In order to determine the effective concentrations of BIO and LiCl, nuclear translocation of β-catenin and induction of Gli1 gene expression have been analyzed in hMADS2 and hMADS3 cells, two stem cell populations isolated from separate donors. As shown in Fig 1A, stimulation of hMADS3 cells with 0.5 μM BIO led to nuclear accumulation of β-catenin, whereas the inactive BIO molecule (MeBio 12) had no effect. Under our culture conditions, 0.1 μM BIO had no effect on β-catenin stabilization and 5 μM led to cell death (not shown). Treatment of cells with 20 mM LiCl had the same effect than 0.5 μM BIO whereas 10 mM LiCl and 20 mM sodium chloride (NaCl) had no effect (not shown). Nucleofection of hMADS3 cells with a reporter gene under the control of Tcf/Lef response-element showed that the β-catenin pathway was activated in hMADS cells after inhibiting GSK3 (see Additional file 1). Gli1 gene expression was also stimulated in hMADS3 and hMADS2 cells that were maintained in the presence of 0.5 μM BIO or 20 mM LiCl (Fig. 1B).

Then, the functional consequence of GSK3 inhibition on proliferation of hMADS cells has been investigated. Morphology of hMADS cells changed after treatment with 0.5 μM BIO or 20 mM LiCl. From a fibroblastic-like morphology, GSK3 inhibitor-treated cells displayed a cuboidal shape (see Additional file 2). We did not observe any change in morphology or any effect on proliferation when cells were treated with 0.1 μM BIO (not shown). In contrast, at a higher concentration of 0.5 μM BIO, treatment of hMADS2 cells for 5 days led to a 40–50% inhibition of proliferation. Effects of BIO on cell morphology and proliferation were reversible indicating that the GSK3 inhibitor did not affect cell viability (not shown). MeBio-treated cells did not display any significant inhibition of proliferation. The inhibitory effect of BIO was also observed after treatment of mesenchymal stem cells isolated from human bone-marrow (hMSCs) (Fig. 2A). BIO-inhibitory effect was detected as early as 3 days after treatment and was reproducible on hMADS3 cells (Fig. 2B). Inhibition on hMADS3 cell proliferation was also observed to a similar extent after treatment with 20 mM LiCl but not with NaCl (Fig. 2C). As shown in Fig. 3, BIO inhibited the ability of hMADS cells and hMSCs to proliferate at the single cell level, a crucial feature of stem cells. This indicates that BIO cannot be used to expand mesenchymal stem cells ex vivo. This is in contrast with a previous report showing that BIO was able to maintain pluripotent embryonic stem cells by activating Wnt pathway 18 and that inhibition of β-catenin promotes proliferation of uncharacterized stromal cells of adipose tissue 19. These discrepancies could be due to a differential effect of Wnt pathway in embryonic versus human adult mesenchymal stem cells or/and to the fact that, as shown above, BIO activated both Wnt and Hh pathways within the same cell population. The effect of Hh pathway on proliferation on hMADS cells remains to be investigated. In addition, as also shown above, GSK3 is involved in several pathways and we cannot rule out that BIO could have other targets than GSK3 in hMADS cells. Altogether, these data show that BIO inhibited the proliferation of human stem cells, strongly suggesting that GSK3 inhibitors could be potent regulators of the pool of adipocyte precursors in adipose tissue.

To investigate the effect of GSK3 inhibition on adipocyte differentiation, hMADS cells were induced to differentiate in the presence of BIO or MeBIO. Adipocytes and osteoblasts share the same mesenchymal precursor cell. Adipogenesis and osteogenesis are processes that respond to a balance in bone marrow and this balance can be disrupted under pathological conditions such as osteoporosis in which adipocytes develop at the expense of osteoblasts 520. We took advantage of hMADS cells that were previously demonstrated to differentiate in vitro into functional adipocytes and osteoblasts 1617, to investigate the effect of GSK3 inhibition on both lineages.

Adipogenic and osteogenic differentiations, as indicated by Oil Red O and Alizarin staining respectively, were dramatically inhibited in the presence of BIO, whereas MeBIO did not display any significant effect (Fig. 4, left panel). BIO-inhibitory effect was quantified using enzymatic activities such as glycerol-3-dehydrogenase (GPDH), which is expressed in adipocytes but not in osteoblasts, and alkaline phosphatase (ALP) activity, which is expressed in osteoblasts but not in adipocytes. As shown in Fig. 4, GPDH activity was inhibited when hMADS cells were maintained for 10 days in adipogenic medium or in a medium allowing simultaneously adipogenic and osteogenic differentiations supplemented with BIO. ALP activity was also inhibited when cells were maintained in osteogenic or adipogenic/osteogenic media supplemented with the GSK3 inhibitor (Fig. 4, right panel). Then, we investigated the impact of BIO treatment during cell proliferation on subsequent differentiation. Quantification of GPDH enzymatic activity indicated that cells previously treated with BIO during proliferation displayed a lower capacity to undergo adipocyte differentiation than untreated- or MeBio-treated cells (see Additional file 3). This result indicates that GSK3 plays a role in maintaining the differentiation potential of undifferentiated hMADS cells. This is reminiscent of our previous work showing the critical role of FGF2 pathway on self-renewal of hMADS cells 14.

Finally, analysis of adipocyte- and osteoblast-gene expression confirmed that treatment of hMADS cells with of BIO during differentiation led to the inhibition of both differentiation programs. As shown in Fig. 5, all of the genes known to be expressed during adipocyte differentiation that we have analyzed were inhibited in the presence of BIO. Genes expressed during osteogenesis were also inhibited by BIO, although at a lower extent compared to adipogenesis-related genes. Altogether, these results are in agreement with previous reports showing that activation of Wnt or Hh pathways precludes differentiation of murine 3T3-L1 preadipose cells into adipocytes 921. Our data indicate that the GSK3 inhibitor was a potent inhibitor of commitment of human preadipocyte precursors into the adipogenic lineage. The effect of Wnt pathway on osteogenesis remains more controversial. Wnt has been reported to enhance osteogenic differentiation of murine mesenchymal stem cells 22. More recently, Cho et al. have shown that Wnt signalling suppresses osteogenic differentiation of stroma cells isolated from human adipose tissue 19. Therefore, the effect of Wnt signalling on bone formation seems different in human versus mouse and our data with BIO-treated hMADS cells are in agreement with a negative effect of Wnt on osteogenesis of human cells.

It has been observed that patients treated with LiCl for bipolar disorder display weight gain and reduced risk of fractures, suggesting that LiCl promotes both adipogenesis and osteogenesis in humans in vivo 2324. However, a central effect of LiCl and an indirect effect on adipocytes and osteoblasts in these patients cannot be excluded.

Then, we analyzed BIO-inhibition on adipogenic and osteogenic differentiations of hMADS cells more precisely in order to address whether a chronic treatment with the inhibitor was required. For that purpose, we treated hMADS cells during the first 3 days of differentiation and analyzed the impact of this treatment on GPDH and ALP enzymatic activities as well as on the expression of specific markers. In regards to adipogenesis, treatment of hMADS cells for the first 3 days only was sufficient to inhibit GPDH activity measured at day 10 (Fig. 6A). Analysis of adipocyte gene expression confirmed that the inhibitory effect of BIO after 3 days of treatment persisted at day 10 for most of the genes analyzed (Fig. 6B). In contrast, a treatment restricted to the first 3 days had no effect on both osteogenic markers such as ALP enzymatic activity and expression of several genes known to play a role in osteogenesis. Altogether, data indicate that treatment of hMADS cells with BIO during the early step of differentiation only led to inhibition of late steps of adipogenesis, whereas it had no effect on osteogenesis. This strongly suggests that the molecular mechanisms underlying inhibition of adipogenic and osteogenic differentiation by BIO are different.

Identification of mechanisms leading to the differential effects on adipogenesis and osteogenesis of hMADS cells could lead to a preferential use of the GSK3 inhibitors on adipocyte differentiation in vivo. In conclusion, our data strongly suggest that GSK3 is a promising pharmacological target to regulate both the number and differentiation of adipocyte precursors in human adipose tissue. However, we have to keep in mind that potential adverse effects could be observed due to the fact that GSK3 is implicated in numerous signalling pathways. Therefore, it would be important in the future to identify the signalling pathway mediating the adipogenic effect of GSK3.

hMADS cells were obtained from the stroma of human adipose tissue as described previously 15. Adipose tissue was collected, with the informed consent of the parents, as surgical scraps from surgical specimen of various surgeries, as approved by the Centre Hospitalier Universitaire de Nice Review Board. The cell populations that have been studied in this work were isolated from the pubic region fat pad of a 5-year old (hMADS2) and of a 4-month old (hMADS3) male donor. Proliferation medium for routine maintenance of hMADS cells is composed of DMEM (low glucose) containing 10% foetal calf serum, 10 mM HEPES, 100 U/ml penicillin and streptomycin and supplemented with FGF2 as previously reported 14. After reaching 80% confluence, adherent cells were dissociated in 0.25% trypsin EDTA and seeded at 4500 cells/cm². Cells have been maintained in low serum concentration for studying effects of GSK3 inhibitors on proliferation. This medium is composed of 60% DMEM low glucose, 40% MCDB-201, insulin (10 μg/ml), transferrine(5 μg/ml), selenium(50 ng/ml), dexamethasone (10^-9M), ascorbic sodium acid (50 μg/ml), 2.5 ng/ml FGF2 and supplemented with 0.5% FCS. This medium containing a low serum concentration allows the maintenance of stem cell features of hMADS2 and hMADS3 cells (not shown). Human mesenchymal stromal cells isolated from bone marrow, termed hBMSC, were purchased from Cambrex and used as recommended by the manufacturer. Cultures were maintained at 37°C in a humidified gassed incubator, 5% CO₂ in air.

Adipocyte differentiation was performed as described previously 16. Basically, confluent cells were cultured in DMEM/Ham's F12 media supplemented with transferrin (10 μg/ml), insulin (0.86 μM), triiodothyronine (0.2 nM), dexamethasone (1 μM), isobutyl-methylxanthine (100 μM) and rosiglitazone (500 nM). Three days later, the medium was changed (dexamethasone and isobutyl-methylxanthine were omitted). Neutral lipid accumulation was assessed by Oil red O staining 25. For osteoblasts, confluent cells were cultured in α-MEM medium containing 10% FCS, L-ascorbic acid phosphate (50 μg/ml), β-glycerophosphate (10 mM) and 100 nM dexamethasone. Alizarin red staining was performed as previously described 25. For simultaneous adipocyte and osteoblast differentiations, confluent cells were cultured in the differentiation medium consisting in 50% adipogenic and 50% osteogenic media.

Cell homogenates for both Glycerol-3-phosphate dehydrogenase (GPDH) and Alkaline phosphatase (ALP) enzymatic activity measurements were prepared in 20 mM Tris-HCl pH 7.5 buffer containing 1 mM EDTA and 1 mM 2-mercaptoethanol. GPDH activity was measured by the absorbance at 340 nm as described previously 26 and ALP activity was assayed using the p-Nitrophenyl Phosphate Liquid Substrate System (Sigma). Absorbance was measured at 412 nm.

Cells were plated onto 12-well plates (10⁴ cells per well). GSK3 inhibitors were added 24 hours after cell plating in order to avoid a potential effect of the compounds on the efficiency in cell attachment. After the appropriate time, cells were trypsinized as mentioned above and counted with a Coulter counter. For each experiment, three wells per condition were counted.

Cells were plated at a density of 10 cells/cm² in 100-mm² dishes. 15 days after plating cells were fixed with 0.25% glutaraldehyde and stained with 0.1% crystal violet. Colonies containing at least 40 cells were enumerated under a light microscope. Medium was changed 3 times a week.

hMADS cells were rinsed twice with PBS at 4°C, and fixed with 4% paraformaldehyde in PBS for 15 min at room temperature. Free aldehydes were quenched by a 15-min incubation with 1 M glycine in PBS. Fixed cells were permeabilized with 0.5% Triton for 10 min, and unspecific reactions blocked with 5% bovine serum albumin in PBS for 30 min. Cells were then incubated for 1 h with an anti β-catenin (Santa Cruz Biotechnology) antibody diluted in 5% bovine serum albumin in PBS, followed by a 488-Alexa Fluor conjugated secondary antibody. Nuclei were counterstained with DAPI. Images were taken on an LSM510 META confocal microscope (Zeiss).

Total RNA was extracted using TRI-Reagent™ kit (Euromedex, France) according to the manufacturer's instructions and RT-PCR analysis was conducted as described previously 14. All primers sequences, designed using Primer Express software (Applied Biosystems, France), see Additional file 4. For quantitative PCR, final reaction volume was 25 μl, including specific primers (0.4 μM), 5 ng of reverse transcribed RNA and 12.5 μl SYBR green master mix (Applied Biosystems, France). Quantitative PCR conditions were as follows: 2 min at 50°C, 10 min at 95°C, followed by 40 cycles of 15 sec at 95°C, 1 min at 60°C. Real-time PCR assays were run on an ABI Prism 7000 real-time PCR machine (Applied Biosystems, France). Normalization was performed using the geometrical average of the housekeeping genes G6PDH, POLR2A and TBP. Quantification was performed using the comparative-ΔCt method. Level of expression was represented using the genesis software 27. This software generates expression images using a colour code according to the expression intensities.

Statistical significance was checked by using t-tests whenever comparing two conditions or a one-way ANOVA followed by t-tests whenever comparing more than two conditions. Asterisks indicate the significance levels as mentioned in the figure legends.