Generation of Insulin-Producing Islet-Like Clusters from Human Embryonic Stem Cells



Recent success in pancreatic islet transplantation has energized the field to discover an alternative source of stem cells with differentiation potential to β cells. Generation of glucose-responsive, insulin-producing β cells from self-renewing, pluripotent human ESCs (hESCs) has immense potential for diabetes treatment. We report here the development of a novel serum-free protocol to generate insulin-producing islet-like clusters (ILCs) from hESCs grown under feeder-free conditions. In this 36-day protocol, hESCs were treated with sodium butyrate and activin A to generate definitive endoderm coexpressing CXCR4 and Sox17, and CXCR4 and Foxa2. The endoderm population was then converted into cellular aggregates and further differentiated to Pdx1-expressing pancreatic endoderm in the presence of epidermal growth factor, basic fibroblast growth factor, and noggin. Soon thereafter, expression of Ptf1a and Ngn3 was detected, indicative of further pancreatic differentiation. The aggregates were finally matured in the presence of insulin-like growth factor II and nicotinamide. The temporal pattern of pancreas-specific gene expression in the hESC-derived ILCs showed considerable similarity to in vivo pancreas development, and the final population contained representatives of the ductal, exocrine, and endocrine pancreas. The hESC-derived ILCs contained 2%–8% human C-peptide-positive cells, as well as glucagon- and somatostatin-positive cells. Insulin content as high as 70 ng of insulin/μg of DNA was measured in the ILCs, representing levels higher than that of human fetal islets. In addition, the hESC-derived ILCs contained numerous secretory granules, as determined by electron microscopy, and secreted human C-peptide in a glucose-dependent manner.

Disclosure of potential conflicts of interest is found at the end of this article.


Islet cell transplantation for patients with type I diabetes has recently shown considerable promise for achieving insulin independence [1, [2]–3]. However, the extreme shortage of cadaver pancreata has challenged the scientific community to use stem cells with the potential to differentiate into pancreatic β cells [4]. Human ESCs (hESCs), which are telomerase-positive, immortal, and capable of both self-renewal and differentiation into all cell types of the body, could potentially supply an unlimited number of insulin-producing β cells for transplantation into diabetic patients [5].

Embryonic stem (ES) cell lines are generated from the inner cell mass of developing blastocysts and are able to proliferate in vitro indefinitely as undifferentiated cells in the presence or absence of a feeder layer [6, 7]. Undifferentiated ES cells can be differentiated in vitro and in vivo to cell types or tissues belonging to the ectoderm, mesoderm, or endoderm lineage [6, 7]. Thus, a continuously accumulating body of scientific literature documents that both mouse and human ES cells can differentiate into diverse cell lineages, including cardiac and skeletal muscles [8, 9], neuronal [10, 11] and oligodendroglial [12] cells, hematopoietic cells [13], endothelial cells [14], and hepatocytes [15].

The adult pancreas is composed of three major tissue types: exocrine cells that produce digestive enzymes, ducts that transport the digestive enzymes, and endocrine islets consisting of cells producing insulin (β cells), glucagon (α cells), somatostatin (δ cells), and pancreatic polypeptide (PP cells) [16]. Pancreatic organogenesis is a highly regulated and complex process controlled by various morphogens, presumably produced in the local microenvironment, as well as by key transcriptional regulators responsible for specifying different cell types present within the pancreas [17, 18]. Among them, Pdx1, a pancreas duodenum homeodomain transcription factor, is expressed throughout pancreas development, including cell type specification and differentiation, with expression persisting in mature β cells [19, 20]. Similarly, Ptf1a/P48, a bHLH transcription factor, is also required for pancreatic specification, although its expression in the mature pancreas becomes restricted to exocrine cells [21, 22]. Ngn3, a transiently expressed transcription factor, is responsible for commitment to all endocrine cell types within the pancreas [23]. During embryogenesis, the pancreas develops as two buds, dorsal and ventral, which are formed from a specialized epithelium located in the primitive gut endoderm [16]. The formation of dorsal pancreatic endoderm, which expresses Pdx1, a key transcription factor for pancreas development, is specified by factors such as activin and fibroblast growth factor (FGF) produced by the notochord [24]. The ventral bud, on the other hand, has the potential to become liver or pancreas depending on the signals received from the mesenchyme [16]. It has been shown that basic fibroblast growth factor (bFGF) signaling from the cardiac mesoderm dictates liver-specific differentiation of ventral endoderm cells, whereas in the absence of bFGF, a pancreatic cell fate is triggered [25]. At later stages of embryogenesis, both the dorsal and ventral buds rotate, and they fuse to become one organ.

Various strategies have been used to derive endoderm as well as insulin-producing β cells from ES cells. It has been shown recently that mouse ESCs (mESCs) could differentiate to mesendoderm and subsequently to endoderm with the use of activin A [26, 27]. Moreover, treatment with activin A in the presence of a low concentration of serum led to definitive endoderm formation from hESCs [28]. Using a genetically modified insulin-secreting cell clone, Soria et al. [29] demonstrated that cells differentiated from mESCs could normalize hyperglycemia in diabetic animals. Subsequently, other studies described differentiation protocols to produce insulin-positive cells from unmodified mESCs or genetically modified mESCs using several different pancreas-specific transcription factor genes [30, [31], [32]–33]. Further characterization of the insulin-positive cells produced using particular protocols suggested that these cells accumulated insulin from the media [34, 35] and did not de novo synthesize endogenous insulin. An initial report investigating spontaneous hESC differentiation demonstrated spontaneous generation of limited numbers of insulin-positive cells [36]. Subsequently, other studies reported the generation of insulin-producing cell clusters from hESCs using more defined culture conditions [37, 38]. More recently, D'Amour et al. reported the generation of pancreatic hormone-expressing endocrine cells which, like fetal β cells, released C-peptide in response to several different secretagogues [39]. However, C-peptide release by these cells was marginal in response to glucose.

Since hESCs are capable of generating all tissues, it is likely that the instructive signals that direct pancreas organogenesis and β cell differentiation during embryonic development could also instruct hESCs to commit to a similar fate in vitro. In this report, we present a multistep differentiation protocol, free of serum, to generate insulin-producing cells. During the 36-day protocol, hESCs were treated with growth factors and signaling molecules to differentiate them through definitive and pancreatic endoderm to endocrine cell populations. Pancreatic endocrine cells were then matured to islet-like clusters (ILCs) that contained insulin-, glucagon-, and somatostatin-producing cells.

Materials and Methods

Undifferentiated hESC Culture

H1, H7, and H9 hESCs were cultured on Matrigel (Invitrogen, Carlsbad, CA, plates under feeder-free conditions using mouse embryonic fibroblast (MEF)-derived conditioned medium as described by Xu et al. [7].

In Vitro Differentiation of ILCs from Undifferentiated hESCs

Differentiation was carried out in four stages, as shown in Figure 1.

Figure Figure 1..

Differentiation scheme for generating insulin-positive islet-like clusters (ILCs) from human ESCs (hESCs). Stage 1, undifferentiated hESCs, grown in mouse embryonic fibroblast-derived conditioned medium supplemented with 8 ng/ml basic fibroblast growth factor (bFGF), were treated with 1 mM Na-butyrate and 4 nM activin A for 1 day and subsequently with a reduced concentration of 0.5 mM Na-butyrate and 4 nM activin A for 6 days to induce endoderm differentiation. Stage 2, the endoderm cells were differentiated in low-adherent culture conditions to form cellular aggregates, and these aggregates were subsequently grown for 14 days in medium containing B27 medium supplemented with 2 ng/ml bFGF, 20 ng/ml EGF, and 100 ng/ml Noggin. Stage 3, hESC-derived aggregates were further cultured for 7 days in the same medium without bFGF. Stage 4, on day 29, cell aggregates were transferred into a maturation medium containing 10 mM nicotinamide and 50 ng/ml IGFII to generate ILCs. Abbreviations: bFGF, basic fibroblast growth factor; EGF, epidermal growth factor; hES, human embryonic stem; IGFII, insulin-like growth factor II; Na-Butyrate, sodium butyrate.

Stage 1: Definitive Endoderm Induction.

Confluent undifferentiated hESCs (uhESCs) were cultured in RPMI 1640 medium (Invitrogen) containing 11.1 mM glucose supplemented with 1× B27 (Invitrogen) (RPMI 1640/B27 medium), 4 nM activin A (R&D Systems Inc., Minneapolis,, and 1 mM sodium butyrate (Sigma-Aldrich, St. Louis, for 1 day. After 24 hours, medium was replaced with fresh RPMI 1640/B27 medium supplemented with 4 nM activin A and 0.5 mM sodium butyrate. The cells were cultured in this medium for another 6 days.

Stage 2: Pancreatic Endoderm Formation.

At the end of stage 1, the cells were dissociated with 200 units/ml collagenase IV (Invitrogen) at 37°C for 3–5 minutes and scraped off the plate in RPMI 1640/B27 medium supplemented with 20 ng/ml epidermal growth factor (EGF) (R&D Systems), 2 ng/ml bFGF (Invitrogen), and 100 ng/ml noggin (R&D Systems). The cells were transferred into ultralow-attachment six-well plates (Corning, St. Louis, at a ratio of 1:1 and cultured in the same medium. The cells were fed with fresh medium every 2–3 days for 2 weeks.

Stage 3: Endocrine Induction.

bFGF was withheld from the culture after 2 weeks, and cell clusters were cultured in suspension in RPMI 1640/B27 medium supplemented with EGF and noggin for 1 week.

Stage 4: ILC Maturation.

The cell clusters were cultured with fresh RPMI 1640 medium containing 0.5% bovine serum albumin, 10 mM nicotinamide (Sigma-Aldrich), and 50 ng/ml insulin-like growth factor (IGF-II) (R&D Systems) for 5 days and without IGF-II for another 2 days.

Microarray Differential Gene Expression Analysis

DNA microarray analysis was performed using a custom oligonucleotide microarray (MWG-Biotech Inc., High Point, NC, containing 752 genes.

Gene Expression by Quantitative Real-Time Polymerase Chain Reaction

Total RNA was isolated using an RNeasy mini kit (Qiagen, Valencia, CA,, and genomic DNA was removed by digestion with deoxyribonuclease I (Invitrogen). Two μg of RNA was reverse-transcribed into cDNA with random hexamers and Superscript II (Invitrogen). The probe and primer sequences used for quantitative polymerase chain reaction (PCR) are summarized in supplemental online Table 1. Quantitative PCR was performed in a total of 25 μl of reaction mix containing 1× TaqMan Master Mix (Applied Biosystems, Foster City, CA,, 300 nM probe, and 200 nM each forward and reverse primers, template cDNA, and 1× cyclophilin (ABI), a housekeeping gene, as internal control. For the probes and primers from ABI, 1× probe and primer mix were added in the reaction mix. PCR amplification was carried out using ABI Prism 7900 HT Sequence Detection System (SDS 2.1; Applied Biosystems) and stopped at 40 cycles (program: 2 minutes at 50°C, 10 minutes at 95°C, and 40 repetitions of 15 seconds at 95°C and 1 minute at 60°C). Relative gene expression of the mRNA was expressed as percentage relative to the corresponding values obtained from RNA from 18–20-week-old human fetal pancreata (Stratagene, La Jolla, CA,


Washed cell clusters were fixed in 4% paraformaldehyde (PFA) at room temperature for 15 minutes, permeabilized in 100% ethanol at room temperature for 1 minutes, and blocked with 5% normal goat serum or 2% fetal bovine serum (FBS)/phosphate-buffered saline (PBS) at room temperature for 1 hour. The cells were then incubated with the primary antibodies overnight at 4°C. After three washes with PBS, the cells were incubated with the secondary antibodies at room temperature for 30 minutes followed by three washes in PBS and mounted with Vectashield Mounting Medium containing 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA, The sources of antibodies and dilutions used are summarized in supplemental online Table 2. Images were captured under a Nikon fluorescent microscope (Nikon, Tokyo, or a Zeiss LSM 510 confocal microscope (Carl Zeiss, Jena, Germany, at the Cell Sciences Imaging Facility at Stanford University Medical Center, Stanford, CA, using Adobe Photoshop software (Adobe Systems Inc., San Jose, CA,

Flow Cytometry

hESCs and hESC-derived cells or cell clusters at various stages of differentiation were washed with PBS and dissociated with 0.25% trypsin-EDTA (Invitrogen) for 10–15 minutes. The cells were washed and fixed with 2% PFA in PBS at room temperature for 15 minutes and permeabilized with IntraPrep permeabilization reagent (Immunotech, Marseille, France, or with 90% methanol in PBS. For single-color assay, 5.0 × 105 cells were incubated with 0.2 μg of monoclonal anti-human C-peptide antibody (Monosan Corp., Uden, The Netherlands,, 0.5 μg of goat anti-Pdx1 antibody (a gift from Dr. Chris Wright, Vanderbilt University, Nashville, TN), or isotype-matched control antibody at room temperature for 30 minutes; washed with PBS; and stained with appropriate secondary antibody conjugated to Alexa Fluor 488 at room temperature for 15 minutes. The cells were then washed, resuspended in 0.5% PFA in PBS. Flow cytometric analysis was performed on a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA, using Cell Quest and/or FlowJo software (FlowJo LLC, Ashland, OR, For two-color staining, day 8 cells were dissociated and stained with 20 μl of allophycocyanin-conjugated CXCR4 antibody (BD Biosciences). The cells were washed, fixed in 2% PFA at room temperature for 15 minutes, and permeabilized in BD Perm/Wash buffer for 30 minutes. The cells were then stained with 0.1 μg of Sox17 or 0.1 μg of Foxa2 antibodies followed by Alexa Fluor 488-conjugated donkey anti-goat IgG. Isotype-matched antibody staining was performed to determine specific staining. Samples were analyzed as described above. For all flow cytometric analysis, cells were stained in a total volume of 100 μl.

Electron Microscopy

Aliquots of hESC-derived day 36 ILCs were examined by transmission electron microscopy to identify cells containing typical islet endocrine cell secretory vesicles, as previously described [40].

ILC Filtration

A Cellmicrosieve (Biodesign Inc. of New York, New York, with a pore size of 200 was placed on top of a 50-ml conical tube (Corning). Day 36 cell clusters were passed through the Cellmicrosieve. Two washes with culture medium were performed to ensure maximal passage of the cell clusters less than 200 μm in diameter.

Glucose-Stimulated Insulin Release and Insulin Content Analysis

hESC-derived clusters at day 36 were plated onto a Matrigel-coated 48-well plate in triplicate, approximately 100–200 clusters per well, and were allowed to attach to the plate overnight at 37°C. Next, the clusters were washed three times with glucose-free RPMI 1640 medium. The cells were first incubated with RPMI 1640 medium containing 2.8 mM glucose at 0.5 ml per well at 37°C for 3.5 hours. The supernatants were collected, and the same ILCs were further treated with 0.5 ml of RPMI 1640 medium containing 20 mM glucose and incubated at 37°C for another 3.5 hours. The supernatants were collected again and both low- and high-glucose-treated ILC supernatants were stored at −20°C until they were analyzed. The C-peptide level was measured using the Mercodia Ultrasensitive Human C-peptide ELISA Kit (ALPCO Diagnostics, Windham, NH,, and the values were normalized to DNA content. For DNA quantitation, the ILCs were lysed in 0.5% Triton X-100 TE buffer and sonicated. A Quant-iT PicoGreen dsDNA Reagent Kit (Invitrogen) was used to determine DNA concentration according to the manufacturer's recommendation. The total amount of C-peptide in the supernatant of each well was normalized to the corresponding amount of DNA in the well. The stimulation index was calculated by determining the ratio of the amount of C-peptide released by the ILCs with 20 mM glucose and the amount of C-peptide released after 2.8 mM glucose treatment. The insulin content of the hESC-derived ILCs was determined by radioimmunoassay (RIA) according to a published protocol [40], and the values were normalized to the DNA content.

Enumeration of Apoptosis by Immunofluorescence

ILCs were prepared for glucose-stimulated insulin release in the manner described above. After collection of the supernatants, the clusters were washed with PBS, fixed in 4% PFA for 20 minutes, and permeabilized by 0.1% Triton X-100 in 0.1% sodium citrate on ice for 2 minutes. Apoptosis was assessed by staining the ILCs with terminal deoxynucleotidyl-mediated dUTP nick end-labeling (TUNEL) using a commercially available kit (Roche Diagnostics, Indianapolis, IN, following the manufacturer's instructions. To determine whether the insulin-producing cells within the ILCs were TUNEL-positive, the clusters were subsequently double-stained with mouse anti-human C-peptide as described above.


Sodium Butyrate and Activin A Induce Definitive Endoderm Formation from hESCs

Undifferentiated hESCs underwent considerable morphological changes during the various stages of differentiation, and the results are described in supplemental online Figure 1. It has been previously demonstrated that activin A treatment of hESCs in the presence of low concentrations of serum induces definitive endoderm development [28]. Our differentiation protocol does not use serum, and hence we compared the combination of sodium butyrate and activin A to activin A treatment alone for the induction of endoderm-specific Sox17, Foxa2 and Hnf4α gene expression (Fig. 2A, upper panels). Sodium butyrate alone was largely ineffective in inducing endoderm formation in H1 cells. Activin A treatment alone caused induction of endoderm-specific genes, but the sodium butyrate and activin A combination induced higher levels of Sox17, Foxa2, and Hnf4α expression in H1 hESCs (Fig. 2A). Similarly, treatment of H7 and H9 cells with sodium butyrate and activin A also generated definitive endoderm formation (supplemental online Fig. 2A). To determine whether the endoderm population generated with sodium butyrate plus activin A could also differentiate to the pancreatic lineage, the various endoderm populations were directly differentiated with EGF, bFGF, and noggin for a week, and Pdx1 gene expression was determined (Fig. 2A, lower panel). In these experiments, definitive endoderm generated using both sodium butyrate and activin A produced the highest levels of Pdx1 expression at day 15 when grown in the presence of B27 medium containing EGF, bFGF, and noggin. hESCs differentiated in all other conditions during the first week of differentiation did not respond well to subsequent growth factor cocktail treatment, as evidenced by the lack of their induction of Pdx1 gene expression (Fig. 2A, lower panel). Thus, sodium butyrate plus activin A treatment of hESCs induced differentiation of definitive endoderm populations, as assessed by marker gene analysis. Microarray expression profile analysis of day 8 endoderm cells identified 28 mRNAs, including Foxa2, Sox17, CXCR4, CK-18, and CK-19, which were induced at least 2.5-fold in activin A- and sodium butyrate-treated H1 hESCs at day 8 (supplemental online Table 3). In addition, other mesoderm/endoderm-related genes, such as Cer1 and Gsc, were also highly expressed in the activin A- and sodium butyrate-treated cells compared with the starting uhESCs. The complete list of genes that were induced at the definitive endoderm stage by microarray analysis is shown in the supplemental online data.

Figure Figure 2..

Characterization of human ESC (hESC)-derived endoderm by gene and protein expression analysis. (A): Effect of Na-butyrate and ActA on endoderm and pancreatic endoderm induction. Human embryonic stem cells (H1) were treated with medium only, medium with Na-butyrate alone (1 mM Na-butyrate for 1 day and then 0.5 mM for 6 days; upper panels), with 4 nM activin A alone, and with a combination of Na-butyrate and ActA for 7 days followed by combination of 2 ng/ml bFGF, 20 ng/ml EGF, and 100 ng/ml noggin for 7 days (lower panel). Total RNA was isolated from cells treated with each condition. Expression of the endoderm genes Sox17, Foxa2, Hnf4a, and Pdx1 was analyzed by quantitative reverse transcription-polymerase chain reaction. All mRNA expression levels were normalized to the housekeeping gene cyclophilin expression. Relative gene expression was determined by normalization to that in human fetal pancreas. (B): Protein expression of transcription factors in hESC-derived endoderm cells. Day 8 hESC-derived cells were immunostained with antibodies against Foxa2 (green), Sox 17 (red), Gata4 (red) (upper panels), CXCR4 (red), or proliferation marker Ki-67 (red), followed by appropriate secondary antibodies conjugated to fluorochromes; a merge from both images shows colocalization of Foxa2 and Ki-67 (lower panels). Blue, DAPI for nuclear staining; green, Alexa Fluorochrome 488; red, Alexa Fluorochrome 594. (C): Two-color flow cytometric analysis of hESC-derived definitive endoderm. Day 8 cells were dissociated and stained with anti-CXCR4 antibody, followed by anti-Sox17 (middle panel) or anti-Foxa2 (right panel) antibody staining. Isotype-matched control antibody staining was performed using the same cells to determine background fluorescence (left panel). Abbreviations: ActA, activin A; APC, allophycocyanin; DAPI, 4′,6-diamidino-2-phenylindole; NaBut, sodium butyrate.

We performed immunocytochemical analysis on the hESC-derived endoderm population (Fig. 2B). The majority of the endoderm cells expressed Foxa2, Sox17, and Gata4 transcription factors, with expression localized within the nucleus. In addition, the chemokine receptor CXCR4, recently shown to be present on definitive endoderm cells [27, 28], was also highly expressed in these cells. The Foxa2+ cells were also assessed for their potential proliferative capacity by staining with an antibody to Ki-67 (Fig. 2B). Many of the Foxa2+ cells coexpressed Ki-67, suggesting that these cells may have replicative potential.

To confirm that this culture condition produced definitive endoderm rather than extraembryonic endoderm, we performed two-color flow cytometry analysis. Double staining of CXCR4 and Sox17 showed that the majority of Sox17+ cells were also positive for CXCR4 (Fig. 2C), whereas approximately between 2% and 3% of the total cells were Sox17+/CXCR−. Similarly, most of the Foxa2+ cells in the endoderm population were also positive for CXCR4 (Fig. 2C). However, a small subset of CXCR4-expressing cells were Sox17−, most likely representing a mesoderm population. Taken together, these data indicate that treatment with activin A and sodium butyrate efficiently induced definitive endoderm, and extraembryonic endoderm cells were rare in our differentiated endoderm population.

Pdx1-Expressing Pancreatic Endoderm Is Evident at Day 15

hESC-derived endoderm cell populations were treated with collagenase at day 8 and transferred into low-attachment plates for aggregate formation in the presence of bFGF, EGF, noggin, and B27 supplement (stage 2). Various sizes of cellular aggregates were produced in these cultures. To identify the pancreatic endoderm population, the cells at day 15 were examined for Pdx1 protein expression by immunocytochemistry (Fig. 3A, a, d, g, and j) in combination with Foxa2 (Fig. 3Ab), Sox17 (Fig. 3Ae), the CXCR4 receptor (Fig. 3Ah), or Ki-67 (Fig. 3Ak). Most Pdx1+ cells also expressed Foxa2 (Fig. 3Ac). Coexpression with Sox17 (Fig. 3Af) and the CXCR4 receptor (Fig. 3Ai) was also seen in some Pdx1+ cells. Many Pdx1+ cells coexpressed Ki-67 nuclear antigen, indicating that these Pdx1+/Ki-67+ pancreatic endoderm cells were proliferative (Fig. 3Al). To quantitate the percentage of Pdx1+ cells in the pancreatic endoderm population, we performed flow cytometric analysis at days 8 and 15 of differentiation (Fig. 3B). Very few Pdx1-expresing cells were detected at the definitive endoderm stage (day 8). In the experiment shown in Figure 3B, at the pancreatic endoderm stage (day 15), 24% of cells expressed Pdx1, ranging from 15% to 25% of the total population in the collective experiments performed to date (Fig. 3B). In addition, gene expression analysis showed that by day 15, pancreatic endoderm cells also started to express other pancreas-related genes, such as HlxB9, Ptf1a, Ngn3, and Nkx6.1 (described below).

Figure Figure 3..

Characterization of human ESC (hESC)-derived pancreatic endoderm by Pdx1 expression. (A): Two-color immunofluorescence analysis was performed on day 15 endoderm cell aggregates. a–c, Pdx1 (a) (red), Foxa2 (b) (green), and a merge of both images (c); d–f, Pdx1 (d) (red), Sox17 (e) (green), and a merge of both images (f). Solid arrow indicates weak expression of Sox17, whereas dashed arrow indicates high expression of Sox17 in cellular aggregates (e). g–i, Pdx1 (g) (red), CXCR4 (h) (green), and a merge of both images (i). Solid arrow indicates weak expression of CXCR4, whereas dashed arrow indicates high expression of CXCR4 on cellular aggregates (h). j–l, Pdx1 (j) (red), Ki-67 (k) (green), and a merge of both images (l). (B): Flow cytometric analysis of Pdx1 expression in hESC-derived definitive endoderm (lower left panel) and day 15 pancreatic endoderm (lower right panel). Isotype-matched control antibody staining was performed to determine background staining for both definitive (upper left panel) and pancreatic (upper right panel) endoderm cell populations.

hESC-Derived ILCs Produce Insulin, Glucagon, and Somatostatin Hormones

Upon further differentiation of Pdx1+ cell clusters to day 36, cells producing the pancreatic hormones insulin (C-peptide), glucagon, and somatostatin were observed. Immunocytochemical staining data from one representative experiment using H1 hESCs are presented in Figure 4, Figure 4.A; these data showed that the majority of the small bud-like structures expressed C-peptide and glucagon (Fig. 4, Figure 4.A, panels a and b). The expression of human C-peptide in these clusters suggests that proinsulin was newly synthesized during the later stages of differentiation, with the insulin mRNA first detected at day 22 of this differentiation protocol (described below). The C-peptide-, glucagon-, and somatostatin-positive cells were predominantly localized in the small bud-like clusters, as well as in some of the smaller ILCs. Three-color staining analyzed by confocal microscopy showed that in addition to C-peptide+ (Fig. 4, Figure 4.Ac) and glucagon+ (Fig. 4, Figure 4.Ad) cells, the ILCs also expressed somatostatin (Fig. 4, Figure 4.Ae), a hormone released by the δ cells of pancreatic islets. A few of the C-peptide-expressing cells coexpressed glucagon (Fig. 4, Figure 4.Af, yellow), whereas another C-peptide+ subset of cells coexpressed somatostatin (Fig. 4, Figure 4.Af, purple). Taken together, these results suggest that similar to islets, hESC-derived ILCs contained insulin-, glucagon-, and somatostatin-producing cells. Although the results for ILC characterization reported in this study were obtained mainly using the H1 cell line, H7 and H9 cells were also tested and found to differentiate to the ILCs that expressed C-peptide and glucagon (supplemental online Fig. 2B).

Figure Figure 4..

Characterization of hESC-derived ILCs. (A): Expression of islet-specific hormones in hESC-derived budding ILCs. ILCs with bud-like structures, indicated by arrows (a), derived from H1 hESCs at day 36 of differentiation were double-stained with anti-human C-peptide (red) and anti-glucagon (green) specific antibodies. The images were taken at magnifications of ×2.5 (a) and ×10 (b). Confocal images for C-peptide (c) (red), glucagon (d) (green), somatostatin (e) (blue), and a merged image of (c–e) (f) are shown. In f, solid arrows indicate cells coexpressing C-peptide and glucagon (yellow), whereas dashed arrows indicate cells coexpressing C-peptide and somatostatin (purple). (B): Coexpression of Pdx1 with pancreas-specific markers. ILCs derived from H1 hESCs at day 36 of differentiation were stained for human C-peptide (a) (green) and Pdx1 (b) (red), and a merge of both images is shown (c); they were double-stained for CK-19 (d) (green) and Pdx1 (e) (red), and a merge of both images is shown (f). These images were analyzed by confocal microscopy. Solid arrows indicate cells positive for both C-peptide and Pdx-1 (c) or cells coexpressing CK19 and Pdx-1 (f), whereas dashed arrows indicate cells with nuclear expression of Pdx-1 (c and f). (C): Electron micrographs of hESC-derived ILCs show granulated cells that are intact and contain secretory granules (magnification, ×1,300). (D): Flow cytometric analysis was performed using an anti-human C-peptide antibody with undifferentiated hESCs, day 8 definitive endoderm cells, and cells dissociated from day 36 ILCs after trypsinization. C-peptide-positive staining was observed only with cells obtained from the ILCs. Abbreviations: hESC, human ESC; ILC, islet-like cluster.

Figure Figure 4..


β Cells in mature human islets also express the pancreas-specific transcription factor Pdx1, which plays an important role in insulin gene regulation. The majority of the C-peptide-expressing cells (Fig. 4, Figure 4.Ba) also expressed Pdx1 (Fig. 4, Figure 4.B, b and c), the expression of which was mainly localized in the nucleus, although some cytoplasmic Pdx1 staining was also observed.

In addition to the endocrine cells, we also found evidence that pancreatic ductal cells were present in the day 36 cultures. In this case, CK-19+ cells were observed upon immunocytochemical analysis (Fig. 4, Figure 4.Bd). Furthermore, some of the epithelial cells showed nuclear expression of Pdx1 (Fig. 4, Figure 4.B, e and f), suggesting that these cells could represent a cell type similar to the ductal progenitor cells of the pancreas.

When further characterized by transmission electron microscopy, the ILCs were also shown to contain some cells with numerous secretory vesicles (Fig. 4, Figure 4.C). Many of these cells contained granules with a morphological appearance conforming to adult islet β cells.

Using flow cytometry, we were able to quantitate the percentage of human C-peptide+ cells in our ILC populations. Although the undifferentiated and the hESC-derived endoderm cells did not show C-peptide expression, more than 4.0% cells within the ILCs were positive for C-peptide (Fig. 4, Figure 4.D). To date, the percentage of C-peptide+ cells has ranged from 2% to 8% in the differentiated cell populations produced.

The majority of the C-peptide+ cells appeared in the smaller cell clusters. As a result, we attempted to enrich the insulin-producing ILCs by size sieving using a cell sieve with a pore size of 200 μm. Results from one representative experiment are shown in Figure 5, where photomicrographs of unseparated, small ILCs (<200 μm) and large ILCs (>200 μm) are presented. Quantitative reverse transcription (RT)-PCR analysis of cells present in these three types of ILCs revealed that the small ILCs were enriched 3–14-fold for the mRNAs for all three islet hormones and Pdx1. The fold increase in the amount of mRNA was calculated with respect to the unseparated ILCs, for which the value was normalized to 1 (Fig. 5D).

Figure Figure 5..

Enrichment and further characterization of ILCs. Human ESC-derived ILCs were enriched at day 36 by passing the clusters through a 200-μm nylon mesh. Some of the unseparated ILCs (A), as well as size-separated small (B) and large (C) ILCs, were used to determine key islet gene expression by quantitative reverse transcription-polymerase chain reaction analysis. (D): For each gene, the level of expression obtained with unseparated ILCs was normalized to 1 to calculate fold increase in the small or large ILCs. Abbreviation: ILC, islet-like cluster.

Pancreatic Development Recapitulated in This Differentiation Protocol

To get a clearer picture of the cell transitions during this protocol, quantitative RT-PCR for a number of genes involved in pancreas development was performed at various stages throughout differentiation (Fig. 6, A and B). Oct3/4 expression decreased immediately after differentiation was initiated with sodium butyrate and activin A (Fig. 6B), showing the rapid onset of differentiation of the hESCs. Sox17 and Foxa2 reached maximum induction at 8 days into the protocol (Fig. 6A). Pdx1 was induced during the second stage of differentiation and was closely followed by the induction of another key pancreas transcription factor, Ptf1a. The endocrine progenitor-specific gene, Ngn3, was also induced between 15 and 22 days of differentiation, and the expression persisted throughout the differentiation process. Other islet-related genes, such as Isl1 and Nkx6.1, were also detected in these cells, and their levels peaked at later stages (stages 3 and 4) of differentiation. Insulin, glucagon, and the glucose-transporter molecule Glut-2 showed maximal expression at stage 4 of the protocol. Similarly, the expression of a pancreatic exocrine marker, amylase, was highly induced at later stages of differentiation. Figure 6B shows a graphical illustration of the transitions in gene expression observed in this protocol. The sequence of gene expression changes observed in these experiments suggests the emergence of definitive endoderm followed by pancreatic endoderm formation, with further differentiation to endocrine and exocrine pancreatic cells during the 36-day protocol.

Figure Figure 6..

Temporal expression of pancreatic lineage genes during human ESC (hESC)-derived islet-like cluster (ILC) generation. (A): H1 cells were differentiated to ILCs as described above. Cell samples were collected at days 1, 8, 15, 22, 29, and 36 and analyzed by quantitative reverse transcription-polymerase chain reaction (RT-PCR) for Sox17, Foxa2, Pdx1, Ptf1a, Ngn3, Isl-1, Nkx6.1, insulin, glucagon, Glut-2, and amylase gene expression. For each sample, relative expression levels were normalized to corresponding levels in human fetal pancreas. (B): Overview of temporal gene expression during ILC derivation from hESC. Quantitative RT-PCR was performed using RNA samples from undifferentiated hESCs or cell aggregates collected at the indicated time points to quantify the levels of Oct3/4, Foxa2, Pdx1, Ptf1a (p48), Hlxb9, Ngn3, Nkx6.1, insulin, and glucagon gene expression. For each gene, the level of expression at each time point was normalized to the maximal level observed, arbitrarily set as 1. The levels presented here from four separate experiments represent the trend over time and not the absolute expression level of a particular gene.

Insulin Content of hESC-Derived ILCs and C-Peptide Release After Glucose Stimulation

We performed insulin content analysis of the ILCs by RIA. The insulin content of the ILCs from three different preparations was 1.7, 15.9, and 70.1 ng of insulin/μg of DNA. The insulin content of two of these ILC preparations exceeded the insulin content of human fetal islets (12 ng of insulin/μg of DNA) [41, 42].

To determine whether the ILCs are capable of insulin secretion in response to glucose stimulation, we tested the ability of hESC-derived ILCs for human C-peptide production after sequential treatment with low and high concentrations of glucose in a static assay (Fig. 7A). In this experiment, the uhESCs and the ILCs were first incubated with 2.8 mM glucose for 3.5 hours to determine the basal level of C-peptide release. The same ILCs were subsequently incubated with 20 mM glucose for an additional 3.5 hours to determine glucose-stimulated C-peptide release. hESC-derived ILCs released C-peptide in a glucose-dependent manner, with a stimulation index of 3.5 (Fig. 7A). The undifferentiated cells did not release a measurable amount of C-peptide into the medium. Five separate hESC-derived ILC preparations were subjected to glucose-stimulated insulin release analysis, and the average stimulation index from these experiments was 2.5 (range, 1.4–3.9). These results indicate that the ILCs generated by our protocol were able to secrete C-peptide (and insulin) in response to high concentration of glucose, as measured by the static assay.

Figure Figure 7..

Glucose-stimulated C-peptide release by ILCs. (A): To determine glucose-dependent release of human C-peptide, uhESC and ILCs were washed in glucose-free medium and incubated in low-glucose (2.8 mM) medium for 3.5 hours. Cell supernatants were collected for C-peptide release analysis. Next, the same uhESCs and the ILCs were incubated for an additional 3.5 hours in high glucose (20 mM) medium. Cell supernatants were also collected after glucose stimulation. The amount of C-peptide released in the supernatants was measured by enzyme-linked immunosorbent assay and normalized to DNA content. (B): Assessment of apoptosis in human ESC-derived ILCs cultured in low and high concentrations of glucose. Fluorescent DNA fragmentation assay (TUNEL) and immunostaining of human C-peptide were performed in ILCs at hour 0 (upper panels), at 3.5 hours after 2.8 mM glucose treatment (middle panels), and after an additional 3.5 hours of 20 mM glucose treatment (lower panels). Fluorescence photomicrographs were taken of individual fields and subsequently merged to detect apoptosis of insulin/C-peptide-expressing cells. Representative fields from each treatment conditions are shown. Abbreviations: hr, hour; ILC, islet-like cluster; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling; uhESC, undifferentiated human ESCs.

Chronic exposure to high glucose is known to cause apoptosis of human islets in culture [43]. To determine whether exposure to a high glucose concentration for 3.5 hours could specifically cause apoptosis of insulin/C-peptide-positive cells and nonspecific release of C-peptide, the ILCs were examined for apoptosis. To perform the experiment, hESC-derived ILCs were sequentially treated with 2.8 mM glucose for 3.5 hours and then with 20 mM glucose for an additional 3.5 hours. ILCs cultured in the maturation medium were treated as controls to determine background apoptosis. In our assay, apoptosis of insulin-positive cells was assessed by TUNEL staining to determine DNA fragmentation, and β-like cells were identified using anti-C-peptide antibodies. After 3.5 hours of exposure to low glucose (2.8 mM; Fig. 7B, middle panels) followed by 3.5 hours of high glucose (20 mM; Fig. 7B, lower panels), no induction of apoptosis as measured by TUNEL staining was observed in the C-peptide-positive cells. Although a few TUNEL-positive cells (Fig. 7B, green) were seen within the ILCs of the various treatment groups, these cells were negative for C-peptide (Fig. 7B, red). These data indicate that nonspecific apoptosis was not the cause of the high glucose-stimulated C-peptide release from the ILCs.


We have described a new protocol to differentiate hESCs through definitive endoderm, pancreatic endoderm, and endocrine populations to ultimately generate ILCs expressing insulin, glucagon, and somatostatin. The hESC lines used in this study were grown under feeder-free conditions, and the differentiation medium was free of serum components. The protocol described here consistently generated 2%–8% human C-peptide containing cells, and similar to islets, the ILCs contain cells with secretory granules. Most importantly, hESC-derived ILCs possess properties of pancreatic islets, including expression of multiple endocrine hormones and production of C-peptide in response to glucose stimulation.

In the current protocol, we chose to use a combination of various factors, including activin A, a transforming growth factor β gene family member [24], as well as EGF and bFGF at specific stages of differentiation. In combination with activin A, we used sodium butyrate, a potent histone deacetylase inhibitor [44], during the early stages of endoderm formation, as it was previously shown to induce differentiation of hESCs to hepatocyte-like cells in vitro [15]. In certain cell types, sodium butyrate treatment has been shown to upregulate target genes for growth factors, signaling molecules, and transcription factors [45]. Although the exact mechanism by which sodium butyrate treatment in combination with activin A resulted in endoderm differentiation is unclear, it is tempting to speculate that exposure of hESCs to sodium butyrate triggered epigenetic changes, making these cells more responsive to activin A-mediated induction of endoderm-specific genes, such as Sox17 and Foxa2, as well as Pdx1, at later stages of differentiation. To date, we have observed that between 50% and 90% of endoderm cells expressed Sox17 and/or Foxa2, as measured by flow cytometric analysis (Fig. 2C). It is important to note that fetal calf serum, a reagent used in other studies [26, 28, 38], was not required in this protocol when activin A and sodium butyrate were used. Selection of growth factors and media supplements during differentiation was chosen based on current knowledge of pancreas development. We reasoned that the addition of activin A followed by bFGF might allow the development of pancreatic endoderm, since both molecules are produced by the notochord during embryonic dorsal pancreas development [24]. The later addition of EGF was designed to allow the growth and differentiation of pancreatic epithelial cells generated from the hESC-derived definitive endoderm. The inclusion of noggin was selected to further promote differentiation of definitive endoderm to pancreatic lineage by enhancing Pdx1 gene expression [46]. Withdrawal of bFGF was done to prevent hepatic development [25] and to facilitate pancreatic development, whereas the addition of nicotinamide [47] and IGF-II [48] induced maturation of pancreatic endocrine cells.

The expression of Pdx1 is instrumental for pancreas development, as both exocrine and endocrine components of the pancreas develop from Pdx1+ cells [19, 20, 49]. In our differentiation protocol, Pdx1 expression was first observed at day 15, and some of these Pdx1+ cells coexpressed Foxa2 and, to a lesser extent, Sox17, suggesting that the differentiation of Pdx1+ cells was preceded by the formation of definitive endoderm. Brolen et al. have characterized a similar Foxa2+/Pdx1+ pancreatic progenitor population, which differentiated in vivo to endocrine cells in the presence of angiogenic or other signals derived from the mouse pancreas [50]. These cells, however, failed to differentiate to insulin-producing cells in vitro. Recently, Lavon et al. determined that overexpression of Pdx1 and Foxa2 in hESC-derived EBs led to the differentiation of pancreatic cells in vitro, although insulin-producing cells were not observed [51]. These studies indicate that additional signals were necessary for pancreatic β-cell generation from hESCs whether direct differentiation or EB derivation methods were used. We observed that under the current culture conditions, hESCs could directly differentiate to definitive endoderm and, later, to pancreatic endoderm cells with Pdx1 expression (Fig. 2A, lower panel). However, this condition was not conducive for ILC generation, and three-dimensional aggregate formation was necessary to generate insulin-producing ILCs in our protocol (data not shown). In suspension cultures, we observed formation of bud-like structures, and the majority of Pdx1+ cells were located in these buds. At later stages of differentiation, insulin-, glucagon-, and somatostatin-expressing cells were also largely localized in these bud-like structures. The differentiation process described in this report is somewhat analogous to pancreatic development in vivo.

Recently, Xu et al. demonstrated that hESCs could spontaneously differentiate in FBS-containing medium to pancreatic progenitors and that some of these cells could further differentiate to insulin-producing cells using an EB-based protocol [38]. However, defined conditions and specific factors were not used to induce the differentiation of insulin-producing cells from hESCs. In fact, until recently, most of the published work on generation of insulin-producing cells from hESCs did not quantitate the percentage of β-like cells in their cultures, and more importantly, the functionality of these cells was not tested. Although direct differentiation can lead to efficient endoderm formation [27, 28] and generation of putative pancreatic progenitors, some of the conditions used were not suitable for generating insulin-secreting cells in vitro. Recently, D'Amour et al. have reported on generating pancreatic hormone-producing cells from hESCs [39]. However, the hESC lines used in the study were maintained on MEF feeder layers, and the differentiation was performed in the presence of fetal bovine serum. Although the hESC-derived insulin-positive cells produced C-peptide in response to multiple stimuli, their response to glucose was marginal [39]. The reason for this differential response to glucose observed with the insulin-producing cells generated by D'Amour et al. [39] and those reported in this publication is not clear. However, there are several notable differences in the methodologies by which these cells were differentiated. First, our protocol was performed largely in suspension culture, producing ILCs with bud-like structures, whereas the other protocol was based on adherent culture during the entire differentiation period. Second, our differentiation time of 36 days for ILC generation is almost twice as long as that reported in the other protocol [39]. Finally, the use of different growth factor combination to induce hESC differentiation by these two studies is likely to influence the characteristics of the insulin-producing cells.

The expression pattern of transcription factor genes during in vitro differentiation of hESCs to ILCs recapitulates some aspects of embryonic pancreas differentiation. A sequential expression of genes linked to pancreas development emerged when the temporal pattern of some of the key transcription factors was analyzed (Fig. 6). The sequential appearance of Sox17, Foxa2, Pdx1, Ptf1a, Ngn3, and Nkx6.1, followed by islet hormones, insulin, glucagon, and somatostatin, clearly suggests that the hESCs followed a progression similar to that of in vivo pancreas development [18]. It has been reported that Ngn3, which is critical for pancreatic endocrine development, is expressed only transiently in vivo [52]. The fact that Ngn3 expression persisted in these cells at the ILC stage suggests that many cells may still be at the endocrine progenitor stage of development and require additional signals to develop into mature β cells. Alternatively, since Ngn3 is also expressed in neuronal cells [53], it is conceivable that the hESC-derived ILCs also contained cells of neuronal lineages. Several types of homeodomain proteins, Pax4 (data not shown) and Nkx6.1, both of which have been shown to be important for differentiation of pancreatic endocrine cells and, especially, of β cells, were expressed during ILC development. In addition, Hlxb9, a transcription factor expressed in early pancreatic dorsal bud and again, later, in mature pancreatic β cells [18], was also expressed in a biphasic manner during the differentiation of ILCs.

hESC-derived ILCs released C-peptide in response to glucose stimulation in vitro in a static assay, suggesting that some of the insulin-secreting cells had acquired glucose sensing machinery and excitation properties. Since the number of insulin-positive cells and the total amount of insulin content in our ILC preparations were considerably lower than adult human pancreatic islets, biphasic insulin secretion is not expected. Addition of theophylline [40], an insulinotropic agent, increased the amount of C-peptide release further (data not shown), suggesting that the transport machinery of the ILCs is similar to β cells. We have observed that some C-peptide positive cells in hESC-derived ILCs coexpressed glucagon or somatostatin. To a much lesser extent, a few cells were found in our cultures that coexpressed all three islet hormones: insulin, glucagon, and somatostatin. Although the functionality of these cells is unknown, the existence of such cells has been reported with both mouse [54] and human [55] fetal pancreas. It is unclear at this time whether these cells represent pancreatic endocrine progenitors or an immature cell type belonging to the fetal stage of pancreas development. We also observed that our day 36 ILCs contained Pdx1-bright cells that did not express C-peptide, whereas the Pdx1 expression in the C-peptide+/Pdx1+ cells was lower (Fig. 4, Figure 4.B). It has been reported recently that during fetal pancreas development, epithelial progenitor cells expressed higher levels of Pdx1 than developing β cells [56]. Some of these Pdx1+ cells coexpressed CK-19, suggesting that our cultures also contained pancreatic ductal cells or their precursors [56]. In addition, ILCs differentiated from hESCs contained cells expressing the pancreatic exocrine enzyme, amylase (Fig. 6A). We believe this is the first report in which hESCs have been differentiated to the pancreatic lineage and in which all three major pancreas tissue-specific cell types are present in the cultures. Obviously, hESC-derived ILCs contained many other cell types, since insulin-producing cells are predominantly localized within the small ILCs. We have examined the expression of other developmentally regulated genes that are associated with hepatocyte (transthyretin [TTR] [15] and α-fetoprotein [AFP] [57]), intestine (Cdx2 [58]), cardiomyocyte (cardiac α-myosin heavy chain [αMHC] [9]), and neural cell (TUBB4 [59] and neurofilament [NF] [60]) development. Not surprisingly, we observed differential levels of expression of TTR, AFP, and Cdx2 genes (all endoderm-related), as well as low levels of NF and TUBB4 (neural) at various stages of ILC generation (data not shown). However, no αMHC (cardiomyocyte) gene expression was detected during the entire course of differentiation (data not shown). Additional studies are in progress to characterize the non-insulin-producing cell populations and to determine whether or not these cells are beneficial for β-cell differentiation from hESC.

In summary, using feeder-free cultures of hESCs in the presence of specific growth factors and signaling molecules, we have developed a reproducible method of generating ILCs with characteristics similar to those of pancreatic β cells. We have also established a method to isolate the clusters that are enriched in insulin-producing cells. Although many questions remain to be answered regarding the biochemical and metabolic characteristics of the ILCs, this protocol represents an important first step toward generating glucose-responsive, insulin-producing cells from hESCs. Thus, results presented in this study provide credence that the differentiation of hESCs to mature pancreatic β cells is indeed possible and can be used as a potential source for transplantation into diabetic patients to regulate glucose homeostasis.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


We thank Drs. Jane Lebkowski and Ray Rajotte (University of Alberta, Edmonton, AB, Canada) for their valuable suggestions during the course of the study and for critical reading of the manuscript, and we thank Drs. Cal Harley, Chunhui Xu, and Bonnie Cooper for critically reviewing the manuscript. We thank Patricia Lovelace for excellent technical assistance with flow cytometric analysis, and we also thank Dr. Joe Gold for helpful suggestions. G.K. is an Alberta Heritage Foundation for Medical Research Senior Scholar.