10b have low bone mass, affected MSCs proliferation and differentiation, and increased propensity of muscle satellite cells to accumulate fat [36,46]. We have demonstrated previously that PPARc2 ligands selective only for pro-adipocytic activity do not affect Wnt10b expression, whereas ligands selective only for anti-osteoblastic activity suppress Wnt10b expression [15]. Here, we have shown that Wnt10b is under the negative control of PPARc2 anti-osteoblastic activity and this control is independent of b-catenin pool regulating PPARc2 proadipocytic activity.
The possibility to activate b-catenin independently of Wnt signaling has been recently demonstrated in respect to the bone marrow response to mechanical stimuli [47,48]. It has been shown that under mechanical stress b-catenin suppresses adipocyte differentiation and PPARc activity through a mechanism which involves inactivation of GSK3b, comprising of mTORC2mediated phosphorylation of Akt protein and resulting in increased b-catenin stability [47,48]. Although not investigated here it would be of interest to examine whether the mechanisms of bcatenin destabilization by TZD-activated PPARc2 employs some of the components which increase its stability and prevent adipogenesis during mechanical stress.
Another important aspect of this study is the regulation of PPARc insulin sensitizing activity through interaction with bcatenin. The results showed here indicate that degradation of bcatenin positively correlates with increased expression of PPARccontrolled markers of insulin signaling, including pAkt, whereas stabilization of b-catenin leads to the loss of this positive regulation even in the presence of Rosi. It is well recognized that one of the adverse effects of anti-diabetic TZDs is weight gain due to increased fat mass, which suggests that TZDs anti-diabetic and pro-adipocytic activities are tied. However, as recently reported these two activities are independently linked to the phosphorylation status of two distinct serines within the PPARc protein [17][18][19]. Although it is highly speculative at this point, our results raise an interesting possibility that b-catenin cross-talk with PPARc, either through direct interaction or through alteration of GSK3b activity, regulates the phosphorylation of both serine 273 and serine 112, which are essential to the anti-diabetic and the proadipocytic activity of this nuclear receptor, and that this interaction is one of the culprits for unwanted effect of TZDs on weight gain.
Currently both clinically approved TZDs, rosiglitazone and pioglitazone, undergo critical evaluation of their clinical use due to adverse cardiovascular, cancer and skeletal effects, nonetheless there is no doubt that PPARc agonists are the most effective among available anti-diabetic drugs [49]. Therefore, better understanding of mechanisms, which regulate multiple activities of PPARc nuclear receptor including anti-osteoblastic activity, is important for the development of new class of PPARc agonists, which will harness selectively the desired insulin sensitizing activity without unwanted effects. Although our studies may not fully reflect functional interaction between PPARc and b-catenin in vivo, because they use a model of U-33/c2 cells which were specifically designed to study PPARc2 pro-adipocytic and antiosteoblastic activities in marrow cells in vitro, however they may suggests that in the quest for efficient and safe anti-diabetic PPARc agonists interaction between b-catenin/PPARc and Wnt10b/ PPARc should be considered. Figure S1 Ectopic expression of PPARc2 under control of elongation factor 1a in U-33/c2 produces basal expression of PPARc2 and upon TZD activation commits cells to terminally differentiated adipocytes. A. Analysis of PPARc1 and PPARc2 transcript levels in U-33/c and U-33/c2 cells. Gene expression is presented as fold difference as compared to PPARc1 levels in U-33/c cells. B. Western blot analysis of total PPARc protein levels in U-33/c and U-33/c2 cells. C. Northern blot analysis of PPARc target gene FABP4/aP2 upon activation with Rosi indicates that its expression transiently upregulated in U-33/c, whereas its expression is sustained in U-33/c2 cells (* p,0.05).

Supporting Information
(TIF) Figure S2 GSK3 b inhibitor 6-bromoindirubin-39-oxime (BIO) protects b-catenin from PPARc2 mediated degradation and suppresses adipogenesis. U-33/c2 cells were pre-treated with 5 mM BIO for 2 h followed by treatment with either 5 mM BIO or in combination of BIO with 1 mM Rosi for 24 h. Non-treated cells or treated with 1 mM Rosi alone were used as controls.