Committing Embryonic Stem Cells to Early Endocrine Pancreas In Vitro


  • Hsun Teresa Ku,

    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    Search for more papers by this author
  • Nan Zhang,

    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    Search for more papers by this author
  • Atsushi Kubo Ph.D.,

    Corresponding author
    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    • Box 1496, Mount Sinai School of Medicine, New York, New York 10029-6574, USA. Telephone: 212-659-8244; Fax: 212-849-2437
    Search for more papers by this author
  • Ryan O'Connor,

    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    Search for more papers by this author
  • Minwei Mao,

    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    Search for more papers by this author
  • Gordon Keller,

    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    Search for more papers by this author
  • Jonathan S. Bromberg M.D., Ph.D.

    1. Department of Gene and Cell Medicine and the Recanati/Miller Transplantation Institute, Mount Sinai School of Medicine, New York, New York, USA
    Search for more papers by this author


A panel of genetic markers was used to assess the in vitro commitment of murine embryonic stem (ES) cells toward the endoderm-derived pancreas and to distinguish insulin-expressing cells of this lineage from other lineages such as neuron, liver, and yolk sac. There are two nonallelic insulin genes in mice. Neuronal cells express only insulin II, whereas the pancreas expresses both insulin I and II. Yolk sac and fetal liver express predominately insulin II, small amounts of insulin I, and no glucagon. We found that ES-derived embryoid bodies cultured in the presence of stage-specific concentrations of monothio-glycerol and 15% fetal calf serum, followed by serum-free conditions, give rise to a population that expresses insulin I, insulin II, pdx-1 (a pancreas marker), and Sox17 (an endoderm marker). Immunohistochemical staining shows intracellular insulin particles, and its de novo production was confirmed by staining for C-peptide. Most, but not all, of the insulin+ or C-peptide+ cells coexpress glucagon, demonstrating a differentiation pathway to pancreas rather than yolk sac or fetal liver. Addition of β-cell specification and differentiation factors activin β B, nicotinamide, and exendin-4 to later-stage culture increased insulin-positive cells to 2.73% of the total population, compared with the control culture, which gave rise to less than 1% insulin-staining cells. These findings suggest that stepwise culture manipulations can direct ES cells to become early endocrine pancreas.


Type I diabetes is marked by a deficiency of endocrine β cells in the pancreatic islets of Langerhans. Daily injection of insulin is the current treatment for the disease. Because insulin injection cannot match the precise timing and dosing of physiological secretion of insulin by islets in response to hyperglycemia, severe side effects develop over time. Transplantation of islets represents a potential cure; however, limitations on the availability of cadaveric organs restrict this procedure to only a small percentage of patients. Pancreatic endocrine stem cells exist in the developing embryonic pancreas. The isolation, expansion, and differentiation of pancreatic endocrine stem cells would provide an unlimited source of islet cells for transplantation and treatment for type I diabetes. A potential source for endocrine stem cells is embryonic stem (ES) cells. Because of their unlimited proliferation and differentiation potentials, ES cells are considered an important source for cell therapies targeted to several diseases, including diabetes.

There are two nonallelic insulin genes (insulin I and II) expressed in multiple sites in mice during development [17]. Neuronal cells [2] express only insulin II, whereas pancreas expresses both insulin I and II. Fetal liver and yolk sac, which lack the expression of glucagon [6], express predominately insulin II [3]. Insulin I is also expressed in fetal liver and yolk sac but at much lower levels [3, 4]. The transcription factor pdx-1 is essential for the earliest stage of pancreatic development [810]. Tissue recombination studies have shown that the endocrine pancreas is derived from endoderm [1113]. Thus, insulin I, when combined with other lineage markers, can be used to trace the development of pancreatic β cells.

In this study, we define culture conditions in which pancreatic lineage insulin-expressing cells can be distinguished from those derived from nonpancreatic origin using gene expression profiling. This culture system provides the foundation for future identification, characterization, and manipulation of pancreatic endocrine stem cells in vitro.

Materials and Methods

Growth and Differentiation of ES Cells

Mouse ES lines R1, E14.1, and CCE [14] were maintained in an undifferentiated state as described previously [15]. In short, ES cells were maintained on irradiated feeder layers in Dulbecco's modified Eagle medium (DMEM) supplemented with 15% fetal calf serum (FCS) (Gemini Bio-Products Inc., Woodland, CA; selected batch), penicillin, streptomycin, 4% conditioned media containing leukemia inhibitory factor (LIF), and 3 × 10−4 M monothioglycerol (MTG; Sigma, St. Louis). Feeder cells were obtained by trypsinization of E14 embryos from which livers were removed. Before the onset of differentiation, ES cells were transferred and grown on gelatinized six-well plates for two rounds to deplete feeder cells. For the generation of embryoid bodies (EBs), ES cells were trypsinized into a single-cell suspension, washed twice, and plated at 5,000 (R1) or 2,000 (E14.1 and CCE) cells/ml into differentiation media containing Iscove's modified Dulbecco's medium (IMDM; Mediatech, Inc., Herndon, VA), 50 μg/ml ascorbic acid (Sigma), 6 × 10−3 M MTG, and 15% FCS. EBs were maintained in 60 × 15-mm Petri dishes (VWR Scientific Products, West Chester, PA) to avoid adherence. On day 2 of differentiation, EBs were washed twice in IMDM containing 0.1% bovine serum albumin (Sigma) and transferred to differentiation media containing IMDM, 50 μg/ml ascorbic acid, 6 × 10−4 M MTG, and either 15% FCS or knockout serum replacement (SR) (Invitrogen, Carlsbad, CA).

On day 6 of EB differentiation, whole EBs were washed, counted, and plated to tertiary culture in six-well plates, with or without precoating of 0.1% gelatin. The basic tertiary culture medium contained DMEM/F-12 (1:1) (Mediatech, Inc.) and 15% SR, with or without 10 ng/ml purified recombinant human basic fibroblast growth factor 2 (FGF2; R&D Systems, Minneapolis). The first change of media occurred on the fifth day of the tertiary culture to allow attachment of most EBs onto the plates. Media was changed every 3–4 days thereafter. In a typical experiment, a total of 25 serum (S) EBs (average, 1.88 ± 0.67 × 105 cells/25 S EBs) were seeded per well. Over a 13-day culture period, cell numbers increased 2.72 ± 0.32-fold/well per 25 S EBs. At the end of tertiary culture, cells from three wells were collected and combined for subsequent analysis. Nicotinamide (Sigma), exendin-4 (Sigma), and activin β B (R&D Systems) were used at the concentration of 10 mM, 0.1 nM, and 10 ng/ml, respectively. Conditioned media for LIF was obtained by culturing Chinese hamster ovary cells that were transgenic for LIF, as described previously [16].

RNA Isolation and Reverse Transcriptase–Polymerase Chain Reaction Analysis

Cells from tertiary culture were incubated with 0.25% trypsin-EDTA for 3 minutes at 37°C, washed, and incubated with 2 mg/ml collagenase D (Roche Diagnostics, Basel, Switzerland) and 10 μg/ml deoxyribonuclease I (Calbiochem, San Diego) for 30 minutes at 37°C to produce single-cell suspension. Total RNA was extracted from cells using RNeasy Mini Kit (Qiagen, Courtaboeuf Cedex, France), and genomic DNA was digested with deoxyribonuclease I (Invitrogen). Equal amounts (1 μg) of RNA were used for reverse transcription (Invitrogen or Qiagen). One twentieth of the cDNA mixture was used for all genes, except one fortieth for preproinsulin II and Pax6 amplification. Oligonucleotides used as amplimers for the polymerase chain reaction (PCR), product sizes, and cycle numbers are summarized in Table 1.

Table Table 1.. List of gene-specific primers
GeneForward primerReverse primerCycle #Product size (bp)
Albumin 15′ gctacggcacagtgcttg 3′5′ caggattgcagacagatagtc 3′35265
Amylase 25′ ttctgctgctttccctcatt 3′5′ cattgttgcaccttgtcacc 3′40299
Brachyury5′ catgtactctttcttgctgg 3′5′ ggtctcgggaaagcagtggc 3′30312
Flk-15′ cacctggcactctccaccttc 3′5′ gatttcatcccactaccgaaag 3′30239
GAPDH5′ tgaaggtcggtgtgaacggat 3′5′ caggggggctaagcagttggt 3′30470
GATA-15′ cattggccccttgtgaggccagaga 3′5′ acctgatggagcttgaaatagaggc 3′30288
Glucagon5′ cattcacagggcacattcacc 3′5′ ccagcccaagcaatgaattcc 3′40206
HNF-3β5′ tggtcactggggacaagggaa 3′5′ gcaacaacagcaatagagaac 3′30288
Insulin I5′ tagtgaccagctataatcagag 3′5′ acgccaaggtctgaaggtcc 3′38288
Insulin I (second pair)5′ ggagcccaaacccacccagg 3′5′ cactgatccacaatgccac 3′38227
Insulin II5′ ccctgctggccctgctctt 3′5′ aggtctgaaggtcacctgct 3′40212
nestin5′ ttcccccttgcctaataccct 3′5′ tacctctgtggctgcttcttt 3′30444
nkx2.25′ aagagccctttctacgacagc 3′5′ gcgtcacctccatacctttct 3′30402
Pax65′ ccatctttgcttgggaaatccg 3′5′ gcttcatccgagtcttctccgttag 3′30312
Pdx-15′ ccaccccagtttacaagctc 3′5′ tgtaggcagtacgggtcctc 3′40324
Rex-15′ cgtgtaacatacaccatccg 3′5′ gaaatcctcttccagaatgg 3′35128
Sox175′ gccaaagacgaacgcaagcggt 3′5′ tcatgcgcttcacctgcttg 3′35210

Quantitative Real-Time PCR

Real-time PCR was performed on a Light Cycler (Roche) in glass capillaries using the Quanti Test SYBR Green PCR kit (Qiagen) as described by the manufacturer. In the same PCR run, each sample was tested with primers specific for insulin I and normalized with the housekeeping gene cyclophilin A. The sequences of primers for cyclophilin A are 5′ agggtggt-gactttacacgc 3′ and 5′ atccagccattcagtcttgg 3′. The results are expressed as the relative expression level compared with cyclophilin A.


Day-6 EBs were grown on cover slides that were precoated with 0.1% gelatin. Cells were fixed in 4% paraformaldehyde and 0.15% picric acid in phosphate buffer, pH 7.5, for 30 minutes and then washed with phosphate buffer twice. Cell membranes were permeablized with 0.1% Triton-X100 (Pierce, Rockford, IL) in phosphate buffer at room temperature for 5 minutes. Protein block serum-free (Dako, Carpinteria, CA) solution was added to inhibit nonspecific staining of the cells. Guinea pig anti-swine insulin and rabbit anti-human glucagon antibodies were purchased from Dako. Guinea pig anti-rat C-peptide II serum (recognizes both C-peptide I and II) was purchased from Linco Research (St. Charles, MO). Control nonimmune guinea pig and rabbit sera were purchased from Jackson Immuno Research Laboratories (Bar Harbor, ME). Primary antibodies were applied to samples and incubated at room temperature for 1 hour. Cy3-conjugated donkey anti-guinea pig and Cy2-conjugated donkey anti-rabbit secondary antibodies were purchased from Jackson Immuno Research Laboratories. Dilution of the antibodies was made according to the manufacturer's recommendations. Samples were washed with phosphate buffer twice before secondary antibodies were applied and incubated at room temperature for 1 hour. Samples were washed again with phosphate buffer twice and subjected to fluorescence microscopy analysis using a Leica DMRA2 microscope. Images were captured as grayscale pictures, and colors were painted subsequently using Openlab software (Improvision Inc., Lexington, MA).

Flow Cytometric Analysis

Cells from tertiary culture were made into a single-cell suspension as described above. Cells were fixed in 4% para-formaldehyde/0.15% picric acid and stained with anti-insulin and anti-glucagon antibodies as described for immunocytochemistry. Nonimmune sera were used for control staining. Cells were analyzed using a FACSCalibur (Becton, Dickinson, San Jose, CA).

Statistical Analysis

Student's t-test was used to determine statistical significance.


Effects of Serum and Serum-Free Culture Conditions on EB Differentiation

Because serum is the primary source of inducing factor in ES cell differentiation cultures, we chose to study the effects of serum and serum-free conditions during EB differentiation. Figure 1 depicts the standard conditions of our culture system. EBs that were grown for 6 days in the presence of serum were designated as S EBs, and EBs that were grown for 2 days in serum followed by 4 days in serum-free culture were designated as SR EBs. EBs from three independent ES cell lines were grown in these conditions and collected every 24 hours, and total RNA was prepared and analyzed by reverse transcriptase (RT)-PCR.

Figure Figure 1..

Protocol for differentiation of ES cells to insulin-expressing cells. Abbreviations: EB, embryoid body; ES, embryonic stem; FGF2, fibroblast growth factor 2; MTG, monothioglycerol; S, serum; SR, serum replacement.

To test whether endoderm differentiation occurs in the EBs, the expression of several genes for the three germ layers was assessed. The markers used for definitive endoderm lineages were Sox17 [17] and HNF-3β [18, 19]. The results (Figs. 2A–2C) show that levels and kinetics of Sox17 expression among the different cell lines were comparable and peaked by days 5 or 6 in both SR and S EBs. A different expression pattern, however, was observed for HNF-3β . SR EBs had prolonged expression of HNF-3β from days 3 to 6 in all three ES lines tested. In contrast, in S EBs, HNF-3β expression peaked on day 4 and decreased by day 6. It is known that HNF-3β is expressed by the definitive endoderm in the developing early embryo, as well as in the notochord and the floor plate of the neurotube [19]. Thus, at this early stage, HNF-3β may be an indicator for both neuronal and endoderm differentiation. Sox17 expression is restricted to visceral and definitive endoderm [17]. The pattern of Sox17 expression suggests that the endoderm program is induced in both S and SR EBs.

Figure Figure 2..

EB differentiation in the presence or absence of serum. Three different ES lines, R1 (A), E14.1 (B), and CCE (C), were cultured in the absence of leukemia inhibitory factor and in the presence of serum for 2 days. Day-2 EBs were transferred to SR media or remained in serum for an additional 4 days. EBs were collected daily, and total RNA was extracted and analyzed for gene expression by reverse transcription–polymerase chain reaction. Abbreviations: EB, embryoid body; ES, embryonic stem; S, serum; SR serum replacement.

For mesoderm lineages, brachyury (T-box gene; mesoderm progenitors) [20], GATA-1 (mesoderm-derived hematopoietic lineages) [21], and flk-1 (endothelial cells) [22] were analyzed. Brachyury expression increased in both sets of conditions; however, GATA-1 and flk-1 were expressed at higher levels in the presence of serum. This result suggests that EBs maintained in serum are enriched for cells of the blood and endothelial lineages compared with EBs transferred to serum-free medium.

For early neuroectoderm lineages, Pax6 [23] and nkx2.2 [24] expression were examined. SR EBs have prolonged expression of both markers, indicating an early induction of the neuroectoderm program. In contrast, expression of both genes was downregulated in S EBs between days 4 and 6 of differentiation. Taken together, these results suggest that EBs cultured in serum-free conditions were relatively enriched for neuroectoderm lineages, whereas serum-containing cultures were enriched for hematopoietic and endothelial lineages. Endoderm is induced in both sets of conditions during EB differentiation.

Insulin I was not expressed in ES cells, SR EBs, or S EBs at these early times (Fig. 3). Pdx-1 was consistently upregulated in S EBs from all three ES cell lines. SR EBs showed variable pdx-1 expression among the experiments (Figs. 2, 3), presumably because of low cell and mRNA copy number in this population. Because the three ES cell lines were similar in their gene expression profiles, the R1 ES cell line was selected for subsequent experiments.

Figure Figure 3..

Generation of insulin I–expressing cells, requirement for integrity of EBs, effects of gelatin coating in tertiary cultures, and specificity of insulin I primers. (A, B):Whole (no Try) day-6 EBs were plated on six-well plates coated with 0.1% gelatin (G+) or left untreated (G–) and cultured in basic serum-free conditions plus 10 ng/ml FGF2 for 11 days. (C): Day-6 S EBs were treated with 0.25% Try/EDTA for 30 seconds, followed by gentle aspiration to generate smaller pieces of EBs (Try 30′'). Total RNA was extracted and analyzed for gene expression by RT-PCR. (D): Pancreatic islets from 12-week-old BALB/c mice, brain from 8-week-old C57BL/6 mice, and S EB–derived population collected on day 13 of the basic tertiary culture (without FGF2) were processed for RT-PCR with the first set of insulin I primers [25] and designated cycle numbers. Abbreviations: EB, embryoid body; ES, embryonic stem; FGF2, fibroblast growth factor 2; G, gelatin; Glu, glucagon; Ins, insulin; RT-PCR, reverse transcription–polymerase chain reaction; S, serum; SR, serum replacement; Try, trypsin.

Emergence of Insulin I–Expressing Cells in Tertiary Culture

To additionally analyze the gene expression patterns of the differentiated cells, both day-6 S and SR EBs were transferred to identical tertiary cultures (Fig. 1) containing serum-replacement media and FGF2 [25] and cultured for 11 days. Total RNA was extracted, and multiplex RT-PCR analysis was performed. Insulin I is expressed after 11 days in tertiary culture. However, it was only detected in cultures established from S EBs (Figs. 3A, 3B). This result was confirmed by the use of a second set of insulin I–specific PCR primers (data not shown). In contrast, insulin II is expressed in both S and SR EB-derived cells. Pdx-1 expression is restricted to cells derived from S EBs. These results suggest that S EB-derived cells may contain pancreatic lineage cells or their progenitors.

To confirm the specificity of the insulin I primers [25] used, expression of insulin I in islets and brain from adult mice was compared (Fig. 3D). It was found that insulin I is only expressed by adult pancreatic islets but not brain, demonstrating that the primers used are specific. In addition, insulin I expression in the S EB-derived cells is in the quantitative range of between 34 and 40 cycles of PCR (Fig. 3D).

Additional analysis revealed that the SR EB-derived population is enriched for nkx2.2- and Pax6-expressing cells (Fig. 3B), whereas the S EB-derived population is enriched for Sox17 and Flk-1. These findings indicate the development of ectoderm in SR EB-derived cells and endoderm and endothelial cells in S EB-derived population, respectively. There is no difference in the expression pattern of glucagon, amylase 2 (a marker for exocrine cells), nestin, HNF-3β, and Rex-1 (a marker for undifferentiated ES cells [26]) between S and SR EB-derived cells in tertiary culture.

Kinetic analysis revealed that higher levels of insulin I expression in S EB-derived cells were achieved after at least 11 days in the tertiary culture (Fig. 4). Pdx-1 expression persisted up to day 15 in S EB-derived cells. In contrast, SR EB-derived cells had minimal or no insulin I expression. Although pdx-1 is expressed in day-6 SR EBs (Figs. 2, 3), by day 7 of tertiary culture, its levels were greatly diminished (Fig. 4). Insulin II expression was maintained throughout the tertiary culture for both S and SR EB-derived cells.

Figure Figure 4..

Kinetics of insulin I, insulin II, and pdx-1 expression in tertiary culture. Day-6 S or SR EBs were initiated in basic tertiary culture conditions (serum-free plus fibroblast growth factor 2). At various time points, total RNA was extracted and analyzed for gene expression by reverse transcription–polymerase chain reaction. Abbreviations: EB, embryoid body; Ins, insulin; S, serum; SR, serum replacement.

The presence of insulin protein in both S and SR EB-derived cells was demonstrated by immunofluorescent staining (Fig. 5) using an antibody that recognizes both insulin I and II. Although under lower magnification both S and SR EB-derived cells demonstrated insulin staining (Figs. 5A, 5B), higher magnification revealed that only S EB-derived cells stained for both insulin and the nuclear-staining 4′,6′-diamidino-2-phenylindole (DAPI; Figs. 5E, 5F). In contrast, insulin staining of SR EB-derived cells was often associated with anucleate cells (Fig. 5I), in agreement with a previous report [27]. Insulin-positive cells were present as groups of approximately four to eight cells (Figs. 5E, 5F) and were scattered around the margin of the EB mass. They tended to be small in size and were distinct from the larger, flat cells that did not stain with the anti-insulin antibody (Fig. 5G). Control nonimmune serum showed the specificity of the insulin staining, and an area of the same small clustered cells is purposely shown to demonstrate this negative staining (Fig. 5H). As another control, porcine insulin was incubated with the anti-insulin antibody. This treatment abrogated insulin staining, demonstrating the specificity of the antiserum (Fig. 5C). Most (Fig. 5D), but not all (Figs. 5E, 5F), of the insulin-stained cells coexpressed glucagon in S EB-derived cultures. Numerous insulin staining particles were detected in SR EB-derived cells (Fig. 5B). However, the fact that almost none of these cells costained with DAPI suggests that this pattern of staining is indicative of dead cells taking up insulin from the media [27]. To demonstrate any live cells that express insulin, SR EB-derived cells were embedded in paraffin and sectioned, and insulin-positive cells with DAPI-staining nuclei could be seen within the EB-derived mass (data not shown). In contrast to the S EB-derived cells, which have an even distribution of insulin throughout cytoplasm, the insulin particles in SR EB-derived cells were more concentrated around the perinuclear space (data not shown). Taken together, the results from gene expression profiling and immunohistochemistry demonstrate that the insulin-expressing cells derived from the two sets of culture conditions likely represent distinct lineages.

Figure Figure 5..

Immunofluorescence staining of total insulin and C-peptide in EB-derived cells. Day-6 S (A, D–G, J, L, M) or SR (B–C, I, K) EBs were plated on gelatin-coated cover slips and cultured in basic tertiary cultures plus fibroblast growth factor 2 for 11–18 days. Cells were fixed in 4% paraformaldehyde/0.15% picric acid and processed for immunofluorescence staining with anti-insulin, anti-glucagon, anti–C-peptide, and control nonimmune sera (total protein concentration matched with respective antisera), as depicted. (C): Porcine pancreatic insulin (Sigma) was added at a final concentration of 100 μg/ml. (N, O): Sections of formalin-fixed, paraffin-embedded adult islets were double stained with anti–C-peptide plus anti-glucagon or control nonimmune sera, respectively. DAPI was used to stain nuclei. Magnification:A–D, J–K, ×100; E–I, L–O, ×1,000. Abbreviations: DAPI, 4′,6′-diamidino-2-phenylindole; EB, embryoid body; GP, guinea pig; N.D., not determined; S, serum; SR, serum replacement; WT, wild type.

The SR supplement in the media of the tertiary culture contains exogenous insulin, which could be taken up into the cells [27]. Accordingly, we tested whether insulin in S EB-derived cells is a result of de novo synthesis or cell uptake by staining the cells with an antiserum that recognizes both C-peptide I and II, the cleavage byproducts of proinsulin. The results showed that S, but not SR, EB-derived cells stained positive for C-peptide (Figs. 5J, 5K), confirming de novo intracellular insulin production by the S EB-derived population. Most of the cells that expressed C-peptide also stained positive for glucagon (Figs. 5L, 5M), identical to the results of insulin and glucagon double staining in S EB-derived cells (Fig. 5D). It should be noted that not all of the S EBs initiated in the tertiary culture gave rise to insulin+ or C-peptide+ staining cells. For those EBs that developed insulin+ or C-peptide+ cells, the frequency of these cells per microscopic field was variable as well. To determine the frequency of insulin staining cells in culture, S EB-derived populations procured from days 11 through 13 of tertiary culture were double-stained with anti-insulin and anti-glucagon antibodies and analyzed by flow cytometry. Although cells stained for both insulin and glucagon from these cultures were not discernable, 1%–2% of the cells were glucagon single-positive, and less than 1% of cells were stained for insulin (data not shown and Fig. 9B). These results show that the conversion efficiency of insulin-expressing cells remained low in these basic cultures.

Figure Figure 9..

β-cell specification and differentiation factors increase insulin-expressing cells in tertiary culture. Day-6 serum embryoid bodies were cultured in basic tertiary media without fibroblast growth factor 2. On days 5, 7, or 9 of the tertiary culture, a mixture of 10 mM nicotinamide, 0.1 nM exendin-4, and 10 ng/ml activin βB was added to the media. On day 13 of the tertiary culture, total RNA was extracted and analyzed for gene expression by quantitative real-time reverse transcription–polymerase chain reaction (A), single cells were stained with insulin- and glucagon-specific antibodies and analyzed by fluorescent flow cytometry (B), or cells grown on cover slides were immunostained for insulin and glucagon (C). (A): Data represent the mean and standard deviation from duplicate samples. Two additional experiments showed similar results. (B): Representative flow cytometric analysis is shown for cells that were cultured with (added on D7−3°) or without (none) β-cell differentiation factors. Cells were stained for intracellular insulin and glucagon (left panels); nonimmune sera were used for control negative staining (right panels). (C): Representative staining of insulin and glucagon-expressing clusters are shown. Magnification of the images is indicated. Abbreviations: DAPI, 4′,6′-diamidino-2-phenylindole; Glu, glucagon; Ins, insulin.

Insulin I is expressed not only by pancreas, but also by yolk sac [1, 3, 4] and fetal liver [3]. The fact that most insulin or C-peptide+ cells in S EB-derived cells coexpress glucagon indicates that they are not yolk sac related, because this tissue does not express glucagon [6]. To test whether mouse liver expresses glucagon, fetal and adult liver were harvested and analyzed by RT-PCR. The results showed that neither fetal nor adult liver express glucagon (Fig. 6). Thus, the insulin+ glucagon+ or C-peptide+glucagon+ double-positive cells that originated from S EB are unlikely to be hepatic in origin. Taken together, these results suggest that our culture conditions for S EBs are directed toward the early pancreatic lineage insulin-expressing cells.

Figure Figure 6..

Lack of glucagon expression in fetal and adult liver. Total RNA isolated from liver of 12-week-old and E14 C57BL/6 mice or purified islets from BALB/c mice was processed for reverse transcription–polymerase chain reaction analysis.

Characterization of Culture Conditions for Insulin I–Expressing Cells

To define the culture requirements for the development of the insulin I–expressing cells, several variations of the culture methods depicted in Figure 1 were tested. In all EB cultures, it should be noted that the concentration of MTG is 10-fold higher on the first 2 days than the following 4 days. To test whether the high concentration of MTG is necessary to induce insulin I expression in tertiary culture, S EBs were incubated with different doses of MTG from days 0 to 2 and then cultured in standard conditions. The findings from this analysis indicated that insulin I expression was dependent on the presence of 1,500–6,000 μM of MTG during the first 2 days of differentiation (Fig. 7).

Figure Figure 7..

Requirement of higher MTG concentration for the first 2 days of EB culture. R1 embryonic stem cells were cultured in EB media containing the designated concentration of MTG. On day 2 of culture, EBs were washed and plated in EB media containing 600 μM MTG. All EB cultures contained 15% fetal calf serum. Day-6 serum EBs were cultured in the basic tertiary media (with fibroblast growth factor 2) for 10 or 20 days before being processed for RT–polymerase chain reaction analysis. Abbreviations: EB, embryoid body; Ins, insulin; MTG, monothioglycerol; RT, reverse transcriptase.

Interaction between different cell types and germ layers in the early embryo is necessary for the generation of pancreatic epithelium [2832]. In our cultures, many S EB-derived masses remained multilayered, and as a consequence the cells within the center of the mass often formed necrotic foci (data not shown). To test whether smaller pieces of the three-dimensional structures are sufficient for the generation of insulin I–expressing cells, day-6 S EBs were subjected to trypsin digestion for 30 seconds before plating into tertiary cultures. This treatment abrogated the generation of insulin I–expressing cells in tertiary cultures, although pdx-1+ cells still developed (Fig. 3C). This result suggests that, at least from days 0 to 11 in tertiary culture, cellular interactions are required for the further development of insulin I–expressing cells.

The effects of gelatin coating of culture wells, which hastens attachment of EBs in tertiary culture, on the development of insulin I–expressing cells were tested. Day-6 S EBs were cultured for 11 days on plates that were either coated with 0.1% gelatin or left untreated. The results show that culture on gelatin did not change the expression pattern of insulin I, pdx-1, or the other genes tested (Fig. 3A).

Our initial tertiary culture medium contained 10 ng/ml FGF2 [25] and serum-free supplements. To test the requirement for FGF2 and the effects of serum in tertiary culture, day-6 S EBs were plated in the presence or absence of either FGF2 or serum. Insulin I expression was examined after 11 days of culture. The results show that insulin I is only expressed in cells cultured with media containing SR. The presence or absence of FGF2 has no effect on insulin I expression (Fig. 8A). To test whether the lack of expression of insulin I was attributable to the inhibitory effects of serum, SR and serum were added together in tertiary culture. The results (Fig. 8B) show that serum is not inhibitory to insulin I expression, suggesting that other components in the serum-free supplements are required for the development of insulin I–expressing cells in the tertiary culture.

Figure Figure 8..

Effects of FGF2, SR, and S on insulin I expression in tertiary culture. Day-6 S (A, B) or SR EBs (A) were initiated in cultures with (+) or without (−) 15% FCS, SR, or 10 ng/ml FGF2. On day 11 of tertiary culture, total RNA was extracted and analyzed for gene expression by RT–polymerase chain reaction. Abbreviations: EB, embryoid body; FCS, fetal calf serum; FGF2, fibroblast growth factor 2; RT, reverse transcriptase; S, serum; SR, serum replacement.

β-Cell Specification and Differentiation Factors Increase Insulin-Expressing Cells Derived from S EBs

Activin βB, which mimics the effects of notochord, is required for the specification of pancreas during development [30]. Nicotinamide [3335] and exendin-4 [36], an agonist of glucagon-like peptide-1 receptor, are implicated in the proliferation and differentiation of fetal pancreatic β cells. To test whether these factors could increase the levels of insulin I expression in the basic cultures, S EB-derived cells were incubated with activin βB, nicotinamide, and exendin-4 and analyzed by quantitative real-time RT-PCR. The results show that addition of these three factors on day 7 of the tertiary culture, around the time when insulin I message begins to emerge (Fig. 4), produces the greatest increase (an average of 33-fold) of insulin I message compared with the controls (Fig. 9A). In contrast, addition of the three factors at an earlier or later time resulted in a lower increase of insulin I message. These results suggest that the addition of these factors may induce differentiation at the expense of the proliferation of the islet progenitors and that maturation may require at least 6 days to complete.

To test whether addition of these three factors increases the frequency of insulin-positive cells, the S EB-derived population was stained with insulin- and glucagon-specific antibodies and analyzed by fluorescent flow cytometry. Control basic culture conditions consistently gave rise to fewer than 1% insulin-positive cells (Fig. 9B, data not shown). Addition of the factors on day 7 of the tertiary culture resulted in a greater percentage (2.73 ± 0.04%) of insulin-positive cells, whereas addition of these factors on day 5 resulted in a smaller increase in insulin-expressing cells (1.17 ± 0.20%; p < .05 compared with day-7 addition). There was no difference in total output of cell number generated between factor-treated cultures (5.5 ± 0.8 × 105/well, day-7 addition; 5.5 ± 1.8 × 105/well, day-5 addition) and control cultures (4.6 ± 2.0 × 105/well). These results suggest that the combination of the three factors induces differentiation of β-cell progenitors.

Immunohistochemical staining for insulin and glucagon on S EB-derived cells that were treated with differentiation factors revealed the presence of large clusters that were stained positive for insulin (Fig. 9C). In contrast to the four to eight cells per group shown in Figure 5, these typically consist of several hundred to thousands of cells, and insulin- and glucagon-expressing cells seem to be intermingled within the clusters. The precise number of insulin-expressing clusters per slide is difficult to determine, because in some areas, several smaller clusters were in close vicinity and could be counted as either one large or several small clusters. In general, there was a range of 6 to 12 large clusters per slide from 25 S EBs cultured in the presence of nicotinamide, exendin-4, and activin βB. Taken together, these results demonstrate that the committed ES-derived cells in these cultures are responsive to known β-cell specification and differentiation factors and suggest that β-cell progenitors may be present in significant numbers in tertiary cultures.


The ancestral rodent insulin II locus gave rise to insulin I as a result of an RNA-mediated duplication-transposition activity [37]. A unique enhancer/promoter region located upstream of the transcription initiation site in the insulin I gene has restricted its expression in several tissues. Both insulin I and II fully compensate the function of the other, and only mice deficient in both genes develop the diabetic phenotype [38]. The differential expression of insulin I and II in various fetal tissues can thus serve as a criterion to separate neuronal lineages, which express only insulin II, from the other lineages.

Recent efforts have been directed at the generation of insulin-producing cells using the ES-EB culture system [25, 3944] to provide islet cells for transplantation and cure of type I diabetes. However, previous ES cell culture studies [25, 3944] did not address nor specify the extent of endoderm and pancreatic differentiation in their cultures. Our results demonstrated that the S EB-derived population expressed both insulin I and II genes, as well as glucagon, pdx-1, and Sox17. Most of the insulin-expressing or C-peptide–expressing cells generated from the basic culture conditions coexpress glucagon. This profiling pattern is consistent with early endocrine pancreas. During embryogenesis, a population that simultaneously expresses insulin and glucagon can be detected in the dorsal pancreatic anlage on E11.5, after the appearance of the first glucagon-expressing cells on E9.5 [45, 46]. It was previously thought that these double-positive cells were the progenitors of endocrine lineages [8, 45, 46]. However, using genetic lineage ablation [47] and tracing [48] approaches, it was found that this first wave of insulin+glucagon+ cells is not the progenitor to most of the β cells that later populate the islets in the second wave. Although the insulin+glucagon+ cells described in our study may not be able to divide and give rise to large numbers of β cells, they indicate that our culture conditions are differentiating ES cells toward pancreatic lineage insulin-expressing cells instead of fetal liver or yolk sac lineages. The pancreatic endocrine progenitors for the second wave of β cells may be present, but specific growth factors that are needed to drive their differentiation and maturation may be limiting in the basic culture condition. In support of this notion, we found that addition of activin βB, nicotinamide, and exendin-4 on day 7 of the tertiary culture increased insulin I expression by 33-fold and insulin+glucagon cells to 2.7% of the total population collected at day 13 of the tertiary culture. The intensity of insulin expression from the factor-treated S EB-derived population is lower than that from adult islets (data not shown). Thus, it is likely that these insulin-expressing cells may not yet be functionally mature. Efforts are underway to further confirm the stage of maturation and function of these insulin-expressing cells.

There are several differences in our culture conditions compared with that reported by Lumelsky et al. [25]. First, high concentration of MTG is used on the first 2 days of EB culture. MTG and its related thiol compounds, including 2-mercaptoethnol, can decrease oxidative stress by increasing intracellular glutathione concentration [49]. Glutathione concentration fluctuates at specific stages during early embryo development [50]. However, the mechanism by which MTG exerts its function in our cultured cells is currently unknown. Second, a longer serum exposure to EBs permits the development of flk-1+ cells, while diminishing the neuronal potential. A recent study demonstrated that pancreatic epithelium require contact from blood vessels to differentiate into insulin-expressing cells [32]. Our results demonstrate that tissue aggregates were required for insulin I expression (Fig. 3) and that flk-1 was expressed only in S but not in SR EB-derived cells (Fig. 3). Sox17-expressing endoderm and pdx-1+ cells were present in day-6 SR EBs (Figs. 2, 3), but they failed to commit to pancreatic β cells. Together, these results suggest that the codevelopment of flk-1–expressing cells may be a critical step in β-cell differentiation in our culture. However, the identity of the flk-1+ cells in our culture as endothelial cells remains to be determined. Third, our culture system does not require FGF2. Last, it should be noted that an unknown component (trade secret; Invitrogen) in the SR media of the tertiary culture may be either directly or indirectly required for the development of insulin I–expressing cells.

In contrast to the S EB-derived population, the SR EB-derived cells were not committed to the β-cell lineage, because no insulin I or pdx-1 was expressed at later time points of the tertiary culture. It is likely that these cells represent the neuronal lineage insulin-expressing cells, as described recently [51]. Interestingly, glucagon-expressing and amylase A2–expressing cells were present in the SR EB-derived cells. This may not be surprising, because the day-6 SR EBs had endodermal (Sox17) and pancreatic (pdx-1) potential. The tertiary culture conditions may have allowed the development of glucagon-expressing endocrine and exocrine cells from SR EBs but failed to support β-cell lineage differentiation.

It is now recognized that incremental advances in directing ES cell differentiation toward the endoderm-derived pancreas are needed to achieve the goal of generating transplantable glucose-responsive insulin-secreting cells. Accordingly, we have dissected out the culture conditions leading to the commitment of ES cells to early pancreatic cells. With the culture method developed in this report, future efforts will be directed at the enrichment of endocrine stem cells and the elucidation of the growth factors required for the proliferation, differentiation, and maturation of these cells.


The authors thank Marion Kennedy for valuable advice on ES cell culture and Dan Chen and Haojiang Zhang for excellent technical assistance. This work was supported in part by the Juvenile Diabetes Research Foundation, International grants 5-2001-889 and 1-2004-12 to H.T.K. and 4-1999-697 to J.S.B.