Mesenchymal stem cells (MSCs) are multipotent progenitors that can be found in many connective tissues, including fat, bone, cartilage, and muscle. We report here a method to reproducibly differentiate human embryonic stem cells (hESCs) into MSCs that does not require the use of any feeder layer. The cells obtained with this procedure are morphologically similar to bone marrow MSCs, are contact-inhibited, can be grown in culture for about 20 to 25 passages, have an immunophenotype similar to bone marrow MSCs (negative for CD34 and CD45 and positive for CD13, CD44, CD71, CD73, CD105, CD166, human leukocyte antigen [HLA]-ABC, and stage-specific embryonic antigen [SSEA]-4), can differentiate into osteocytes and adipocytes, and can be used as feeder cells to support the growth of undifferentiated hESCs. The ability to produce MSCs from hESCs should prove useful to produce large amounts of genetically identical and genetically modifiable MSCs that can be used to study the biology of MSCs and for therapeutic applications.
Mesenchymal stem cells (MSCs) are multipotent progenitors that can be found in many connective tissues including fat, bone, cartilage and muscle. Because of this differentiation potential, MSCs are generally considered to have a large therapeutic potential, particularly in the areas of cell therapy and regenerative and reconstructive medicine [1, , , –5].
MSCs were first recognized as distinct cell populations in the 1990s and were isolated by a variety of procedures. Cell preparations with similar morphology and properties have been termed bone marrow stromal cells, mesenchymal stem cells, skeletal stem cells, or multipotent adult progenitor cells by different investigators. Although these different names may designate cells with different differentiation potential, for the sake of simplicity, we will use the term mesenchymal stem cells in a generic manner in this report.
MSCs are very rare cells that were first isolated from the bone marrow. The classic method to isolate MSCs from bone marrow relies on their capacity to adhere to plastic and their resistance to trypsinization during passage in culture in simple culture medium (i.e., Dulbecco's modified Eagle's medium [DMEM] plus 10% fetal calf serum). Lodie et al. have characterized human bone marrow MSCs prepared by four different methods and found only subtle differences between the different preparations . In addition to bone marrow, MSCs have been isolated from multiple organs, including synovial fluid, cartilage, deciduous teeth, cord blood, amniotic fluid, placenta, and adipose tissue [7, , , , , , –14]. Recently, MSCs have also been obtained by differentiation of hESCs . Since no unique marker of MSCs has been identified, investigators have relied on a series of functional and morphological criteria to identify them. These criteria include growth on plastic, resistance to trypsin, presence of specific cell surface antigens, and potential to differentiate into adipocytes, chondrocytes, and osteocytes [16, –18]. We report here a novel method to reproducibly differentiate H1 (WA01)  human embryonic stem cells (hESCs) into MSCs.
Materials and Methods
Culture of hESCs
The H1 hESCs were cultured on γ-irradiated (80 Gray) feeder cells (mouse embryonic fibroblasts [MEFs] or hES cell-derived MSCs [ES-MSCs]) plated at 75,000 cells per cm2 at 37°C and 5% O2 and 7.5% CO2. hESCs medium contained DMEM/Ham's F-12, 20% Knockout Serum Replacer (KSR), 2 mM l-glutamine, minimal essential medium nonessential amino acid solution (NEAA), 0.1 mM penicillin-streptomycin 1% (all from Gibco, Grand Island, NY, http://www.invitrogen.com), 4 ng/ml basic fibroblast growth factor l (R&D Systems Inc., Minneapolis, http://www.rndsystems.com; or ProSpect-Tany, Technogene, Rehovot, Israel, http://www.prospec.co.il) and 0.1 mM 1-thioglycerol (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com). The culture medium was changed daily, and the cells were passaged once a week.
Microarray design, RNA extraction, probe labeling, and data processing and analysis were as described in Rybicki et al. .
Fluorescence-Activated Cell Sorting Analysis
Cells were harvested using 0.05% trypsin-0.53 mM EDTA (Gibco) and after neutralization resuspended in staining buffer (Dulbecco's phosphate-buffered saline [DPBS] + 5% KSR) at a concentration of 106 cells per ml.
For each antibody used, 105 cells were stained. Antibodies for CD13, CD71, CD105, human leukocyte antigen (HLA)-ABC, isotype control for IgG1, IgG2a, IgG3, and fluorescein isothiocyanate (FITC) rat anti-mouse IgG(H+L) were from eBioscience; CD34 FITC, CD45PE, CD73PE, isotype control IgG1K FITC, IgG1K PE, and FITC rat anti-mouse IgG1 were from BD Biosciences (San Diego, http://www.bdbiosciences.com); CD44, SSEA-4, and TRA 1-85 were from the Developmental Studies Hybridoma Bank (Iowa City, IA, http://www.uiowa.edu/∼dshbwww). At least 10,000 events were acquired for each sample using a FACSCalibur (BD Biosciences). Dead cells were gated out using propidium iodide staining (1 μg/ml).
Functional Differentiation of ES-MSCs
To induce osteogenic differentiation, cells were seeded at 3,000 cells per cm2 in D10 medium (DMEM with 10% fetal bovine serum [FBS]). At 50%–70% confluency, growth medium was supplemented with 100 nM dexamethasone (Decadron; Merck & Co., Whitehouse Station, NY, http://www.merck.com), 50 μM ascorbic acid-2-phosphate, and 10 mM β-glycerophosphate (all from Sigma-Aldrich). The medium was replaced every 3–4 days for 21 days. Cultures were washed twice with phosphate-buffered saline (PBS), fixed in a solution of ice-cold 70% ethanol for 1 hour, and stained for 10 minutes with 1 ml of 40 mM Alizarin red (Sigma-Aldrich) .
For adipogenic differentiation, cells were seeded at 104 cells per cm2, and two methods were used. In the first, at confluence, cells were put in D10 medium supplemented with 1 μM dexamethasone, 0.2 mM indomethacin, 10 μg/ml insulin, and 0.5 mM 3-isobutyl-1-methyl-xanthine (all from Sigma-Aldrich). In the second, at confluence, cells were put in hES medium in partial hypoxia (5% O2). In both cases, medium was replaced every 3–4 days for 21 days. Cells were washed three times with PBS, fixed in 10% formalin for 1–2 hours, and stained for 15 minutes with fresh oil red O solution (Sigma-Aldrich) .
Total RNAs were isolated using RNeasy Lipid Tissue Mini Kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) and treated with DNase I (Roche Molecular Biochemicals, Manheim, Germany, http://www.roche-applied-science.com). Total RNA isolated from normal human adipose tissue (a gift from Dr. M. Hawkins) was used as a positive control for adipocyte-specific gene expression. Reverse transcription was performed with Superscript III first strand kit (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) with approximately 5 μg of RNA and random decamer primers. For each sample, 20-μl reactions were set up in duplicate, each with 10 μl of Quantitect SYBR Green PCR Kit (Qiagen), 0.5 pmol of each primer, and 2 μl of cDNA (equivalent of 50 ng of RNA). In the case of Oct-4 detection, detection was performed using the single-tube Quantitect SYBR Green reverse transcription-polymerase chain reaction (RT-PCR) kit (Qiagen).
The following primers were used (gene symbol is given in square brackets).
To control for the amount of cDNA in each reaction, β2M was used as an internal standard. The amplifications were performed on a Roche LightCycler using three-temperature cycling consisting of denaturation step at 95°C for 8 seconds, an annealing step at 58°C for 18 seconds, and an extension step at 72°C for 20 seconds. At the end of each reaction, melting curves were acquired during cooling from 95°C to 65°C. Levels of expression were normalized to the level of B2M expression in each cells and to the level of expression in the CD45-positive cell controls using the following formula: 2exponent (–(Ct[tested gene in tested cells] – Ct[B2M in tested cells])/ (Ct[tested gene in CD45 cells]– Ct[B2M in CD45 cells]), where Ct is the cycle threshold at which PCR products were first detected.
Spectral karyotyping was performed as described before .
Population doubling time was calculated following the formula (logN/log2)/t, where N is the number of cells at confluence divided by the initial number of cells, and t is the number of hours in culture. The growth curve was established by multiplying the initial number of cells by the amplification fold for each passage.
Immunostaining with SSEA-4 Antibody
hESC colonies were fixed in situ with formalin for 15 minutes. After three rinses with PBS, cells were incubated overnight at 4°C with mouse anti-human SSEA-4 antibody (DHSB) or isotype control (eBioscience, San Diego, http://www.ebioscience.com) (approximately 2 μg per 106 cells) in PBS + 5% KSR. Antibody localization was performed using rat anti-mouse IgG (H+L) immunoglobulin conjugated to fluorescein isothiocyanate (eBioscience).
Differentiation of hESCs into hematopoietic cells was performed as previously described . Briefly, hESCs were cocultured with fetal hepatocytes (FH-B-hTERT)  in DMEM + 20% FBS + 1% penicillin/streptomycin + 1% nonessential amino acids. Half of the medium was replaced every 3 days. After 2 weeks, cells were harvested and plated on methylcellulose. Colonies were counted 17 days later.
The Raclure Method
Spontaneously differentiated cells often appear in hESC cultures in the center or at the edges of the colonies (Fig. 1A). These cells are mechanically eliminated from the culture before each passage because they tend to promote the differentiation of the undifferentiated cells remaining in the dish. We observed that these scraped cells (“raclures” in French) remain multipotent since after replating they can differentiate into multiple cell types, including neurons and cardio-myocytes, depending on the culture conditions (data not shown). We report here a procedure to differentiate these raclures into MSCs.
To obtain MSCs, raclures are first plated in D10 medium (DMEM + 10% FBS + 1% penicillin/streptomycin + 1% nonessential amino acids) until a thick, multilayer epithelium of cells develops. This requires at least 4 weeks, with weekly medium changes. Once the epithelium is formed, it can survive in culture for many months with minimal medium change (once every 3 weeks). The MSCs are then isolated by dissociation of this epithelium using a mixture of trypsin, collagenase type IV, and dispase for 4–6 hours and replating of the cells in D10 medium. Two or 3 days after the passage, the population of cells obtained is morphologically uniform, and cells exhibit a characteristic regular fingerprint-like morphology when they are observed at confluence by phase-contrast microscopy (Fig. 1B). The resulting cultures can then be passaged at least 20 times.
Since the hESCs were cultured on MEFs, it was important to determine whether MEFs constituted a significant proportion of the raclures. To assess the percentage of irradiated MEFs that were scraped with the raclures, fluorescence-activated cell sorting analysis was performed on the day of the scraping and 1 week later using TRA 1-85, a human-specific antibody, to differentiate the MEFs from the human cells. This analysis revealed that about 5% of the raclures consisted of cells of murine origin and that a week after plating the raclures, more than 99.5% of the cells were human (data not shown). As discussed below, in the established MSC cultures, no cells of mouse origin could be detected, suggesting that the MEFs present in the raclures were completely mitotically inactivated by the irradiation.
The procedure to isolate MSCs is reproducible, since 15 independent experiments yielded similar populations of cells. The dissociation of the epithelium could be performed from 50 to more than 180 days after initial plating of the raclures without noticeable differences in the type of cells obtained. We report here on the detailed characterization of two of these isolates, P37R and P51R.
P37R was produced by dissociation of the epithelium 180 days after plating of raclures of H1 cells at passage 37. P51R was produced by dissociation of the epithelium 50 days after plating of raclures of H1 cells at passage 51. The P37R cells are small and spindle-shaped (Fig. 1C). The P51R cells are larger (Fig. 1C). P51R is strictly contact inhibited and forms a monolayer. P37R cells are also contact inhibited but form a bilayered or trilayered epithelium. The morphology of the other isolates varied between these two extremes.
To determine the growth characteristics of the P37R and P51R isolates, cultures at passage 5 were trypsinized every 3 days, counted, and replated at 10,000 cells per ml. As shown in Figure 2A, the growth rate of the cells started to decrease at passage 17, and the cultures reached senescence between passage 25 and passage 30.
Spectral karyotype analysis of the P37R and P51R isolates at passage 13 revealed no abnormalities, suggesting that the karyotype of these cells is stable (data not shown).
As a first step toward the characterization of the P37R cells, we performed preliminary expression analysis using a custom-made cDNA glass slide micro-array produced by the Albert Einstein College of Medicine Micro-Array Facility. These arrays contain 6,000 human cDNAs. Two comparisons were performed. The first comparison was between MSC-P37R cells and undifferentiated hESCs. The second was between the MSC-P37R cells and mesenchymal cells termed human embryonic palatal mesenchymal (HEPM) that we obtained from the American Type Culture Collection (Manassas, VA, http://www.atcc.org) . More than 40% of the spots on the array were upregulated or downregulated at least 1.5-fold when the MSC-P37R cells were compared with undifferentiated hESCs. By contrast, this number was only 12% when the MSC-P37R were compared with the HEPM cells, demonstrating that MSC-P37R resemble HEPM cells much more than undifferentiated hESCs.
We then examined the most highly expressed genes in the MSC-P37R cells as compared with the hESCs and found that out of the 25 most highly expressed genes in the MSC-P37R, 15 had been identified previously in a serial analysis of gene expression analysis [28, 29] as genes that were highly expressed in bone marrow and cord blood mesenchymal cells, suggesting that the hESC-derived cells had a profile of expression similar to primary MSCs. Table 1 illustrates genes in the hESC-derived MSC-P37R cells that are the most highly induced compared with hESCs.
Table Table 1.. Most highly expressed genes in P37R as compared to hESCs
Flow Cytometry Analysis
MSCs isolated by different methods present a somewhat variable profile of antigen expression and share many characteristics with endothelial, epithelial, and muscle cells, complicating the task of finding a simple identifying marker. Nevertheless, there is a general consensus that MSCs are negative for CD45 and CD34 but positive for markers such as SH2, SH3, and SH4 [1, 17]. To characterize the surface antigen profiles of P37R and P51R, we used the above-mentioned antibodies plus others that have been reported to be expressed on only some MSC populations. As controls, we used the HEPM cell line. The results of this analysis are shown in Figure 3. P37R, P51R, and HEPM were positively stained for CD44, CD71 (transferrin receptor), CD73 (SH3), CD105 (endoglin, SH2), CD166, and HLA-ABC and negative for CD34 and CD45 (Fig. 3). Staining for CD13 revealed a very weak expression of this epitope for P37R and P51R as compared with HEPM. Interestingly, staining for SSEA-4 , an antigen expressed on hESCs and in early human embryos, was high for P37R and P51R but negative for HEPM. The human-specific antibody TRA 1-85  was used to exclude the possibility that the cells were not of human origin. The immunophenotypes of P37R and P51R were therefore compatible with the hypothesis that these cells were MSCs. To determine whether there was any contamination by cells of murine origin that could originate from MEFs that would have survived the irradiation, we also tested the MSCs with TRA 1-85, a human-specific antibody. No cells of mouse origin could be detected (Fig. 3).
Functional Differentiation of P37R and P51R
To confirm that P37R and P51R were MSCs, we performed functional differentiation assays. We focused on osteogenesis and adipogenesis since MSCs derived from multiple organs have the capacity to differentiate along these pathways,
Osteogenic differentiation was performed using the β-glycerophosphate method as previously described for primary adult MSCs . After 3 weeks of differentiation, a majority of the cells had differentiated into osteoblasts, as demonstrated by calcium deposition in the matrix visualized with alizarin red staining (Fig. 4A). HEPM cells were used as a positive control .
Adipogenic differentiation was performed by two different methods. At first, attempts were made by the classic 3-isobutyl-1-methylxanthine (IBMX) method [17, 21, 33]. This resulted in adipocytic differentiation, but the lipid vesicles obtained were small (Fig. 4B). Much larger vesicles were obtained using an alternate method of adipogenic differentiation that we developed in the laboratory.
This new method relies on the serendipitous observation that P37R cells grown in serum-free medium (KSR + DMEM/F-12 + l-glutamine + NEAA) accumulates small cytoplasmic vesicles that are lightly stained by oil red O, on the finding that hypoxia enhances lipid accumulation , and on the report that FGF enhances PPAR-γ ligand-induced adipogenesis of MSC . As seen in Figure 4B, incubation of MSCs in these conditions resulted in the differentiation of most of the culture into cells that contain large cytoplasmic vesicles that are brightly stained with oil red O. Another advantage of this new serum withdrawal/hypoxia (SWH) method is that it is much less sensitive to initial plating density (data not shown) .
To characterize the adipocytes produced from hESC-derived MSCs, we performed a real-time RT-PCR analysis on RNA extracted from MSC-P37R and MSC-P51R either undifferentiated or differentiated using the IBMX and SWH methods. As controls, we also tested undifferentiated hESCs, purified adult hematopoietic cells (CD45+), human primary breast adipocytes, and HEPM cells. Primers specific for transcription factors involved in adipogenesis (PPAR-γ2 and SREBf1c), for proteins involved in lipid droplets (perilipin and adipophilin) and lipid metabolism (lipoprotein lipase and GAPDH), or for cytokines produced by adipocytes (adiponectin, PGAR, and leptin) were used as described by Fink et al. .
This analysis revealed that both methods induce the expression of adipocytic markers but that there were differences between the two differentiation protocols (Fig. 5).
In the case of MSC-P37R, PPAR-γ2, a highly specific adipocyte marker, was expressed at relatively high levels (5% of fully mature adipocytes) prior to differentiation. The SWH method led to further induction of PPAR-γ2 to approximately 15% of the level in mature adipocytes. The SWH protocol also led to a large increase in the level of perilipin, led to a moderate increase in adipophilin production, and had little or no effect on SREBf1c expression. By contrast, the IBMX method led to a small decrease in PPAR-γ2 expression but to dramatic increases in SREBf1c, adipophilin, and PGAR expression, three genes that are expressed at high levels in adipocytes. Neither treatment induced the production of adiponectin, leptin, or lipoprotein lipase. Results with MSC-P51R were similar except that the induction of PPAR-γ2, perilipin, and adipophilin were not as strong and that trace amounts of lipoprotein lipase could be detected.
We conclude from the oil red O stain and from these expression data that both the P37R and the P51R isolates have adipocytic potential, that the SWH method is more efficient than the IBMX method for the induction of adipocytic differentiation, and that the two methods lead to slightly different types of adipocytes or to adipocytes at different stage of maturation.
hESC-Derived MSCs Support the Growth of Undifferentiated hESCs
Recent reports have shown that bone marrow-derived MSCs can support the growth of hESCs  and hematopoietic stem cells (HSCs) . To test whether hESC-derived MSCs can support the growth of hESCs, undifferentiated H1 cells were passaged on either irradiated P51R, irradiated P37R, or irradiated MEFs, and the resulting H1 cells were compared by a variety of assays. After 32 successive passages, the H1 cells had similar undifferentiated morphologies regardless of the feeder used (Fig. 6A, 6B). Immunostaining with SSEA-4  antibodies (Fig. 6C–6H) and real-time RT-PCR analysis for Oct-4 (Fig. 6I) revealed that the level of expression of these markers, which are known to be expressed at high levels in H1 cells, was high regardless of the feeder used.
To functionally characterize H1 cells grown on P37R and P51R, we differentiated them into hematopoietic cells by coculture with FH-B-hTERT (a human fetal hepatocyte cell line) cells as described in Qiu et al. . After 2 weeks of coculture, the percentage of CD34-positive cells was determined by flow cytometry. These experiments revealed that similar numbers of CD34-positive cells were produced whether MEF, P37R, or P51R were used as feeders (Fig. 6J). Methylcellulose assays confirmed and extended these results since they revealed that hESCs grown on the P51R feeders yielded erythroid and myeloid colonies in relatively large numbers (Fig. 6J; ). Together, these data show that H1 cells grown on P37R and P51R retained their morphology, the expression of hESC markers, and their ability to differentiate into hematopoietic cells. These data suggest that these feeders are able to support the growth of undifferentiated H1 cells that can be differentiated in hematopoietic cells. Additional studies will be required to determine whether hESCs grown on P37R and P51R retain their ability to differentiate into other cell types.
We describe here a simple method to derive bipotent MSCs from hESCs. The cells obtained grow very robustly, have a stable karyotype, are contact inhibited, senesce after about 20 passages, have an immunophenotype similar to that of MSCs derived from other sources, have adipocytic and osteocytic potential, and can support the growth of hESCs and hematopoietic progenitors.
The cells that we obtained appear similar to the hESC-derived MSCs described recently by the Studer laboratory , but our method has the advantage of not requiring any feeder layer of animal origin that could complicate the use of these cells for clinical purposes. Whether the bipotency of the P37R and P51R isolates reflects a true bipotency of individual cells or the presence of two types of progenitor cells in the isolates remains unproven since we have not yet characterized clonal population of cells. However, microscopic observations showing an almost complete differentiation into adipocytes favor the hypothesis that the P37R and P51R cells are truly bipotent.
Derivation of mesenchymal cells from human ES cells should help us understand the specification events that occur during early human development and could have useful clinical applications, since these cells have a large therapeutic potential, particularly in the areas of cell therapy and regenerative and reconstructive medicine.
From an experimental point of view, derivation of MSCs from hESCs has specific advantages over derivation from other sources. In the first place, a virtually unlimited amount of genetically identical MSCs can be produced from hESCs. Second, the ability to genetically manipulate the hESCs before the production of the MSCs should provide unparalleled experimental opportunities to study the molecular basis of the totipotency of the MSCs cells, as well as pathways that lead to human osteogenic and adipocytic differentiation. Finally, the ability of the hESC-derived MSCs to support the growth of hESCs is also of practical importance [39, 40] since they represent an almost unlimited source of autogenic feeder that eliminates the risks of transmission of xenogeneic pathogens and since they can be conveniently cultured for more than 20 passages, allowing the production of large amounts of cells.
The authors indicate no potential conflicts of interest.
E.E.B. is supported by NIH Grants R01-DK56845, P01-HL55435, and P20-GM075037. A.C.R. is supported by NIH Grants HL-68962 and HL-38655. The monoclonal antibodies for CD44, SSEA-4, and TRA1-85 (developed by J.T. August and J.E.K. Hildreth, D. Solter, and P.W. Andrews, respectively) were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development and maintained by the Department of Biological Sciences, University of Iowa, Iowa City, IA. We thank Dr. C. Montagna of the Albert Einstein College Of Medicine Genome Imaging Facility for the spectral karyotyping analyses, Dr. Meredith Hawkins and Dr. Wei Jie for the generous gift of RNA from human primary adipocytes, and Dr. Philip Scherer for useful discussions.