Glycogen synthase kinase-3α/β inhibition promotes in vivo amplification of endogenous mesenchymal progenitors with osteogenic and adipogenic potential and their differentiation to the osteogenic lineage

Abstract

Small molecules are attractive therapeutics to amplify and direct differentiation of stem cells. They also can be used to understand the regulation of their fate by interfering with specific signaling pathways. Mesenchymal stem cells (MSCs) have the potential to proliferate and differentiate into several cell types, including osteoblasts. Activation of canonical Wnt signaling by inhibition of glycogen synthase kinase 3 (GSK-3) has been shown to enhance bone mass, possibly by involving a number of mechanisms ranging from amplification of the mesenchymal stem cell pool to the commitment and differentiation of osteoblasts. Here we have used a highly specific novel inhibitor of GSK-3, AR28, capable of inducing β-catenin nuclear translocation and enhanced bone mass after 14 days of treatment in BALB/c mice. We have shown a temporally regulated increase in the number of colony-forming units–osteoblast (CFU-O) and –adipocyte (CFU-A) but not colony-forming units–fibroblast (CFU-F) in mice treated for 3 days. However, the number of CFU-O and CFU-A returned to normal levels after 14 days of treatment, and the number of CFU-F was decreased significantly. In contrast, the number of osteoblasts increased significantly only after 14 days of treatment, and this was seen together with a significant decrease in bone marrow adiposity. These data suggest that the increased bone mass is the result of an early temporal wave of amplification of a subpopulation of MSCs with both osteogenic and adipogenic potential, which is driven to osteoblast differentiation at the expense of adipogenesis. © 2011 American Society for Bone and Mineral Research.

Introduction

Mesenchymal stem cells (MSCs) are multipotent adult cells capable of extended in vitro proliferation and differentiation to multiple lineages including osteoblasts.1 In principle, they can be isolated from a variety of tissues, expanded in vitro, gene-modified, and administered to the patient. Alternatively, in vivo stimulation of MSCs is thought to provide clinical means to enhance the endogenous stem cell pool in situations where this is diminished, such as in osteoporosis.2 For this reason, MSCs hold great potential for the correction and regeneration of tissues such as bone and to support hematopoietic stem cells when numbers are suboptimal.3, 4 However, clinical developments have been limited by the lack of knowledge of MSC biology, particularly the signaling pathways regulating their self-renewal and maintenance during tissue regeneration and lack of specific ways to manipulate them. Determining those pathways is critical to designing novel therapeutic strategies.

Canonical Wnt signaling has been shown to be involved in the self-renewal of stem cells such as hematopoietic stem cells (HSCs).5 Activation of Wnt signaling is mediated through the inhibition of glycogen synthase kinase 3 (GSK-3) and prevents the degradation of β-catenin, the main downstream effector of Wnt.6 In vivo administration of GSK-3 inhibitors has been shown to improve the self-renewal ability of HSCs in nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice,5 suggesting that administration of GSK-3 inhibitors could be used to enhance expansion of stem cells with repopulation capacity. The effect of blocking GSK-3 in MSCs is less clear.

Activation of Wnt signaling with Wnt 3a has been shown to preserve MSC proliferation capacity and to suppress their differentiation.7, 8 Moreover, the overexpression of LRP5, a key coreceptor specifically involved in canonical Wnt signaling, was reported to increase proliferation of MSCs.9 Activation of Wnt signaling by lithium chloride, a nonspecific GSK-3 inhibitor, was shown to suppress dexamethasone-induced osteogenesis.10 However, other studies have reported that canonical Wnt signaling inhibits MSC proliferation.11 The use of a Dkk1 blocker induced MSCs to reenter the cell cycle through antagonizing canonical Wnt signaling.12 Moreover, Wnt/β-catenin signaling was reported to shift MSC cell fate toward osteoblastogenesis at the expense of adipogenesis13 and increases bone mass in osteopenic rats.14

There are a number of possible reasons behind the controversial findings. In vitro studies have used MSC-like cell lines13–15 or primary human MSCs isolated and expanded in vitro under different culture conditions,9, 10, 12 mostly in the absence of environmental signals. In vivo studies relied on loss- and gain-of-function approaches,13, 16 leading to the presence or absence of factors well beyond physiologic levels, or relied on the use of nonselective GSK-3 inhibitors such as lithium chloride.10, 16 In this study we hypothesized that activation of canonical Wnt signaling by an inhibitor of GSK-3 promotes in vivo bone formation as a result of expansion of the mesenchymal progenitor cell compartment. To address this hypothesis, we used a novel selective inhibitor of GSK-3α/β, AR28, and studied the effects on proliferation and differentiation of mesenchymal progenitors in vivo.

Material and Methods

Reagents and animals

BALB/c mice were obtained from Charles River (Margate, UK) and used at 5 to 7 weeks of age. All protocols were carried out according to the approved home office license. The GSK-3α/β inhibitor AR28 was provided by AstraZeneca (Sodertalje, Sweden) and dissolved in DMSO for in vitro studies. Bone marrow cells (BMCs) were flushed from the long bones and used for progenitor assays and alkaline phosphatase assay in the presence of AR28, as described below. For in vivo studies, AR28 was formulated in 15% hydroxypropyl-β-cyclodextrin in 0.05 M phosphate buffer (pH 6) and 0.4% glycerol and administered at 30 mg/kg twice a day by subcutaneous injection for 3 and 14 days. BMCs flushed from the femurs were used for progenitor cell assays; tibias were used for micro–computed tomographic (µCT) and histomorphometric analysis. Six and 2 days before termination of the experiment (following 14 days of treatment), mice were injected with calcein (30 mg/kg, i.v.; Sigma-Aldrich, St Louis, MO, USA) to assess bone-formation rates.

β-Catenin stabilization assay and kinase activity

Murine C3H10T1/2 MSCs (LGC Standards, Teddington, UK) were cultured in Eagles' basal medium (BME) plus 10% fetal bovine serum (FBS) and 2 mM GlutaMax (Invitrogen, Paisley, UK). Cells were seeded in BME plus 5% FBS and 2 mM GlutaMax into 96-well plates (5000 cells/well) 18 hours before treatment with AR28 (6 nM to 20 µM). After 24 hours, cells were fixed in 4% paraformaldehyde for 30 minutes at room temperature. Cells were washed three times in PBS before being incubated in Block buffer (1.1% bovine serum albumin in 0.2% Triton-X-PBS) for 1 hour at 4°C. Cells were incubated in 1.25 µg/mL of mouse anti-β-catenin primary antibody (BD Transduction Laboratories, Oxford, UK) in Block buffer at 4°C overnight. C3H10T1/2 cells were washed and incubated in Block buffer for 1 hour at room temperature before incubation with 4 µg/mL of Alexa fluor 647 donkey anti-mouse IgG secondary antibody (Invitrogen) for 2 hours at room temperature. Cells were washed three times with PBS and stored under 100 µL/well of PBS prior to image analysis using an IN Cell Analyzer 3000 (GE Healthcare, Cardiff, UK). Determination of kinase inhibition by AR28 was performed using the KinaseProfiler Service (Millipore, Watford, UK). The IC₅₀ values for the 18 human kinase enzymes were determined from 10-point inhibition curves from assays using recombinant kinases according to Millipore instructions.

Progenitor cell assays

Colony-forming unit–fibroblasts (CFU-F)

BMCs were plated at densities of 5 × 10⁵ and 10⁶ cells/well (6-well plates) in duplicate in murine MSC complete medium (Stem Cell Technologies, Inc., Grenoble, France). The plates were incubated for 14 days at 37°C in 5% CO₂ in air. Colonies were visualized by Giemsa's stain solution (VWR, Leicestershire, UK). Colonies with a minimum of 50 cells were considered as one CFU-F.

Colony-forming unit–osteoblast (CFU-O)

BMCs were plated at a density of 5 × 10⁵ and 10⁶ cells/well in duplicates in Dulbecco's modified Eagle medium (DMEM) plus 10% fetal calf serum (FCS) supplemented with 100 nM dexamethasone (Sigma-Aldrich), 10 mM β-glycerophosphate (Sigma-Aldrich), and 0.05 mM L-ascorbic acid (Sigma-Aldrich). Cells were maintained in culture for 14 days and fed twice a week. Cells were stained for alkaline phosphatase (ALP) activity using the 86 R Alkaline Phosphatase Kit (Sigma-Aldrich). The colonies consisting of at least 20 cells positive for ALP were considered as one CFU-O.

Colony-forming unit–adipocyte (CFU-A)

BMCs were plated at limiting dilutions (serial 1:2 dilution, 8 wells/cell concentration, range 10⁵ to 6.25 × 10³ cells) in DMEM plus 10% FCS (Hyclone; Fisher Scientific, Loughborough, UK) in a 96-well culture plate and fed twice a week for 2 weeks. After 2 weeks, culture were supplemented with rosiglitazone (5 µM) and cultured at 37°C in CO₂ in air for 2 further weeks. To detect lipid vacuoles, cultures were stained for oil red O as described previously.17 A well was considered positive if it contained more than 20 cells with red lipid vacuoles. The number of CFU-A was calculated following the Poisson distribution and using the formula F_o = e^−x, where F_o is the fraction of colony-negative wells, e is a constant whose value is 2.71, and x is the number of colony-forming units per well.18

Hematopoietic progenitor assay

BMCs were plated at a density of 10⁵ cells/mL in triplicate in 1% methylcellulose in Iscove's modified Dulbecco medium (IMDM) supplemented with 15% FBS, 1% bovine serum albumin (BSA), 10 mg/mL of recombinant human (rh) insulin, 200 mg/mL of human transferrin, 10⁻⁴ M 2-mercaptoethanol, 2 mM L-glutamine, 50 ng/mL of recombinant murine (rm) stem cell factor, 10 ng/mL of rm interleukin-3, 10 ng/mL of rh interleukin 6, and 3 units/mL of rh erythropoietin (Methocult; Stem Cell Technologies Inc.) and incubated at 37°C in 5% CO₂ in air. The colony-forming unit–granulocyte-macrophage progenitors (CFU-GM) were scored after 14 days under an inverted light microscope.

Quantification of gene expression during differentiation

BMCs at a density of 5 × 10⁶ cells were cultured in DMEM plus 10% FCS and osteogenic supplements as described earlier for 14 days in the presence or absence of AR28. For detection of ALP, cells were lysed in 500 µL of lysis buffer, and ALP activity was determined according to the manufacturer's instructions (Sigma-Aldrich). Optical density was measured using a Spectramax plate reader (Molecular Devices, Wokingham, UK) set at 405 nm at 5-minute intervals for up to 60 minutes. The measured ALP activity was normalized to the amount of DNA present in the cell lysate, measured by PICO Green Assay Kit according to the manufacturer's instructions (Molecular Probes, Paisley, UK). To quantify osteocalcin (Ocn) gene expression, total RNA was extracted from differentiated cells using RNAqueous 4PCR Kit (Ambion, Warrington, UK). Reverse transcription was carried out after treatment with DNAse using the 1st Strand cDNA Synthesis Kit (GE Healthcare, Amersham, UK) and 2 µg of total RNA. Quantitative real-time PCR was performed using SYBR Green PCR Master Mix (Eurogentec, Romsey, UK), and the following primers were used at 0.1 µM: Ocn, forward: 5'-CCT ACA AAC GCA TCT ACG GTA TCA C-3' and reverse: 5'-AAA GCC GAG CTG CCA GAG T-3'; L32, forward: 5'-GAC AAC AGG GTG CGG AGA AG-3' and reverse: 5'-TGT TGC TCC CAT AAC CGA TGT-3' on an ABI7900HT thermocycler (Applied Biosystems, Warrington, UK). The data were collected and analyzed using SDS 2.0 software (Applied Biosystems). To assess adipogenic differentiation, 5 × 10⁶ BMCs were cultured in DMEM plus 10% FCS for 2 weeks and then induced to differentiate by the addition of rosiglitazone at 5 µM for 2 weeks in the presence or absence of AR28. Expression values of lipoprotein lipase (LPL) and proliferator-activated receptor γ (PPARγ) were detected by quantitative real-time PCR as described earlier using the following primers LPL, forward: 5'-CCA GCT GGG CCT AAC TTT GA-3' and reverse: 5'-AAG ACA TCT ACA AAA TCA GCG TCA TC-3'; and PPARγ, forward: 5'-TGC CAG TTT CGA TCC GTA GAA-3' and reverse: 5'-TCA AGG TTA ATG AAA CCA GGG ATA TT-3'.

Micro–computed tomography (µCT)

For the µCT imaging, excised tibias from the treated mice were scanned using a µCT scanner (Model 1172; Skyscan, Aartselaar, Belgium) at 50 kV and 200 µA with a 0.5-mm aluminium filter using a detection pixel size of 4.3 µm. Images were captured every 0.7 degrees through 180 degrees of the bone. The scanned images were reconstructed using the Skyscan Recon software and analyzed using Skyscan CT analysis software. A standard trabecular bone volume of interest was chosen starting 0.5 mm from the growth plate and included all the trabeculae in 1 mm³ of the bone.

Histologic analysis

Tibiae were fixed in 10% neutral buffered formalin, decalcified in EDTA, and embedded in paraffin, and 3 µm sections were cut using a Leica Microsystems microtome (Leica Microsystems, Milton Keynes, UK). The sections were stained with either hematoxylin and eosin (H&E) or tartrate-resistant acid phosphatase (TRACP) to identify osteoclasts and counterstained with Gill's hematoxylin. The sections were examined by light microscopy (Leica Microsystems). The numbers of osteoblasts and osteoclasts per millimeter were measured on 6.5 mm of the corticoendosteal surfaces starting 0.25 mm from the growth plate using the Osteomeasure analysis software (Osteometrics, Decatur, GA, USA). To determine bone-formation rates, the left tibias were fixed in 70% ethanol and embedded in LRwhite medium-grade resin (Taab Laboratories, Reading, UK), and undecalcified sections were cut onto Superfrost Plus slides (VWR International, Lutterworth, UK). Sections were examined under ultraviolet (UV) illumination using a DMRB microscope (Leica Microsystems). The proportion of corticoendosteal bone surface undergoing mineralization and the separation between the two fluorescent labels were measured using the Osteomeasure system in an area measuring 0.75 mm², 0.25 mm from the growth plate. The mineralized surface (MS/BS) was calculated as [(dLS + 0.5sLS)/BS], where dLS = double-label surface, sLS = single-label surface, and BS = total bone surface. Mineral apposition rate (MAR) was calculated as Ir.L.Th/Ir.L.t, where Ir.L.Th = the separation between two labels and Ir.L.t = the time separating the two labels. Finally, the bone-formation rate was calculated as the product of MAR × MS. Bone marrow adiposity was determined by measuring the area occupied by the adipocytes, recognized as white holes and defined by the light-blue staining of the cell wall starting 2.25 mm from the growth plate (in the diaphysis) and measuring over 0.75 mm².

Statistical analysis

All experiments were analyzed using t tests or one way ANOVA–Dunnett's post test for multiple comparisons. All results are expressed as the mean ± SEM. Significant p values were less than .05 with *p < .05, **p < .01, and ***p < .001.

Results

The GSK-3 inhibitor increased functional β-catenin protein expression in the nucleus of murine mesenchymal C3H10T1/2 cells

In these studies, we used the specific GSK-3α/β inhibitor AR28, which demonstrated from 70- to greater than 6000-fold selectivity over a panel of other kinases and an IC₅₀ of 5 nM (Supplemental Fig. S1A–C). In the absence of a Wnt signal, GSK-3 acts to bind and phosphorylate β-catenin, identifying it for proteasomal degradation. Activation of the Wnt/β-catenin signaling cascade or GSK-3 inhibition leads to hypophosphorylation of β-catenin and its subsequent cytosolic stabilization and translocation to the nucleus, where it binds T-cell factor/lymphoid enhancer–binding factor (TCF/LEF) transcription factors, resulting in differential expression of downstream Wnt target genes.19 The stabilization of β-catenin in the nuclei of murine C3H10T1/2 cells, determined as normalized nuclear/cytoplasmic ratio of β-catenin fluorescence intensity, was increased after 24 hours of treatment with AR28 (6 nM to 20 µM), and this increase in nuclear protein expression was concentration-dependent (Supplemental Fig. S1D, E). These results confirm that AR28 acts to inhibit GSK-3 in murine cells and indicate activation of the canonical Wnt/β-catenin signaling cascade.

The GSK-3 inhibitor AR28 enhances the clonogenic ability of mesenchymal progenitors with osteogenic and adipogenic potential in vitro

To determine whether AR28 promoted an increase in the number of mesenchymal progenitors, murine bone marrow cells were exposed initially to AR28 at 1 µM, 300 nM, and 100 nM, and the number of CFU-F and CFU-O were measured. Signs of toxicity were observed when AR28 was given to the cells every 3 days, with no growth at 1 µM, an average 2% to 4.5% growth at 300 nM, and 76% to 83% growth at 100 nM (n = 4; data not shown). A further reduction in the dose at 50, 10, and 1 nM resulted in a significant increase in the number of CFU-F at 50 nM (Fig. 1A; p = .024, n = 8). This effect was more pronounced when fresh compound was added every 3 days for 2 weeks (Fig. 1B; p = .018, n = 8). Similarly, a significant increase was seen in the number of mesenchymal progenitors with osteogenic potential (CFU-O) and adipogenic potential (CFU-A). However, this was seen at 50 and 10 nM for CFU-O (Fig. 1C; p = .001, n = 7) and at 10 nM for CFU-A (Fig. 1D, p = .009, n = 10). Representative examples of CFU-F, CFU-O, and cultures used to determine CFU-A are shown in Supplemental Fig. S2.

Figure 1. GSK-3α/β inhibitor AR28 enhances the clonogenic ability of mesenchymal progenitors with osteogenic and adipogenic potential. Murine bone marrow cells were cultured in presence of AR28 at 50, 10, and 1 nM or in presence of only dimethylsulfoxide (DMSO) as control. The number of colony-forming units–fibroblasts (CFU-F) was counted in cultures that were exposed to AR28 only once at the start of the culture (A, n = 8) or every 3 days for 2 weeks (B, n = 8). The number of colony-forming units–osteoblasts (CFU-O) was obtained by culturing murine bone marrow cells in the presence of osteogenic differentiation supplements and AR28 at 50, 10, and 1 nM for 2 weeks (C, n = 7). The number of colony-forming units–adipocytes (CFU-A) was enumerated by culturing murine bone marrow cells by limiting dilution for 2 weeks in mesenchymal medium followed by 2 weeks in adipogenic differentiation medium. AR28 was administered at 50, 10, and 1 nM for the entire duration of the cultures. The percentage of wells negative was used to calculate the frequency of CFU-A, as described in “Materials and Methods” (D, n = 10). All results are expressed as percentage of the number of colonies obtained in the presence of DMSO and are represented as mean ± SEM. Data were analyzed by one-way ANOVA matched observations–Dunnet post test for multiple comparisons. *p < .01; **p < .001.

The GSK-3 inhibitor AR28 enhances the differentiation ability of mesenchymal progenitors to the osteogenic but not adipogenic lineage in vitro

To determine whether AR28 had an effect on the differentiation of MSCs to the osteogenic lineage, we cultured bone marrow cells in MSC medium with the addition of osteogenic supplements and evaluated the amount of ALP, an early marker of osteoblast differentiation, and Ocn gene expression after 14 days. A significant increase in the amount of ALP was observed when cells were treated with AR28 at 10 and 50 nM (Fig. 2A; n = 3, p < .001). Similarly, an increase in Ocn expression was observed in all cultures when they were differentiated in the presence of AR28 (Fig. 2B; n = 4, DMSO vs AR28 (50 µM), p < .05). These data suggests that inhibition of GSK-3 enhanced differentiation to the osteogenic lineage.

Figure 2. GSK-3α/β inhibitor AR28 enhances the differentiation ability of mesenchymal progenitors to the osteogenic but not adipogenic lineage. Murine bone marrow cells were cultured in MSC medium in the presence of osteogenic supplements and AR28 at 50 or 10 nM or dimethilsulfoxide (DMSO) alone for 2 weeks, at the end of which cultures were assessed for alkaline phosphatase activity normalized to the amount of DNA present in the cultures (A, n = 3) and osteocalcin gene expression relative to the expression of ribosomal gene L32, used as housekeeping gene (B, n = 4). Murine bone marrow cells were cultured in MSC medium for 2 weeks, followed by 2 weeks in the presence of adipogenic differentiation supplements. AR28 at 50, 10, and 1 nM was added either in the initial 2 weeks of MSC expansion (C, n = 9) or during the incubation with the adipogenic supplements (D, n = 7). In panels C and D, results are expressed as percentage of the number of colonies obtained in the presence of DMSO. PPARγ (E, n = 5) and LPL (F, n = 5) gene expression was assessed on cultures exposed to adipogenic supplements in the presence of AR28 at 50 and 10nM during the differentiation phase. Data were analyzed by one-way ANOVA matched observation–Dunnet post test for multiple comparisons *p < .05; **p < .01; ***p < .001.

The enumeration of CFU-A is based on a limiting dilution assay, and it consists of expansion of the cells for 2 weeks in MSC medium followed by 2 weeks in the presence of adipogenic supplements for differentiation. To determine whether the increase in the number of CFU-A seen when bone marrow cells were exposed to AR28 was due to an increase in proliferation and/or survival of mesenchymal progenitors or to an enhancement of differentiation, AR28 was added either during the expansion phase or during the differentiation phase. A significant increase in the number of CFU-A was seen at 10 nM only when AR28 was added during the expansion phase (Fig. 2C; n = 9, p = .021). No significant difference was seen when AR28 was added during the differentiation phase (Fig. 2D; n = 7, p = .532). A similar trend was seen when markers of adipogenic differentiation LPL and PPARγ were measured following adipogenic differentiation in the presence of AR28 (Fig. 2E, F; n = 5, ns). These data suggest that inhibition of GSK-3 in vitro increases the number of mesenchymal progenitors with osteogenic and adipogenic potential but drives their differentiation only to the osteogenic lineage.

GSK-3 inhibitor AR28 stimulates an increase in an initial wave of mesenchymal progenitors with osteogenic and adipogenic potential and drives their differentiation to the osteogenic lineage in vivo

To determine the effects of activation of canonical Wnt signaling on mesenchymal progenitors with time and in the presence of environmental signals, AR28 was injected subcutaneously at 30 mg/kg twice a day for 3 and 14 days in BALB/c mice at 5 to 7 weeks of age, and the numbers of CFU-F, CFU-O, and CFU-A in the bone marrow were counted. A significant increase in the numbers of CFU-O (Fig. 3A; n = 10, p = .025) and CFU-A (Fig. 3B; n = 10, p = .024) but not CFU-F (Fig. 3C; n = 9, p = .09) was observed following administration of the compound for 3 days. This was not seen after 14 days, when, in contrast, an average 66% decrease in CFU-F was observed (Fig. 3D; n = 7, p = .039) compared with animals injected with vehicle, and no significant difference in the numbers of CFU-O (Fig. 3E; n = 10, p = .967) and CFU-A (Fig. 3F; n = 10, p = .618) was found. These data suggest that inhibition of GSK-3 produces an initial transient increase in committed progenitors with osteogenic and adipogenic potential but with time inhibits the proliferation or even drives to exhaustion mesenchymal progenitors of the CFU-F population.

Figure 3. GSK-3α/β inhibitor AR28 stimulates the amplification of an initial wave of mesenchymal progenitors with osteogenic and adipogenic potential. BALB/c mice were injected subcutaneously twice a day with 30 mg/kg of AR28. Bone marrow was harvested after 3 days (A–C) or 14 days of treatment (D–F), and the number of mesenchymal progenitors was counted. An increase in the number of colony-forming units-osteoblasts (CFU-O) was observed in the bone marrow of mice treated with AR28 for 3 days. (A) The average number of CFU-O/10⁶ cells from 10 mice bone marrows. Similarly, a significant increase in the number of colony-forming units-adipocytes (CFU-A) was observed in the bone marrow of mice treated with AR28 for 3 days (B, n = 10). In contrast, no significant difference in the number of colony-forming units–fibroblasts (CFU-F) was observed after 3 days of treatment (C, n = 9), but this was seen to decrease significantly after 14 days of treatment (D, n = 7). After 14 days of treatment with AR28, no significant difference was observed in number of CFU-O (E, n = 10) and CFU-A (F, n = 10) in the mice. Data are expressed as the number of CFUs/10⁶ bone marrow cells plated. All data were analyzed by unpaired t test. *p < .05.

To determine whether the increase in progenitors with osteogenic and adipogenic potential translated into an increase in mature cell types, the number of osteoblasts was evaluated at the endocortical surface. No significant difference in the number of osteoblasts (Fig. 4A; n = 10, p = .856) was observed after 3 days of treatment. In contrast, a significant increase in the number of osteoblasts was seen by day 14 (Fig. 4C and Supplemental Fig. S3A; n = 10, p = .002). To assess whether the increase in differentiation of mesenchymal progenitors to the osteoblast lineage inhibited differentiation to adipocytes, the percentage of bone marrow area occupied by adipocytes' lipid content was determined. Interestingly, this was significantly reduced already after 3 days of treatment (Fig. 4B; p = .005) and persisted after 14 days (Fig. 4D and Supplemental Fig. S3B; n = 10, p = .006). These data suggest that the initial wave of increased mesenchymal progenitors differentiates to the osteogenic lineage at the expense of adipogenesis.

Figure 4. GSK-3α/β inhibitor AR28 enhances the differentiation ability of mesenchymal progenitors to osteoblasts and inhibits adipogenic differentiation. (A) No difference in the number of osteoblasts over 6.5 mm of tibial corticoendosteal surface was found in mice treated with AR28 (30 mg/kg) twice a day for 3 days compared with mice treated with vehicle. (B) A significant decrease in the bone marrow area occupied by adipocytes was found in mice treated with AR28 for 3 days. (C) A significant increase in osteoblasts number over 6.5 mm of tibial corticoendosteal surface for mice treated with vehicle or AR28 for 14 days. (D) The average percentage of bone marrow area covered by adipocytes in BALB/c mice treated with vehicle and AR28 for 14 days. Data are presented as means ± SEM and analyzed by unpaired t test with **p < .01.

GSK-3 inhibitor AR28 enhances the proliferation of committed hematopoietic progenitors and their differentiation to the osteoclast lineage but does not prevent an overall increase in bone mass

Since osteoblasts were found to increase significantly following administration of AR28 for 14 days compared with animals injected with vehicle, bone volume, trabecular number, and trabecular thickness were measured on days 3 and 14 of administration by µCT to determine the overall effect on bone mass. As expected, quantification of a 1-mm³ region of trabecular bone in the tibia did not reveal any change in bone mass, trabecular number, and trabecular thickness after 3 days of treatment (Fig. 5A–D). However, after 14 days, it revealed over 50% increase in bone volume/tissue volume (Fig. 5E, F; n = 10, p < .001). This increase in bone mass was the result of an increase in trabecular number (Fig. 5G; n = 10, p = .001) and trabecular thickness (Fig. 5H; n = 10, p < .001). Bone-formation and mineral apposition rates were assessed following administration of calcein and also were increased significantly (Supplemental Fig. S4; n = 10). To exclude that the increase in bone mass was the result of a decreased maturation and activation of osteoclasts, as described previously,20 the numbers of CFU-GM, the precursor of osteoclasts, and TRACP⁺ osteoclasts were measured on day 3 and 14. Similar to what was with mesenchymal progenitors and in agreement with the study of Trowbridge and colleagues,5 a significant increase in the number of hematopoietic progenitors was seen after 3 days of administration of AR28 (Fig. 6A; n = 8 vehicle and n = 9 AR28, p = .046), and it was back to normal levels after 14 days (Fig. 6B; n = 10, p = .553). Moreover, the number of osteoclasts at the endocortical surface was found to increase significantly only after 14 days of treatment with AR28 compared with animals treated with vehicle only (Fig. 6C, D; n = 10, p < .03). These data suggest that the induction of bone anabolism is the result of enhanced numbers of osteoblasts, which overrides the increased bone catabolism by osteoclasts and leads to enhanced bone mass.

Figure 5. GSK-3α/β inhibitor AR28 enhances bone formation after 14 days of treatment. (A) A representative example of a horizontal section of tibia after X-ray µCT scanning of a BALB/c mouse treated for 3 days with vehicle or AR28. X-ray µCT scanning of tibia of BALB/c mice treated with AR28 (30 mg/kg) twice daily for 3 days showed no significant difference in bone volume in the cancellous bone area (B), trabecular number (C), or trabecular thickness (D). (E) A representative example of a horizontal section of tibia after X-ray µCT scanning of a BALB/c mouse treated for 14 days with vehicle or AR28. A significant increase in bone volume in the cancellous bone area (F) owing to an increase in both number (G) and thickness of the trabeculae (H) was found. All data are presented as means ± SEM and analyzed by unpaired t test. **p < .01; ***p < .001.

Figure 6. GSK-3α/β inhibitor AR28 enhances the differentiation of myeloid progenitors and osteoclasts. Treatment of BALB/c mice with AR28 at 30 mg/kg twice daily subcutaneously transiently enhances the number of myeloid progenitors with a significant increase in the number of colony-forming units-granulocytes-machrophages (CFU-GM) after 3 days of treatment (A, vehicle n =8 and AR28 n = 9) and return to control levels by 14 days (B, n = 10). Data are expressed as the number of CFU-GM/10⁵ bone marrow cells plated. The number of osteoclast measured over 6.5 mm of corticoendosteal surface showed no significant difference after 3 days of treatment (C) but a significant increase after 14 days of treatment with AR28 in BALB/c mice compared with BALB/c mice treated with vehicle (D). All data are presented as means ± SEM and analyzed by unpaired t test. *p < .05.

Discussion

Stem cells hold considerable promise for the therapy of degenerative disease. However, when thinking about stem cells, most people think about cell-replacement approaches, which are costly and present with considerable challenges in relation to engraftment, tissue integration, and function. Small molecules could be an attractive alternative to amplify the stem cells pool or direct their differentiation by targeting specific signaling pathways. Moreover, they can be used to understand basic mechanisms regulating stem cell self-renewal and differentiation. Elucidating how Wnts act as instructive cues for the recruitment, maintenance, and maturation of MSCs and their differentiated progenies is of primary interest given the potential use of these cells in regenerative medicine. Prior studies have implicated canonical Wnt signaling in the regulation of bone metabolism at several levels, in amplifying the undifferentiated MSC pool,7, 8 commitment of undifferentiated MSCs to the osteoblast lineage,13, 14 and stimulation of their differentiation.14 However, differences in cell preparations, use of models that did not take into account the cellular context, or use of nonphysiologic loss- and gain-of-function approaches are some of the reasons for the emergence of contrasting data. Moreover, the presence of large numbers of Wnts, Wnt receptors, coreceptors, and soluble inhibitors with loosely defined actions has made it difficult to define the function of Wnt signaling in MSCs. Here we have addressed some of these issues. Since the bone marrow is a complex and highly organized structure and represents a difficult organ to reconstruct in vitro, we designed experiments that considered the cellular context of MSCs in their in vivo environment, minimizing the loss of functionality that has been associated with the long-term expansion of MSCs. First, in in vitro experiments where Wnt signaling was activated with the GSK-3α/β inhibitor AR28, we have used bone marrow–derived cells that were cultured directly without further fractionation onto tissue culture plastic. Second, we have studied the effect of AR28 following direct injection in BALB/c mice and counted the number of progenitors and their differentiated mature cells. Moreover, to reduce the complexity of the system, we have used a selective inhibitor of GSK-3, leading to enhanced translocation of β-catenin to the nucleus to activate canonical Wnt signaling. Both in vitro and in vivo studies showed an increase in the number of mesenchymal progenitors with osteogenic and adipogenic potential. Of interest is the difference in response in the number of CFU-F between in vitro and in vivo studies. While an increase in the number of CFU-F was seen in vitro, this was not observed in vivo, where a significant decrease was observed after prolonged exposure to AR28. Even more interesting is the temporal regulation of Wnt signaling on MSCs in vivo. An increase in the number of mesenchymal progenitors is seen only at the earlier time point of 3 days. This effect seems to be lost after 14 days of treatment, when the number of osteoblasts is increased significantly. Our in vitro data are in agreement with what has been shown by Baksh and colleagues21 following addition of Wnt3a, a factor that also activates canonical Wnt signaling, to suspension-grown cells. This leads to an increase in the number of CFU-F, CFU-O, and CFU-A in vitro. There are a number of reasons for the lack of increase in the number of CFU-F in vivo seen at early time points followed by a significant decrease with time. One possibility is that mesenchymal progenitors contained in the more primitive CFU-F population have been recruited to commit to progenitors with osteogenic and/or adipogenic potential earlier than 3 days and are driven to exhaustion as a result of the continuous commitment and differentiation process. An alternative or additional explanation holds that activation of canonical Wnt signaling amplifies a more committed subpopulation of mesenchymal progenitors with osteogenic and/or adipogenic potential that are driven to differentiate to the osteogenic lineage at the expense of adipogenic differentiation. The increase in osteoblast numbers and in bone mass would provide a negative-feedback signal to the progenitors, including the CFU-F, causing their inhibition to proliferate. In support of this, differentiated osteoblasts or osteocytes have been found to produce Wnt inhibitors such as DKK1, which is strongly upregulated during the late phases of osteoblast differentiation22 and has been shown to inhibit activation of canonical Wnt signaling and osteoblast differentiation.23 The presence of a feedback mechanism in vivo would explain the discrepancy between our in vivo and in vitro results. In vitro during the CFU-F assay, cells are seeded at low density, and only those with high proliferative capacity are selected for survival. Signals required to provide a negative feedback would not be present owing to the lack of differentiating osteoblasts and to the clonogenic nature of the assay. In vivo, the proliferative effect on the CFU-F compartment is possibly short-lived and masked by the process of commitment, differentiation, and negative feedback. The presence of a negative-feedback mechanism would reconcile some of the contrasting findings on whether activation of canonical Wnt signal had a positive or negative effect on proliferation and differentiation to the osteogenic lineage because this would depend on the subpopulation of mesenchymal progenitors considered, the length of time the cells were exposed to the signal, and the cellular context the target cells were in at the time point analyzed.

While there is agreement that inhibition of GSK-3 leads to enhanced bone mass,14, 16 the mechanism by which this occurs is controversial—whether this is due to enhanced drive to differentiation to osteoblasts and/or inhibition of osteoclast differentiation. We have shown that inhibition of GSK-3 blocks preadipocyte differentiation and enhances osteoblast differentiation in vivo despite progenitors with both potentials being amplified. This is in agreement with in vitro studies on the preadipocyte cell line ST215 and is in line with studies on sustained expression of β-catenin by expression of Wnt10b in preadipocytes or following stimulation by mechanical loading, where adipogenesis is blocked in favor of osteogenesis.13, 24, 25 Of interest is our data on osteoclast maturation and activation. In contrast to what was reported by Glass and colleagues,20 where it was shown that Wnt signaling promoted the ability of differentiated osteoblasts to inhibit osteoclast differentiation, we have shown a transient increase in the number of TRACP⁺ osteoclasts, most likely owing to an increase in the number of hematopoietic progenitors, from which osteoclasts are derived. Our findings are consistent with hematopoietic stem and progenitor cells described previously to increase in the presence of BIO, an inhibitor of GSK-3.5

Of interest is the multilayer of osteoblasts observed in some areas at the endocortical surface in mice treated with AR28. Although it is difficult to completely exclude endosteal fibrosis, similar to that seen in humans following stimulation with parathyroid hormone,26 this is more likely to reflect a feature of osteoblast formation in young mice. It is not unusual to see areas, at the endocortical surface, with multilayer of osteoblasts in untreated mice at a young age, although this is usually limited to two or three layers. Moreover, a parallel study with AR28 in a model of myeloma bone disease carried out in mice of older age showed an increase in osteoblast numbers, but they were not disposed in multilayers.27

In conclusion, our data are compatible with inhibition of GSK-3 acting on proliferation and commitment of MSCs with osteo- and adipogenic potential, which is driven to osteogenic differentiation at the expense of adipogenic differentiation. It also highlights the complex network of responses taking place with time and occurring in different cellular types at specific stages of commitment. It draws attention to the powerful effect that amplifying stem and progenitor cells in vivo may have on bone mass. Future work is required to exclude the possibility of a reduction in stem cell numbers following prolonged treatment. This may require a more detailed study on the type of stem/progenitors affected and whether those are the more primitive one defined by their ability to engraft following transplantation assay. Moreover, it will be important to determine the best dose and schedule of administration and to dissect the molecular events mediating the feedback mechanism to guarantee a prolonged and sustained effect of this GSK-3 inhibitor as a bone anabolic treatment.

Disclosures

IB and PIC received funding from AstraZeneca, AG was employed on research funds granted by AstraZeneca, and PJOS, JP, MS, MD, WK are employees of AstraZeneca. AR28 is manufactured by AstraZeneca.

Acknowledgements

We are very grateful to Orla Gallagher, Darren Loth, Julia Hough, Les Coulton, and Holly Evans for expert technical assistance. This work was supported by funding from AstraZeneca to IB and PIC.