TGFβ1-Induced Differentiation of Human Bone Marrow-Derived MSCs Is Mediated by Changes to the Actin Cytoskeleton

TGFβ is a potent regulator of several biological functions in many cell types, but its role in the differentiation of human bone marrow-derived skeletal stem cells (hMSCs) is currently poorly understood. In the present study, we demonstrate that a single dose of TGFβ1 prior to induction of osteogenic or adipogenic differentiation results in increased mineralized matrix or increased numbers of lipid-filled mature adipocytes, respectively. To identify the mechanisms underlying this TGFβ-mediated enhancement of lineage commitment, we compared the gene expression profiles of TGFβ1-treated hMSC cultures using DNA microarrays. In total, 1932 genes were upregulated, and 1298 genes were downregulated. Bioinformatics analysis revealed that TGFβl treatment was associated with an enrichment of genes in the skeletal and extracellular matrix categories and the regulation of the actin cytoskeleton. To investigate further, we examined the actin cytoskeleton following treatment with TGFβ1 and/or cytochalasin D. Interestingly, cytochalasin D treatment of hMSCs enhanced adipogenic differentiation but inhibited osteogenic differentiation. Global gene expression profiling revealed a significant enrichment of pathways related to osteogenesis and adipogenesis and of genes regulated by both TGFβ1 and cytochalasin D. Our study demonstrates that TGFβ1 enhances hMSC commitment to either the osteogenic or adipogenic lineages by reorganizing the actin cytoskeleton.


Introduction
Fat and bone tissues both originate from bone marrow progenitor cells called skeletal stem cells, also known as bone marrow-derived multipotent stromal cells or mesenchymal stem cells (MSCs). The formation of these tissues is regulated throughout an organism's lifetime by homeostatic mechanisms within the marrow cavity. It has been suggested that an imbalance between osteogenic and adipogenic lineage commitment and differentiation is responsible for age-related impairment of bone formation, and a number of therapeutic interventions targeting and activating MSCs, thus enhancing bone mass, have been proposed. Indeed, the identification of novel strategies to steer human skeletal (mesenchymal) stem cell differentiation towards the production of osteoblastic cells, thus increasing bone formation, is very topical in the bone biology field.
The transforming growth factor (TGF) superfamily consists of over 40 members, including activins, inhibins, bone morphogenetic proteins (BMPs), growth and differentiation factors (GDFs), and TGFβs [1]. TGF family members are multifunctional regulators of cell growth and differentiation, playing pivotal roles during embryonic development, organogenesis, and tissue homeostasis [2]. The cytokine TGFβ1 is among the most abundant in bone matrix [3] and is secreted by endothelial cells, epithelial cells, fibroblasts, smooth muscle cells, and most immune cells [4]. TGFβ1 is deposited in bone matrix as an inactive, latent complex with latencyassociated protein (LAP), the binding of which masks the receptor domains of active TGFβ1. During bone formation, osteoclast-mediated bone resorption activates TGFβ1 by cleaving LAP and releasing it from bone matrix, thus creating a transient gradient of active TGFβ1 that attracts MSCs to bone remodeling sites, where they undergo osteoblastic differentiation [5]. Furthermore, TGFβ1 is known to regulate the proliferation and differentiation of osteoprogenitor cells [6].
Actin microfilaments are composed of polymers of actin, the most abundant cellular protein which also forms the thinnest part of the cytoskeleton, and are primarily responsible for skeletal structure [7]. Cellular actin exists in two forms, filamentous polymerized actin (F-actin) and globular/monomer depolymerized actin (G-actin), and transitions between these forms during highly dynamic intracellular polymerization and depolymerization processes [8]. In mammals, actin polymerization factors regulate actin polymerization and depolymerization [9]. While the stiffness of actin is lower than that of microtubules, actin molecules form a highly organized structural network, supported by a large number of interacting cross-linking proteins, which together confer a substantial amount of mechanical strength [10]. The cytoskeleton is known to be important for determining cell morphology and for mediating changes in adhesion and differentiation [11]. Indeed, during human MSC (hMSC) lineage commitment, cells undergo significant morphological changes and actin cytoskeletal reorganization which contribute to the determination of cellular fate [7,12].
In this study, we investigated the effect of TGFβ-induced actin cytoskeleton modifications on the potential of hMSCs to differentiate into osteogenic and adipogenic lineages, as well as the effect of the actin polymerization inhibitor cytochalasin D (CYD). Our data suggest that TGFβ-induced actin cytoskeleton reorganization is a prerequisite for hMSC differentiation into osteocytic or adipocytic lineages.

TGFβ1
Treatment Enhanced the Osteogenic Differentiation of hMSCs. A single treatment with TGFβ1 (10ng/ml, for 2 days) enhanced hMSC osteogenic differentiation, as shown by the increased mineralized matrix formation made evident by alizarin red S staining (Figures 1(a) and 1(b)). Conversely, when TGFβ1 signaling was blocked with the inhibitor SB-431542 (10 μM), significantly lower mineralized matrix formation was observed (Figures 1(a) and 1(b)). Consistent with this, higher expression of the osteoblastic genes alkaline phosphatase (ALP), runt-related transcription factor 2 (RUNX2), and osteocalcin (OCN) was observed in hMSCs undergoing osteogenic differentiation in the presence of TGFβ1, while treatment with the TGFβ1 inhibitor SB-431542 severely inhibited this expression (Figure 1(c)).

TGFβ1
Treatment Enhanced the Adipogenic Differentiation of hMSCs. Next, we examined the effect of treating hMSCs with a single dose of TGFβ1 (10 ng/ml, for 2 days) on adipogenic differentiation. We found that adipogenic differentiation was enhanced following TGFβ1 treatment, as shown by an increase in the number of lipid-filled adipocytes (Figures 1(d) and 1(e)). Similarly, the expression of several adipogenic gene markers, including lipoprotein lipase (LPL), peroxisome proliferator-activated receptor gamma 2 (PPARG-2), adipocyte protein 2 (aP2), and ADIPOQ, was upregulated following TGFβ1 treatment, while treatment with SB-431542 reversed these effects (Figure 1(f)).

TGFβ1 Stimulation Has No Effect on hMSC Viability or
Proliferation. The effect of TGFβ1 on hMSC cell viability was assessed using alamarBlue assay reagent. No significant effect on viability was observed after 4 days of treatment ( Figure 2(a)). To investigate the effect of TGFβ1 on cellular proliferation, we used the xCELLigence RTCA DP® cell proliferation assay system, which allows the continuous monitoring of cell numbers over time. As shown in Figure 2(b), there was no measurable difference in hMSC proliferation in the presence or absence of TGFβ1.

Molecular Phenotype of TGFβ1-Treated hMSCs.
To understand the molecular mechanisms underlying the TGFβl-mediated regulation of hMSC differentiation, we compared global gene expression in TGFβl-treated hMSCs and vehicle-treated control cells using microarray analysis. In total, 1932 gene transcripts were significantly upregulated, and 1298 were significantly downregulated following TGFβl treatment. Significant changes were defined as a fold change ≥ 2, p < 0 05 and are listed in Supplementary Tables S1 and S2. Hierarchical clustering of differentially expressed genes revealed a clear distinction between TGFβl-treated and control samples (Figure 3(a)). Next, we used performed gene ontology analysis to identify the biological processes that were favored following TGFβl treatment. We found that the genes that were significantly altered in TGFβl-treated MSCs were enriched within several skeletal and extracellular matrix categories, including extracellular matrix (53 genes), extracellular matrix organization (51 genes), and proteinaceous extracellular matrix (Supplementary Table S3). Furthermore, pathway analysis of significantly changed genes revealed the significant enrichment of several signaling pathways in TGFβl-treated hMSCs. Among these, the most  enriched pathways were "regulation of actin cytoskeleton," "MAPK signaling," "focal adhesion," "TGFβ1 signaling," "adipogenesis," "endochondral ossification," and "osteoblast signaling" (Figure 3(b)). Table 1 lists the genes within osteogenesis-and adipogenesis-related signaling pathways that were upregulated in TGFβ1-treated cells. A selected panel of genes known to be involved in cell differentiation and TGFβ signaling that were significantly changed in the microarray data were examined by qRT-PCR. In general, a good degree of concordance was observed between the microarray and qRT-PCR data (Figure 3(c)).

Actin Microfilaments in MSCs
Are Altered following Treatment with TGFβ1 or the Actin Polymerization Inhibitor CYD. Our molecular phenotyping analysis of TGFβltreated hMSCs revealed a significant enrichment of genes associated with cytoskeletal changes. Based on this, and on our previous observations that TGFβl treatment triggers significant morphological changes in hMSCs, we examined the effect of TGFβl on the cytoskeleton using transmission electron microscopy (TEM), which has the power to reveal structural changes in actin microfilaments. Actin microfilament polymerization was found to be inhibited in cells treated with either the potent actin polymerization inhibitor CYD or the TGFβ inhibitor SB-431542. In contrast, TGFβ1 treatment was associated with a prominent distribution of actin filaments, organized as bundles/aggregates, in the perinuclear area and at one cell pole ( Figure 4). The ultrastructural characteristics of the cells under the various treatment conditions are summarized in Supplementary Table S4. 2.6. CYD Regulates Osteogenic and Adipogenic Differentiation in the Presence of TGFβ1. To confirm that TGFβ1 regulates actin cytoskeletal dynamics, hMSCs undergoing either osteogenic or adipogenic differentiation were treated with TGFβ1 in the absence or presence of the actin polymerization inhibitor CYD. CYD treatment significantly inhibited hMSC osteogenic differentiation in both the presence and absence of TGFβl, as shown by reduced mineralization (Figure 5(a)). Similarly, expression of the osteogenic marker genes ALPL, RUNX2, and OCN was inhibited by CYD treatment, with and without TGFβl ( Figure 5(b)). Conversely, CYD treatment enhanced hMSC adipogenic differentiation, as shown by a greatly increased number of lipid-filled mature adipocytes and the increased expression of the adipogenic marker genes LPL and PPARG-2. These effects were maintained when cells were treated concomitantly with TGFβl (Figures 5(c) and 5(d)).

Molecular
Phenotype of CYD-Treated Cells. The data presented above suggest that CYD and TGFβ1 target similar molecular pathways during hMSC osteogenic and adipogenic differentiation. In order to investigate this further and to elucidate the molecular mechanisms underlying the CYD- mediated effects on hMSC differentiation, microarray analysis was performed to establish global gene expression profiles for CYD-treated and controls cells. In total, 10,855 genes were significantly upregulated, and 2523 genes were significantly downregulated following CYD treatment. Genes were defined as significantly changed if they had a fold change ≥ 2 and p < 0 05 and are listed in Supplementary Tables S5 and  S6. As was seen with TGFβ1 treatment, hierarchical clustering of the differentially expressed genes revealed a clear distinction between untreated and CYD-treated hMSCs ( Figure 6(a)). Pathway analysis of these genes revealed several molecular pathways that were enriched upon CYD treatment ( Figure 6(b)). Among the most significant were pathways involved in the regulation of the actin cytoskeleton, focal adhesion signaling, endochondral ossification, TGFβ1 signaling, regulation of the microtubule cytoskeleton, and MAPK signaling (Figure 6(b)). The genes that are associated with these pathways that were upregulated in CYD-treated cells are listed in Table 2. Forty-two genes that are involved in adipogenesis-related pathways were significantly enriched in CYD-treated cells (Table 3). Interestingly, 218 genes were both upregulated in TGFβ1-treated hMSCs and downregulated in CYD-treated hMSCs ( Figure 6(c)), showing that the molecular signature on CYD treatment is the inverse of that seen with TGFβ1 treatment and suggesting that these genes may be involved in TGFβ-mediated cytoskeletal reorganization (Table 4).

Discussion
TGFβ is a potent regulator of various biological functions in many cell types, but its effects on hMSC differentiation are, to   (c) qRT-PCR validation of selected genes that were upregulated in the microarray data (n = 3, * p < 005; * * * p < 0001). Cells treated with vehicle (DMSO) were used as controls.
date, poorly understood. In the present study, we contribute to this understanding and demonstrate that TGFβ can enhance both osteoblastic and adipocytic lineage commitment by modulating changes to the actin cytoskeleton. TGFβ1 is known to regulate the proliferation and differentiation of osteoprogenitor cells [6,[13][14][15], and it reportedly stimulates bone matrix apposition and bone cell replication [16]. Several studies have demonstrated that TGFβ1 promotes bone formation in vitro by recruiting osteoblast progenitors and inducing bone matrix formation at early stages of differentiation. In addition to this direct regulation of bone formation, TGFβ1, along with BMPs, enhances RUNX2 expression at early differentiation stages [17]. This is consistent with our finding that TGFβ1 promoted osteogenesis and was associated with the upregulation of the osteogenic genes ALPL, RUNX2, and OCN.
Furthermore, we showed that TGF-β1 treatment enhanced the in vitro adipocytic differentiation of hMSCs. This is consistent with several previously reported studies which demonstrate that TGFβ1 has a positive effect on adipogenic differentiation under specific culture conditions [18,19]; an early study considering rat brown adipocytes showed an upregulation of lipogenic enzymes following TGFβ1 treatment [19].
Our results showed that TGFβ1 treatment did not affect MSC cell growth in vitro. Previously, conflicting results have been published; some studies reported that TGFβ1 regulated osteoprogenitor proliferation in vitro [13,20], whereas Yu et al. reported that TGFβ1 treatment strongly inhibited the proliferation of human lung epithelial cells [21]. The mitogenic effects of TGFβ on cells are reportedly variable; while progressive mitogenesis was stimulated in confluent cells following treatment with 0.15-15 ng/ml TGFβ, in sparse cultures 0.15 ng/ml TGFβ exhibited inhibitory effects. However, at all cell densities, 15 ng/ml TGFβ stimulated collagen synthesis, with this effect being most pronounced when DNA synthesis was declining [22]. Most of the published data on TGFβ has shown a mitogenic effect on osteoprogenitors [16,[23][24][25][26], but relatively few studies have examined the growth inhibitory effect of this cytokine on osteoblast-like cells [27,28]. It is likely that these contradictory observations reflect the fact that the effect TGFβ has on cellular proliferation is dependent upon TGFβ concentration, culture conditions including cell density, the cell model system (tumorigenic versus nontumorigenic), the differentiation stage of the target cell population, and/or the presence of other growth factors.

Endochondral ossification
On the other hand, adipocytic differentiation is associated with the morphological change from fibroblast-like cells to spherical cells filled with fat droplets [33]. These morphological alterations are also associated with cytoskeletal changes and   actin reorganization, which takes place in the early lineage commitment stage, prior to the upregulation of many adipocytic-specific gene markers [34]. The differentiation of hMSCs into the adipocytic lineage in vitro is known to be influenced by the cytoskeletal tension that results following actin reorganization [32]. Furthermore, TGFβ1 Ca 2+ signaling is known to regulate osteoblast adhesion through enhanced α5 integrin expression, the formation of focal contacts, and the mediation of cytoskeleton reorganization [35,36]. Additionally, the TGFβ1-mediated stimulation of DNA synthesis in mouse osteoblastic cells is reportedly associated with morphological changes and is accompanied by the enhanced synthesis and polymerization of cytoskeletal proteins [37]. Consistent with this, our data suggests that TGFβ1 enhances hMSC lineage commitment by regulating the morphology of the actin cytoskeleton, focal adhesion, and endochondral ossification, via the TGFβ1 and MAPK signaling pathways.
Also consistent with our results are reports that CYDmediated reductions in actin polymerization stimulate adipogenesis, but inhibit osteogenesis [30], suggesting that cytoskeletal modification is a prerequisite for cell fate determination. Our gene expression profiling revealed that the genes FGF1, FGF2, and KRAS, which commonly regulate actin cytoskeleton reorganization, were upregulated and downregulated in TGFβ1-and CYD-treated cells, respectively, suggesting that they are involved in the actin polymerization-mediated differentiation of MSCs.
We showed that during osteogenesis, TGFβ1 treatment reorganized the cytoskeleton, but this reorganization, and thus osteogenesis, could be disturbed by CYD treatment. Conversely, treatment with either TGFβ1 or CYD promoted adipogenesis. This observation can potentially be explained by considering that TGFβ1 and CYD promote the formation of different cytoskeleton patterns, both of which support adipogenesis. Alternatively, it is possible that cytoskeletal reorganization leading to adipogenesis can be promoted by both TGFβ1-dependent and -independent mechanisms, and that CYD-mediated cytoskeletal reorganization cannot override the TGFβ1-independent mechanism. We propose a model wherein TGFβ1 regulates cytoskeletal organization by modulating actin cytoskeleton-related genes, leading to enhanced hMSC differentiation into both osteoblasts and adipocytes (Figure 7). We propose that CYD enhances adipogenesis and inhibits osteogenesis by regulating the expression of a number of key candidate genes, including FGF2, TGFβ2, Plat, EGR2, MEF2D, and IRS1. These genes were modulated by both TGFβ1 and CYD and  are thus heavily implicated in the determination of hMSC fate. In summary, our study provides novel molecular insights into the role of the intracellular TGFβ signaling pathway in bone and bone marrow adipose tissue formation. This signaling involves the reorganization of the actin cytoskeleton in order to control the lineage-specific differentiation of hMSCs.

Cell
Culture. An hMSC-TERT cell line was created previously to serve as a model of human primary MSCs by overexpressing human telomerase reverse transcriptase (hTERT) in normal human bone marrow MSCs [38]. This cell line has been extensively characterized and exhibits a similar cellular and molecular phenotype to primary MSCs [39]. For the current experiments, we used a previously characterized subline derived from hMSC-TERT cells, termed hMSC-TERT-CL1 [40]. For ease, this cell line is referred to as "hMSC" for the remainder of the manuscript. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 4500 mg/l D-glucose, 4 mM L-glutamine, 110 mg/l sodium pyruvate, 10% fetal bovine serum (FBS), 1× penicillin/streptomycin (pen/strep), and nonessential amino acids. All reagents were purchased from Gibco, USA.

OsteoImage Mineralization
Assay. The formation of mineralized matrix in vitro was quantified using an Osteo-Image mineralization assay kit according to manufacturer's instructions (Cat. number PA-1503; Lonza, USA). Briefly, culture medium was removed, and cells were washed once with PBS and then fixed with 70% cold ethanol for 20 minutes. Next, diluted staining reagent was added at a level recommended by the manufacturer, and plates were incubated in the dark for 30 minutes at room temperature. The cells were then washed, and staining was quantified using a fluorescence plate reader (Molecular Devices Co., Sunnyvale, CA, USA) with excitation and emission wavelengths of 492 and 520 nm, respectively.   Briefly, cells were cultured in 96-well plates in 100 μl of the appropriate medium before 10 μl alamarBlue substrate was added at the indicated time points. Plates were then incubated in the dark at 37°C for 1 hour. AlamarBlue fluorescence was then detected using a Synergy II microplate reader (Bio-Tek Inc.) with excitation and emission wavelengths of 530 nm and 590 nm, respectively.

RTCA Cell Proliferation
Assay. An xCELLigence RTCA (real-time cell analysis) DP system (ACEA Biosciences Inc., San Diego, CA) was used to measure the rate of cellular proliferation according to manufacturer's protocol. Briefly, 100 μl DMEM supplemented with 10% FBS was loaded onto each well of an E-plate 16 chamber slide, which was then placed inside the humidified incubator of the RTCA DP analyzer for 1 hour at 37°C to allow the membrane surface and medium to equilibrate. After 1 hour, background measurements were performed. Next, 5000 cells/100 μl DMEM + 10% FBS were added per well, and measurements were recorded at 15-minute intervals for various total durations, depending on the experimental setup.

Transmission Electron Microscopy (TEM).
For TEM, cells were trypsinized, washed with PBS, pelleted, and then fixed in 2.5% glutaraldehyde (Cat. number 16500; Electron Microscopy Sciences) in 0.1 M phosphate buffer (pH 7.2) at 4°C for 4 hours. Next, the cells were washed in 0.1 M phosphate buffer (pH 7.2) 3 times for 30 minutes each and then treated with 1% osmium tetroxide (OsO 4 ) in 0.1 M phosphate buffer (pH 7.2) for 2 hours. Cells were then dehydrated in increasing concentrations of ethanol (10%, 30%, 50%, 70%, 90%, and 100%) for 15 minutes each, before being resuspended in acetone and incubated for 15 minutes. The resulting cell suspension was then aliquoted into BEEM® embedding capsules and infiltrated firstly with a 2 : 1 acetone : resin mixture for 1 hour and secondly with a 1 : 2 acetone : resin mixture for 1 hour. Following infiltration, the BEEM capsules were centrifuged at 2500 rpm for 5 minutes and embedded in pure resin for 2 hours. The resin was then polymerized by baking in an oven at 70°C for 12 hours. Semithin sections (0.5 μm thickness) were prepared and stained with 1% toluidine blue. Ultrathin sections (70 nm thickness) were prepared and mounted on copper grids and then stained firstly with uranyl acetate (saturated ethanol solution) for 30 minutes, rinsed with double distilled water and then stained with Reynold's lead citrate solution for 5 minutes before a final rinse with distilled water. The contrasted ultrathin sections were examined and photographed under a JEOL 1010 transmission electron microscope (JEOL, Tokyo, Japan).

Statistical Analysis.
All results are presented as the mean ± SD of at least 3 independent experiments. Differences between groups were assessed using Student's t-test, and p values < 0.05 were considered statistically significant.

Conflicts of Interest
The authors declare that they have no conflicts of interest.