Wnt6, Wnt10a and Wnt10b inhibit adipogenesis and stimulate osteoblastogenesis through a β-catenin-dependent mechanism

Abstract

Wnt10b is an established regulator of mesenchymal stem cell (MSC) fate that inhibits adipogenesis and stimulates osteoblastogenesis, thereby impacting bone mass in vivo. However, downstream mechanisms through which Wnt10b exerts these effects are poorly understood. Moreover, whether other endogenous Wnt ligands also modulate MSC fate remains to be fully addressed. In this study, we identify Wnt6 and Wnt10a as additional Wnt family members that, like Wnt10b, are downregulated during development of white adipocytes in vivo and in vitro, suggesting that Wnt6 and/or Wnt10a may also inhibit adipogenesis. To assess the relative activities of Wnt6, Wnt10a and Wnt10b to regulate mesenchymal cell fate, we used gain- and loss-of function approaches in bipotential ST2 cells and in 3T3-L1 preadipocytes. Enforced expression of Wnt10a stabilizes β-catenin, suppresses adipogenesis and stimulates osteoblastogenesis to a similar extent as Wnt10b, whereas stable expression of Wnt6 has a weaker effect on these processes than Wnt10a or Wnt10b. In contrast, knockdown of endogenous Wnt6 is associated with greater preadipocyte differentiation and impaired osteoblastogenesis than knockdown of Wnt10a or Wnt10b, suggesting that, amongst these Wnt ligands, Wnt6 is the most potent endogenous regulator of MSC fate. Finally, we show that knockdown of β-catenin completely prevents the inhibition of adipogenesis and stimulation of osteoblast differentiation by Wnt6, Wnt10a or Wnt10b. Potential mechanisms whereby Wnts regulate fate of MSCs downstream of β-catenin are also investigated. In conclusion, this study identifies Wnt10a and Wnt6 as additional regulators of MSC fate and demonstrates that mechanisms downstream of β-catenin are required for Wnt6, Wnt10a and Wnt10b to influence differentiation of mesenchymal precursors.

1 Introduction

Mesenchymal stem cells (MSCs) give rise to numerous cell types, including adipocytes and osteoblasts. The dysregulation of adipogenesis or osteoblastogenesis is implicated in the pathogenesis of diseases such as obesity, type 2 diabetes and osteoporosis. Hence, elucidating mechanisms that regulate MSC fate may facilitate the development of treatments for these diseases. One established regulator of MSC fate is the Wnt signaling pathway.

The Wnts are a family of secreted glycoproteins that consist of at least 19 members in mammals, and which mediate autocrine and paracrine effects by binding to frizzled (Fzd) receptors and LDL-related protein 5/6 (LRP5/6) coreceptors [1]. In the Wnt/β-catenin pathway, Wnt ligands mediate downstream effects by stabilizing β-catenin, a multifunctional protein involved in cell adhesion and transcriptional regulation. In the absence of Wnt stimulation, cytoplasmic β-catenin is localized within a multi-protein ‘destruction complex,’ consisting of the scaffold proteins Axin and adenomatous polyposis coli (APC), and the kinases CKIα and GSK-3β. Within this complex, β-catenin is phosphorylated by CKIα and GSK-3β, allowing its polyubiquitination and subsequent proteasomal degradation. Binding of Wnt ligands to Fzd and LRP5/6 promotes dissociation of the destruction complex and thereby prevents β-catenin degradation. Consequently, cytoplasmic β-catenin accumulates and translocates into the nucleus where it coactivates the T-cell factor (TCF)/lymphoid enhancer factor family of transcription factors to regulate Wnt/β-catenin target genes [1], which typically encode proteins associated with cell fate regulation.

Research over the past decade has established Wnt/β-catenin signaling as an important regulator of MSC fate. The first study linking Wnt signaling to adipogenesis demonstrated that expression of Wnt10b decreases during adipogenesis in vitro and that ectopic expression of Wnt10b inhibits adipogenesis by suppressing expression of the adipogenic transcription factors, peroxisome proliferator-activated receptor-γ (PPARγ) and CCAAT/enhancer binding protein-α (C/EBPα) [2]. Based on this initial report, numerous additional studies have focused on Wnt10b as the candidate Wnt family member that regulates MSC fate. For example, Wnt10b expression also decreases during brown adipogenesis in vitro [3] and white adipogenesis in vivo [4]. Moreover, transgenic mice that overexpress Wnt10b in adipose tissue (FABP4-Wnt10b mice) have reduced adiposity and are resistant to obesity [5, 6]. This demonstrates that Wnt10b can inhibit white adipose tissue (WAT) development in vivo. Unexpectedly, FABP4-Wnt10b mice were also found to have increased bone mass [7], as do mice expressing Wnt10b in osteoblasts from the osteocalcin promoter [8]. These observations led to the identification of Wnt10b as a stimulator of osteoblast differentiation and mineralizing activity [8, 9]. Conversely, mice lacking Wnt10b have decreased trabecular bone [7]. This demonstrates that Wnt10b acts as an endogenous regulator of osteoblastogenesis and bone formation in vivo. In contrast, ablation of Wnt10b in mice does not overtly affect adipogenesis or adipose tissue mass, exerting only a mild effect on lipid accumulation and adipocyte gene expression in regenerating myofibers [10]. This suggests that other Wnts may act as endogenous inhibitors of adipogenesis in vivo, thereby compensating for the absence of Wnt10b. Although Wnt10a has been suggested as an endogenous inhibitor of brown adipogenesis [11, 12], whether Wnt10a or other Wnt ligands act as endogenous negative regulators of white adipogenesis has yet to be reported. Moreover, the mechanisms through which Wnts regulate MSC fate remain poorly understood.

In this manuscript, we identify Wnt10a and Wnt6 as additional Wnt ligands that inhibit adipogenesis and stimulate osteoblastogenesis. We show that Wnt6 and Wnt10a expression decreases during adipogenesis in vitro and in vivo, and that ectopic expression of Wnt6 or Wnt10a inhibits adipogenesis to a similar extent as Wnt10b. Although ectopic Wnt6 stimulates osteoblastogenesis to a weaker extent than Wnt10a or Wnt10b, we show that knockdown of Wnt6 is associated with the greatest stimulation of adipogenesis and impairment of osteoblastogenesis. This suggests that Wnt6 is a more potent endogenous regulator of MSC fate than Wnt10a or Wnt10b, at least in vitro. Finally, we demonstrate that β-catenin is absolutely required for the regulation of adipogenesis and osteoblastogenesis by Wnt6, Wnt10a or Wnt10b. Thus, mechanisms downstream of β-catenin are responsible for the regulation of MSC fate by these Wnt ligands.

2 Materials and Methods

Cell Culture

3T3-L1 cells were maintained in Dulbecco’s modified Eagle’s medium containing 8% bovine calf serum (Denville Scientific, South Plainfield, NJ, USA), 100 units/mL penicillin, 100 μg/mL streptomycin, 2 mM L-glutamine and 1 mM sodium pyruvate in a 10% CO₂ incubator. For adipogenesis, cells were grown to confluence in the above medium containing 10% fetal bovine serum (Gemini Bio-products, Sacramento, CA, USA) in place of bovine calf serum. At two days post-confluence, adipogenesis was induced with methylisobutylxanthine, dexamethasone and insulin (MDI) as described previously [13]. Sub-maximal induction of 3T3-L1 adipogenesis with dexamethasone and insulin (DI) or dexamethasone (Dex) only was performed as follows: at two days post-confluence, cells were treated with DI or Dex rather than MDI. Two days later, cells were fed with fresh medium supplemented with 5 μg/mL insulin and 10% fetal bovine serum. For inhibition of adipogenesis by Wnt3a, recombinant murine Wnt3a (R&D Systems, Minneapolis, MN) was included in the adipogenic medium throughout the differentiation procedure. ST2 cells were maintained and differentiated in ST2 medium (⟨-minimal essential medium that contained 10% fetal bovine serum, 100 units/mL penicillin, 100 μg/mL streptomycin and 2 mM L-glutamine) in a 5% CO₂ incubator. For adipogenesis, cells that had been confluent for one day were fed with fresh ST2 medium supplemented with 5 μg/mL insulin, 0.5 mM methylisobutylxanthine, 1 μM dexamethasone and 5 μM troglitazone (day 0). Cells were subsequently fed on day two with fresh ST2 medium plus 5 μg/mL insulin and 5 μM troglitazone, and re-fed with fresh ST2 medium every two days thereafter. To induce osteoblastogenesis, ST2 cells were grown to confluence and fed with osteogenic medium (ST2 medium supplemented with 10 mM β-glycerophosphate and 25 μg/mL ascorbic acid-2-phosphate). Cells were fed with fresh osteogenic medium every two days thereafter. Where indicated, osteoblastogenesis was enhanced by supplementing the osteogenic medium with 3 μM CHIR99021 [4] (Stemgent, Cambridge, MA), as described previously [9]. Accumulation of neutral lipids in adipocytes was evaluated with Oil Red-O staining [13]. The degree of mineralization in osteoblasts was determined with Alizarin Red staining and was quantified by assaying calcium content, both as described previously [9].

Animals

Epididymal adipose tissue was isolated from 16-week-old C57BL/6 mice (The Jackson Laboratory, Bar Harbor, ME) and separated into stromovascular and adipocyte fractions for RNA purification, as described previously [4]. All animal procedures were approved by the University of Michigan committee on the use and care of animals, with daily care of mice overseen by the unit for laboratory animal medicine.

Retroviral Infection and Constructs

Genes were stably introduced into 3T3-L1 and ST2 cells by retroviral infection as described previously [14]. pLNCX was purchased from Clontech (Mountain View, CA, USA). To generate pLNCX-Wnt6, a 1162 bp insert containing the full coding sequence (−19 bp to +1140 bp, relative to start codon) of murine Wnt6 was cloned into the ClaI (5′) and HindIII (3′) restriction sites of pLNCX. To generate pLXSN-Wnt10a, we subcloned Wnt10a from pBluescript-Wnt10a [15] into the EcoRI (5′) and XhoI (3′) restriction sites of modified pLXSN [2]. pLXSN-Wnt10b was described by Ross et al [2]. Wnt6, Wnt10a and Wnt10b were stably knocked down by expression of shRNAs from the pSuperior.retro.puro vector (OligoEngine, Seattle, WA, USA). The corresponding shRNAs were designed according to the manufacturer’s instructions to target the following transcript sequences: Wnt6 (5′-GCATTGGTGCAACTGCACAAC-3′); Wnt10a (5′-GAATGAGACTCCACAACAACC-3′); Wnt10b (5′-GCTGTGTGATGAGTGTAAAGT-3′). β-catenin was stably knocked down by expression of an shRNA from the pSiren-RetroQ vector (Clontech, Mountain View, CA, USA), as described previously [13]. This β-catenin shRNA plasmid, and the control shRNA plasmid expressing shRNA against firefly luciferase, were both kindly provided by Jaswinder K. Sethi (University of Cambridge, UK).

Cell Lysates and Immunoblotting

For whole-cell lysates, cells were washed once with phosphate-buffered saline, scraped into lysis buffer (1% SDS, 12.7 mM EDTA, 60 mM Tris-HCl; pH 6.8) and homogenized when necessary by passing through a 26 gauge needle five times. Lysates were then centrifuged at 20,000 rcf for 15 min at 4 °C and supernatants transferred to fresh tubes and stored at −80 °C. Cytosolic protein lysates were prepared as described previously [4]. Protein concentration in cell lysates was estimated using the BCA protein assay (Thermo Scientific, Waltham, MA, USA). For SDS-PAGE, lysates were diluted to equal protein concentration in lysis buffer plus 1X NuPage LDS buffer (Invitrogen, Carlsbad, CA, USA) supplemented with 2.5% 2-mercaptoethanol. Samples were boiled for 5 min, cooled on ice for 1 min, vortexed, and equal protein amounts separated on gradient polyacrylamide gels (Invitrogen, Carlsbad, CA, USA). Samples were then transferred to Immobilon PVDF membranes (Millipore, Billerica, MA, USA). Equal protein loading between lanes was confirmed by Ponceau staining of membranes after transfer. Membranes were blocked in 5% milk (1 h at room temperature) and then immunoblotted with the indicated primary antibodies (each in 5% BSA), and HRP-conjugated secondary antibody (1:5000 dilution in 5% milk; IgG-peroxidase; GE HealthCare, Waukesha, WI, USA) was visualized with Super Signal enhanced chemiluminescence (Pierce, Rockford, IL, USA). Mouse monoclonal PPARγ antibody (catalog # MAB3872) was from Millipore (Billerica, MA, USA). Rat monoclonal α-tubulin antibody (catalog # MA1-80017) was from Thermo Scientific (Waltham, MA, USA). Rabbit monoclonal ERK1/2 antibody (catalog # 4695) was from Cell Signaling Technology (Danvers, MA, USA). Rat monoclonal FABP4 antibody (catalog # MAB1143) was from R&D systems (Minneapolis, MN, USA). Mouse monoclonal β-catenin antibody (catalog # 610154) was from BD transduction laboratories (San Jose, CA, USA).

Real-time qPCR

One μg of total RNA was reverse-transcribed to cDNA using TaqMan RT reagents (Applied Biosystems, Carlsbad, CA, USA). Quantitative PCR (qPCR) was performed using Platinum Taq polymerase (Invitrogen, Carlsbad, CA, USA), with SYBR green I used to monitor amplification of DNA on the I-Cycler thermal cycler and IQ real-time PCR detection system (Bio-Rad, Hercules, CA, USA). Prior to use, all primers were validated with a cDNA titration and product specificity was confirmed via melting curve analysis and electrophoresis of qPCR products [16]. Expression of each gene was calculated based on a cDNA titration within each plate, and was then normalized to the expression of TBP mRNA or 18S rRNA. The corresponding primer sequences are shown in Table 1.

Table 1.

– Sequences of primers for qPCR

Transcript	Primer sequence (5′-3′)
18S rRNA [9]	(F) CGATGCTCTTAGCTGAGTGT (R) GGTCCAAGAATTTCACCTCT
Tbp	(F) ACCTTATGCTCAGGGCTTGG (R) GCCATAAGGCATCATTGGAC
Wnt6 (set 1)	(F) GCGGAGACGATGTGGACTTC (R) ATGCACGGATATCTCCACGG
Wnt6 (set 2)	(F) ATGGATGCGCAGCACAAGCG (R) ATGGATGCGCAGCACAAGCG
Wnt10a	(F) CCACTCCGACCTGGTCTACTTTG (R) TGCTGCTCTTATTGCACAGGC
Wnt10b (set 1)	(F) ACGACATGGACTTCGGAGAGAAGT (R) CATTCTCGCCTGGATGTCCC
Wnt10b (set 2)	(F) GCTGACTGACTCGCCCACCG (R) AAGCACACGGTGTTGGCCGT
Pparg [3]	(F) GGAAAGACAACGGACAAATCAC (R) TACGGATCGAAACTGGCAC
Fabp4 [3]	(F) TGGAAGCTTGTCTCCAGTGA (R) AATCCCCATTTACGCTGATG
Cebpa [3]	(F) TGGACAAGAACAGCAACGAG (R) TCACTGGTCAACTCCAGCAC
Nr2f2 (COUP-TFII)	(F) TTCACCCATGTCAGCCGAC (R) GGCCTTGAGGCAGCTATACTC
Alpl	(F) ACACCAATGTAGCCAAGAATGTCA (R) GATTCGGGCAGCGGTTACT
Ocn	(F) CCACCCGGGAGCAGTGT (R) CTAAATAGTGATACCGTAGATGCGTTTG
Runx2	(F) TTTAGGGCGCATTCCTCATC (R) TGTCCTTGTGGATTAAAAGGACTTG
Osterix	(F) TCTCAAGCACCAATGGACTCCT (R) GGGTAGTCATTTGCATAGCCAGA
Twist1	(F) ACGCAGTCGCTGAACGAGGC (R) GTCAGGGAAGTCGATGTACC
Dlx5	(F) TGACAGGCGTGTTTGACAGAAGAG (R) CGGGAACGGAGCTTGGA
Igf1	(F) CTACAAAAGCAGCCCGCTCT (R) CTTCTGAGTCTTGGGCATGTCA
Id2	(F) GCATCCTGTCCTTGCAGGCATCTG (R) AGTCCAGGCCGGAGAACGACA
β-catenin [13]	(F) CCGTTCGCCTTCATTATGGA (R) GGCAAGGTTTCGAATCAATCC
Tle3 [17]	(F) TGGTGAGCTTTGGAGCTGTT (R) CGGTTTCCCTCCAGGAAT

Statistical analyses

Statistical significance was determined using a two-tailed Student’s t-test assuming equal variances, and is indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001.

3 Results

3.1 Expression of Wnt6 and Wnt10a decreases during adipogenesis

Endogenous inhibitors of adipocyte differentiation, such as Wnt10b, are typically downregulated during adipogenesis [18]. Therefore, to identify additional Wnt ligands that might act as endogenous inhibitors of adipogenesis, we first profiled Wnt ligand expression in the adipocyte and stromovascular (preadipocyte-containing) fractions of WAT. As expected, Wnt10b mRNA was markedly reduced in adipocytes relative to stromovascular cells, whereas expression of the adipocyte genes, FABP4 and PPARγ, was enriched in the adipocyte fraction (Fig. 1A). Among the other Wnt ligands, Wnt6 and Wnt10a were decreased in adipocytes relative to stromovascular cells to a similar extent as Wnt10b (Fig. 1A). Based on this expression profile, we investigated whether Wnt6 or Wnt10a are also suppressed during in vitro adipogenesis of bipotential ST2 cells or 3T3-L1 preadipocytes. For both cell types, adipogenesis was confirmed by Oil Red-O staining for neutral lipid accumulation (Fig. 1B, C) and by elevated expression of PPARγ and FABP4 (Suppl. Fig. 1). As shown in Figures 1D and 1E, both Wnt6 and Wnt10a mRNAs were suppressed to a similar extent as Wnt10b during both ST2 and 3T3-L1 adipogenesis. These data reveal that expression of Wnt6 and Wnt10a, like that of Wnt10b, is decreased in the adipocyte fraction of WAT in vivo and during white adipogenesis in vitro, suggesting that Wnt6 and Wnt10a might also repress adipogenesis.

Figure 1. Wnt6, Wnt10a and Wnt10b mRNA expression decreases during adipogenesis in vivo and in vitro.

(A) Epididymal WAT from C57BL/6J mice was separated into stromovascular (SVF) and Adipocyte fractions. Total RNA was isolated from each fraction, and the expression of FABP4, PPARγ, Wnt10b, Wnt10a and Wnt6 was analyzed by qPCR and normalized to 18S rRNA. Data are presented relative to the maximum expression for each transcript as mean ± SD (n = 5 or 6). (B-E) ST2 bipotential cells (B, D) or 3T3-L1 preadipocytes (C, E) were induced to differentiate into adipocytes. (B, C) Adipogenesis was assessed by staining with Oil Red-O at day 0 or days 6-8 post-induction. Micrographs of stained cells, representative of at least three independent experiments, are shown. (D, E) Total RNA was isolated at the indicated time points and the transcript expression of Wnt6, Wnt10a and Wnt10b was analyzed by qPCR and normalized to 18S rRNA. Data are presented relative to expression at day 0 as mean ± SD of three independent experiments. Statistically significant differences between SVF and Adipocytes (A) or compared to transcript levels at day 0 (D, E) are indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001.

3.2 Ectopic expression of Wnt10a or Wnt6 inhibits adipogenesis

To investigate whether Wnt6 or Wnt10a inhibit preadipocyte differentiation, we retrovirally expressed Wnt6 or Wnt10a, or an empty vector control (EV), in ST2 cells and 3T3-L1 preadipocytes. Wnt10b-expressing cells were similarly generated to allow comparison to the effects of ectopic Wnt6 or Wnt10a. Quantitative PCR confirmed increased expression of Wnt6, Wnt10a or Wnt10b in each cell line, relative to EV cells (Figs. 2A, B). Ectopic Wnt expression was associated with increased levels of free cytosolic β-catenin, albeit to a lesser extent in the Wnt6-expressing cells than in cells expressing Wnt10a or Wnt10b (Figs. 2C, D). In some cases, ectopic expression of one Wnt was associated with decreased endogenous transcripts for other Wnts (Figs. 2A, B), although this was not consistently observed through all experiments (see Fig. 6). Ectopic Wnt10a or Wnt10b suppressed expression of FABP4, PPARγ and C/EBPα in ST2 cells (Fig. 2A), and all three Wnts suppressed transcripts for these genes in 3T3-L1 preadipocytes (Fig. 2B). Thus, Wnt6, Wnt10a and Wnt10b suppress the expression of adipocyte genes, even before adipogenesis is induced.

Figure 2. Wnt6, Wnt10a and Wnt10b inhibit adipogenesis.

ST2 cells or 3T3-L1 preadipocytes were infected with retroviruses for stable expression of Wnt6, Wnt10a, Wnt10b or an EV control. (A, B) Cells were grown to confluence, total RNA was isolated and the expression of Wnt6, Wnt10a, Wnt10b, PPARγ, FABP4 and CEBPα was analyzed by qPCR. (C, D) β-catenin was analyzed by immunoblotting in cytosolic lysates of confluent EV- and Wnt-expressing cells; α-tubulin was used as a loading control. (E-H) EV and Wnt-expressing ST2 cells (E, G) or 3T3-L1 preadipocytes (F, H) were induced to differentiate into adipocytes. At day 6 (ST2) or day 8 (3T3-L1) post-induction, extent of adipogenesis was assessed by Oil Red-O staining (E, F) and analysis of adipocyte gene expression (G, H). Transcript expression in (A), (B), (G) and (H) was normalized to 18S rRNA and is presented relative to the maximum expression of each transcript as mean ± SD of at least three independent experiments. Statistical significance compared to transcript expression in EV cells is indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001. Immunoblots and images of Oil Red-O-stained cells are representative of at least three independent experiments.

Figure 6. β-catenin is required for suppression of PPARγ by Wnt6, Wnt10a or Wnt10b.

Wnt-expressing ST2 cells or 3T3-L1 preadipocytes were infected with retroviruses expressing shRNAs against β-catenin (shβ-cat) or a control shRNA against firefly luciferase (shControl). (A) Total RNA was isolated from confluent cells and expression of the indicated transcripts was analyzed by qPCR and normalized to TBP mRNA (ST2s) or 18S rRNA (3T3-L1s). Data are presented relative to the maximum expression of each mRNA as mean +/− SD of three independent experiments. Statistically significant differences in transcript expression are indicated as follows: Wnt-expressing cells vs EV, * = P < 0.05; ** = P < 0.01; *** = P < 0.001. shControl vs shβ-catenin cells, # = P < 0.05; ## = P < 0.01; ### = P < 0.001. (B) Knockdown of β-catenin was confirmed by immunoblotting of whole-cell lysates prior to induction of differentiation; α-tubulin or ERK1/2 were used as loading controls. Immunoblots are representative of three independent experiments.

Effects of ectopic Wnts on adipogenesis were then investigated. Quantitative PCR confirmed maintenance of ectopic Wnt expression throughout adipogenesis (Figs. 2G, H). The EV ST2 and 3T3-L1 cells differentiated into adipocytes, as assessed by Oil Red-O staining (Figs. 2E, F) and adipocyte gene expression (Figs. 2G, H). In contrast, ectopic Wnt6, Wnt10a or Wnt10b completely prevented neutral lipid accumulation and markedly suppressed PPARγ, C/EBPα and FABP4 in both cell types (Figs. 2E-H). Although each of these Wnts inhibited 3T3-L1 and ST2 adipogenesis, the effects of Wnt6 were slightly weaker than those of Wnt10a or Wnt10b. These results demonstrate that, like Wnt10b, both Wnt6 and Wnt10a can stabilize β-catenin and inhibit adipogenesis.

3.3 Stable expression of Wnt10a or Wnt6 stimulates osteoblastogenesis

Because Wnt10b also stimulates osteoblast differentiation [7-9], we next investigated the expression and function of Wnt6, Wnt10a and Wnt10b in the context of ST2 osteoblastogenesis. We analyzed the expression of these Wnt ligands during osteoblastogenesis in ST2 cells. Differentiation into osteoblasts was confirmed by staining for matrix mineralization with Alizarin red (Suppl. Fig. 2A), and by increased expression of alkaline phosphatase and osteocalcin, two osteoblast marker genes (Suppl. Fig. 2B, C). Expression of Wnt6, Wnt10a and Wnt10b was detectable during osteoblastogenesis; however, the degree of expression did not change during differentiation (Suppl. Fig. 2D-F). These data suggest that, in contrast to adipogenesis, transcripts for these Wnt ligands are not regulated during ST2 osteoblastogenesis.

Nevertheless, given that Wnt10b stimulates osteoblast differentiation [7-9], we next investigated whether ectopic Wnt6 or Wnt10a also promote osteoblastogenesis. To do so, we first analyzed whether ectopic Wnts affect expression of genes related to osteoblastogenesis prior to the induction of differentiation. As shown in Figure 3A, ectopic Wnt10a or Wnt10b potently stimulated expression of alkaline phosphatase in ST2 cells. Ectopic Wnt6 also increased alkaline phosphatase expression, albeit to a far lesser extent than ectopic Wnt10a or Wnt10b (Fig. 3A). Each of the Wnt-expressing cells also displayed upregulation of Twist1 (Fig. 3A), a transcription factor that modulates osteoblastogenesis [19]. However, Wnt6, Wnt10a or Wnt10b did not significantly affect expression of several other genes associated with osteoblast differentiation or activity (Fig. 3A). These cells were then induced to differentiate into osteoblasts and the extent of differentiation was determined by analyses of matrix mineralization. This revealed that Wnt10a or Wnt10b strongly stimulate osteoblastogenesis, with marked increases in Alizarin red staining and calcium content relative to EV cells (Figs. 3B, C). Wnt6 also stimulated osteoblastogenesis; however, effects were weaker than those of Wnt10a or Wnt10b (Figs. 3B, C). These data demonstrate that Wnt6 and Wnt10a, like Wnt10b, can stimulate osteoblast differentiation.

Figure 3. Wnt6, Wnt10a and Wnt10b stimulate osteoblastogenesis.

(A) Expression of alkaline phosphatase (Alpl), osteocalcin (Ocn), Osterix (Osx), Twist1, Dlx5 and Runx2 in RNAs from ST2 cells (Fig. 2A) was analyzed by qPCR. Transcripts were normalized to 18S rRNA and are presented relative to maximum expression as mean ± SD of at least three experiments. (B, C) EV- or Wnt-expressing ST2 cells were induced to differentiate into osteoblasts. The extent of osteoblastogenesis was assessed at day nine post-induction by staining for matrix mineralization with Alizarin red (B) and by quantification of matrix calcium content (C). Micrographs and wells of stained cells (B) are representative of at least three independent experiments. Calcium content (C) was normalized to cellular protein content and is presented as mean ± SD of four biological replicates per cell type, representative of two experiments. Statistically significant differences in transcript expression or calcium content between Wnt-expressing and EV cells are indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001.

3.4 Knockdown of Wnt6, Wnt10a or Wnt10b enhances adipogenesis and impairs osteoblastogenesis

The above findings demonstrate that ectopic expression of Wnt6, Wnt10a or Wnt10b inhibits adipogenesis and stimulates osteoblastogenesis. However, whether endogenous expression of these Wnt ligands also modulates fate of mesenchymal precursors remained to be determined. To investigate this possibility, we generated ST2 cells with shRNA-mediated knockdown of Wnt6, Wnt10a or Wnt10b. Each of these Wnt ligands was significantly suppressed by expression of their respective shRNAs (Fig. 4A). Wnt10b expression was also significantly reduced in the shWnt6 and shWnt10a cells, and Wnt6 expression was 80% lower in the shWnt10b cells (Fig. 4A), consistent with mutual cross-regulation of Wnt expression [20]. We encountered several technical difficulties in assessing Wnt knockdown in these cell lines. Using our original Wnt6 qPCR primers (Table 1; primer set 1), we could not detect Wnt6 knockdown in the shWnt6 ST2 cells (data not shown). However, Wnt6 mRNA knockdown was consistently detectable in these cells (Fig. 4A) using qPCR primers that flank the Wnt6 shRNA target site (Table 1; primer set 2). The extent of Wnt10b knockdown was also greater when assessed using qPCR primers that flank the Wnt10b shRNA target site (Table 1; primer set 2). These observations are consistent with a previous study showing that qPCR primer position can affect the efficacy of detecting shRNA-mediated knockdown by qPCR [21]. Additionally, knockdown of Wnt10a in the shWnt10a cells was only detectable in the first passage of cells selected after retroviral infection (Fig. 4A). In subsequent passages of these cells, knockdown of Wnt10a mRNA was no longer apparent, regardless of qPCR primer position (data not shown). Nevertheless, β-catenin protein was consistently lower in each Wnt knockdown cell line (Fig. 4B), suggesting functional knockdown of each of these Wnt ligands in ST2 cells. We therefore investigated effects of the Wnt knockdowns on ST2 adipogenesis.

Figure 4. Wnt knockdown enhances adipogenesis.

(A-D) ST2 cells were infected with retroviruses for expression of shRNAs against Wnt6, Wnt10a, Wnt10b or scrambled control (shControl). (E-H) 3T3-L1 preadipocytes were infected with retroviruses for expression of shWnt6 or shControl. (A, E) Total RNA was isolated from confluent cells and the expression of the indicated transcripts was analyzed by qPCR. (B, F) β-catenin was analyzed by immunoblotting in whole-cell protein lysates of confluent cells; α-tubulin was used as a loading control. (C, D) shRNA-expressing ST2 cells were induced for adipogenesis with MDI +/− TZD (5 μM). At day 6 post-induction, cells were stained with Oil Red-O (C) or total RNA was isolated for analysis of adipocyte gene expression by qPCR (D). (G, H) shControl and shWnt6-expressing 3T3-L1 preadipocytes were induced for adipogenesis as indicated. At day 8 post-induction, cells were stained with Oil Red-O (G) or whole-cell protein lysates were immunoblotted for PPARγ and FABP4; ERK1/2 was used as a loading control (H). Data in (A), (D) and (E) were normalized to 18S rRNA or TBP mRNA and are presented relative to the maximum expression of each transcript as mean +/− SD of 3-4 biological replicates representative of one to four independent experiments. Statistical significance compared to transcript expression in shControl cells is indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001. Immunoblots (B, F, H) and images of Oil Red-O-stained cells (C, G) are representative of at least three independent experiments.

In confluent ST2 cells before inducing adipogenesis, knockdown of Wnts generally increased the expression of FABP4, PPARγ and Id2 (Inhibitor of DNA binding 2), a transcription factor that stimulates PPARγ expression and adipogenesis [22] (Fig. 4A). In contrast, knockdown of Wnt6 or Wnt10b was associated with decreased expression of TLE3 (transducin-like enhancer of split 3; Fig. 4A), a transcriptional co-regulator that enhances PPARγ activity [17]. Induction of adipogenesis with MDI only was associated with relatively weak differentiation in shControl cells (Fig. 4C, D). However, MDI-induced adipogenesis was enhanced in each Wnt knockdown cell line, with shWnt6 cells displaying the greatest increases in adipocyte marker gene expression (Fig. 4C, D). Including TZD in the differentiation cocktail further increased adipogenesis in shControl cells (Fig. 4C, D). However, even with TZD, lipid accumulation and adipocyte marker genes tended to be higher in each Wnt knockdown cell line, with shWnt10b cells showing the strongest effects (Fig. 4C, D). These data suggest that endogenous Wnt6, Wnt10a, and Wnt10b inhibit ST2 adipogenesis.

We further investigated effects of Wnt knockdown on 3T3-L1 adipogenesis. Wnt6 was knocked down by over 60% in shWnt6-expressing 3T3-L1 preadipocytes (Fig. 4E). However, both Wnt10a and Wnt10b mRNAs were also significantly reduced in these cells, consistent with the mutual cross-regulation observed with Wnt knockdown in ST2 cells (Fig. 4A). Reduced expression of Wnt6, Wnt10a, and Wnt10b in shWnt6 3T3-L1 preadipocytes was associated with decreased total β-catenin protein (Fig. 4F) and elevated FABP4 mRNA (Fig. 4E). In contrast, decreased Wnt expression did not affect PPARγ, C/EBPα or TLE3 mRNAs, and Id2 expression was over 80% lower in shWnt6 relative to shControl preadipocytes (Fig. 4E). Induction of adipogenesis with full adipogenic cocktail (MDI) or under limiting conditions (DI or Dex) revealed a dramatic enhancement of adipogenesis in the shWnt6-expressing cells (Fig. 4G, H). The largest differences in lipid accumulation or expression of PPARγ and FABP4 were apparent in response to induction of adipogenesis with DI or Dex only. However, even with MDI, shWnt6 cells accumulated more lipid and expressed higher levels of PPARγ than shControl cells (Figs. 4G, H). These findings confirm that endogenous Wnt6, Wnt10a and Wnt10b repress 3T3-L1 preadipocyte differentiation.

The effects of Wnt knockdown on ST2 osteoblastogenesis were next assessed. Alkaline phosphatase expression was suppressed by over 90% in each of the shWnt cell lines prior to exposure to osteogenic media (Fig. 5A). The shControl and Wnt knockdown cells were then induced to differentiate into osteoblasts in the absence or presence of CHIR99021, a GSK3 inhibitor that stabilizes β-catenin and thereby enhances osteoblastogenesis [9]. In the absence of CHIR99021, osteoblastogenesis was impaired in each of the Wnt knockdown cells, as assessed by Alizarin red staining and quantification of matrix calcium content (Figs. 5B, C). Although CHIR99021 markedly enhanced osteoblast differentiation in the shControl cells (Figs. 5B-C), this effect was blunted in the shWnt10a cells and completely blocked in shWnt6 and shWnt10b cells (Figs. 5B-C). These findings suggest that endogenous Wnt6, Wnt10a and Wnt10b are required for ST2 osteoblastogenesis.

Figure 5. Wnt knockdown impairs osteoblastogenesis.

(A) Alkaline phosphatase expression in RNAs from confluent ST2 cells (Fig. 4A) was analyzed by qPCR and normalized to TBP expression and is presented relative to expression in shControl cells as mean +/− SD of four biological replicates. (B, C) shControl, shWnt6, shWnt10a and shWnt10b ST2 cells were induced to differentiate into osteoblasts. At day 6 post-induction, the extent of osteoblastogenesis was assessed by staining with Alizarin red (B) and by quantification of matrix calcium content (C). Plates and micrographs of Alizarin red-stained cells (B) are representative of three independent experiments. Calcium content (C) was normalized to cellular protein content and is presented as mean +/− SD of four biological replicates per cell type, representative of two experiments. Statistically significant differences in Alp expression or calcium content between shControl and shWnt cells are indicated by *** (= P < 0.001).

3.5 Wnt6, Wnt10a and Wnt10b inhibit adipogenesis and stimulate osteoblastogenesis through a β-catenin-dependent pathway

We next investigated the mechanisms underlying regulation of MSC fate by Wnt6, Wnt10a and Wnt10b. Forced stabilization of β-catenin inhibits adipogenesis [2, 13] and β-catenin is required for osteoblast differentiation and mineralization [23-26]. Therefore, given that β-catenin levels are elevated by ectopic Wnt expression (Fig. 2) and reduced by Wnt knockdown (Fig. 4), it is highly likely that β-catenin mediates the effects of Wnt6, Wnt10a and Wnt10b on adipogenesis and osteoblastogenesis. To investigate this possibility, we stably knocked down β-catenin in Wnt-expressing ST2 and 3T3-L1 cell lines. Quantitative PCR confirmed knockdown of β-catenin by 60% in ST2 cells and by over 75% in 3T3-L1 preadipocytes (Fig. 6A). Knockdown of β-catenin did not affect expression of endogenous Wnt6, Wnt10a or Wnt10b, and ectopic expression of these Wnts was apparent in both shControl and shβ-catenin cells (Fig. 6A). Indeed, ectopic Wnt6, Wnt10a or Wnt10b stabilized β-catenin protein in shControl cells, whereas β-catenin protein was undetectable in shβ-catenin ST2 or 3T3-L1 cells (Fig. 6B). Ectopic Wnt6, Wnt10a or Wnt10b also increased β-catenin transcript expression in the shControl ST2 cells; however, this effect was not observed in 3T3-L1 preadipocytes (Fig. 6A).

We next investigated effects of β-catenin knockdown on the inhibition of adipogenesis by Wnt6, Wnt10a, or Wnt10b. Consistent with results in Figure 2, ectopic Wnt6, Wnt10a or Wnt10b significantly suppressed PPARγ mRNA in shControl cells, even before the induction of adipogenesis (Fig. 6A). Knockdown of β-catenin prevented these effects and also significantly increased PPARγ mRNA in EV cells (Fig. 6A). After inducing adipogenesis, ectopic Wnt6, Wnt10a or Wnt10b robustly suppressed lipid accumulation (Fig. 7A) and expression of PPARγ and FABP4 (Fig. 7B) in shControl cells. Knockdown of β-catenin completely prevented these effects (Figs. 7A, B) and alone enhanced ST2 adipogenesis, with shβ-catenin EV cells expressing more PPARγ and FABP4 than the shControl EV cells (Fig. 7B). Finally, β-catenin knockdown completely prevented the inhibition of 3T3-L1 adipogenesis by Wnt3a (Suppl. Fig. 3). These results conclusively show that β-catenin is required for the inhibition of adipogenesis by Wnt10b, Wnt10a, Wnt6 and Wnt3a.

Figure 7. β-catenin is required for suppression of adipogenesis by Wnt6, Wnt10a or Wnt10b.

Wnt-expressing shControl or shβ-catenin ST2 cells and 3T3-L1 preadipocytes were induced to differentiate into adipocytes and the extent of differentiation at day 6 (ST2) or day 8 (3T3-L1) was assessed by Oil Red-O (A) or by immunoblotting for PPARγ and FABP4 in whole-cell lysates, with ERK1/2 used as a loading control (B). These data are representative of at least three independent experiments.

The effects of β-catenin knockdown on osteoblast differentiation were then studied. Consistent with results in Figure 3, ectopic Wnt6, Wnt10a or Wnt10b markedly increased alkaline phosphatase expression in shControl ST2 cells before induction of osteoblastogenesis, with Wnt10a or Wnt10b again exerting a more potent effect than Wnt6 (Fig. 8A). β-catenin knockdown significantly decreased alkaline phosphatase expression by 70% in EV cells, and completely prevented the induction of alkaline phosphatase by Wnt6, Wnt10a or Wnt10b (Fig. 8A). We then induced osteoblastogenesis in each of these cell lines in the absence or presence of CHIR99021. As expected, ectopic Wnt6, Wnt10a, Wnt10b or CHIR99021 stimulated matrix mineralization in shControl ST2 cells, with Wnt6 again showing the weakest activity (Figs. 8B-C). β-catenin knockdown completely prevented these effects (Figs. 8B-C), conclusively showing that β-catenin is required for the stimulation of osteoblastogenesis by Wnt10b, Wnt10a, Wnt6, or by inhibition of GSK3.

Figure 8. β-catenin is required for stimulation of osteoblastogenesis by Wnt6, Wnt10a or Wnt10b.

(A) Alkaline phosphatase expression in RNAs from Wnt-expressing shControl or shβ-catenin ST2 cells (Fig. 6A) was analyzed by qPCR and normalized to TBP mRNA. Data are presented as mean +/− SD of three biological replicates from a representative experiment. (B, C) Wnt-expressing shControl or shβ-catenin ST2 cells were induced to differentiate into osteoblasts in the absence or presence of CHIR99021 (3 μM). At day 8 post-induction, the extent of osteoblastogenesis was assessed by staining with Alizarin red (A) and by quantification of matrix calcium content. Plates and micrographs of Alizarin red-stained cells (A) are representative of three independent experiments. Calcium content (B) was normalized to cellular protein content and is presented as mean ± SD of four replicate samples per cell type, from one of two independent experiments. Statistically significant differences in mRNA expression or calcium content are indicated as follows: Wnt-expression or CHIR99021 treatment vs EV cells, * = P < 0.05; ** = P < 0.01; *** = P < 0.001. shControl vs shβ-catenin cells, # = P < 0.05; ## = P < 0.01; ### = P < 0.001.

3.6 Mechanisms of Wnt-induced MSC fate regulation downstream of β-catenin

We next investigated whether previously identified regulators of adipogenesis are targeted by Wnts in a β-catenin-dependent manner. As a positive control, we first analyzed expression of IGF-1, which we previously identified as a Wnt1 target gene in 3T3-L1 preadipocytes [27]. As shown in Figure 9A, Wnt6, Wnt10a and Wnt10b each increased IGF-1 mRNA. β-catenin knockdown prevented this effect and alone was sufficient to suppress IGF-1 expression by over 35% in EV cells. This finding confirmed the utility of these cell lines for the identification of Wnt/β-catenin target genes. The transcription factor COUP-TFII (Chicken ovalbumin upstream promoter transcription factor 2; also called Nr2f2) inhibits adipogenesis by suppressing PPARγ expression [28, 29]. Okamura et al reported that Wnt3a increases COUP-TFII expression, and that β-catenin knockdown decreases basal amounts of COUP-TFII protein [28]. Thus, they proposed that COUP-TFII mediates the inhibition of adipogenesis by Wnt signaling. In contrast, we found no effect of β-catenin knockdown on COUP-TFII mRNA in control 3T3-L1 or ST2 cells (Figs. 9A, B). Furthermore, ectopic Wnt6, Wnt10a and Wnt10b each suppressed COUP-TFII expression in 3T3-L1 preadipocytes in a β-catenin-dependent manner (Fig. 9A), but did not affect COUP-TFII expression in ST2 cells. These data suggest that, under our experimental conditions, Wnts do not stimulate COUP-TFII expression in mesenchymal precursors. Additionally, we found that sustained suppression of COUP-TFII during 3T3-L1 adipogenesis is not observed until after day 4 of preadipocyte differentiation (Suppl. Fig. 1E), consistent with a previous study [29]. In contrast, Wnt/β-catenin signaling is rapidly suppressed upon induction of 3T3-L1 adipogenesis (Fig. 1E) [2, 4, 30]. These observations are not consistent with COUP-TFII mediating the inhibition of adipogenesis by Wnt signaling.

Figure 9. Expression of adipogenic regulators in Wnt-expressing shControl and shβ-catenin 3T3-L1 preadipocytes and ST2 cells.

Expression of the indicated mRNAs was analyzed in RNA samples from Wnt-expressing shControl or shβ-catenin 3T3-L1 preadipocytes or ST2 cells (Fig. 6A) by qPCR and normalized to 18S rRNA (A) or TBP mRNA (B). Data are presented relative to the maximum expression for each transcript as mean +/− SD of three independent experiments. Statistically significant differences in transcript expression are indicated as follows: Wnt-expressing cells vs EV, * = P < 0.05; ** = P < 0.01; *** = P < 0.001. shControl vs shβ-catenin cells, # = P < 0.05; ## = P < 0.01; ### = P < 0.001.

As mentioned above, Id2 promotes adipogenesis by stimulating PPARγ expression [22]. Given that Wnt signaling suppresses Id2 expression [22], downregulation of Id2 might contribute to the repression of adipogenesis by Wnt signaling. Consistent with this hypothesis, we found that Wnt6, Wnt10a and Wnt10b decreased Id2 expression in 3T3-L1 preadipocytes in a β-catenin-dependent manner (Fig. 9A). However, these Wnts did not regulate Id2 expression in ST2 cells, although β-catenin knockdown was associated with increased Id2 mRNA in Wnt-expressing ST2 cells (Fig. 9B). Thus, suppression of Id2 by Wnt signaling may not be a universal mechanism for influencing fate of mesenchymal precursors.

Given that Wnt knockdown in ST2 cells was associated with suppression of TLE3 (Fig. 4A), we also investigated whether ectopic Wnts or β-catenin deficiency affected TLE3 expression. In shControl ST2 cells Wnt10a and Wnt10b each increased TLE3 expression, whereas Wnt6 had no effect (Fig. 9B). Although these effects of Wnt10a and Wnt10b were β-catenin-dependent, knockdown of β-catenin did not influence TLE3 expression in EV- or Wnt6-expressing ST2 cells (Fig. 9B). Similarly, TLE3 expression was not consistently regulated by Wnt expression or β-catenin knockdown in 3T3-L1 preadipocytes (Fig. 9A). These data suggest that Wnt6, Wnt10a or Wnt10b likely inhibit adipogenesis independently of effects on TLE3 mRNA expression.

4 Discussion

4.1 Wnt6 and Wnt10a as regulators of MSC fate

Although 19 Wnt ligands have been identified in mammals [1], few of these have been studied in the context of MSC fate. In addition to Wnt10b, ectopic Wnt1 and recombinant Wnt3a each suppress adipogenesis in vitro [2, 31], and Wnt5a has been reported to inhibit adipogenesis [32]. Conversely, other studies report stimulation of adipogenesis by Wnt5a, as well as by Wnt4 and Wnt5b [33-35]. Nishizuka et al also reported suppression of Wnt6 mRNA during adipogenesis; however, they did not investigate whether Wnt6 regulates adipogenesis [34]. Similarly, Wnt10a has been proposed as an endogenous inhibitor of brown adipogenesis [3, 11, 12], but this has not been empirically demonstrated. Thus, the present study is the first to show that Wnt6 and Wnt10a regulate fate of mesenchymal precursors.

Disruption of Wnt/β-catenin signaling promotes spontaneous adipogenesis in vitro [2, 4, 13], supporting the notion that endogenous Wnt ligands inhibit adipogenesis. Wnt10b has long been touted as the endogenous inhibitory Wnt; however, no published studies have conclusively demonstrated this. Although knockdown of pro-adipogenic Wnt4 or Wnt5a impairs adipogenesis [34], to our knowledge no previous studies have used stable Wnt knockdown to investigate endogenous anti-adipogenic Wnts. Our attempts to knock down Wnt6, Wnt10a or Wnt10a individually were complicated by technical difficulties in detecting Wnt knockdown in ST2 cells. The robust knockdown of β-catenin protein suggests that our Wnt knockdowns may be more apparent if assessed at the protein level, because the almost-total knockdown of β-catenin protein is far greater than the 60-75% knockdown detected for β-catenin mRNA (Fig. 6). Unfortunately, lack of reliable antibodies against Wnt6, Wnt10a or Wnt10b undermined our attempts to detect these proteins (data not shown). Nevertheless, our Wnt knockdown cells consistently display decreased β-catenin protein, enhanced adipogenesis and impaired osteoblastogenesis, suggesting functional Wnt knockdown in each of these cell lines.

Another observation from our shWnt-expressing cell lines is that, in all cases, Wnt knockdown is associated with decreased expression of other Wnts. This indicates potential positive feedback between Wnts, consistent with our previous finding that Wnt1 stimulates expression of Wnt4 and Wnt5a in preadipocytes [20]. Although the mechanisms underpinning such cross-regulation remain unclear, β-catenin is unlikely to be involved because knockdown of β-catenin does not affect endogenous expression of Wnt6, Wnt10a or Wnt10b (Fig. 6A). In any case, that knockdown is not limited to the shRNA target-Wnt also partially confounds our ability to use these cell lines to determine the actions of endogenous Wnt6, Wnt10a or Wnt10b individually. However, comparing results across cell lines allows informative conclusions to be drawn. In the ST2 cells, the greatest anti-osteogenic and pro-adipogenic effects are observed in the shWnt6 and shWnt10b cells, which are distinguished from the shWnt10a cells by having knockdown of Wnt6 but not of Wnt10a. Hence, amongst these three Wnts, only Wnt6 knockdown is uniquely associated with the strongest effects on MSC fate. Endogenous Wnt6 may therefore exert more potent effects on mesenchymal precursors than endogenous Wnt10a or Wnt10b, although potential synergy between combined Wnt6 and Wnt10b knockdown cannot be excluded. Conversely, we found that ectopic Wnt6 exerts weaker effects on β-catenin stabilization and MSC fate than ectopic Wnt10a or Wnt10b. However, we believe that this is likely a consequence of the weaker degree of relative overexpression of Wnt6 (compared to the relative overexpression of Wnt10a or Wnt10b), rather than reflecting inherent differences in the biological potency of each of these Wnts per se. Ultimately, approaches such as gene targeting may be required to corroborate our in vitro studies and to more firmly elucidate the relative individual abilities of endogenous Wnt6, Wnt10a or Wnt10b to regulate fate of mesenchymal precursors in vivo.

4.2 Regulation of Wnt expression

We investigated Wnt6 and Wnt10a as regulators of MSC fate, with less focus on mechanisms regulating Wnt6 or Wnt10a expression. Signaling via insulin receptor substrate-1 decreases Wnt10a and Wnt6 expression in brown adipogenesis [11, 12], suggesting that insulin may promote suppression of these Wnts in white adipogenesis. CREB activation can also decrease Wnt10a mRNA [11], consistent with the cyclic AMP-mediated suppression of Wnt10b in 3T3-L1 adipogenesis [4, 30]. However, which components of the adipogenic induction cocktail suppress Wnt6 or Wnt10a remains to be established. Although transcripts for these Wnts do not change during osteoblastogenesis (Suppl. Fig. 2), β-catenin is clearly required for osteoblast differentiation (Fig. 8) [23-26]. Therefore, osteoblastogenesis may be associated with increased Wnt/β-catenin signaling at a level independent of Wnt transcript expression, such as through regulation of Wnt secretion or expression of modulators of this pathway.

In addition to regulation during adipogenesis, physiological or pathophysiological conditions modulate Wnt expression in WAT and in brown adipose tissue (BAT). For example, cold exposure decreases expression of Wnt10b, but not of Wnt10a, in BAT [3]. However, effects of cold exposure on Wnt6 expression in BAT remain unaddressed. Additionally, obesity, TZD treatment, or feeding status modulate Wnt10b expression in WAT [36], which may link metabolic status to the regulation of adipogenesis in vivo. Whether nutritional signals also regulate WAT expression of Wnt10a and/or Wnt6 therefore remains an intriguing possibility.

4.3 Wnts in human disease

Mutations in genes encoding Wnt ligands have been associated with bone mass defects or susceptibility to metabolic diseases in humans, underscoring the importance of the Wnt pathway in the regulation of MSC fate. For example, polymorphisms in the WNT10B gene associate with bone mineral content [37, 38] or abdominal adiposity [39] in some human populations, and mutations in WNT10B have been associated with obesity [40]. Additionally, variants of WNT5B strongly associate with susceptibility to type 2 diabetes [41]. Given the clear impact of Wnt6 and Wnt10a on mesenchymal precursor fate in vitro, variants in these genes might also impact bone mass or metabolic disease in humans. Future studies should explore this possibility.

4.4 Role of β-catenin in modulation of adipogenesis and osteoblastogenesis by Wnts

Although it has long been assumed that Wnts inhibit adipogenesis primarily by targeting β-catenin, the present study is the first to conclusively demonstrate that β- catenin is required for Wnts to suppress adipocyte differentiation, at least for Wnt6, Wnt10a, Wnt10b and Wnt3a. In contrast, Wnt5a reportedly inhibits ST2 adipogenesis independently of β-catenin [32], and Wnt signaling via Fzd2 may also inhibit 3T3-L1 adipogenesis through a β-catenin-independent mechanism [31]. Additionally, β-catenin is implicated in the stimulation of adipogenesis by other Wnt ligands. Thus, Wnt5b promotes adipogenesis by antagonizing Wnt/β-catenin signaling [33, 35], which might also underlie the stimulation of adipogenesis by Wnt5a [34, 42, 43]. In contrast, Wnt4 reportedly stabilizes β-catenin [44], which is inconsistent with the suggestion that Wnt4 stimulates adipogenesis [34]. Ultimately, the requirement for β-catenin in Wnt-mediated MSC fate regulation could be more firmly established by investigating whether β-catenin knockdown affects the ability of Wnts to modulate adipogenesis or osteoblastogenesis. Indeed, β-catenin knockdown attenuates the inhibition of adipogenesis by mechanical strain [45] or by tumor necrosis factor-α [13]. Thus, our β-catenin knockdown cell lines serve as useful tools for assessing the β-catenin-dependency of Wnt ligands and other reported regulators of MSC fate.

4.5 Mechanisms downstream of β-catenin in MSC fate regulation

Even without ectopic Wnt expression, it is clear that β-catenin dramatically impacts MSC fate. That β-catenin knockdown enhances ST2 adipogenesis is consistent with the pro-adipogenic effects of β-catenin ablation reported previously [13, 46]. The requirement of β-catenin for osteoblast differentiation has also been firmly established [23-26]; hence, it is not surprising that our shβ-catenin ST2 cells are incapable of osteoblastogenesis. A remaining question regards how β-catenin impacts fate of mesenchymal precursors. Our identification of alkaline phosphatase as a β-catenin-dependent Wnt target gene may explain why β-catenin is necessary for osteoblastogenesis, because alkaline phosphatase is required for osteoblast matrix mineralization [47, 48]. Additionally, we show that endogenous β-catenin suppresses PPARγ expression in 3T3-L1 preadipocytes and ST2 cells (Figs. 6, 7). This likely also contributes to the requirement of β-catenin for osteoblast differentiation, because PPARγ suppresses osteoblastogenesis [9, 49, 50].

How Wnt/β-catenin signaling suppresses PPARγ is not thoroughly understood. We found that ectopic Wnt6, Wnt10a and Wnt10b signal through β-catenin to suppress Id2 expression in 3T3-L1 preadipocytes; however, knockdown of these Wnts also suppresses Id2 expression in this cell type (Figs. 4, 6). Moreover, in ST2 cells Wnt knockdown increases Id2 mRNA, whereas ectopic Wnts or β-catenin knockdown do not affect Id2 expression (Figs. 4, 6). Thus, although the downregulation of Id2 may contribute to the inhibition of 3T3-L1 adipogenesis by ectopic Wnt6, Wnt10a or Wnt10b, the suppression of Id2 is clearly not necessary for Wnt-induced anti-adipogenesis per se. Our data further demonstrate that Wnt6, Wnt10a or Wnt10b are unlikely to modulate MSC fate through effects on COUP-TFII or TLE3 transcript expression; however, it remains possible that Wnts target COUP-TFII or TLE3 activity post-transcriptionally to impact mesenchymal precursors.

Another unexplored possibility is that β-catenin directly inhibits adipogenic gene expression. One recent study shows that β-catenin binds to the FABP4 promoter in preadipocytes, but that this association decreases during adipogenesis [17]. Given that β-catenin can directly repress transcription [1], β-catenin might inhibit adipogenesis by directly repressing transcription from the promoters of adipocyte genes. Approaches such as ChIP-Seq [51, 52] could be used to identify β-catenin binding sites in preadipocytes and thereby further investigate this possibility.

5 Conclusions

In conclusion, we have identified Wnt10a and Wnt6 as endogenous regulators of adipogenesis and osteoblast differentiation in mesenchymal precursors. β-catenin is absolutely required for the inhibition of adipogenesis and stimulation of osteoblastogenesis by Wnt6, Wnt10a and Wnt10b. Moreover, each of these Wnts signal via β-catenin to suppress PPARγ and induce alkaline phosphatase expression, changes that contribute to their effects on fate of mesenchymal precursors. However, the mechanisms through which β-catenin impacts MSC fate remain incompletely understood and should be investigated further.

Supplementary Material

01

Supplemental Figure 1 – Expression of PPARγ, FABP4 and COUP-TFII during ST2 and 3T3-L1 adipogenesis.

Expression of PPARγ, FABP4 or COUP-TFII mRNAs from the ST2 and 3T3-L1 adipogenesis timecourses (Fig. 1) was analyzed by qPCR and normalized to 18S rRNA. Data are presented relative to expression at day 0 as mean ± SD of three independent experiments. Statistical significance compared to mRNA expression at day 0 is indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001.

02

Supplemental Figure 2 - Expression of Wnt6, Wnt10a and Wnt10b during ST2 osteoblastogenesis.

ST2 bipotential cells were induced to differentiate into osteoblasts (A) The extent of osteoblastogenesis was assessed prior to induction (day 0) and at day 20 post-induction by staining for matrix mineralization with Alizarin red. Whole wells and micrographs of stained cells, representative of three independent experiments, are shown. (B-F) Total RNA was isolated at day 0, 1, 2, 4, 8, 12, 16 and 20 post-induction and the transcript expression of alkaline phosphatase (Alpl), Osteocalcin (Ocn), Wnt6, Wnt10a and Wnt10b was analyzed by qPCR and normalized to 18S rRNA. Transcript expression (B-F) is reported relative to expression at time 0 as mean ± SD of three independent experiments. Statistical significance compared to mRNA expression at day 0 is indicated as follows: * = P < 0.05; ** = P < 0.01; *** = P < 0.001.

03

Supplemental Figure 3 - β-catenin is required for inhibition of adipogenesis by Wnt3a.

3T3-L1 preadipocytes stably expressing shControl or shβ-catenin were induced to differentiate with MDI in the absence or presence of the indicated concentrations of recombinant Wnt3a. At day 8 post-induction, the cells were stained with Oil Red-O to assess lipid accumulation. Plates and micrographs of stained cells, representative of four independent experiments, are shown.

Highlights.

> Overexpression of Wnt6, Wnt10a or Wnt10b blocks adipogenesis and stimulates osteoblastogenesis.

> Stable knockdown of these Wnts enhances adipogenesis and impairs osteoblastogenesis.

> β-catenin is required for effects of these Wnts on fate of mesenchymal precursors.

Abbreviations

MSC

mesenchymal stem cell

PPARγ

peroxisome proliferator-activated receptor-γ

FABP4

fatty acid-binding protein 4

C/EBPα

CCAAT/Enhancer-binding protein-α

WAT

white adipose tissue

BAT

brown adipose tissue

Id2

inhibitor of DNA binding 2

TLE3

transducin-like enhancer of split 3

Fzd

frizzled receptor

LRP

low density lipoprotein receptor-related protein

GSK3

glycogen synthase kinase-3

TCF

T-cell factor