Induction of Differentiation of Embryonic Stem Cells into Insulin-Secreting Cells by Fetal Soluble Factors



Cell signals produced during pancreas embryogenesis regulate pancreatic differentiation. We show that the developing pancreas releases soluble factors responsible for in vitro endocrine pancreatic differentiation from embryonic stem cells (ESCs). A mouse D3 ESC line was transfected with a human insulin promoter/βgeo/phosphoglycerate kinase–hygromycin-resistant construct. To direct differentiation, cells were cultured for 7 days to form embryoid bodies and then plated for an additional 7 days. During this 14-day period, besides eliminating leukemia inhibitory factor, cells were cultured in low serum concentration with the addition of conditioned media from embryonic day–16.5 pancreatic buds. Islet cell differentiation was studied by the following: (a) X-gal staining after neomycin selection, (b) BrdU (bro-modeoxyuridine) studies, (c) simple and double immunohistochemistry for insulin, C-peptide, and glucose transporter 2 (Glut-2), (d) reverse transcription–polymerase chain reaction for insulin and pancreas duodenum homeobox 1 (PDX-1), (e) insulin and C-peptide content and secretion assays, (f) intraperitoneal glucose tolerance test, (g) electrophysiology (patch-clamp studies in inside-out configuration), and (h) transplantation of differentiated cells under the kidney capsule of streptozotocin-diabetic mice. The differentiated ESCs showed the following: changes in the mRNA levels of insulin and PDX-1; coexpression of insulin, C-peptide, and Glut-2; glucose and tolbutamide-dependent insulin and C-peptide release; K-channel activity regulated by ATP; and normalization of blood glucose levels after transplantation into diabetic mice and hyperglycemia after graft removal. In this study, we establish a battery of techniques that could be used together to properly characterize islet cell differentiation. Moreover, identification of factors released by the developing pancreas may be instrumental in engineering β cells from stem cells.


In recent years, several studies have described the differentiation of insulin-containing and -secreting cells from mouse embryonic stem cells (ESCs) [16]. Some of the protocols described so far have met with the following problems: (a) low insulin content and release [36], (b) high apoptotic rate [4], and (c) no capacity to normalize blood glucose in diabetic animals [3]. Thus, these methods are of questionable value for the large-scale production of differentiated islet cells.

One of the fundamental problems of past protocols has been their dependence on a limited number of factors, namely nicotinamide [1] and extrinsic factors known to expand the nestin-progenitors [3] and promote β-cell proliferation and differentiation. However, no new factors are being added to the list. Mature islet cells with regulated insulin release should possess glucose-sensing mechanisms, secretory machinery, and the capacity to synthesize and store proinsulin and process proinsulin to insulin. To meet these criteria, we believe it is critical to identify novel differentiation factors for use in the design of a protocol that combines several strategies for the generation of islet cells. In this regard, it is well known that pancreas development involves cell interactions with surrounding tissues, as well as between pancreatic precursor cells [7, 8]. In this paper, we tested for signals intrinsic to the pancreas that have been shown to be important for pancreatic development. We focus on the ability of soluble factors released during pancreas embryogenesis to induce β-cell differentiation from ESCs.

Materials and Methods

Generation of Vector and Transfection Protocol

A DNA molecule containing the human insulin promoter/βgeo (HIP/βgeo) gene and a phosphoglycerate kinase–hygromycin-resistant (pGK-hygror) gene in a pBSII-SK (Stratagene, La Jolla, CA, common vector was constructed. HIP gene was obtained by digestion with NcoI (Roche Applied Science, Madrid, Spain, from pHins300 cloned in pBSII-SK. Right orientation was verified after ligation. The βgeo fragment was obtained after double digestion of the pSA-βgeo plasmid with HindIII-XhoI and subcloned in pBS-HIP digested with HindIII-XhoI. This new construct (GB2) was then digested with XhoI and blunted. Finally, the construct was digested with KpnI to insert the pGK-hygro (SmaI-KpnI). The HIP/βgeo/pGK-Hygror construct was inserted into the SmaI-KpnI binding site that exists inside the LacZ gene of the pBluescriptks vector. The construct is depicted in Figure 1A. The plasmid was linearized with KpnI digestion and transfected in D3 ESCs by electroporation. Transfected cells were selected for growth in the presence of 200 μg/ml hygromycin (Calbiochem, San Diego,

Generation of Pancreatic Rudiment–Conditioned Medium

Pancreatic rudiments (PBs) were dissected from OF1 mouse embryos on embryonic day 16.5 (e16.5). PBs (n = 8) were then cultured for 10 days in 5 ml of D3 cell medium without leukemia inhibitory factor (LIF) and supplemented with 3% fetal bovine serum (FBS). Conditioned medium was then collected, filtered, and stored at −80°C. A pool of supernatants was used for each experiment after dilution (1:1) in D3 cell medium.

Cell Culture and In Vitro Differentiation Procedure

Undifferentiated GB2 transfected ESCs (D3 cell line) were cultured in Dulbecco's modified Eagle's medium (Gibco, Grand Island, NY, supplemented with 15% FBS (Gibco), nonessential amino acids (1%), 2-mercapto-ethanol (0.1 mM), L-glutamine (4 mM), sodium pyruvate (1 mM), penicillin (100 IU/ml), and streptomycin (0.1 mg/ml). The undifferentiated state was maintained by 1,000 U/ml recombinant LIF (Gibco). To direct the differentiation, hygromycin-resistant cells were grown for 7 days in suspension (1.5 × 106 cells/ml) in nonadherent Petri dishes to allow formation of embryoid bodies (EBs). During this period, FBS concentration was reduced to 3% and LIF was withdrawn as indicated (Fig. 1B). The EBs were then plated for an additional 7 days, and FBS was increased to 10%. During this 14-day period, the cells were treated with conditioned media from e16.5 PBs (dilution 1:1) (Fig. 1B). Finally, for ES-Ins/βgeo selection, the differentiated cultures were grown in the same differentiation medium in the presence of 2.3 mg/ml G418 (Gibco).

X-gal Staining

Cells were fixed in 4% paraformaldehyde for 5 minutes and washed with phosphate-buffered saline (PBS). After fixation, cells were incubated overnight at room temperature in the X-gal reaction solution and then washed with PBS.

Bromodeoxyuridine Incorporation

For bromodeoxyuridine (BrdU) staining, the cells were incubated at 37°C with 10 βM BrdU (Sigma, St. Louis, for 20 hours. Cells were then fixed with 4% paraformaldehyde for 5 minutes and washed with PBS, and the DNA was denatured with 2 M HCL for 30 minutes at room temperature. The rest of the protocol is a standard immunocytochemistry protocol. BrdU mouse monoclonal 1:500 (Sigma) was used as primary antibody, and anti-mouse tetra-methylrhodamine isothiocyanate (TRITC) 1:300 (Sigma) was used as secondary antibody. Nuclear staining was performed by adding 300 nM 4′,6-Diamidino-2-phenylindole (DAPI) (Sigma) for 5 minutes at room temperature before visualization.

Reverse Transcription–Polymerase Chain Reaction Analysis

RNA was isolated from differentiated ESCs as described by Chomczynski [9]. Total RNA (1 μg) was reverse-transcribed and cDNAs were amplified using the Superscript one-step reverse transcription–polymerase chain reaction (RT-PCR) kit (Gibco). Reactions (25 μl) containing mRNA-specific primers, 5 pmol each, were incubated at optimal annealing temperatures and subjected to 35 cycles of amplification. The PCR products were separated using 2% agarose gels and stained with ethidium bromide. The oligonucleotide pairs used for PCR and the size of the amplified products were as follows: insulin (55°C) TCCT-GCCCCTGCTGGCCCTGC (sense) and AGTTGCAGTAGT-TCTCCAG (anti-sense) (312 bp); pancreas duodenum homeobox 1 (PDX-1) (55°C) GACCAAGATTGTGCGGTGACC (sense) and GACCCCAGGTTGTCTAAATTGG (anti-sense) (451 bp). All experiments were carried out in triplicate, and the reproducibility of the observations was confirmed in three to four independent experiments.


A standard immunocytochemistry protocol was used. Cells were fixed with 4% paraformaldehyde for 4 minutes, washed with PBS, and permeabilized with 0.02% Triton X-100 overnight. Primary antibodies and dilutions were as follows: insulin mouse monoclonal (1:250; Sigma), C-Peptide guinea pig polyclonal (1:100; Linco Research, St. Charles, MO,, and glucose transporter 2 (Glut-2) rabbit poly-clonal (1:250; Chemicon International, Temecula, CA, Primary antibody localization was done using anti-mouse TRITC or fluorescein isothiocyanate (FITC) (1:200; Sigma), anti–guinea pig FITC (1:50; DakoCytomation, Glostrup, Denmark, and anti-rabbit TRITC (1:125; Sigma). Proper controls for secondary antibodies revealed no nonspecific staining. Cells were counter-stained with 300 nM DAPI (Sigma) for 5 minutes before visualization. Grafts were fixed with 4% paraformaldehyde for 2 hours at room temperature, embedded in sucrose gradients, and frozen in octamer binding transcription factor (OCT). Immunohistochemistry was performed in 7-μm tissue sections prepared by microtomy (Leica, Heerbrugg, Switzerland, using standard methods. Primary and secondary antibodies and dilutions were the same. The pancreata were fixed overnight in 4% paraformaldehyde at 4°C, embedded in OCT, and frozen in liquid nitrogen. Ten-micrometer tissue sections were taken. The endogenous peroxides activity was quenched by 0.3% hydrogen peroxide methanol for 60 minutes at room temperature. After washing with PBS, the slides were incubated with 3% normal goat serum (Sigma) for 60 minutes at room temperature. The slides were then incubated with polyclonal guinea pig anti-insulin (DakoCytomation) overnight at 4°C. After washing with PBS, the biotinylated anti-rabbit Immunoglobulin G (H + L) antibody (Vector Laboratories, Burlingame, CA, was added and incubated for 60 minutes at room temperature followed by streptavidin-biotin solution for another 30 minutes at room temperature. Finally, the slides were washed with PBS and stained with diaminobenzidine substrate kit (DakoCytomation) at room temperature for 1–2 minutes. The color reaction was finished by repeated washes in distilled water. The slides were counterstained with Meyer's hematoxylin for 10 minutes and then dehydrated and mounted with coverslips. Samples were analyzed using fluorescence (Olympus, Tokyo, and laser scanning (LSM-510; Carl Zeiss, Jena, Germany, microscopy.

Insulin and C-Peptide Secretion Assays

Secretion studies were performed as previously described [1]. In summary, 2.5 × 105 cells were cultured overnight in D3 culture medium supplemented with 10% FBS and PB–conditioned media in 24-well dishes. The cells were then washed three times with Krebs buffer for 5 minutes each and incubated for 4 hours in 500 μl of fresh modified Krebs buffer with 0.1% bovine serum albumin (BSA) and 3 mM glucose. The temperature of the Krebs buffer was held constant at 37°C and was continuously gassed with a mixture of O2 (95%) and CO2 (5%) for a final pH of 7.4. Afterward, the supernatant was discarded and the cells were incubated for 1 hour in 250 μl of the same Krebs buffer, at which point the culture supernatants were collected. Cells then received a final 1-hour incubation in 250 μl of the same fresh modified Krebs buffer, but containing 22 mM glucose. At the end of the incubation period, the buffer was collected. In the last two incubations, buffer supernatants were centrifuged at 3,000 rpm for 5 minutes. Insulin and C-peptide were assayed by radioimmunoassay (RIA) using two different Coat-a-Count kits (Diagnostics Products Corporation, Los Angeles, One kit was used to detect insulin and the other to detect C-peptide. All values were determined against a standard curve prepared with rat insulin or C-peptide. For measurement of total insulin and C-peptide cell content, cell pellets were sonicated in 1 mM acetic acid containing 0.1% BSA. In addition, cellular extract was also determined using RIA. Secretion was normalized for cell number by measuring total protein in each experiment with the method of Bradford.


Standard patch-clamp methods were used to record whole-cell currents and single-channel currents from inside-out membrane patches. Patch pipettes were pulled from Harvard Apparatus Ltd. (Kent, U.K., borosilicate glass capillaries using a two-stage puller (Mecanex SA, Nyon, Switzerland, with resistances in the range of 8–12 MΩ. During whole-cell experiments, the bath contained extracellular solution (values in mM): NaCl 140.3, KCl 5.4, MgCl2 1, CaCl2 2.5, and HEPES-NaOH, pH 7.4. The pipette was filled with an intracellular solution (values in mM): KCl 140, MgCl2 1, EGTA 1, ATP 0.3, and HEPES 10, pH 7.2. For inside-out patch recordings, the pipette was filled with (values in mM) 5.4 KCl, 140.3 NaCl, 10 HEPES, 2.5 CaCl2, and 1 MgCl2, pH 7.4. Bath solution contained (values in mM) 140 KCl, 1 MgCl2, 10 HEPES, and 1 EGTA, pH 7.2. Solutions containing ATP dipotassium salt were applied through an RSC-100 rapid solution changer (Bio-Logic–Science Instruments SA, Claix, France, KATP channel unitary currents were registered from excised membrane patches in the inside-out configuration [10]. Currents were measured using an Axopatch 200 amplifier (Axon Instruments/Molecular Devices Corp., Union City, CA, and stored in a tape recorder (DAT, DTR-1202; Bio-Logic–Science Instruments SA) for subsequent analysis using custom written software. Experiments were filtered through an eight-pole Bessel filter (Frequency Devices) at 3 kHz and sampled at 10 kHz by a Digidata 1200 (Axon Instruments/Molecular Devices Corp.). Pipette potential was held at 0 mV throughout the recording. In whole-cell experiments, cell membrane potential was held at −70 mV during the delivery of 10-mV hyper and depolarizing voltage pulses of 200 ms duration applied alternatively every 2 seconds. The experiments were carried out at room temperature (20°C–24°C).

Animal Transplantation Studies

Male Swiss albino (OF1) mice (B&K Universal Ltd, Hull, U.K.,, aged 8–12 weeks, were used as recipients of the implants. Animals were made diabetic 3–4 days before transplantation by a single i.p. injection of streptozotocin (STZ) (Sigma) 180 mg/kg of body weight freshly dissolved in citrate buffer (pH 4.5). Before implantation, diabetes was confirmed by the presence of blood glucose at concentrations higher than 300 mg/dl. Every 2 days, between 9 and 11 a.m., blood glucose was measured from the snipped tail of mice under fasting conditions using a portable glucose meter (Química Farmaceútica Bayer, Barcelona, Spain, For cell implantation, 5 × 106 insulin-secreting cells or undifferentiated ESCs were washed and resuspended in 25 μl of D3 culture medium supplemented with 10% FBS and, in the case of differentiated cells, PB–conditioned medium. Mice were anesthetized with an i.p. injection containing Ketamine hydrochloride 50 mg/kg, atropine 0.8 mg/kg, and diazepam 4 mg/kg. The left kidney was exposed through a lumbar incision, and cells were transferred under the kidney capsule using a blunt 30-gauge needle (Hamilton Bonaduz AG, Bonaduz, Switzerland, Grafts were removed after 13 days. The grafts were analyzed using immunohistochemistry to detect the presence of insulin-producing cells. For intraperitoneal glucose tolerance test (IPGTT), mice were given an i.p. injection of glucose (2 g/kg of body weight). Whole venous blood was obtained from the tail vein at 0, 30, 60, 90, 120, 180, and 210 minutes after the glucose injection. All experimental procedures involving animals were approved by the Miguel Hernández University Institutional and Animal Care Committee and performed in accordance with the guidelines set by the European Animal Care and Use Guidelines.


Insulin Is Not the Soluble Proteinic Factor Responsible for ESC Differentiation

To assess the differentiation potential of soluble factors released by embryonic pancreas, we cultured undifferentiated transfected D3 ESCs for 7 days to form EBs and then plated the cells for an additional 7 days. During this 14-day period, besides eliminating LIF, cells were cultured in low serum concentration with the addition of conditioned media from e16.5 PBs (Fig. 1B). After the differentiation protocol, insulin-producing cells were neomycin-selected and X-gal staining revealed that 95% of the cells were β-galactosidase–positive (Fig. 2E) (n = 4). Undifferentiated transfected D3 ESCs did not survive neomycin selection (n = 4). These data indicated that our cell-trapping system allowed for a good selection of the insulin-containing cells.

To demonstrate that soluble factors released by the PBs were proteins, undifferentiated transfected D3 ESCs were cultured with the same protocol, but with the conditioned media heated at 65°C for 90 minutes. In this condition, the cells did not survive neomycin selection (n = 4). Because insulin is a growth factor, it was important to study whether the differentiation potential of the conditioned media was due to the insulin released by the e16.5 PBs. Insulin release from PBs was 0.78 ± 0.14 ng/ml (n = 4). When undifferentiated transfected D3 ESCs were cultured with the same protocol (but with the addition of mouse insulin to the culture media) at the same concentration as the values found in the conditioned media, the cells did not survive neomycin selection. This suggests that insulin is not the soluble proteinic factor responsible for ESC differentiation, although we cannot exclude the possibility that released insulin may act in combination with other factors. Finally, e16.5 PBs do not contain or release C-peptide.

PB Soluble Factors Decrease Cell Proliferation

When searching for a candidate protocol for the production of insulin-releasing cells to be used for cell therapy, it is important to assess cell proliferation. The data provided by BrdU (Figs. 2A, 2B) incorporation indicated that D3 ESCs differentiated with conditioned media from PBs resulted in a 40% decrease in cell proliferation when compared with undifferentiated D3 embryonic stem transfected cells (Fig. 2C). Both D3 and PB cells were taken at passage 26. BrdU incorporation in native islet cells was 0.2%.

Co-culture with PB–Conditioned Medium Results in the Presence of Transcripts and Proteins Found in Normal β-Cell

Looking at the first level of our characterization, we investigated the expression of β-cell markers by RT-PCR analyses and immunofluorescence. RT-PCR analyses indicated that differentiated cells showed an increase in the expression levels of insulin and PDX-1 when compared with undifferentiated D3 transfected cells (Fig. 2D). Both markers were also expressed in mature β cells (Fig. 2D).

Insulin (Figs. 3A, 3B) and C-peptide (Fig. 3C) antibody stainings were present in 96% and 93% of the differentiated cells, respectively (n = 3). Moreover, 91% of differentiated cells coexpressed insulin and C-Peptide (n = 3) (Fig. 3E). In addition, 54% of the cells were positive for Glut-2 (Fig. 3D) (n = 3), and 48% of the cells coexpressed insulin and Glut-2 (Fig. 3F) (n = 3). The percentage of insulin-positive cells after culture with conditioned medium and before G418 selection was 18%. When undifferentiated D3 cells were cultured in the same conditions, but in the absence of conditioned medium, the proportion of insulin positive cells was 5%.

Differentiated ESCs Show Regulated Insulin and C-Peptide Release, Together with Functional K-Channel Activity Regulated by ATP

To further investigate the functional status of insulin-positive cells, we analyzed glucose-dependent insulin and C-peptide release. At the in vitro functional level, we first examined the differentiated cells for regulated insulin and C-peptide release. After neomycin selection, insulin and C-peptide content studies demonstrated that differentiated cells contained 2.7 ± 0.1 and 1.5 ± 0.3 ng/μg protein, respectively (n = 3). When we looked for regulated secretion, we observed that in the presence of 3 mM glucose, differentiated cells had a constitutive insulin release (n = 7) (Fig. 4A). Moreover, in the presence of 22 mM glucose, differentiated cells secreted insulin at 547 ± 71 pg/μg protein per 60 minutes (n = 7) (Fig. 4A). In addition, in response to 25 μM tolbutamide, in the presence of 3 mM glucose, insulin secretion was 416 ± 20 pg/μg protein per 60 minutes (n = 4) (Fig. 4A). Finally, in the presence of 22 mM glucose, differentiated cells secreted C-peptide at 606.8 ± 49.9 pg/μg protein per 60 minutes (n = 7). No insulin or C-peptide content or secretion was observed in undifferentiated D3 transfected cells (n = 5). The tolbutamide-induced insulin release suggested the possibility of the existence of functional ATP-regulated K channels, which is very important because it plays a crucial role in the coupling of electrical activity to energy metabolism in β cells [11]. To test this hypothesis, we performed whole-cell, as well as excised, patch recordings in differentiated cells. To study the inhibitory effect of ATP on channel activity, we used a protocol of alternating test solutions containing 2 mM ATP (15–30s) with a 60s exposure to a control solution (with no ATP). Figure 4B shows a sample of current changes in a membrane patch containing a minimum of two channels recorded shortly after patch isolation. In both our differentiated cells and in normal mouse pancreatic β cells (n = 10), ATP (2 mM) applied at the beginning of the record completely blocked K+ channel activity. When ATP was withdrawn, open channel probability recovered after 50s. On return to 2 mM ATP, channel activity was again inhibited (n = 8). No single-channel activity was observed in undifferentiated cells (n = 14) (Fig. 4B). In addition, the contribution of the ATP-sensitive K channels to the resting conductance was estimated by the addition of 100 μM tolbutamide. This antidiabetic agent diminished the resting conductance of differentiate cells from 541 ± 63 pS to 265 ± 40 pS (n = 8, p = .01). These data demonstrate the presence of functional ATP-sensitive K channels in our differentiated cells.

Transplantation Assays Maintain Glucose Levels in Diabetic Mice

The last part of the characterization of differentiated cells focused on in vivo experiments. To evaluate the capacity of differentiated cells to maintain normal glucose homeostasis, transplantation assays were performed. As is indicated in “Materials and Methods,” 5 × 106 undifferentiated transfected cells (○) (n = 9) and differentiated insulin-secreting cells (•) (n = 6) were transplanted under the kidney capsule of STZ-diabetic mice. All STZ-induced diabetic mice transplanted with undifferentiated D3 transfected cells had blood glucose levels higher than 450 mg/dl and died 12–15 days after STZ treatment (Fig. 5A). Two days after transplantation, all mice transplanted with differentiated cells showed blood glucose levels under 150 mg/dl. Normal blood glucose values were maintained throughout the study (Fig. 5A). This result is due to the presence of our transplanted differentiated cells, as no pancreas regeneration was observed in STZ-diabetic mice (Fig. 5G; n = 3). In addition, some animals (▴) (n = 3) that were kept transplanted for 6 weeks were normoglycemic for the entire 6-week period (Fig. 5A). Although D3 cells come from 129/Sv+c mice and are transplanted into OF1 mice, our preliminary data indicate that after differentiation these cells have no expression of MHC-1 (major histocompatibility complex class 1) molecules on their surface (data not shown). This could explain the lack of graft rejection. All nondiabetic animals transplanted with undifferentiated D3 transfected cells presented tumors 15–20 days after engraftment (n = 10). In contrast, none of the mice transplanted with differentiated cells developed tumors, even 6 weeks after engraftment. Two days after graft removal, blood glucose levels increased to values over 350 mg/dl (n = 6), further indicating that rescue of the diabetic phenotype was due to transplanted cells (Fig. 5A). To assess the ability of the transplanted mice to dispose of a glucose load, an IPGTT test was performed. As noted in Figure 5B, fasting plasma glucose levels were the same for both control nondiabetic and transplanted mice. Upon glucose challenge, transplanted mice showed no significantly higher plasma glucose levels than control nondiabetic animals. Recovery of euglycemia was the same in transplanted mice with respect to nondiabetic mice. Finally, we studied the presence of insulin-positive cells in undifferentiated and differentiated transplanted cells. Immunohistological studies showed 80% insulin-positive cells in removed grafts from differentiated cells (n = 6). In contrast, no insulin-positive cells were observed in grafts from undifferentiated cells (n = 5). Finally, 81% of the cells from removed grafts tested positive for β-galactosidase (data not shown).


In this study, we demonstrate that PBs release soluble factors responsible for in vitro endocrine pancreatic differentiation from ESCs. The differentiated cells produced using this strategy showed a significant upregulation of certain genes involved in β-cell development. In addition, the protocol generated a large quantity of cells coexpressing insulin and C-peptide. Moreover, half of the cells obtained coexpressed two proteins essential for the stimulus-secretion coupling process: Glut-2 and insulin. Finally, differentiated cells also expressed the transcription factor PDX-1, which is very important for stimulus-secretion coupling. The observation that differentiated cells coexpress insulin and C-peptide is very important, as has been previously indicated by other authors [12]. The fact that the substitution of insulin for conditioned medium in the differentiation protocol does not result in differentiated cells suggests that our differentiation effect is mostly due to soluble factors released by PBs.

To continue the characterization of differentiated cells, we show that the insulin and C-peptide content of differentiated cells is enough to ensure a good insulin and C-peptide release: 14% (for insulin) and 8% (for C-peptide) of the levels found in normal mouse pancreatic islets [13]. Taking into account that β cells release approximately 0.5%–1% of their insulin content [14], the differentiated cells have enough insulin and C-peptide to effectively resolve in vivo control of glucose homeostasis. In contrast with our previous reports [1, 2], these cells showed a better secretagogue-induced insulin and C-peptide response, with the values observed perfectly matching the glucose and tolbutamide insulin and C-peptide responses found in native β cells. Two facts support the idea that these differentiated cells have a truly regulated release instead of merely insulin uptake from the media: (a) the cells have C-peptide release, and (b) they showed tolbutamide-induced insulin secretion.

These cells additionally are characterized by the presence of functional ATP-dependent K channels, which are known to play an important role in the secretory response after glucose entry and metabolism in the β cell. These data show an improvement in these differentiated cells with respect to our previous published results [1, 2] and suggest a fine-tuned, in vivo, regulated insulin release. In fact, the in vitro glucose-induced insulin release of these cells was twice what was obtained in previous reports.

All these findings indicate that the processes mediating the sensing and secretion of insulin are well developed in these differentiated cells. For all these reasons, transplantation of differentiated cells is able to maintain the glucose homeostasis, even upon a glucose challenge. Moreover, tumor formation is an important aspect to consider when thinking about cell therapy based on ESCs. The fact that no tumors were observed, together with the decrease in the BrdU incorporation, indicates that the cells were well differentiated and that the modifications in the cell-trapping system were very efficient. Finally, all these data demonstrate, as we previously reported [15], that once the cells are differentiated, they survived, produced insulin, and remained differentiated.

Recent studies [12, 16] have called into question some of the criteria used to accept differentiated cells as islet cells and have suggested the adoption of new criteria for these processes. Therefore, we studied (a) at the structural level, the presence of transcripts and proteins found in the β cell, (b) at the in vitro functional level, the nutrient and non–nutrient-induced insulin and C-peptide release, together with the existence of functional K-channel activity regulated by ATP, and (c) at the in vivo level, transplantation assays for β-cell function, as well as graft studies. Thus, phenotypic characterization of islet β cells should include at least the characterization of glucose-sensing mechanisms, the secretory machinery, and the synthesis, storage, and processing of proinsulin. Finally, the absence of tumors, together with in vivo insulin release, consistently explains blood glucose normalization.

Ultimately, we anticipate that a combination of factors, including soluble factor from fetal PBs, may allow for the development of in vitro differentiation protocols to generate islet cells from ESCs.

Figure Figure 1..

In vitro differentiation-selection protocol. (A): Structure of the HIP/βgeo/pGK-Hygror construct. (B): Scheme showing the differentiation-selection protocol. Abbreviations: EB, embryoid body; FCS, fetal calf serum; HIP, human insulin promoter; Hygror, hygromycin-resistant; LIF, leukemia inhibitory factor; Monol, monolayer; Neor, neomycin-resistant; PB, pancreatic rudiment; pGK, phosphoglycerate kinase.

Figure Figure 2..

D3 embryonic stem differentiated neomycin-selected cells were less proliferate. Fluorescent images showing nuclear staining with DAPI (A) and BrdU (B). Arrows indicate negative cells for BrdU. Scale bars = 20 7μm. (C): Percentage of BrdU incorporation in undifferentiated transfected cells (D3) and differentiated cells (PB). (D): Reverse transcription–polymerase chain reaction analysis of pancreas (P), undifferentiated transfected cell (D3) and differentiated cell (PB) gene expression for insulin (ins) and PDX-1. (E): Neomycin-selected cells showing X-gal staining (n = 4). Scale bar = 50 μm. Each value represents mean ± SEM of three experiments. Statistical significance was tested by Student's t-test: *p < .01 when compared with differentiated cells (PB). Abbreviations: BrdU, bromodeoxyuridine; DAPI, 4′,6-Diamidino-2-phenylindole; PB, pancreatic rudiment; PDX-1, pancreas duodenum homeobox 1.

Figure Figure 3..

Differentiated cells express several pancreatic β-cell markers. Confocal images of cells positive for insulin (A, B), C-peptide (C), and glucose transporter 2 (Glut-2) (D). Double immunofluorescence staining for insulin (red) and C-peptide (green) (E) and for insulin (green) and Glut-2 (red) (F). Scale bars = 20 μm.

Figure Figure 4..

Differentiated cells show physiological pancreatic β-cell properties. (A): Insulin secretion of undifferentiated transfected cells (D3) and differentiated cells (PB) in response to 3 mM (open bars) and 22 mM (black bars) glucose. For PB+T group, differentiated cells were incubated with 3 mM glucose (open bar) and 3 mM glucose plus 25 μM tolbutamide (black bars). Each value represents mean ± SEM of four to seven experiments. Statistical significance was tested by Student's t-test: *p < .01 when compared with 3 mM glucose. (B): Inhibition of single-channel currents by ATP. Single-channel currents were recorded from an inside-out patch at a membrane potential of 0 mV. The arrowheads indicate current level when the channel is closed. Top: a representative trace from an undifferentiated cell is shown (n = 14). Middle: two representative traces from a differentiated cell are shown (n = 8). Bottom: a representative trace from a mouse pancreatic β cell is shown (n = 10). Top and middle panels have the same scale. Abbreviations: IRI, immunoreactive insulin; PB, pancreatic rudiment; T, tolbutamide.

Figure Figure 5..

Transplantation of differentiated insulin-secreting cells into STZ-treated diabetic mice normalize blood glucose levels. (A): Shown are overnight fasted blood glucose levels in STZ-treated mice transplanted, under the kidney capsule, with 5 × 106 undifferentiated trans-fected cells (○) (n = 9) and 5 × 106 differentiated insulin-secreting cells (•) (n = 6). Left arrowhead indicates the transplantation day and right arrowhead indicates the graft removal day. Transplants were maintained for 6 weeks in three animals (6). (B): Intraperitoneal glucose tolerance test in overnight fasted nondiabetic mice (○) (n = 6) and mice 12 days after transplant with 5 × 106 differentiated insulin-secreting cells (•) (n = 6). Values are means ± SEM. Fluorescent images of grafts from differentiated cells showing nuclear staining with DAPI (C) and insulin staining (D). Fluorescent images of grafts from undifferentiated cells showing nuclear staining with DAPI (E) and unspecific staining (F). Scale bars = 50 μm. (G): Immunohistochemistry for insulin expression in pancreas from transplanted mice after graph removal (n = 3). Scale bar = 100 μm. Abbreviations: DAPI, 4′,6-Diamidino-2-phenylindole; STZ, streptozotocin.


This work was supported by the Generalitat Valenciana (grant no. CTIDIB/2002/130), the Ministerio de Ciencia y Tecnología (grant nos. GEN2001-4748-C05-05, PM 99-412, and SAF2003-03307), the Instituto de Salud Carlos III (grant nos. G03/171, G03/210, and G03/212), and European Community (grant no. QOL-2001-3). We thank N. Illera for her technical assistance, S. Ingham for editing the figures, and M. Harvey for text revision. G.B. is supported by Fundació La Marató de TV3, and P.V. by a fellowship from the European Foundation for the Study of Diabetes. P.V. and F.M. contributed equally to this work.


The authors indicate no potential conflicts of interest.