Division of Stem Cell Biology, Institute of Molecular Embryology and Genetics, Kumamoto University, Kumamoto, Japan
21st Century Center of Excellence, Kumamoto University, Kumamoto, Japan
Global Center of Excellence, Kumamoto University, Kumamoto, Japan
Division of Stem Cell Biology, Department of Regeneration Medicine, Institute of Molecular Embryology and Genetics, Kumamoto University, Honjo 2-2-1, Kumamoto 860-0811, Japan. Telephone: 81-96-373-6620; Fax: 81-96-373-6807
The generation of specific lineages of the definitive endoderm from embryonic stem (ES) cells is an important issue in developmental biology, as well as in regenerative medicine. This study demonstrates that ES cells are induced sequentially into regional-specific gut endoderm lineages, such as pancreatic, hepatic, and other cell lineages, when they are cultured directly on a monolayer of mesoderm-derived supporting cells. A detailed chronological analysis revealed that Activin, fibroblast growth factor, or bone morphogenetic protein signals are critical at various steps and that additional short-range signals are required for differentiation into Pdx1-expressing cells. Under selective culture conditions, definitive endoderm (47%) or Pdx1-positive pancreatic progenitors (30%) are yielded at a high efficiency. When transplanted under the kidney capsule, the Pdx1-positive cells further differentiated into all three pancreatic lineages, namely endocrine, exocrine, and duct cells.
Disclosure of potential conflicts of interest is found at the end of this article.
The definitive endoderm is derived from the extreme anterior end of the primitive streak in mice . Studies in other animal models demonstrate that the definitive endoderm and mesoderm arise from a common bipotent progenitor, the mesendoderm. The differentiation of the definitive endoderm involves a process in which either an endoderm or a mesoderm fate is selected. Information on the genes that influence the formation of definitive endoderm in the mouse comes largely from the analysis of gene-manipulated embryos . However, the destruction of genes that function in the formation of mesoderm or endoderm often results in defective gastrulation and failure in the formation of endoderm and mesoderm . Furthermore, the small size and limited number of cells in early murine embryos make interventional studies difficult; therefore, in vitro studies are needed to reveal the molecular mechanisms.
Embryonic stem (ES) cells can be cultured indefinitely in an undifferentiated state and stimulated to differentiate into various cell types. These cells can be used for the in vitro dissection of early inductive processes and therefore provide a useful tool for developmental biological studies. In addition, ES cells are regarded as a potential source of specific cell populations for cell replacement therapy. Previously, several reports claimed the generation of insulin-producing cells from ES cells via the differentiation of progenitors that express nestin [4, 5]. Other researchers reported that this insulin staining resulted from an insulin uptake from the medium after using the same procedure to generate nestin-expressing cells . This highlights the importance of generating pancreatic cells from ES cells by following normal developmental processes. Although several groups have reported the in vitro generation of definitive endoderm cells [7, , , , –12] and insulin-producing cells  from mouse and human ES cells, the molecular mechanisms for each inductive process are still not known.
Pdx1 (Pancreatic and duodenal homeobox gene 1) is a regional endoderm marker whose expression marks the dorsal and ventral pancreatic buds, as well as a part of the stomach and duodenal endoderm . In this study, we established a novel procedure for inducing Pdx1-expressing gut regional-specific cells from ES cells, following a pathway that appeared to mimic the in vivo developmental processes in vitro. The high efficiency for generating Pdx1-expressing cells makes it feasible to elucidate the underlying molecular mechanisms of the inductive processes, thereby enabling manipulation of the direction of ES cell differentiation. The results also show that the addition of Activin and basic fibroblast growth factor (bFGF) increased the yield of the definitive endoderm (47%) and Pdx1-expressing cells (30%) in the culture. These Pdx1-expressing cells further differentiated to express insulin, glucagon, pancreatic polypeptide, somatostatin, amylase, or Dolichos biflorus agglutinin (DBA) when grafted under a kidney capsule, demonstrating the potential of these cells to turn into all pancreatic lineages, namely endocrine, exocrine, and duct cells. Therefore, it is feasible that this procedure can be used for regenerative medicine in the future.
Materials and Methods
ES Cell Lines
The ES cell line SK7, containing a Pdx1 promoter-driven green fluorescent protein (GFP) reporter transgene, was established by culturing blastocysts obtained from transgenic mice homozygous for the Pdx1/GFP gene . The SK7 ES cell line was maintained on mouse embryonic fibroblast (MEF) feeders in Glasgow minimum essential medium (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) supplemented with 1,000 units/ml leukemia inhibitory factor (LIF; Chemicon, Temecula, CA, http://www.chemicon.com), 15% Knockout Serum Replacement (KSR; Gibco, Grand Island, NY, http://www.invitrogen.com), 1% fetal bovine serum (FBS; HyClone, Logan, UT, http://www.hyclone.com), 100 μM nonessential amino acids (NEAA; Invitrogen), 2 mM l-glutamine (l-Glu; Invitrogen), 1 mM sodium pyruvate (Invitrogen), 50 units/ml penicillin and 50 μg/ml streptomycin (PS; Invitrogen), and 100 μM β-mercaptoethanol (β-ME; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com). The ES cell lines R1, J1, and Pdx1/lacZ ES  were maintained on MEF feeders in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) supplemented with LIF, 10% FBS, NEAA, l-Glu, PS, and β-ME.
Supporting Cell Lines
The mesonephric cell line M15, which expresses WT1, was established from mouse mesonephros overexpressed with the large T protein of polyoma virus under the control of the early viral enhancer. M15  were kindly provided by Dr. T. Noce (Mitsubishi Kagaku Institute of Life Science, Tokyo, Japan) and Dr. M. Rassoulzadegan (University of Nice-Sophia Antipolis, Antipolis, France). ST2  and OP9  were kindly provided by Dr. S. Nishikawa (Center for Development Biology [CDB], RIKEN, Kobe, Japan). PA6  was kindly provided by Dr. Y. Sasai (CDB), respectively. MEF was isolated from embryonic day (E) 12.5–14.5 mouse embryo. M15 and PA6 were grown in DMEM supplemented with 10% FBS. ST2 were grown on RPMI medium supplemented with 5% FBS and 100 μM β-ME. OP9 were grown in α-modified Eagle's medium (Invitrogen) supplemented with 10% FBS. M15 and MEF were treated with mitomycin C (Sigma-Aldrich) at 10 μg/ml for 2.5 hours before use. M15, OP9, and PA6 were seeded on gelatin-coated 24-well plates at 2 × 105 cells per well. ES cells were then plated on the supporting cells when the latter were confluent.
Differentiation of ES Cells
For the differentiation studies, ES cells were transferred to gelatin-coated plates without MEF and cultured for 2 days to remove the MEF. With Pdx1/lacZ ES cells, embryoid bodies (EB) were generated for 2 days prior to plating on supporting cells. For other ES cell lines, EB formation was omitted, unless otherwise mentioned. ES cells were seeded at 5,000 cells per well in 24-well plates plated with supporting cells. The cells were then cultured in differentiation medium (DMEM supplemented with 10% FBS, NEAA, l-Glu, PS, and β-ME) for up to 12 days. Media were replaced every 2 days with fresh differentiation medium. In the case of PA6, differentiation medium containing 10% KSR instead of FBS was used.
For the transfilter assays, the ES cells were seeded at 20,000 cells per well in gelatin-coated six-well plates and cultured either with or without a Cell Culture Insert (BD Biosciences, San Diego, http://www.bdbiosciences.com) that had been previously plated with M15. For differentiation studies under serum-free conditions, ES cells were cultured in ITS medium (DMEM supplemented with 10 μg/ml insulin [Sigma-Aldrich], 5.5 μg/ml transferrin [Sigma-Aldrich], 6.7 pg/ml selenium [Sigma-Aldrich], and 0.25% Albmax [Invitrogen]). For differentiation studies using fixed M15, a monolayer of M15 treated with 4% paraformaldehyde (PFA) for 30 minutes at room temperature and washed several times with phosphate-buffered saline was used. For long-term culture, ES cells were grown on M15 in presence of 10 ng/ml Activin and 5 ng/ml bFGF on days 0–8 and then switched to low-glucose (1,000 mg/ml) medium containing 10% KSR, 10 ng/ml Activin, 5 ng/ml bFGF, and 10 mM nicotinamide (Sigma-Aldrich) at day 8.
Growth Factors and Inhibitors
The following concentrations were used: recombinant human Activin-A (R&D Systems Inc., Minneapolis, http://www.rndsystems.com), 10 or 100 ng/ml; human bFGF (Peprotech, Rocky Hill, NJ, http://www.peprotech.com), 5 ng/ml; recombinant human bone morphogenetic protein 7 (BMP7; R&D Systems), 25 ng/ml; SB431542 (Sigma-Aldrich), 1 or 5 μM; SB203580 (Calbiochem, San Diego, http://www.emdbiosciences.com), 1 μM; SU5402 (Calbiochem), 10 mM; recombinant mouse Noggin/Fc chimera (R&D Systems), 100 ng/ml; U0126 (Sigma-Aldrich), 10 μM; LY294002 (Sigma-Aldrich), 3 μM; retinoic acid (Sigma-Aldrich), 10−6 M; LE540 (Wako Chemical, Osaka, Japan, http://www.wako-chem.co.jp/english), 1 or 10 μM.
RNA was extracted from ES cells using TRI Reagent (Sigma-Aldrich) or the RNeasy micro-kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) and then treated with DNase (Sigma-Aldrich). Three micrograms of RNA was reverse-transcribed using Moloney Murine Leukemia Virus reverse transcriptase (Toyobo, Osaka, Japan, http://www.toyobo.co.jp/e) and oligo(dT) primers (Toyobo). The primer sequences and numbers of cycles are shown in supplemental online Table 1. The polymerase chain reaction (PCR) conditions for each cycle were as follows: denaturation at 96°C for 30 seconds, annealing at 60°C for 2 seconds, and extension at 72°C for 45 seconds. Reverse transcription (RT)-PCR products were separated by 5% nondenaturing polyacrylamide gel electrophoresis, stained with SYBR Green I (Molecular Probes, Eugene, OR, http://probes.invitrogen.com), and visualized using a Gel Logic 200 Imaging System (Kodak, Rochester, NY, http://www.kodak.com).
The following antibodies were used: goat anti-Amylase (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com), goat anti-albumin (Sigma-Aldrich), mouse anti-βIII-tubulin (Sigma-Aldrich), mouse anti-Cdx2 (BioGenex, San Ramon, CA, http://www.biogenex.com), guinea pig anti-C-peptide (Linco Research, St. Charles, MO, http://www.lincoresearch.com), biotin-conjugated DBA lectin (Sigma-Aldrich), goat anti-Foxa2 (Santa Cruz Biotechnology), mouse anti-GFP (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com), rabbit anti-GFP (MBL International Corp., Woburn, MA, http://www.mblintl.com), mouse anti-glucagon (Sigma-Aldrich), rabbit anti-Nanog (ReproCELL, Tokyo, http://www.reprocell.com/en), rabbit anti-Nkx2.1 (Santa Cruz Biotechnology), mouse anti-Oct3/4 (Santa Cruz Biotechnology), rabbit anti-pancreatic polypeptide (Dako, Glostrup, Denmark, http://www.dako.com), rabbit anti-Pdx1 (Transgenic, Kobe, Japan, http://www.transgenic.co.jp), rabbit anti-P-Smad1 (Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com), rabbit anti-P-Smad2/3 (Santa Cruz Biotechnology), goat anti-Sox17 (Santa Cruz Biotechnology), mouse anti-SSEA1 (R&D Systems), biotin-conjugated anti-SSEA4 (R&D Systems), goat anti-somatostatin (Santa Cruz Biotechnology), goat anti-T (Santa Cruz Biotechnology), and rabbit anti-T (Santa Cruz Biotechnology). Secondary antibodies used were Alexa 568-conjugated, Alexa 488-conjugated antibodies (Molecular Probes) and biotin-conjugated antibodies (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). For Pdx1 staining, the ImmunoPure ABC Kit (Pierce, Rockford, IL, http://www.piercenet.com) and the TSA System (PerkinElmer Life and Analytical Sciences, Boston, http://www.perkinelmer.com) were used for amplification of signals. Cells were counterstained with 4,6-diamidino-2-phenylindole (DAPI) (Roche Diagnostics).
Flow Cytometry Analysis
The following antibodies were used: either biotin- or Alexa 488-conjugated anti-E-cadherin monoclonal antibody (mAb) ECCD2, phycoerythrin-conjugated anti-Cxcr4 mAb (BD Biosciences, San Diego, http://www.bdbiosciences.com), biotin-conjugated anti-platelet-derived growth factor receptor α (PDGFRα) mAb, and streptavidin-allophycocyanin (BD Biosciences). The stained cells were analyzed with a FACSCanto (BD Biosciences) or purified with a FACSAria (BD Biosciences). Data were recorded with the BD FACSDiva program (BD Biosciences) and analyzed using the FlowJo program (Tree Star, Ashland, OR, http://www.treestar.com).
Transplantation Under Kidney Capsule
All animal studies were performed in accordance with the guidelines of the Kumamoto University Experimental Animal Institution. The ES cells were cultured for 8 days on M15 with Activin and bFGF. After 8 days culture, the ES cells were dissociated, and either the whole population or the Pdx1/GFP+ cells sorted out by flow cytometry on day 8 were used. The dissociated ES cells were replated onto methacryloyloxyethylphosphorylcholine-coated dishes, cultured for an additional day, harvested, and grafted under the left subcapsular renal space of 10-week-old male C.B-17/Icr-scid/scid Jcl mice (CLEA Japan, Tokyo, http://www.clea-japan.com) at approximately 2 × 106 cells per mouse. Transplantations were performed in triplicate for each condition. Two weeks after transplantation, the grafts were recovered, fixed, and analyzed.
Quantitative Analysis of C-Peptide-Positive Cells in Grafts
In the case of grafts of Pdx1+ cells, the grafts were approximately 500 μm thick (Fig. 7A, GFP+). Grafts were sectioned to 10 μm thick, and a total of approximately 50 sections were constructed per graft. Next, every 10th section (for a total of 5 sections per graft) was stained with anti-GFP or anti-C-peptide antibodies, respectively. Sections were counterstained with DAPI, and total cells per section (approximately 1 × 104 cells) were counted manually. The percentages of GFP+ (single-positive) or GFP+/C-peptide+ (double-positive) cells in each graft (Fig. 7C) were calculated by dividing the sum of the positively stained cell numbers by the sum of total cell numbers and are shown in Figure 7D.
GeneChip Expression Analysis
The biotinylated cRNAs from M15, OP9, and PA6 were hybridized with a MOE430A series of probe array (GeneChip; Affymetrix, Santa Clara, CA, http://www.affymetrix.com). The fluorescence intensity of each probe was quantified using the GeneChip Analysis Suite 5.0 computer program (Affymetrix). Data from individual arrays were normalized, and expression analysis was performed with the GeneSpring GX program, version 7.3 (Agilent Technologies, Palo Alto, CA, http://www.agilent.com). Data declared absent in M15 were removed. The filtered genes, which showed an over two-fold increase in the expression level in M15 cells in comparison with PA6 and OP9, were analyzed for functional gene clusters using NetAffx Analysis Center and GeneSpring. The Gene Ontology (GO) Mining Tool, used in the NetAffx Analysis Center, matches GeneChip probe sets to the annotated genes.
M15, a Mesoderm-Derived Cell Line, Directs ES Cells to Differentiate into Regional-Specific Cells of the Definitive Endoderm Lineages
Our previous studies have described a cocultivation method using a pancreatic rudiment or a pancreatic mesenchyme that directs ES cells to differentiate into Pdx1-expressing regional-specific gut endoderm . However, the requirement for embryonic materials allows a limited production and analysis of the ES cell-derived differentiated cells. Given that the embryonic endoderm requires signals from the adjacent germ layers for subsequent regionalization into specific endoderm organs , a cultured cell line derived from the mesoderm could probably substitute for the embryonic materials and support the differentiation of the ES cells into pancreatic tissue. To screen primary culture cells or cell lines of mesodermal origin for activities promoting differentiation of ES cells into Pdx1-expressing cells, an ES cell line with the lacZ reporter gene inserted into the Pdx1 locus was used . Stromal cell lines, such as OP9 , ST2 , and PA6 , which are known to promote mesoderm or neural differentiation of ES cells, were also tested. Among the cell lines tested, a mesonephric cell line, M15 , consistently demonstrated the strongest ability to induce Pdx1/lacZ-expressing cells from ES cells (Fig. 1A; supplemental online Fig. 1). MEF, bone marrow stromal cell line 9-15c , and ST2 showed moderate ability, whereas COS, STO, OP9, and PA6 showed no promoting ability (supplemental online Fig. 1). The activity of MEF in supporting endoderm differentiation seems consistent with a previous report . However, in light of the extent and the reproducibility of the supporting activities, M15 was used for the subsequent differentiation experiments.
To monitor the expression of Pdx1 in living ES cells, an ES cell line, SK7, was established from a transgenic mouse line, P#48.9 , bearing a GFP reporter driven by the promoter for Pdx1. Pdx1-expressing cells from the embryo, 10.5 days post coitum (E10.5), are considered to be pancreatic progenitor cells; they have consistently been demonstrated to give rise to both the exocrine and endocrine pancreas and the duct . In this transgenic mouse line, the mouse Pdx1 promoter, which recapitulates the endogenous Pdx1 expression, was used to drive the expression of GFP . SK7 expressed undifferentiated ES cell-specific markers such as Oct 3/4, Nanog, SSEA-1, and E-cadherin (supplemental online Table 2). Karyotype analysis indicated that SK7 had normal 40 diploid chromosomes with no apparent abnormalities, thus demonstrating its diploid nature. These characteristics were consistent with those previously reported for mouse ES cell lines. Then, SK7 was used in the following differentiation assays.
Figure 1B shows a representative image of the differentiated SK7 ES cells grown on M15 on day 8 after the initiation of differentiation. Strong GFP expression was visible on the edge of the colony (Fig. 1B). RT-PCR analysis showed that the Pdx1 transcript was detected on days 8 and 12 (Fig. 1C) of differentiation in SK7 ES cells, as well as two other mouse ES cell lines, R1 and J1, when cultured on M15 (Fig. 1D).
Immunocytochemical studies revealed the presence of Nkx2.1+ lung and thyroid cells, albumin+ hepatocytes, and Cdx2+ intestinal cells after day 10 of differentiation. These lineage-specific cells were detected in colonies distinct from the Pdx1/GFP+ colonies (Fig. 1E). These results suggest that ES cells are directed to differentiate into regionally specific, definitive endoderm-derived organs, namely pancreas, lung, liver, and intestine.
ES cells grown on M15, OP9, or PA6 were assayed for the expression of early germ layer-specific markers using RT-PCR (Fig. 1F). The ES cells are small and could be separated from the M15, OP9, and PA6 by flow cytometry (FCM) (supplemental online Fig. 2B). ES cells differentiated for 8 days were purified by FCM to exclude all M15, OP9, or PA6 and were used for the RT-PCR analysis. Definitive endoderm molecular markers, such as Sox17, Mixl1, Hnf4, Afp, and Ttr, were either induced specifically or expressed at a higher level in ES cells grown on M15 than those grown on OP9 or PA6 (Fig. 1F). RT-PCR controls of the feeder cells confirmed that most molecular markers either are not expressed at all or are expressed at much lower levels. Supplemental online Figure 2A shows that the Sox17 protein was confirmed to be expressed in differentiated ES cells but not in M15. And all Pdx1/GFP-positive cells expressed Sox17. The results indicated that M15 specifically direct ES cells to differentiate into definitive endoderm lineages. This is the first evidence that a specific cell line is capable of supporting the generation of definitive endoderm cell lineages from ES cells. All the following differentiation assays were performed on M15.
Differentiation of ES Cells into Definitive Endoderm Lineages is a Multistep Process
A detailed chronological analysis by immunocytochemistry was carried out to determine the time course of the generation of germ layer-specific cells (Fig. 2). First, the ES cells grown on M15 were stained for Foxa2, an endoderm marker, and T, a marker of mesoderm progenitors. On day 3, Foxa2+/T+ (F+/T+) cells appeared. On day 4, most of these F+/T+ cells, which are suggested to represent mesendoderm cells, segregated into Foxa2+/T− (F+/T−) cells and Foxa2-/T+ (F−/T+) cells, with a concurrent decrease of F+/T+ cells. FCM was performed to define the F+/T+ and F+/T− populations by detecting E-cadherin+/PDGFRα+ mesendoderm  and E-cadherin+/Cxcr4+ definitive endoderm  (Fig. 2C, 2D). FCM analysis revealed that E-cadherin+/PDGFRα+ mesendoderm (5%) first became detectable on day 3 (Fig. 2C) and that E-cadherin+/Cxcr4+ definitive endoderm (6%) arose on day 4 (Fig. 2D), which coincided with the appearance of the F+/T+ and F+/T− populations. On day 6, F+/T− cells remained, whereas F−/T+ cells almost disappeared (Fig. 2A). The expression of Pdx1/GFP first appeared on day 6. The Pdx1/GFP+ cells increased and reached a plateau on day 8. Almost all the Pdx1/GFP+ cells were also positive for Foxa2 (Fig. 2B). FCM analysis revealed that on day 8, E-cadherin+/Cxcr4+ definitive endoderm cells increased up to 10% of total cells, and 25% of these cells were Pdx1/GFP-positive (Fig. 2D). Almost all Pdx1/GFP-positive cells were derived from E-cadherin+/Cxcr4+ cells, thereby indicating their definitive endoderm origin (N.S., unpublished results).
The above results showed that M15 induced ES cells to differentiate into definitive endoderm cells and then Pdx1-expressing cells. Based on the expression patterns of these molecular markers, the entire process of induction was subdivided into three critical phases of fate decision events: an early phase (days 0–3), when the ES cells differentiate into bipotential mesendoderm; a middle phase (days 3–4), when the mesendoderm segregates into endoderm; and a late phase (days 4–8), when the decision to differentiate into gut regional-specific endoderm derivatives is made.
To investigate the molecular mechanism of pancreatic induction by M15, microarray analysis was performed with M15, PA6, and OP9. Typical transforming growth factor β (TGFβ) and fibroblast growth factor (FGF) family members expressed in M15 are summarized in supplemental online Fig. 3 and supplemental online Table 2. These results suggest the possible involvement of growth factors, including Activin, bFGF, and BMP.
Activin is known to be a key signal for mesendoderm and definitive endoderm differentiation [8, 9]. FGF4 was reported to be candidate signals secreted from the adjacent germ layers that patterned early endoderm and induced pancreatic differentiation . Moreover, Activin and bFGF were shown to be maintenance signals secreted from notochord to sustain Pdx1 expression in the prepancreatic endoderm [26, 27]. The TGFβ signaling-related molecules expressed in feeder cells were focused. Activin consists of two β inhibin subunits (A and B). M15 expressed both Inhibinβa and Inhibinβb. OP9 and PA6 expressed only Inhibinβa. Moreover, Follistain, an Activin antagonist, was expressed at a relatively high level in OP9 and PA6 but not in M15. These results suggest an overall high Activin activity in M15, which accounts for its strong endoderm-inducing activities. Similarly, another TGFβ family gene, Bmp4, was expressed in M15, PA6, and OP9. However, only M15 expressed a high level of Gremlin1, a BMP antagonist. Retinoic acid (RA) has been suggested to play a role in pancreatic differentiation in zebrafish . M15 expressed a high level of the RA synthesizing enzyme Aldh1a1 (Raldh1) but did not express Cyp26a1, which specifically degrades RA. Next, ES cell differentiation experiments were performed with the addition of Activin, FGF, BMP, and RA or their inhibitors at specific time windows, depicted as blue lines in Figures 3, Figure 4., Figure 5., Figure 6.–7.
Signaling Through Activin and bFGF in the Formation of Mesendoderm
In the early phase (days 0–3), undifferentiated ES cells become mesendoderm. As Activin has been shown as a key molecule in mesendoderm differentiation , the effect of Activin in our present M15 system was first examined. An analysis by FCM of E-cadherin+/ PDGFRα+ cells, which are characteristic of mesendoderm cells , showed that approximately 5% of the ES cell culture grown on M15 consisted of mesendoderm cells (Fig. 3A, control). Figure 3A shows that Activin (10 or 100 ng/ml) increased mesendoderm formation in a dose-dependent manner, consistent with previous reports . SB431542, a specific inhibitor for TGFβ superfamily type I Activin receptor-like kinase receptors ALK4, ALK5, and ALK7 , significantly blocked the formation of mesendoderm. Other growth factors and inhibitors were then tested. The addition of bFGF, BMP7, or Noggin, a BMP antagonist, had little effect on mesendoderm, but SU5402, a specific inhibitor of FGF receptor 1, strongly inhibited mesendoderm differentiation similarly to SB431542. TGFβ2 and Nodal showed effects similar to those of Activin (N.S., unpublished results).
The effects of Activin and FGF inhibitors were further characterized. To allow the detection of differentiation markers such as βIII-tubulin, ES cells were assayed on day 4. Inhibition of FGF signals by SU5402 substantially increased the number of Nanog-positive cells (Fig. 3B) or Oct3/4-positive cells (N.S., unpublished data). Similar effects were observed with U0126, a specific inhibitor of ERK1/2 mitogen activated protein kinase (MAPK), but not with LY294002, an inhibitor of phosphoinositide 3-kinase (PI3K; Fig. 3B). These results demonstrate that FGF signaling through the extracellular signal regulated kinase (ERK)1/2 MAPK pathway but not the PI3K pathway is crucial for the initial differentiation from the inner cell mass-like or epiblast-like ES cells. In contrast, treatment with SB431542 or SB203580, a potent p38 MAPK inhibitor, increased the number of βIII-tubulin+ neuroectoderm cells without affecting the expression of Nanog (Fig. 3B). This suggests that the inhibition of the Activin or p38 MAPK signal triggered neuroectoderm differentiation, consistent with previous studies . The role of Activin and BMP7 was further examined by adding them to ES cells cultured on M15 on days 0–4. Figure 3C demonstrates that Activin, as well as BMP, inhibited neuroectoderm differentiation. BMP inhibited SB431542-induced but not SB203580-induced neuroectoderm differentiation (Fig. 3C). These results are consistent with previous reports that BMP and signaling sustain self-renewal of murine ES cells . Activin did not inhibit SB203580-induced neuroectoderm differentiation. As one study indicated that TGFβ regulates p38 MAPK via the TAK1 pathway , it was suggested that MAPK signaling runs parallel with TGFβ signals to mediate inhibition of the neuroectoderm pathway. The above results demonstrate that Activin signals affect the fate of divergence into mesendoderm versus prospective neuroectoderm at a step later than bFGF and that the Activin signal promotes mesendoderm differentiation through inhibition of neuroectoderm differentiation.
The Role of Activin, bFGF, and BMP Signals in Fate Determination During Definitive Endoderm Differentiation
Next, the middle phase (days 3–4), in which the bipotential mesendoderm is converted to a definitive endoderm or mesoderm fate, was examined (Fig. 4A). On day 4, a mixed population of F+, T+, and F+/T+ was observed in control ES cells grown on M15. As Activin is shown to promote endoderm formation [9, 12], the addition of Activin was tested first. Activin turned out to increase the proportion of Foxa2-expressing cells at 10 ng/ml in our present M15 system (Fig. 4A). No additional effects were observed with an increased dose of Activin during this time window (unpublished data). SB431542 added at 1 μM decreased the proportion of Foxa2-expressing cells but increased that of the T-positive cells; however, when added at 5 μM, SB431542 abolished both Foxa2-expressing and T-expressing cells. FCM analysis revealed that the addition of Noggin, Activin, or bFGF increased the proportion of E-cadherin+/CXCR4+ definitive endoderm cells, whereas addition of BMP7, SB431542, or SU5402 decreased these cells (Fig. 4B). Nodal, which activates signaling components similar to those activated by Activin, has been shown to induce mesoderm differentiation at a low concentration and endoderm differentiation at a high concentration . Therefore, the dose-dependent inhibition of SB431542 suggests that the present ES cell in vitro differentiation system mimics normal developmental processes. On the other hand, addition of bFGF induced E-cadherin+/CXCR4+ definitive endoderm formation, whereas inhibition of FGF by SU5402 did not promote mesoderm differentiation but rather kept the cells at a mesendoderm state.
The above data, together with those shown in Figure 2D, suggest that most of the Foxa2+ cells represented definitive endoderm cells and T+ cells represented mesoderm populations on day 4 here. Smad1 and Smad2/3 are known to transduce signals downstream of BMP and Activin, respectively. Figure 4C shows that phosphorylated Smad2/3 (P-Smad2/3) was localized in Foxa2+ cells (upper panels), whereas P-Smad1 was observed in Foxa2− cells (middle panels) and localized in T+ cells (lower panels). These findings strongly suggest that Activin promotes formation of definitive endoderm by activating Smad2/3, whereas BMP7 induces Smad1 phosphorylation and promotes mesoderm formation, and that Activin and BMP signals are mutually antagonistic because of interaction at the level of Smads.
The differentiation potential of the E-cadherin+/PDGFRα+ cells and E-cadherin+/Cxcr4+ cells to give rise to Pdx1/GFP+ cells was confirmed by cell sorting experiments (supplemental online Fig. 4). E-cadherin+/PDGFRα+ mesendoderm (day 3) and E-cadherin+/Cxcr4+ definitive endoderm cells (day 4) were sorted out by FCM, confirmed with molecular marker expressions by RT-PCR (supplemental online Fig. 4A) [8, 9], and recultured on a monolayer of M15. Both populations gave rise to Pdx1/GFP+ cells, which first appeared on day 6, increased, and reached a plateau on day 8 (supplemental online Fig. 4), in exactly the same timing as the above direct differentiation protocol (Fig. 2).
Activin, bFGF, RA, and Short-Range Signals Promote the Differentiation of Pdx1-Expressing Pancreatic Progenitor Cells
To determine whether the entire induction processes is mediated solely by soluble molecules, a transfilter system was used. M15 and ES cells were grown separately in the upper and lower chambers, respectively (Fig. 5A). The ES cells differentiated in this system produced E-cadherin+/Cxcr4+ definitive endoderm in the same way as those that differentiated in direct contact with the M15. However, Pdx1/GFP+ cells were not generated using this transfilter system. These results suggest that factors secreted by M15 are not sufficient to direct ES cells to differentiate into a gut regional-specific endoderm lineage. Short-range signals or intercellular interactions with supporting cells are required for the induction of Pdx1 expression.
To examine the late phase (days 4–8), growth factors were added on days 4–8, and the ES cells were assayed by FCM on day 8. The addition of Activin or bFGF increased the proportion of Pdx1/GFP+ cells, which was almost abolished by addition of their inhibitors, SB431542 or SU5402, respectively. Exogenous RA had no effect, but LE540, a potent RA receptor antagonist, inhibited Pdx1/GFP+ cell differentiation (Fig. 5B). These results indicate that Activin, bFGF, and RA are required for the induction of Pdx1/GFP+ cells. Notably, the effect of RA on the formation of Pdx1/GFP+ was phase-specific. At the early and middle phases, addition of RA promoted neuroectoderm differentiation, which then resulted in the inhibition of definitive endoderm differentiation (supplemental online Fig. 5). Similarly, the addition of LE540 during these phases resulted in opposite effects: inhibition of neuroectoderm differentiation and potentiation of definitive endoderm differentiation (supplemental online Fig. 5). This suggests that RA is secreted by M15 and that in the coculture system, a locally high concentration of RA is formed, which is responsible for the differentiation of Pdx1+ cells at late phase.
A Remarkable Increase in Pdx1-Expressing Cells by Simultaneous Treatment with Activin and bFGF Throughout All Phases
These results suggest that Activin and bFGF promote ES cell differentiation at all phases of induction (Figs. 3, Figure 4.–5). In an attempt to obtain a maximum yield of Pdx1-expressing cells, Activin and/or bFGF were added throughout the entire processes of ES differentiation in our M15 systems, namely days 0–8, and FCM was carried out to evaluate the effects. The simultaneous treatment of Activin and bFGF resulted in an increase in both the number and extent of expanded ES colonies and a burst increase in the number of Pdx1/GFP-expressing cells (Fig. 6A). The GFP expression was confirmed to mimic that of Pdx1 protein by using anti-Pdx1 antibody. The GFP expression was detected in both the nuclei and cytoplasm, because of a lack of a nuclear localizing signal in most GFP-expressing cells (Fig. 6B). The exogenous addition of Activin and bFGF resulted in a sixfold increase in the proportion of E-cadherin+/Cxcr4+ definitive endoderm cells (8% to 47%), and a threefold increase of Pdx1/GFP+ cells (22% to 65%) within the definitive endoderm population. This meant an approximately 16-fold increase in the proportion of Pdx1/GFP+ cells within the total differentiated ES cells (2% to 31%; Fig. 6C). Increasing the Activin concentration from 10 to 100 ng/ml gave similar results, indicating that 10 ng/ml Activin was sufficient (unpublished data).
ES cell-derived definitive endoderm cells with or without Activin and bFGF treatment for 8 days were recovered by FCM and characterized (Fig. 6D). After performing Activin and bFGF treatment, the endocrine lineage markers, such as Insulin2 (Ins2), Glucagon (Gcg), Pancreatic polypeptide (Ppy), Somatostatin (Sst), and Nkx6–1, were turned on (Fig. 6D, DE, +Act, bFGF). These data suggest that Activin and bFGF promoted mature endocrine lineages, including β cells. Since the expression level of Neurog3 was low in day 8 definitive endoderm, its expression was traced with time. Neurog3 was expressed in the Pdx1(+) cells on day 11 and then downregulated on day 18 (Fig. 6D, d11, d18 Pdx1/GFP+). Neurog3 expression was detected in undifferentiated ES cells, which was as previously reported [4, 33].
To test the effects of growth factors on pancreatic differentiation in other ES cell lines, R1 ES cells were used (Fig. 6E). RT-PCR analysis revealed that Activin and bFGF increased the level of Sox17 and Pdx1 expression on day 8. Therefore, Activin- and bFGF-dependent potentiation of endoderm and pancreatic differentiation is common among different mouse ES cell lines. Next, to determine whether the M15 system is also applicable under serum-free (SF) conditions, the cells were cultured in DMEM supplemented with ITS and albumin. ES cells grown on M15 cultured under SF conditions yielded 5% definitive endoderm and approximately 1% Pdx1/GFP+ cells (Fig. 6F, ITS), which increased to approximately 23% definitive endoderm and 12% Pdx1/GFP+ cells after the addition of Activin and bFGF (Fig. 6F). Next, ES cells were differentiated on PFA-fixed M15 under serum-free conditions (supplemental online Fig. 6). PFA-fixed M15 showed weaker but significant activities in supporting the differentiation into definitive endoderm and Pdx1+ cells, and then the addition of Activin and bFGF was observed to promote both endoderm and pancreatic differentiation. The weaker supporting activity in PFA-fixed M15 might be due to a lack of growth factors secreted from living M15. These results demonstrated that cell-cell interaction is important for pancreatic differentiation and that the effect of Activin and bFGF on ES cells is direct, rather than via indirect effects on the M15. The results of in vitro differentiation are summarized in Figure 6G. The text in blue (Fig. 6G) indicates the signaling molecules involved in each inductive process clarified in the present study.
ES Cell-Derived Pdx1+ Cells Progressed into Mature Pancreatic Endocrine, Exocrine, and Duct Cells Under In Vivo Conditions
To test the differentiation potential, ES cells were grafted under the kidney capsule of scid (severe combined immunodeficiency) mice. Undifferentiated ES cells, differentiated ES cells (whole [−/+] Activin and bFGF), or Pdx1/GFP+ cells (GFP+) were used for the grafts. Pdx1/GFP+ cells (GFP+) used for transplantation were negative for anti-C-peptide antibody (unpublished data). Two weeks after transplantation, the kidneys that had received grafts were excised. Pdx1/GFP+ grafts showed the strongest GFP fluorescent signals (Fig. 7A, arrows). Under these conditions, teratoma formation was observed in grafts of undifferentiated ES cells and unsorted differentiated ES (whole) cells but not of Pdx1/GFP+ cells (Fig. 7A). Teratoma derived from unsorted cells contained various tissues, including adipose tissue, cartilage, epidermis, gut-like epithelium, muscle, βIII-tubulin-positive neural ectoderm, and platelet endothelial cell adhesion molecule-1-positive endothelial cells (supplemental online Fig. 7), which were not observed in grafts of Pdx1/GFP + cells (N.S., unpublished data).
To examine the in vivo differentiation properties of Pdx1/GFP+ cells, RT-PCR and immunocytochemistry analysis were performed with the grafts of Pdx1/GFP+ cells. RT-PCR analysis revealed that the endocrine lineage markers, such as Neurogenin3 (Neurog3), Insulin1 (Ins1), Glucagon (Gcg), and Pancreatic polypeptide (Ppy), as well as the exocrine marker Amylase (Amy), were turned on, and expression of Somatostatin (Sst) and duct marker Cytokeratin 19 (CK19), along with expression of other β cell markers, such as Islet amyloid polypeptide (Iapp) , Kir6.2 , and Glut2, increased after transplantation (Fig. 7B, after).
Immunohistochemical analysis revealed that C-peptide was detected in the Pdx1/GFP-positive cell after transplantation (Fig. 7C). Quantitative analysis was carried out using three grafts. Approximately 17% were GFP-positive, and 2%–5% were C-peptide+/GFP+ pancreatic β cells within the total differentiated cells (Fig. 7D). Moreover, other pancreatic endocrine hormones (glucagon, pancreatic polypeptide, and somatostatin) were expressed in the graft. Amylase-positive exocrine cells and DBA-positive duct cells were also differentiated from Pdx1/GFP+ cells after transplantation (Fig. 7C). Taken together, ES cell-derived Pdx1/GFP+ cells differentiated under our present in vitro differentiation system have the potential to differentiate into all pancreatic lineages, namely endocrine, exocrine, and duct cells.
An induction procedure was established that efficiently and sequentially produced mesendoderm, definitive endoderm, and finally regional-specific definitive endoderm-derived organs in vitro, in a manner that mimics early embryonic inductive events in vivo. This method is a useful tool for the study of definitive endoderm and pancreatic differentiation, because it allows the generation of Pdx1-expressing cells at a high efficiency. Growing ES cells on M15 in the presence of Activin and bFGF induced approximately 31% of the ES cells to differentiate to Pdx1-expressing cells in serum-containing medium or 12% in SF medium (Fig. 6C, 6F). To compare the efficiency of the generation of Pdx1-expressing ES cells in the present procedure with that reported previously, the methods described by Kubo et al.  were used with the SK7 ES cell line. Following this procedure, EB was formed, and a high concentration of Activin (100 ng/ml) was added, but only approximately 1% of the differentiated ES cells expressed Pdx1 (supplemental online Fig. 8). This indicates that the present procedure is 30-fold more efficient in inducing the differentiation of Pdx1-expressing cells. Moreover, it should be noted that when EB formation is adopted, the total cell yield is 20-fold lower compared with ES cells cultured in monolayer. Therefore, starting from the same ES cell number, the present procedure generates 600-fold more Pdx1-expressing cells.
In our present differentiation procedure, serum conditions were adopted since they gave a better yield than SF conditions (Fig. 6C, 6F). Attempts to differentiate SK7 ES cells under other published serum-free procedures [9, 11] were also carried out; however, the low yield of Pdx1-expressing cells confirmed previous results [8, 9]. The low yield of Pdx1/GFP-positive cells might be partly due to a low surviving rate of SK7 ES cell line under serum-free conditions (N.S., unpublished results). In contrast, our M15 system is applicable to most ES cell lines for generating specific endoderm lineages and thus is a useful system.
In the early phase (Fig. 3), a bFGF signal promotes the initial differentiation of ES cells from an epiblast state through the ERK pathway. This result is consistent with a recent report that FGF stimulation of the ERK signaling cascade triggers transition from self-renewal to lineage commitment . Next, Activin and/or p38 MAPK induced divergence into the mesendoderm lineage versus a presumptive neuroectoderm lineage. These results are consistent with previous reports showing that neuroectoderm specification requires the attenuation of TGFβ signaling  and that p38 MAPK negatively regulates neural differentiation [32, 38]. BMP and Activin/Nodal signaling was reported to promote self-renewal in mouse ES cells [31, 39]. Addition of BMP7, similar to addition of Activin, also resulted in suppression of neuroectoderm differentiation (Fig. 3); this suppression was due to the promotion of self-renewal by BMP and Activin. Studies with mutant mice have shown that Nodal plays a role in anterior-posterior patterning [40, 41]. During the early phase, exogenously added Activin promoted mesendoderm formation in a dose-dependent manner, which agreed with our present results that a low concentration of an Activin inhibitor, SB431542 (1 μM), completely abolished mesendoderm differentiation and that a high concentration of 100 ng/ml Activin was optimal to promote mesendoderm differentiation, as previously reported .
In the middle phase (Fig. 4), the divergence of the mesendoderm into endoderm and mesoderm uses a well-known mechanism of mutually antagonistic P-Smad1 and P-Smad2/3 signaling . BMP is thought to act as a multifunctional regulator of morphogenesis during early development. A null mutation of the Bmp4 gene in mice causes lethal defects in extraembryonic and posterior/ventral mesoderm formation . In the present procedure, serum enhanced rather than inhibited endoderm differentiation, probably because of the expression of BMP inhibitory factors, such as Gremlin1, in M15 (supplemental online Fig. 3). Interestingly, the differentiation into mesoderm is strongly enhanced by the exogenous addition of BMP7, but not BMP2 or BMP4 (Fig. 4; unpublished data). Moreover, neural differentiation was also inhibited with BMP7 but not other BMPs in the early phase (Fig. 3; unpublished data). BMPs exist in vivo as heterodimers showing enhanced biological activity, and the strongest synergy was observed with BMP2/7 or 4/7 heterodimers . However, BMP2 or BMP7 was not detected in M15 (supplemental online Table 3). Therefore, the exogenous BMP2 or BMP7 increases the formation of heterodimers with the BMP4 expressed in the differentiated ES or M15, which in turn inhibited neural differentiation and induced mesoderm formation in the early and middle phases, respectively.
In the late phase, treatment with both Activin and bFGF promoted the differentiation into Pdx1-expressing cells at the expense of hepatic differentiation, as indicated by a decreased expression of Afp (Fig. 6D). Activin and bFGF were previously reported to be potent pancreatic-inducing molecules secreted from the notochord that permit dorsal pancreatic morphogenesis and maintain Pdx1 expression [26, 27]. The synergistic activity of Activin and bFGF has been described to be mediated by the Smad3 and ERK1/2 MAPK signaling pathways during the development of tyrosine hydroxylase-positive neurons . It would be interesting to investigate the underlying mechanism of the synergistic effect of Activin and bFGF on Pdx1 expression. On the other hand, RA has been identified as an important signal in Xenopus, zebrafish, and mouse pancreatic development [28, 46, 47]. In the present M15 system, LE540 inhibited pancreatic differentiation, but the addition of RA had little effect (Fig. 5B). This suggests that the RA secreted from the M15 was sufficient to drive the differentiation from definitive endoderm cells into Pdx1/GFP+ cells. Although a previous report, using an ES cell line with the GFP reporter gene inserted into the Pdx1 locus, has described the potentiation of Pdx1 expression by RA , this is the first report showing the stage-dependent effects of RA.
Moreover, the results of transfilter assay (Fig. 5A) and differentiation study using PFA-fixed M15 (supplemental online Fig. 6) suggested that a short-range cue is essential for pancreatic differentiation. To examine the short-range signals from M15, a list of genes categorized as extracellular region expressed proteins, which are expressed in M15 at an over twice higher level than in OP9 and PA6, are shown in supplemental online Table 4. Further insight into the short-range cues provided by the M15 would require functional studies of the candidate molecules expressed by M15 and might provide additional information that leads to the future establishment of a cell-free procedure.
The gene expression patterns of the definitive endoderm populations isolated by FCM suggest that in the absence of exogenous Activin and bFGF, definitive endoderm cells resemble those of the immature pancreatic epithelium. In the presence of exogenously added Activin and bFGF, ES cells differentiated toward a mature endocrine lineage on day 8 (Fig. 6D). However, the low expression level of Neurog3 suggested that the number of endocrine progenitor cells at day 8 is still low. A prolonged in vitro culture promoted the cells toward a more matured endocrine state, which strongly expressed Neurog3 and Insulin2 (Fig. 6D).
The present observation that ES cells differentiated into C-peptide-expressing cells after being grafted under the kidney capsule suggests that Pdx1/GFP+ cells yielded by the present differentiation procedure in vitro have the potency to further progress toward mature pancreatic β cells in vivo (Fig. 7). Anti-C-peptide antibody used in this study reacts to both C-peptide I and C-peptide II, the cleavage by-products of proinsulin protein. Although Pdx1/GFP+ cells differentiated in vitro expressed transcripts of Insulin 2 to some extent, the amount of insulin protein produced de novo from these cells was too low to be detected (unpublished data). The transplantation studies indicated that Pdx1/GFP+ cells differentiated under in vitro culture conditions have the potential to differentiate into all three pancreatic cell lineages, namely, endocrine, exocrine, and duct cells (Fig. 7). Differentiated ES cells were also transplanted into streptozotocin-induced diabetic mice; however, blood glucose levels were not lowered (unpublished data). This may be due to the fact that that the minimum concentration of islets required for insulin independence is reported to be approximately 9,000 islets per kilogram .
In summary, this study describes an approach by which to induce ES cells to differentiate efficiently into definitive endoderm and Pdx1-expressing cells. The use of M15 as supporting cells for the differentiation of ES cells is applicable to different ES cell lines and allows us to study molecules involved in endoderm and pancreatic differentiation (Fig. 6G). With this knowledge, combined with the literature of the molecular mechanisms of pancreatic development obtained through mutant mice studies, the controlled differentiation of mature β cells will be possible in near future.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank Drs. Douglas A. Melton (Harvard University) and Guoqiang Gu (Vanderbilt University) for providing the Pdx1/GFP transgenic mice; Drs. Christopher Wright (Vanderbilt University) and Maureen Gannon for Pdx1/lacZ ES cells; Drs. Andras Nagy and En Li for R1 and J1 ES cells; Drs. Hiroaki Kodama, Shinichi Nishikawa, Yoshiki Sasai, Toshiaki Noce, and Minoo Rassoulzadegan for providing culture cell lines, Dr. Akira Nagafuchi for anti-E-cadherin mAb ECCD2. We also thank Dr. Minetaro Ogawa for valuable advice on analysis by FCM. We thank members of the Center for Animal Resources and Development at Kumamoto University and the Gene Technology Center for technical assistance. This work was supported by Grant 17045026 from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) Japan (to S.K.); a grant from the Project for Realization of Regenerative Medicine from MEXT (to S.K.); a 21st Century Center of Excellence (COE) grant; and a Global COE grant from MEXT. N.S. and T.Y. are research associates of the 21st Century COE. This work was also supported in part by a grant from the New Energy and Industrial Technology Development Organization (to S.K.) and by grants from the Uehara Foundation and the Takeda Science Foundation.