Enhanced differentiation of human embryonic stem cells to mesenchymal progenitors by inhibition of TGF-β/activin/nodal signaling using SB-431542

Abstract

Directing differentiation of human embryonic stem cells (hESCs) into specific cell types using an easy and reproducible protocol is a prerequisite for the clinical use of hESCs in regenerative-medicine procedures. Here, we report a protocol for directing the differentiation of hESCs into mesenchymal progenitor cells. We demonstrate that inhibition of transforming growth factor β (TGF-β)/activin/nodal signaling during embryoid body (EB) formation using SB-431542 (SB) in serum-free medium markedly upregulated paraxial mesodermal markers (TBX6, TBX5) and several myogenic developmental markers, including early myogenic transcriptional factors (Myf5, Pax7), as well as myocyte-committed markers [NCAM, CD34, desmin, MHC (fast), α-smooth muscle actin, Nkx2.5, cTNT]. Continuous inhibition of TGF-β signaling in EB outgrowth cultures (SB-OG) enriched for myocyte progenitor cells; markers were PAX7⁺ (25%), MYOD1⁺ (52%), and NCAM⁺ (CD56) (73%). DNA microarray analysis revealed differential upregulation of 117 genes (>2-fold compared with control cells) annotated to myogenic development and function. Moreover, these cells showed the ability to contract (80% of the population) and formed myofibers when implanted intramuscularly in vivo. Interestingly, SB-OG cells cultured in 10% fetal bovine serum (FBS) developed into a homogeneous population of mesenchymal progenitors that expressed CD markers characteristic of mesenchymal stem cells (MSCs): CD44⁺ (100%), CD73⁺ (98%), CD146⁺ (96%), and CD166⁺ (88%) with the ability to differentiate into osteoblasts, adipocytes, and chondrocytes in vitro and in vivo. Furthermore, microarray analysis of these cells revealed downregulation of genes related to myogenesis: MYH3 (−167.9-fold), ACTA1 (−161-fold), MYBPH (−139-fold), ACTC (−100.3-fold), MYH8 (−45.5-fold), and MYOT (−41.8-fold) and marked upregulation of genes related to mesoderm-derived cell lineages. In conclusion, our data provides a simple and versatile protocol for directing the differentiation of hESCs into a myogenic lineage and then further into mesenchymal progenitors by blocking the TGF-β signaling pathway. © 2010 American Society for Bone and Mineral Research

Introduction

It is envisioned that stem cells, after differentiation to a specific cell lineage, can be employed to treat a large number of musculoskeletal diseases, including bone and cartilage diseases (eg, localized bone and cartilage defects, nonhealing fractures, and systemic age-related degenerative conditions such as osteoporosis and osteoarthritis)1 and diseases in skeletal muscle and myocardium (eg, muscle dystrophies, myocardial injury following ischemia, and cardiomyopathies).2 Several adult (tissue-specific) stem cells appear to be suitable for clinical applications. For example, bone marrow–derived3, 4 or adipose tissue–derived5, 6 stromal (mesenchymal) stem cells (MSCs) or muscle-derived satellite cells7, 8 are multipotent stem cells that can differentiate, under proper ex vivo culture conditions, into osteoblasts, adipocytes, chondrocytes, or myocytes.7, 9 In addition, preliminary proof-of-concept studies suggest functional improvement of the affected organ after cell transplantation.10, 11 However, a number of factors limit the clinical use of adult stem cells in therapy (eg, the inability to obtain sufficient numbers of cells for therapy owing to inadequate numbers of cells obtained from tissue biopsies, poorly defined stem cell phenotype, and impaired growth and function owing to replicative cellular senescence during ex vivo expansion).12

Human embryonic stem cells (hESCs) represent a good alternative to adult stem cells as a source of an unlimited supply of well-defined pluripotent stem cells that can generate clinical-grade, transplantable, lineage-specific cells.13 Recent publications have demonstrated the development of clinically suitable ex vivo differentiation protocols for obtaining insulin-producing β cells14, 15 and neuronal cells16, 17 from hESCs. Several protocols have been reported in which hESCs were directed to differentiate into musculoskeletal tissue: osteoblastic cells,18–22 chondroyctic cells,23 cardiomyocytes,24–28 and skeletal myoblasts.29 However, most of these protocols relied on using coculture with differentiated cells or complex culture conditions to induce differentiation. In addition, enrichment for a specific cell type appears to be limited, and thus the resulting cell population is comprised of cells at various stages of lineage maturation and commitment.

Members of the transforming growth factor β (TGF-β) superfamily are pleiotropic cytokines that are involved in inducing cell differentiation.30 Activin signaling is known to induce meso/endoderm and to further enhance definitive endoderm in monolayer culture of hESCs.31 Moreover, a recent study by Kitamura and colleagues32 showed that inhibition of TGF-β/activin/nodal signaling by a small molecule inhibitor SB-431542 enhanced expression of mesodermal lineage marker genes and promoted cardiomyogenesis in mouse embryoid body outgrowth cultures.32 In the same study, the authors demonstrated that the TGF-β superfamily ligands, such as activin A/B, nodal, and TGF-β1, activated TGF-β signaling and inhibited cardiomyogensis, suggesting that TGF-β signaling is important for skeletal cell lineage specification of ex vivo expanded ESCs. Similarly, we have reported recently that activation of TGF-β signaling by activin B induced endodermal differentiation of ex vivo cultured hESCs and promoted pancreatic and liver lineage specification.33 In that study, inhibition of TGF-β/activin/nodal signaling by SB-431542 abolished endodermal differentiation and enhanced neuroectodermal differentiation.33

In this study we report an ex vivo culture protocol for directing hESCs into mesenchymal progenitors by inhibiting TGF-β signaling with SB-431542 (SB) under serum-free culture conditions. Detailed cellular and molecular analysis revealed the differentiation of SB-treated hEBs into muscle progenitor cells (MPCs). When these cells were cultured in medium containing fetal bovine serum (FBS), they differentiated further into mesenchymal progenitors that subsequently could develop into osteoblast, chondrocyte, and adipocyte lineages both in vitro and in vivo.

Materials and Methods

Cell culture

The hESC lines: Harvard HUES-934 (kindly provided by Dr D Melton) and in-house hESC line KMEB235 were maintained undifferentiated using γ-irradiated (266.7 rads/min for 15 minutes) primary mouse embryonic fibroblast feeder cells seeded at 20,000 cells/cm² in plates (NUNC, Roskilde, Denmark) precoated with gelatin (0.1%) (G1393, Sigma, St. Louis, MO, USA). Undifferentiated hESCs were cultured with KO-Medium (KO-DMEM, 85%; 10829-018, Invitrogen, Taastrup, Denmark) supplemented with knockout serum replacement (KO-SR, 15%; 10828-028, Invitrogen), Glutamax (1%; 35050-038, Invitrogen), MEM nonessential amino acids (1%; 11140-035, Invitrogen), β-mercaptoethanol (0.1 mM; M7522, Sigma-Aldrich, Brondby, Denmark), penicillin/streptomycin [5000 U/mL (5000 µg/mL), 15070, Invitrogen], human serum albumin (0.5%, SSI, 8409), and basic fibroblast growth factor (bFGF, 5 ng/mL; 13256-029, Invitrogen). Cells were passaged at a 1:6 ratio every 5 to 6 days using trypsin/EDTA [0.1% (1 mM); Invitrogen]. The passage numbers of hESCs used in this study were between 19 and 35, and cells were karyotyped at the beginning and at the end of the study to verify the absence of any chromosomal abnormalities. Colony formation was visible within 2 to 3 days of passaging. Half-medium change was performed daily after the first 48 hours in culture.

Establishment of embryoid body (EB) outgrowth cultures

EB formation: To induce differentiation of hESCs into embryoid bodies (EBs), hESCs were disaggregated using trypsin/EDTA [0.1% (1 mM); 25300-0549, Invitrogen] into small clumps containing 5 to 10 cells and transferred to low-adhesion plastic petri dishes (Costar Ultra Low Attachment, Corning Life Sciences, Amsterdam, The Netherlands) in KO-Medium without addition of basic fibroblast growth factor (bFGF) and β-mercaptoethanol. The cells were cultured in the presence of SB (10 µM; Sigma) or control vehicle (DMSO, 0.01%). EBs were formed after 3 days in culture, and medium changes were performed on days 3, 6, and 8 by gravity sedimentation for 10 minutes at 37°C and 5% CO₂.

EB-explant outgrowth culture: On day 10, hEBs (both SB-treated and control) were transferred to plates and precoated with fibronectin (10 µg/mL; Sigma) in chemically defined medium (CDM) that allowed cellular outgrowth and formation of a monolayer. CDM was composed of DMEM/F-12 (Invitrogen) supplemented with bovine serum albumin (BSA, 0.5%; Fraction V BSA, Sigma), penicillin/streptomycin (1%; Invitrogen), lipids (1%; Invitrogen), and Insulin-Transferrin-Selenium (ITS) (10%; Invitrogen). The cultures were incubated in presence of SB (1 µM) or vehicle (control) (for the stepwise protocol, see Supplemental Fig. 1 A). The hEB-derived outgrowth cultures (OGs) in the presence of SB (termed here SB-OG) or vehicle (control-OG) were allowed to grow to confluency; subsequent passaging was performed using trypsin/EDTA (0.05% trypsin + 0.53 mM EDTA). The SB-OG cells were passaged in excess of 12 passages without evidence of replicative senescence.

Differentiation of EB-derived OG cultures

These studies were performed on hESC-derived OG cultures maintained in CDM between passages 4 and 11. The cells were plated at a density of 10⁴ cells/cm² in the presence of 10% FBS (PAA A15-043, Invitrogen) in α minimum essential medium (α-MEM) for 20 days. The cells were exposed to two-lineage differentiation protocols known to be successful in mesenchymal stem cell differentiation.4

For osteoblast differentiation, cells were seeded at a density of 10⁴ cells/cm² in α-MEM containing 10% FBS. When the cells were between 70% and 80% confluent, the medium was changed to normal culture medium (control) or medium containing osteogenic medium (OS); β-glycerophosphate (10 mM), L-ascorbic acid 2-phosphate (50 to 100 µg/mL), and dexamethasone (10 nM) (Sigma-Aldrich). Medium was renewed every 2 to 3 days throughout the experimental period of 20 days. For adipogenic differentiation, confluent cells were incubated with adipocyte induction medium (AIM) containing FBS (10%), dexamethasone (10 nM), 1-methyl-3-isobutylxanthine (IBMX, 450 µM; Sigma-Aldrich), rosiglitazone (1 µM; BRL49653, kindly provided by Novo Nordisk, Bagsvaerd, Denmark), and human recombinant insulin (3 µg/mL; I9278, Sigma-Aldrich) in DMEM high glucose (Invitrogen). Cells were induced for 15 days, with medium renewal every 2 to 3 days.

Real-time quantitative polymerase chain reaction (RT-qPCR)

Total RNA was extracted using the 6100 Nucleic Acid Prep Station (Applied Biosystems, Carlsbad, CA) using the RNA cell program. A high-capacity cDNA Archive Kit (Applied Biosystems) was used for reverse transcription of 1 µg RNA using random hexamer primers according to manufacturer's instructions. Real-time qPCR was performed on a MyiCycler thermal cycler (Bio-Rad, Copenhagen, Denmark) using iQ SYBR Green supermix (Bio-Rad) according to manufacturer's instructions. Quantification of target and reference genes (β-actin) was performed in duplicate. Following normalization to the β-actin gene, expression levels for each target gene were calculated using the comparative threshold cycle (C_T) method (1/2^ΔCT, where ΔC_T is the difference between C_T target and C_T reference). Data were analyzed using Optical System Software Version 3.1 (Bio-Rad) and Microsoft Excel 2000 (Seattle, WA, USA) to generate relative expression values. Primers used in this study are listed in Supplemental Table 1 employing an annealing temperature of 60°C. Data are presented as the mean ± SD from three independent experiments.

Western blotting

Human ES cells and hEBs were washed in PBS and lysed in RIPA buffer (Invitrogen) supplemented with protease inhibitors. After 1 hour of incubation at 4°C, samples were sonicated 3 × 20 seconds and centrifuged for 10 minutes (4°C, 14,000 rpm). Protein concentration was determined with a BCA kit (Bio-Rad), and 50 µg was loaded on a polyacrylamide gel (10%). Blotted nitrocellulose membranes were incubated overnight with anti-β-actin antibody (1 + 2500 dilution, 45 kDa; Cell Signaling, Danvers, MA), rabbit anti-Smad2/3 antibody, and rabbit anti-phospho-Smad2/3 antibody (1 + 2500 dilution, 58 kDa; Cell Signaling). The blots were developed after binding of the secondary anti-rabbit horseradish peroxidase–labeled antibody (1/500; Santa Cruz Biotechnology, Santa Cruz, CA, USA) using ECL technology and Kodak films.

Immunocytochemical staining

For immunostaining, cells were grown to 80% to 90% confluence on 2-, 4- and 8-well Permanox chamber slides (Nunc, Roskilde, Denmark) coated with fibronectin (Sigma-Aldrich). The slides were fixed in 4% formaldehyde for 10 minutes and then washed three times in PBS^+/+. Nonspecific immunoglobulin G binding sites were blocked for 20 minutes with 5% bovine serum albumin (BSA) in PBS or serum-free block solution (Zymed, San Francisco, CA). The cells were labeled with primary antibodies (Supplemental Table 2), incubated for 1 hour, and then stained with Alexa-Fluor 488– or Alexa-Fluor 555–conjugated (1:400 dilution) secondary antibodies.

Flow cytometric (FACS) analysis

Single cells from hEBs or OG cultures at different time points were harvested using trypsin and, following neutralization in 10% serum, resuspended in FACS buffer [phosphate-buffered saline (PBS^−/−; Invitrogen) + 0.5% BSA] at a concentration of 10⁶ cells/mL. For each antibody used, 10⁵ cells were stained. Antibodies for CD34FITC (DAKO), CD44PE, CD56PE, CD73PE, CD146PE, CD166PE, isotype control IgG1-FITC, and IgG1-PE all were obtained from BD Biosciences (San Diego, CA, USA; http://www.bdbiosciences.com). Before staining, cells were treated with either Fc-blocking buffer (BD Biosciences) or 5% FBS, 0.5% BSA, and 0.05% normal human serum in PBS (blocking buffer) for 10 minutes. Staining with the appropriate dilution of the antibody was performed for 30 minutes on ice in blocking buffer. After two washes in FACS buffer, the cells were resuspended in FACS buffer, and a minimum of 10,000 events were acquired for each sample using a FACSCalibur or FACSCan (BD Biosciences). The analysis was performed with FlowJo software (http://www.flowjo.com/, Tree Star Inc., Ashland, OR).

Histologic analyses

Cell or tissue blocks were sectioned at 4 µm. Immunohistochemical staining was performed on hEBs and implants using DAKO En Vision+ and PowerVision according to the manufacture's instructions (DAKO, Glostrup, Denmark). Briefly, paraffin sections were incubated for 1 hour at room temperature with primary antibodies diluted in ChemMate (DAKO). A list of primary antibodies can be found in Supplemental Table 2. Sections were washed subsequently in Tris-buffered saline (TBS, 0.05 M, pH 7.4), incubated for 30 minutes with secondary anti-mouse Ig/HRP-conjugated polymers (K4001, En Vision + , DAKO), and visualized with 3,30-diaminobenzidine tetrahydrochloride (DAB, S3000, DAKO) according to manufacturer's instruction. Controls were performed without addition of primary antibodies and processed under identical conditions.

Cytochemical staining

Staining for alkaline phosphatase (ALP): Cells undergoing osteoblast differentiation for 10, 15, and 21 days were stained for ALP as described previously36 using incubation in an ALP substrate solution containing naphthol AS-TR phosphate (0.2 mg/mL in water) mixed with fast red TR (0.417 mg/mL in 0.1 M Tris buffer, pH 9.5) for 1 hour at room temperature.

Alizarin red S staining for ex vivo–formed mineralized matrix: After 10, 15, and 21 days of osteoblast differentiation, as described previously, cell cultures were stained for the presence of a mineralized matrix using 40 mM alizarin red-S (AR-S, Sigma), pH 4.2, for 10 minutes at room temperature, washed thoroughly with PBS^−/−, and counterstained with Mayer's hematoxylin; mineralization was stained red.36

Oil red-O staining for lipids: Cells were induced to adipocyte differentiation as described earlier. Lipid-filled adipocytes were visualized using oil red-O staining solution (prepared by dissolving 0.5 g oil red-O powder in 60% isopropanol) and counterstained with Mayer's hematoxylin. Mature adipocytes containing lipid droplets stained red.

Microarray analysis

Both control-OG and SB-OG cells were cultured until 90% to 100% confluence. Total RNA was isolated using RNeasy Mini Kit (Qiagen, Sollentuna, Sweden) for gene expression analysis. Gene expression profiles were determined using Affymetrix HGU133A Plus 2.0 GeneChip (Affymetrix, Copenhagen, Denmark) according to the manufacturer's recommendations.37 Gene ontology classifications for all differentially expressed genes by SB in hESCs were performed using DAVID 2.0 software38 (http://david.abcc.ncifcrf.gov, NIADI, NIH, Bethesda, MD) and IPA Ingenuity pathway analysis. Microarray data have been submitted to the GEO (Gene Expression Omnibus, NCBI, NIH) repository and are found by the following link: www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE15553. Comparing the molecular signature of SB-OG versus SB-OG in 10% FBS (passage 3) was performed on an Illumina Human WG-6 Version 3 GeneChip (Gene Logic, Gaithersburg, MD, USA). RNA was isolated using the same isolation procedure as mentioned earlier; the gene array was performed in triplicate.

In vivo implantation studies

Intramuscular injection of hESC-derived OG cells: All animal experimental procedures were approved by the animal care and use committees of the University of South Denmark. Cells were harvested via trypsinization, washed in PBS, and resuspended in PBS or Matrigel (354234, BD Biosciences). A skin incision was performed in the thigh, and the cells were injected directly into the central part of the biceps femoris muscle. Approximately 5 × 10⁵ hESC-derived OG cells in 50 µL of Matrigel per injection site were used (n = 6). Cells were injected into the muscle using a Hamilton syringe (Reno, NV) with a 30G needle; control injections were performed in parallel with PBS and Matrigel in the same animals. After 4 weeks, the biceps femoris muscles were removed surgically, fixed in 4% formaldehyde, and embedded in paraffin.

Subcutaneous implantation of differentiated EB-derived OG cultures: As discussed previously, EB-derived OG cultures were incubated with FBS (10%) in α-MEM or vehicle (control) for 20 days. Then 5 × 10⁵ cells were mixed with 40 mg of hydroxyapatite–tricalcium phosphate ceramic powder per each implant (HA/TCP, Zimmer Scandinavia, Albertslund, Denmark) and implanted subcutaneously into the dorsal surface of 8-week old female NOD/SCID mice (NOD/LtSz-Prkdcscid), as described in more detail previously.39 After 8 weeks, the implants were recovered, fixed in 4% paraformaldehyde, decalcified using formic acid solution (0.4 M formic acid and 0.5 M sodium formate), and embedded in paraffin.

Results

SB treatment enhances gene expression of meso/ectodermal markers

In order to design a protocol for directing the differentiation of hESCs into mesenchymal progenitor cells, we examined the effect of blocking TGF-β signaling using SB, a specific inhibitor of TGF-β/activin/nodal signaling that blocks ALK-4, -5, and 740 during the induction of mesoderm in the hEB model. Consequently, for this, hESCs were cultured as hEB in serum-free medium in the presence of SB (10 µM) or vehicle for 10 days. Treatment with SB showed inhibition of Smad2 phosphorylation in a dose-dependent manner (Fig. 1A). Interestingly, as shown in Fig. 1B, SB-treated hEBs displayed a dose-dependent increase in gene expression of mesoderm differentiation markers: paraxial mesoderm and somite markers, for example, T-box transcription factor 6 (TBX6) (31-fold) and TBX5 (6-fold). In addition, mesoderm induction was associated with the upregulation of early myogenesis transcription factor Myf5 (40-fold) and myocyte commitment markers [smooth muscle myosin (smMHC, 6-fold) and skeletal muscle myosin (skMHC, 2-fold)] and cardiomyocyte committed markers [NK2 transcription factor–related locus 5 (Drosophila) (Nkx2.5, 32-fold) and troponin T type 2 (cardiac) (cTNT, 20-fold)], as assessed by real-time PCR. Similarly, the immunohistochemical staining of SB-treated hEBs revealed increases in the number of positively stained cells for myocyte markers NCAM (CD56), ASMA, tropomyosin, slow myosin heavy chain (sMHC), and cTNT (data not shown). We also observed upregulation of some neuroectodermal markers β-III-tubulin and NEUROD1 (8- and 14-fold, respectively). Conversely, blocking TGF-β signaling showed inhibition of the endodermal lineage commitment in hEBs, as demonstrated by downregulation of the endoderm-specific marker genes SRY (sex-determining region Y) box 17 (SOX17) and hepatocyte nuclear factor-1β (HNF-1β). Immunohistochemical staining was consistent with gene expression data, which showed low expression of endoderm-specific proteins (data not shown). These data demonstrate enhanced differentiation of EBs into meso/ectoderm (neural crest–like) lineage41 and expression of cellular markers known to be associated with muscle progenitor cells.

Figure 1. Characterization of SB-treated hEBs. (A) Embryoid bodies were treated with three different concentrations of small-molecule SB (0.1, 1, and 10 µM) for 10 days. (Upper panel) Western blot analysis on dose-dependent treatment of EBs with SB using active phospho-Smad2-specific antibody, total Smad2/3 antibody, or total β-actin antibody for equal loading. (B) Total RNA was extracted and real-time PCR analysis was performed on the undifferentiated HUES9 hESCs(–) and 10-day-old EBs treated with different concentrations of SB (0.1, 1, and 10 µM, respectively). Expression levels of target genes are represented as fold induction over undifferentiated hESCs. Gene expression analysis by real-time PCR for markers of paraxial mesoderm (Myf5, TBX6, smMHC, skMHC), cardiac mesoderm (TBX5, Nkx2.5, cTNT), ectoderm (β-III-tubulin, NeuroD1), and endoderm (HNF1β, Sox17). Expression of each target gene was normalized to β-actin. Data are shown as mean ± SD of three independent experiments. ^ap < .05; ^bp < .01 versus control cells (undifferentiated hESCs).

Enrichment of muscle progenitor cells (MPCs)

To selectively enrich for the MPC population, we established monolayer outgrowth cultures (OGs) derived from hEBs (for a stepwise protocol, see Supplemental Fig. 1A). Ten-day-old hEBs, were allowed to adhere to fibronectin-coated tissue culture plates and were grown as a monolayer in CDM medium in the presence of SB (1 µM; SB-OG) or vehicle (control-OG). During ex vivo passaging, SB-OG cell cultures exhibited areas of contracting myocytes (at this stage, cells were arranged as myotubes and not as cell aggregates; Fig. 2A), with greater than 80% of the cells exhibiting a contracting phenotype. The SB-OG cultures contained cells with various morphologies, including spindle shape and cuboidal morphology, as well as round cells. These cellular morphologies are reminiscent of MPCs42 (Fig. 2A, black arrow, white arrow, and black arrowhead, respectively). Control-OG cells exhibited a very large flattened heterogeneous cell population (Fig. 2A).

Figure 2. Characterization of hESC-derived MPC population. (A) Morphology of EB-derived outgrowth cells (OG) grown as a monolayer in the presence or absence of SB431542 (1 µM). Under SB treatment, cells exhibited a myocyte and satellite phenotype (spindle-elongated black arrow, round black arrowhead, and tri/multiangular white arrow) compared with control (DMSO treated) cells (flat). Scale bar = 200 µm. (B) Real-time PCR analysis performed on undifferentiated hESCs and control outgrowth and SB-treated outgrowth cells for smooth muscle myosin (smMHC), cardiac muscle (cTNT, NKX2.5), skeletal muscle myosin (skMHC), myogenic marker gene (Myf5), and neuroectoderm marker neurogenic differentiation 1 (NeuroD1). Expression of each target gene was normalized to β-actin. Data are shown as mean ± SD of three independent experiments. ^ap < .05; ^bp < .01 versus control cells (undifferentiated hESCs). (C) Immunofluorescence staining of SB-OG cells. (D) Double immunofluorescence staining of SB-OG cells (original magnification ×400). (E) Immunostaining for neuroectoderm marker Pax6 (I) in control-OG cells and (II) SB-OG cells (original magnification ×100). (F) Flow cytometric analysis of SB-OG cells for various mesoderm-specific adhesion molecules, CD34, and CD56. At least 30,000 cells were counted; the data shown are a representative picture of five independent experiments.

Real-time PCR expression analysis of myogenic markers43 in SB-OG cells revealed marked upregulation of smooth muscle markers smMHC (12-fold), markers of differentiating cardiomyocytes [cTNT (8-fold), NKX2.5 (26-fold),44 and skeletal myosin heavy chain 2a (skMHC, 64-fold)], and makers of myogensis [Myf5 (161-fold)] (Fig. 2B). In contrast, the expression of neuroectoderm marker NEUROD1 was downregulated in SB-OG cells by 100-fold (Fig. 2B).

In addition, immunohistochemical analysis of the SB-OG cells demonstrated an enrichment of cells expressing specific myogenic markers: NCAM⁺ (86% ± 10%), MyoD⁺ (52% ± 6%) desmin⁺ (75% ± 3.5%), cTNT⁺ (31% ± 5%), tropomyosin⁺ (8% ± 1%), and skeletal myosin [fast myosin heavy chain (fMHC⁺), 16 ± 2%], α-smooth muscle cell (ASMA⁺, 38% ± 3%) (Fig. 2C, D). Furthermore, NCAM⁺ cells were found to stain positive for the myogenic markers desmin⁺ (80% to 90%), ASMA⁺ (30% to 40%), and SERCA1⁺ (SR-1, 19% ± 3%)45 (Fig. 2D) and were negative for paired box transcription factor 7 positive (Pax7⁺, 25% ± 4%)46 (Fig. 2D). Conversely, the control cultures stained positive for Pax6 (between 80% and 90%), suggesting more commitment of these nontreated cells into neuroectodermal differentiation (Fig. 2E). Furthermore, FACS analysis of the SB-OG cultures revealed that more than 30% of the cells are CD34⁺ and above 70% are CD56⁺ (NCAM⁺) (Fig. 2F), signifying the presence of a CD34⁺ muscle progenitor cell population.

Further characterization of SB-OG cultures using DNA microarray

To identify the transcriptional pathways underlying the effects of SB treatment on hESCs, we compared the basal gene expression pattern between SB-OG cells and control-OG cells using Affymatrix DNA microarrays. In response to SB treatment, a total of 3383 probe sets corresponding to 2470 genes/expressed sequence tags were found to be differentially expressed. A total of 810 genes were shown to be upregulated in SB-OG cells by more than 2-fold (p < .001), whereas 1171 genes were downregulated by less than 2-fold (p < .001). Annotation of the upregulated genes using the DAVID database and Ingenuity pathway analysis revealed 21% (9.4 × 10⁻³⁰ < p < 4.5 × 10⁻⁴) of the transcripts to be clustered as muscle-development-related genes, including known myosins (MYBPH, MYH8, MYH2, MYH3, MYL1, MYH11, and MYL4) and troponins (TNNI2, TNNC2, TNNI1, TNNC1, TNNT3, TNNT1, and TNNT2). Activated satellite cell and MPC markers such as MyoD1 (6.9-fold), MEF2C (6.9-fold), M-cadherin (6.7-fold), NCAM (4.5-fold), MEGF10 (3.3-fold), and Pax3 (2.2-fold) were highly expressed in SB-OG cells. From the skeletal muscle development, several specific genes were upregulated: CACNA1S (7.1-fold), CAPN3 (4.1-fold), FOXP2 (3.2-fold), and HSPB2 (6.7-fold), as well as genes found in smooth muscle cells: MYH11 (5.4-fold), AOC3 (4.3-fold), and MYL6B (2.1-fold), confirming the induction of mesoderm and myogenic differentiation. For a detailed list of upregulated genes categorized into muscle development, muscle-contraction-related genes, and myogenesis, see Table 1.

Table 1. Microarray Data Presenting the Upregulated Genes in SB-OG Cells

GeneBank accession no.	Gene name	Gene symbol	Fold change SB/control	Fold change FBS/SB
Note: This table lists significantly upregulated genes that show the highest fold induction of SB-OG versus control-OG cells. N.D. = not determined by the microarray; N.S. = not significant downregulated.
Muscle development
NM_006790	Myotilin	MYOT	9	−41.8
NM_004997	Myosin binding protein H	MYBPH	8.8	−139.6
NM_002472	Myosin, heavy polypeptide 8, skeletal muscle, perinatal	MYH8	8.6	−45.5
NM_001103	Actinin, α2	ACTN2	8.6	−7.6
NM_017534	Myosin, heavy polypeptide 2, skeletal muscle, adult	MYH2	8.3	N.D.
NM_004320	ATPase, Ca^2+ transporting, cardiac muscle, fast twitch 1	ATP2A1	7.7	−2.3
NM_003282	Troponin I type 2 (skeletal, fast)	TNNI2	7.5	−15.6
NM_002152	Histidine–rich calcium–binding protein	HRC	7.5	−19.5
NM_003279	Troponin C type 2 (fast)	TNNC2	7.4	−91.0
NM_000023	Sarcoglycan, α (50 kDa dystrophin–associated glycoprotein)	SGCA	7.1	−11.2
NM_000069	Calcium channel, voltage-dependent, l type, α1s subunit	CACNA1S	7.1	N.S.
NM_001100	Actin, α1, skeletal muscle	ACTA1	7.0	−161.1
NM_004533	Myosin-binding protein C, fast type	MYBPC2	7	−3.6
NM_014332	Small muscle protein, X-linked	SMPX	7	−27.0
NM_006258	Protein kinase, CGMP-dependent, type I	PRKG1	6.9	N.S.
NM_003281	Troponin I type 1 (skeletal, slow)	TNNI1	6.8	N.S.
NM_002470	Myosin, heavy polypeptide 3, skeletal muscle, embryonic	MYH3	6.6	−167.9
NM_001232	Calsequestrin 2 (cardiac muscle)	CASQ2	6.5	−37.8
NM_079420	Myosin, light polypeptide 1, alkali, skeletal, fast	MYL1	6	−91.1
NM_002465	Myosin-binding protein C, slow type	MYBPC1	6.0	−15.0
NM_003275	Tropomodulin 1	TMOD1	5.9	−4.6
NM_005031	FXYD domain–containing ion transport regulator 1 (phospholemman)	FXYD1	5.7	N.S.
NM_006063	Kelch repeat and BTB (POZ) domain–containing 10	KBTBD10	5.7	−92.1
NM_003280	Troponin C type 1 (slow)	TNNC1	5.5	−72.5
NM_020402	Cholinergic receptor, nicotinic, α1 (muscle)	CHRNA1	5.5	−39.3
NM_001927	Desmin	DES	5.4	−14.9
NM_022844	Myosin, heavy polypeptide 11, smooth muscle	MYH11	5.4	−3.7
NM_003319	Titin	TTN	5.3	−2.4
NM_000540	Ryanodine receptor 1 (skeletal)	RYR1	5.2	−11.5
NM_003803	Myomesin 1 (skelemin), 185 kDa	MYOM1	5.2	−46.3
NM_006757	Troponin T type 3 (skeletal, fast)	TNNT3	5.1	−22.0
NM_011618	Troponin T type 1 (skeletal, slow)	TNNT1	5.1	−15.1
NM_000364	Troponin T type 2 (cardiac)	TNNT2	5	−20.3
NM_000751	Cholinergic receptor, nicotinic, δ	CHRND	4.7	−4.3
NM_000366	Tropomyosin 1 (α)	TPM1	4.4	N.S.
NM_000727	Calcium channel, voltage-dependent, γ subunit 1	CACNG1	4.4	−3.5
X04201	Tropomyosin 3	TPM3	4.2	N.S.
NM_001957	Endothelin receptor type A	EDNRA	4.1	−5.7
NM_021098	Calcium channel, voltage-dependent, α1 h subunit	CACNA1H	3.9	−5.3
NM_002476	Myosin, light polypeptide 4, alkali, atrial, embryonic	MYL4	3.9	−81.1
NM_006073	Triadin	TRDN	3.9	N.S.
NM_021021	Syntrophin, β1 (dystrophin-associated protein a1, 59 kDa, basic component 1)	SNTB1	3.7	N.S.
NM_003494	Dysferlin, limb girdle muscular dystrophy 2b (autosomal recessive)	DYSF	3.7	−5.4
NM_012282	KCNE1-like	KCNE1L	3.5	−4.9
AI125337	CUG triplet repeat, RNA-binding protein 2	CUGBP2	3.3	−2.9
NM_001885	Crystallin, αb	CRYAB	3.1	−19.0
NM_000775	Cytochrome P450, family 2, subfamily j, polypeptide 2	CYP2J2	3.1	N.S.
NM_000747	Cholinergic receptor, nicotinic, β1 (muscle)	CHRNB1	3.0	−10.5
NM_001957	Endothelin receptor type A	EDNRA	2.9	−5.7
NM_005159	Actin, α, cardiac muscle	ACTC	2.7	−100.3
NM_007159	Sarcolemma-associated protein	SLMAP	2.5	N.S.
NM_000238	Potassium voltage-gated channel, subfamily H (EAG-related), member 2	KCNH2	2.4	N.S.
NM_004393	Dystroglycan 1	DAG1	2.4	N.S.
NM_001390	Dystrobrevin, α	DTNA	2.3	N.S.
NM_020164	Aspartate β-hydroxylase	ASPH	2.2	N.S.
NM_004137	Potassium large conductance calcium-activated channel, subfamily M, β member 1	KCNMB1	2.2	−14.3
M87313	Dystrophia myotonica-protein kinase	DMPK	2.2	N.S.
NM_003970	Myomesin (M-protein) 2, 165 kDa; myomesin (M-protein) 2, 165 kDa	MYOM2	2.1	−6.0
X04201	Tropomyosin 3	TPM3	2.1	N.S.
NM_002667	Phospholamban	PLN	2	N.S.
Myoblast, skeletal, and striated muscle development
NM_001824	Creatine kinase, muscle	CKM	10.4	−23.2
NM_016599	Myozenin 2	MYOZ2	9	N.S.
NM_182645	Vestigial like 2 (Drosophila)	VGLL2	8.2	−3.8
NM_002478	Myogenic differentiation 1	MYOD1	6.9	−7.4
NM_002397	Mads box transcription enhancer factor 2, polypeptide C (myocyte enhancer factor 2c)	MEF2C	6.9	−15.9
NM_002479	Myogenin (myogenic factor 4)	MYOG	6.8	−18.8
NM_004933	Cadherin 15, M-cadherin (myotubule)	CDH15	6.7	−51.5
NM_002206	Integrin, α7	ITGA7	5.4	N.S.
NM_000615	Neural cell adhesion molecule 1	NCAM1	4.5	N.S.
NM_000210	Integrin, α6	ITGA6	4.4	N.S.
NM_003174	Supervillin	SVIL	4.3	N.S.
NM_000885	Integrin, α4 (antigen CD49D, α4 subunit of VLA-4 receptor)	ITGA4	3.9	−3.4
NM_006500	Melanoma cell adhesion molecule	MCAM	3.6	N.S.
NM_032446	Multiple EGF-like domains 10	MEGF10	3.3	−6.7
NM_016132	Myelin expression factor 2	MYEF2	3.2	N.S.
NM_000214	Jagged 1 (alagille syndrome)	JAG1	3.2	−2.5
NM_013942	Paired box gene 3 (waardenburg syndrome 1)	PAX3	2.2	N.S.
NM_012237	Sirtuin (silent mating type information regulation 2 homologue) 2 (S. cerevisiae)	SIRT2	2.1	N.S.
NM_001731	B-cell translocation gene 1, antiproliferative	BTG1	1.9	N.S.
NM_014048	MKL/myocardin-like 2	MKL2	1.7	−2.5
NM_002207	Integrin, α9	ITGA9	1.6	−2.0
NM_003476	Cysteine- and glycine-rich protein 3 (cardiac LIM protein)	CSRP3	1.5	N.S.
Muscle development
NM_001234	Caveolin 3	CAV3	8.9	−5.4
NM_005592	Muscle, skeletal, receptor tyrosine kinase	MUSK	7.8	N.S.
NM_004543	Nebulin	NEB	6.7	−17.8
NM_012278	Integrin β1-binding protein (melusin) 2	ITGB1BP2	5.8	N.S.
NM_002206	Integrin, α7	ITGA7	5.4	N.S.
NM_013259	Transgelin 3	TAGLN3	5.0	N.S.
NM_005259	Growth differentiation factor 8	GDF8	4.9	N.D.
NM_024344	Calpain 3 (p94)	CAPN3	4.1	N.S.
NM_147147	Blood vessel epicardial substance	BVES	4.0	N.S.
NM_000337	Sarcoglycan, δ (35 kDa dystrophin-associated glycoprotein)	SGCD	3.6	−2.8
NM_003213	Tea domain family member 4	TEAD4	2.9	−2.4
NM_015719	Collagen, type V, α3	COL5A3	2.3	N.S.
NM_003205	Transcription factor 12 (HTF4, helix-loop-helix transcription factors 4)	TCF12	2.1	−2.9
NM_004430	Early growth response 3	EGR3	1.9	N.S.
NM_005982	Sine oculis homeobox homologue 1 (Drosophila)	SIX1	1.7	−2.2
Development of somites and limb
NM_001427	Engrailed homologue 2	EN2	7.8	N.S.
NM_005924	Mesenchyme homeobox 2	MEOX2	5.5	−3.5
NM_005618	Delta-like 1 (Drosophila)	DLL1	5.5	−4.4
NM_003888	Aldehyde dehydrogenase 1 family, member a2	ALDH1A2	5.0	−13.9
NM_001040168	Lunatic fringe homologue (Drosophila)	LFNG	4.5	N.S.
NM_017780	Chromodomain helicase DNA-binding protein 7	CHD7	4.0	−4.8
NM_001103184	Formin 1	FMN1	3.0	N.D.
NM_012486	Presenilin 2 (Alzheimer disease 4)	PSEN2	2.4	N.S.
NM_023105	Fibroblast growth factor receptor 1 (FMS-related tyrosine kinase 2, Pfeiffer syndrome)	FGFR1	2.2	N.S.
NM_014000	Vinculin	VCL	2.1	N.S.
NM_001453	Forkhead box C1	FOXC1	2.3	N.S.
NM_004527	Mesenchyme homeobox 1	MEOX1	2.1	N.S.

Many cell adhesion molecules, such as integrin-α4, -α6, -α7, and -α9, dystroglycan, and Mcam, were found to be differentially expressed, offering a possible explanation for fusion ability and the generation of myotubes. The high upregulation of integrin-α7 has been proposed as highly expressed in undifferentiated embryonic myoblasts.47

A number of transcription factors were upregulated, including the proliferation/differentiation regulator Fos (2.1-fold).48 Additionally, the myoblasts showed an upregulation in Pax3, which is known to control the delamination and migration of somatic muscle progenitors to the limb bud49 and, along with Myf5, is able to activate MyoD expression in the embryonic body.50 Interestingly, several genes whose expression was reported previously to be directly or indirectly linked to Pax3 expression were also relatively highly expressed (Table 1), for example, Meox1.51 In addition, the microarray analysis revealed also that Meox2 (5.5-fold) was upregulated; this has been shown to be an important regulator of limb myogenesis52 and is able to physically interact with Pax3 in vitro.53 The expression of an established Pax3 target, Six-1,54 also was highly upregulated (1.9-fold). Another gene, Mef2C (6.9-fold), which regulates muscle differentiation, was found to be highly upregulated in SB-OG cells, together with several molecules that are known to be involved in skeletal muscle development or somatogenesis such as EN2, DLL1, ALDH1A2, LFNG, CHD7, FMN1,and FoxC1 (see Table 1).

The rest, 79% of genes, were assigned to different biologic categories, including regulation of biologic processes (31%), intracellular signaling (9%), cytoskeleton organization and biogenesis (7%), morphogenesis (6%), cell differentiation and nervous system development (both contain 4% of the upregulated genes), cell adhesion (3%), and development of somites and limbs (2%), and 1% of the genes were categorized in cell proliferation (Supplemental Fig. 1B).

Nine of the highly downregulated genes (<6-fold downregulated) were TGF-β-responsive genes or endoderm development or cell-cell signaling genes (Supplemental Table 3), thus confirming the inhibition of TGF-β signaling by SB.

Pathway analysis identified 24 upregulated signaling pathways and 20 downregulated signaling pathways using the Ingenuity pathway analysis program (Table 2). The microarray data clearly revealed that SB-OG cells constitute a myoblast cell population that contains myoblast precursor cells and myoblasts that express the correct set of genes employed in embryonic development during skeletal development for maturation to myotubes.

Table 2. Signaling Pathways that Are Upregulated in SB-OG Culture

Signaling pathway	Log (p value 0.0001)	Percent upregulated genes
Calcium signaling	10.70	20
Axonal guidance signaling	4.07	12.4
Integrin signaling	4.03	15.6
Actin cytoskeleton signaling	3.93	14.2
Cardiac β-adrenergic signaling	3.88	16.2
Notch signaling	3.41	24.4
Tight junction signaling	3.30	15.1
Neurotrophin/TRK signaling	3.16	17.8
Cell cycle: G1/S checkpoint regulation	3.13	20.0
Aryl hydrocarbon receptor signaling	2.69	13.2
cAMP-mediated signaling	2.51	13.8
Insulin receptor signaling	2.36	13.5
Ephrin receptor signaling	2.08	12.0
VEGF signaling	2.07	14.1
Synaptic long-term potentiation	1.88	13.5
PTEN signaling	1.86	14.1
G protein–coupled receptor signaling	1.77	11.6
Inositol phosphate metabolism	1.72	9.8
Regulation of actin-based motility by rho	1.69	13.0
Apoptosis signaling	1.65	13.8
ERK/MAPK signaling	1.58	11.7
Nitric oxide signaling in the cardiovascular system	1.56	11.8
IGF-1 signaling	1.54	13.0
Wnt/β-catenin signaling	1.53	11.4

In vivo implantation of SB-OG cells

To further examine the ability of SB-OG cells to be engrafted into host skeletal muscle in vivo, cells were injected intramuscularly into the hind limbs of immune-deficient NOD/SCID mice (5 × 10⁵ cells/limb, n = 6). Histologic analysis of implant sections shows an infiltration into existing skeletal muscle bundles by SB-OG cells (Fig. 3A), and immunohistochemical staining revealed coexpression of a human-specific antibody TRA-1-85 with desmin and myogenic markers: NCAM, fMHC, Pax7, dystrophin, and SERCA-1 (Fig. 3B). However, there was no fusion between human cells and host skeletal tissues. We additionally implanted the SB-OG cells with Matrigel (5 × 10⁵ cells/site, n = 6) subcutaneously in NOD/SCID mice to exclude the possibility of teratoma formation from these cells. As shown in Fig. 3C, 4 weeks after injection, the sections of implants were stained positively for human TRA-1-85 in combination with all the above-mentioned myocyte markers without any sign of teratoma formation. In contrast to the intramuscular implants, the subcutaneously implanted cells had matured to myocytes containing sarcomeres when stained with fMHC43 (Fig. 3C, I and I'). In addition, Pax7⁺ cells did not form sarcomas, had a rounded morphology, and were located adjacent to mature myocytes, thus suggesting that the Pax7⁺ cells had formed a pool of myogenic progenitor cells (Fig. 3C, III and III').

Figure 3. In vivo differentiation and maturation of hESC-derived SB-OG cells in NOD/SCID mice. Human ESC-derived SB-OG cells were injected into the hind limb of four NOC/SCID mice; 5 × 10⁵ cells were injected at each site in combination with Matrigel for 4 weeks. As a negative control, a comparable amount of Matrigel was injected. (A) Representative immunohistochemical staining of the area of cell transplantation with anti-human Tra-1-85, verifying the human phenotype of the transplanted cells (original magnification ×40). (B) Higher-magnification images of the transplanted cells with their muscle phenotype and stained with skeletal muscle–specific markers (original magnification ×200). (Insets) Corresponding section with higher magnification (×100).

Differentiation of SB-OG cells into mesenchymal progenitors

Previous studies on hESCs have shown that serum induces mesenchymal differentiation of ESCs.55 To test the ability of SB-OG cells to differentiate into other mesoderm-lineage cells found during skeletal development, we cultured SB-OG cells in the presence of FBS (10%). Interestingly, under these culture conditions, SB-OG cells showed a high ability for plastic adherence and morphologic changes into large cuboidal, flattened, and spindle-shaped fibroblast-like cells (Fig. 4A). The morphologic changes were associated with marked reduction in the number of the cells expressing the specific myogenic markers, for example, NCAM⁺ (5% ± 2%), fMHC⁺ (3% ± 1%), MyoD⁺ (3% ± 1%), cTNT⁺ (0%), and desmin⁺ (20% ± 6%) in FBS-treated cultures (Fig. 4A). In addition, these cells started to express the mesenchymal intermediate filament protein vimentin (Fig. 4A). Consistent with these data, FACS analysis revealed a decreased number of CD56⁺ cells (3% ± 3%) and CD34⁺ cells (1% ± 1%) in FBS-treated cultures (Fig. 4C). Furthermore, an enrichment for cells expressing hMSC surface markers, including, CD44⁺ cells (99.2%), CD73⁺ cells (98%), CD146⁺ cells (96%), and CD166⁺ cells (88%), was observed56 (Fig. 4C). In addition, real-time qPCR analysis revealed marked downregulation of the expression of muscle marker genes Myf5, Nkx2.5, and skMHC in parallel with the upregulation of MSC genes Runx2/CBFA1, Alx4, and Msx2 (Fig. 4B). To confirm that downregulation of myocyte gene expression was initiated by transferring the cells from CDM medium containing SB to medium containing serum, we compared the gene expression pattern between FBS-treated cells and SB-OG cells using microarray analysis. Serum stimulation of SB-OG cells showed an increase of 574 genes (>2-fold, p < .001) and a decrease in 752 genes (<2-fold, p < .001). Annotation of the differentially expressed downregulated genes based on their biologic function revealed that 68% (3.23 × 10⁻²⁷ < p < 3.7 × 10⁻²) of the genes were clustered to muscle development and muscle contraction, including some of the key myocyte genes found upregulated in SB treatment: MYH3 (−167.9-fold), ACTA1 (−161-fold), MYBPH (−139-fold), ACTC (−100.3-fold), MYH8 (−45.5-fold), and MYOT (−41.8-fold), among others (Table 1; myocyte-related downregulated genes are depicted in boldface). Furthermore, genes related to calcium signaling, integrin signaling, actin cytoskeleton signaling, and Notch and tight-junction signaling were highly downregulated (Supplemental Table 4). Conversely, annotation of the upregulated genes revealed that 11% of the upregulated genes were annotated as mesoderm developmental genes, including SFRP1 (11-fold), VEGFC (7.4-fold), TIMP1 (4.5-fold), EFNB2 (4.2-fold), CEBPB (3.5-fold), VCAM1 (3.2-fold), COL18A1 (2.49-fold), VEGFA (2.48-fold), and PPARG (2.15-fold) (Supplemental Table 5). In addition, 7% (1.54 × 10⁻³⁰ < p < 4.41 × 10⁻⁴) of the transcripts are clustered to skeletal development, including known genes such as IL8 (129-fold), CCND1 (14-fold), CSF2 (13-fold), CD44 (11.8-fold), RAC2 (10.4-fold), FOSL1 (9.5-fold), and Sox9 (5-fold).

Figure 4. Directed differentiation of hESC-derived SB-OG cells to multipotent mesenchymal cells. (A, I) Morphology of hESC-derived mesenchymal progenitors after 20 days of stimulation with 10% FBS. Bar = 200 µm. (II) Immunostaining used to monitor NCAM expression shows a marked decreased after 20 days of serum stimulation (original magnification ×10X). (III) Characterization of hESC-derived SB-OG cells with antibodies against fast myosin (fMHC), cardiac troponin-T (cTNT), MyoD1, desmin, and mesenchymal markers vimentin and Runt-related transcription factor-2 (Runx2) nuclear staining with DAPI (blue) (original magnification ×400). (B) Real-time PCR analysis performed on hESC-derived MPCs in CDM medium (CDM); same cells treated with 10% FBS for 20 days (FBS) for Myf5, Nkx2.5, skMHC, Runx2, and homeobox transcription factors (Alx4 and Msx2). Expression of each target gene was normalized to β-actin. Data are shown as mean ± SD of three independent experiments. *p < .05; **p < .01 versus control cells (hESC-derived MPCs) induced with SB. (C) Flow cytometric analysis of hESC-derived mesenchymal progenitors with MSC-specific cell surface markers. These mesenchymal progenitors were negative for CD34 and CD56. The data shown are a representative picture of five independent experiments.

Further analysis of signaling pathways revealed the upregulation of pathways that play essential roles during bone formation and mesenchymal differentiation, such as Wnt/β-catenin signaling,57 ERK/MAPK signaling,58 epidermal growth factor (EGF) signaling,59 and vascular endothelial growth factor (VEGF) signaling60 (Supplemental Table 4).

To further characterize the potential of these cells to differentiate into mesenchymal-specific cell types, we employed culture conditions known to induce human mesenchymal (stromal) stem cells (MSCs) to differentiate to osteogenic and adipogenic lineages.61 As shown in Fig. 5A, the cells were able to differentiate to the osteoblastic lineage, as assessed by increasing ALP activity and the formation of mineralized matrix (Fig. 5A). Moreover, the expression of osteoblastic markers Runx2/CBFA1, osterix (Osx), ALP, osteopontin (OPN), osteonectin (ON), and osteocalcin (OC) was markedly upregulated during osteoblast differentiation of these mesenchymal progenitors (Fig. 5B).

Figure 5. In vitro osteogenic and adipogenic differentiation of hESC-derived MSC-like cells. (A) hESC-derived MSCs were differentiated into osteoblasts by osteogenic mixture (OS) for 20 days. (Upper panel) Mineralized calcium deposition was determined by alizarin red S staining. (Lower panel) ALP cytochemical staining on days 10, 15, and 20 after OS induction. (B) Real-time PCR analysis performed on hESC-derived MSCs in OS medium for osteogenic markers, including Runx2, OPN, ALP, OSX, ON, and OC. (C) Cells were induced in adipocyte induction medium (AIM) for 10 and 15 days and stained by oil red O for adipocytes containing lipid droplets (upper panel). (Lower panel) Total RNA was extracted and RT-PCR performed for adipocyte markers, including human aP2, LPL, and PPARγ-2. Expression of each target gene was normalized to β-actin. Data are shown as mean ± SD of three independent experiments. *p < .05; **p < .01 versus day 0 cells. (D) In vivo maturation and differentiation of hESC-derived MSCs. hESC-derived MSCs were implanted with HA/TCP subcutaneously into NOD/SCID mice. Alcian blue staining of paraffin-embedded implants revealed cartilage formation (×40). Serial sections stained with human-specific Tra-1-85 (×40). Serial sections stained with Sox9 and vimentin (original magnification ×100). The histology of in vivo–formed bone was examined with H&E staining (original magnification ×200). Serial sections showing ON staining of bone and surrounding tissue in combination with Sox-9 staining (×200). b = bone; H&E = hematoxylin and eosin.

We then induced the cells to differentiate into adipocytes using our previously reported adipocyte induction mixture (AIM)36 for 15 days. Gene expression analysis revealed significant upregulation of adipogenic gene marker expression: adipocyte fatty-acid-binding protein (aP2), lipoprotein lipase (LPL), and nuclear factor peroxisome proliferator–activated receptor γ (PPARγ2) (Fig. 5C). Also, cytoplasmic accumulation of fat droplets, stained positive with oil-red O, was detected (Fig. 5C).

To examine the in vivo differentiation capacity of mesenchymal progenitors, cells were mixed with HA/TCP and implanted subcutaneously into NOD/SCID mice.39 Histologic analysis of the implants revealed the formation of ectopic cartilage, as assessed by alcian blue staining and Sox9 protein expression, as well as ectopic bone, as assessed by expression of osteonectin protein (Fig. 5D). To further elucidate the differentiation ability of hESC-derived MSCs, we tested the cells both at early and late passage for in vitro and in vivo differentiation, as mentioned earlier. The results were promising and showed that the cells maintain their differentiation ability during successive passaging. The differentiation ability was tested only up to passage 10 after serum stimulation (Supplemental Figure 4).

Discussion

In this study we have described a simple and reproducible culture method for hESCs that promotes the development of cell populations with a differentiation potential to a number of mesodermal cell lineages, preferentially those found during skeletal development: myocytes, osteoblasts, chondrocytes, and adipocytes. Our culture method is based on inhibition of Smad2/3 phosphorylation in the hESC-EB model by SB followed by establishing hEB-outgrowth cultures in the presence of SB and under defined culture conditions.

Culturing ESCs as EBs activates well-conserved cascades of genes that govern the earliest events during gastrulation and germ-layer formation62 that may be important for subsequent differentiation. In support of this notion, several studies have used the hEB model to differentiate hESCs into mesenchymal osteoblastic cells22 and other cell types (neural-crest-like cells).63 While some studies succeeded in differentiating hESCs directly from monolayer cultures19, 20 our study demonstrates some advantages of using the EB formation step.

Blocking of TGF-β signaling during hEB formation using SB led to selective upregulation of several markers involved in mesoderm induction and myogenic differentiation, as evidenced by enhanced gene expression of TBX6 and Myf5. Expression of TBX6 is restricted to the primitive streak and newly recruited paraxial mesoderm. Later in development, TBX6 expression is restricted to presomitic mesoderm.64 Moreover, genes known to be expressed by cardiomyocytes (TBX5, NKX2.5, and cTNT)65 and myogenic genes (sMHC and NCAM) and tropomyosins66 were highly expressed. Interestingly, it appears that SB treatment can additionally enhance neuroectoderm differentiation depending on culture conditions, as reported recently by Smith and colleagues.67 In this report, the authors blocked TGF-β signaling using either SB (20 µM) or overexpression of Lefty and Creb-S (in hEBs for 14 to 16 days), which enhanced the formation of epithelial rosettes expressing known neuronal genes, for example, Sox1, Sox3, NESTIN, β-III-tubulin, and NGN2, and inhibited endodermal differentiation.67 However, they did not report the expression of mesodermal markers and did not establish EB-outgrowth cultures.

Explant cultures of hEBs in the presence of continuous treatment with SB (SB-OG) under serum-free conditions led to enrichment of cells with a myogenic phenotype. This finding is in agreement with the two-cell-lineage model, which appears during skeletal muscle formation.68, 69 The phenotype of cells obtained from hESCs ranged from immature (Pax7⁺) to more mature myocytic (SERCA1⁺fMHC⁺MyoD⁺) cells. The expression of myogenic markers included the key MPC maker Pax7 together with myogenic precursor and satellite cell markers CD34, MyoD1, Myf5, desmin, and NCAM.42 The high percentage of cells (25%) retaining Pax7 expression during myogenic differentiation suggests the presence of a potential reservoir of muscle progenitor cells (MPCs)70 in these cultures. The implication of the observed Myf5⁺/Pax7⁺ heterogeneity is that most satellite cells, at some stage of muscle development, can revert to a quiescent satellite cell state. Indeed, even in cell culture, activated satellite cells can revert to a Pax7⁺/Myf5⁻ state.71, 72 The satellite cells can assume a number of phenotypes suggestive of stem cells, committed progenitors, or dedifferentiated myoblasts.

Microarray analysis of SB-OG cells confirmed that these cultures contain cells committed to the myogenic phenotype and demonstrated the presence of enhanced expression of markers of known myocytes (MYH2, TNNI2, TNNC2, MYH3, DES,ACTN2, ATP2A1, CACNA1S, CASQ2, MEF2C, TMOD1, TNNT2, and TPM3) as well as muscle stem cell markers (NCAM1 and PAX3) (for details, see Table 1). Culture heterogeneity also has been observed by other investigators. In addition, the muscle contraction in our model was supported by upregulation of the calcium signaling pathway–related genes (see Table 2), including ryanodine receptor 1, phospholamban, calcium channel voltage dependent L-type α1S subunit, ATP2a1, ATP2a2, Cacna1S, Cacna1H, and troponin C.73, 74 Interestingly, 20% of the genes from the calcium signaling pathway are significantly upregulated by SB treatment of hESCs. The gene for calcium channel voltage-dependent L type α1S subunit75 is 7.1-fold upregulated following treatment (Table 2 and Supplemental Fig. 2).

Induction of mesoderm and myogenic cell differentiation by blocking SMAD and therefore TGF-β signaling can be mediated by a number of mechanisms. A recent study has demonstrated that blocking TGF-β signaling led to activation of Notch signaling and consequent upregulation of cell cycle genes [e.g., cyclin-dependent kinase (CDK) and inhibitors p15, p16, p21, and p27].76 Our microarray data corroborate these findings. SB treatment enhanced Notch signaling and ephrin receptor signaling (which is a direct target of Notch) acting upstream of NRG1 (Table 2). Notch signaling regulates embryonic cell fate determination, differentiation, and patterning. A potential link between Notch signaling and the transmembrane protein Megf10 (upregulated 3.3-fold in SB-OG; Table 1), a novel marker of quiescent and activated satellite cells, has been suggested in the control of satellite cell proliferation.77 The ability of SB-OG cells to grow in the absence of serum or other exogenous growth factors suggests that autocrine/paracrine pathways are involved in their proliferation. Other possible signaling pathways may be involved in myogenic differentiation of hESCs. Studies have demonstrated that myocyte cell proliferation is reduced by inhibiting either PI-3 kinase or Akt kinase and enhanced by signaling through the insulin-like growth factor 1 (IGF-1) receptor.78 In microarray analysis, we found that genes related to the PI-3 and Akt kinase signaling pathway were upregulated, suggesting a role in induction of myogenic differentiation of hESCs (for detailed information regarding molecules involved in the above-mentioned signaling, see Supplemental Fig. 2). It is thus plausible that targeting several signaling pathways through a combination of small molecules can direct the differentiation of hESCs to particular lineages. Some studies are being conducted in the literature along these lines, where chemical inhibitors have promoted differentiation into the cardiomyocyte lineage.79

The phenotype of the SB-OG cell population containing mesodermal stem cells equates to MPCs originating within the somites.80 Muscle progenitor cells have been characterized as expressing the transcription factor Pax7 and are positive for the cluster differentiation markers CD34, CD44, CD56, and CD146.45 The high expression of different mesodermal and myogenic markers in the SB-OG cell population also was observed when skeletal myocytes were isolated from different tissues of the body.45 Therefore, this study confirms the presence of an early mesoderm stem cell population that has a multipotential differentiation ability. In addition, in vivo transplantation of SB-OG cells showed the ability of the cells to form muscle bundles that expressed high levels of skeletal muscle markers. However, since no lesion was created, the microenvironment of the host skeletal muscle did not fully mature the myoblasts. This is in contrast to cells that were injected subcutaneously under the skin and demonstrated maturation of myocytes containing sarcomeres. These, in addition, contained a pool of Pax7⁺ MPCs. These data indicate that hESC-derived MPCs can provide a safe cell population for transplantation that not only repairs the damaged tissue but also reverts to a resting MPC population. Future studies with a muscle-injury model are required to assess the therapeutic potential of SB-OG cells.

We demonstrated stromal (skeletal) stem cell (MSC) differentiation of SB-OG cells by the use of FBS in the culture medium. Our data corroborate the recent findings published by Hashimoto and colleagues,9 who have shown that human adult skeletal muscle–derived myogenic progenitors expressing known myogenic markers such as Pax7 and CD34 have the ability to differentiate ex vivo into pure MSC-like cells expressing key marker proteins for MSCs (eg, vimentin, CD44, CD73, CD146, and CD166) in combination with the key osteogenic transcription factor Runx2. This finding was confirmed by microarray analysis of hESC-derived MSCs that clearly showed downregulation of all skeletal muscle and myocyte genes (Table 1) and upregulation of skeletal- and mesodermal-specific genes (Supplemental Fig. 3 and Supplemental Table 4). In addition, these hESC-derived MSCs possess the ability to differentiate into mature adipocytes and osteoblasts. Previous studies have demonstrated the ability of hESCs to generate mesenchymal precursors that differentiated to adipocytes and osteoblasts using a number of methodologies. Barberi and colleagues55 used the approach of coculturing of hESCs with mouse stromal OP9 cells followed by FACS sorting to achieve 5% of CD73⁺ cells. Oliver and colleagues81 have shown that isolation and culture of spontaneously differentiated cells from the periphery of the hESC colonies led to differentiation of hESCs in serum-containing medium into an MSC-like population.

Lineage progression under the experimental protocol described here is illustrated in a flow diagram showing different stages of hESC differentiation: first into meso/ectoderm and further to muscle cells and finally to MSC-like cells that differentiate into adipocytes and osteoblasts (Fig. 6). The protocol described here has several advantages over those published previously. It generates differentiated cells based on manipulating a specific signaling pathway and is, additionally, under defined culture conditions. The phenotype of differentiated cells was demonstrated, in contrast to other studies, in vivo based on muscle, bone, or cartilage formation. Additionally, the differentiated cells did not form teratoma and responded to the microenvironmental signals that enhanced their maturation. Thus this procedure can be adapted for generating clinical-grade cells suitable for therapy. Functional studies remain to be performed to prove the ability of the cells, generated ex vivo, to be engrafted and repair damaged tissues in vivo.10

Figure 6. Schematic diagram for the protocol used in this study to direct the differentiation of hESCs to multipotent mesodermal stem cells.

Disclosures

All the authors state that they have no conflicts of interest.

Acknowledgements

We thank Lone Christiansen for excellent technical assistance with the immunohistochemistry. This study was supported by a grant from the Lundbeck Foundation, a grant from the Velux Foundation, and a grant from the region of Southern Denmark. AM has received a PhD fellowship from the Danish Postgraduate School of Molecular Metabolism, University of Southern Denmark.