Differentiation of Embryonic Stem Cells Conditionally Expressing Neurogenin 3


  • Nathan R. Treff,

    1. University of Wisconsin-Madison School of Medicine and Public Health, Department of Surgery, Madison, Wisconsin, USA
    2. William S. Middleton Memorial Veterans Administration Hospital, Madison, Wisconsin, USA
    Search for more papers by this author
  • Robert K. Vincent,

    1. University of Wisconsin-Madison School of Medicine and Public Health, Department of Surgery, Madison, Wisconsin, USA
    2. William S. Middleton Memorial Veterans Administration Hospital, Madison, Wisconsin, USA
    Search for more papers by this author
  • Melisa L. Budde,

    1. University of Wisconsin-Madison School of Medicine and Public Health, Department of Surgery, Madison, Wisconsin, USA
    2. William S. Middleton Memorial Veterans Administration Hospital, Madison, Wisconsin, USA
    Search for more papers by this author
  • Victoria L. Browning,

    1. University of Wisconsin-Madison School of Medicine and Public Health, Department of Surgery, Madison, Wisconsin, USA
    2. William S. Middleton Memorial Veterans Administration Hospital, Madison, Wisconsin, USA
    Search for more papers by this author
  • Joseph F. Magliocca,

    1. University of Wisconsin-Madison School of Medicine and Public Health, Department of Surgery, Madison, Wisconsin, USA
    2. William S. Middleton Memorial Veterans Administration Hospital, Madison, Wisconsin, USA
    Search for more papers by this author
  • Vivek Kapur,

    1. University of Minnesota, Departments of Veterinary and Biomedical Sciences and Microbiology, St. Paul, Minnesota, USA
    Search for more papers by this author
  • Jon S. Odorico M.D.

    Corresponding author
    1. University of Wisconsin-Madison School of Medicine and Public Health, Department of Surgery, Madison, Wisconsin, USA
    2. William S. Middleton Memorial Veterans Administration Hospital, Madison, Wisconsin, USA
    3. WiCell Research Institute, Madison, Wisconsin, USA
    • University of Wisconsin Hospital, N4/756 CSC, 600 Highland Avenue, Madison, Wisconsin 53792, USA. Telephone: 608-263-4768; Fax: 608-262-6280
    Search for more papers by this author


Expression of the proendocrine gene neurogenin 3 (Ngn3) is required for the development of pancreatic islets. To better characterize the molecular events regulated by Ngn3 during development, we have determined the expression profiles of murine embryonic stem cells (mESCs) uniformly induced to overexpress Ngn3. An mESC line was created in order to induce Ngn3 by adding doxycycline to the culture medium. Genome-wide microarray analysis was performed to identify genes regulated by Ngn3 in a variety of contexts, including undifferentiated ESCs and differentiating embryoid bodies (EBs). Genes regulated by Ngn3 in a context-independent manner were identified and analyzed using systematic gene ontology tools. This analysis revealed Notch signaling as the most significantly regulated signaling pathway (p = .009). This result is consistent with the hypothesis that Ngn3 expression makes the cell competent for Notch signaling to be activated and, conversely, more sensitive to Notch signaling inhibition. Indeed, EBs induced to express Ngn3 were significantly more sensitive to γ-secretase inhibitor-mediated Notch signaling inhibition (p < .0001) when compared with uninduced EBs. Moreover, we find that Ngn3 induction in differentiating ESCs results in significant increases in insulin, glucagon, and somatostatin expression.


With the success of cell-based therapies to reverse insulin dependence in type I diabetics, pursuit of an abundant, renewable source of replacement cells is highly relevant. Directed differentiation of murine embryonic stem cells (mESCs) into pancreatic β cells suitable for transplantation requires a better understanding of the molecular networks regulated during β-cell development. The proendocrine basic helix-loop-helix protein neurogenin 3 (Ngn3) plays a critical role in the development of the hormone-producing islets within the pancreas, including the insulin-producing β cell. Deletion of Ngn3 in mice results in complete loss of pancreatic endocrine cells as well as development of diabetes [1]. In addition, Ngn3 overexpression in the developing pancreas results in an increase in pancreatic endocrine cell mass [2, [3]–4]. Finally, ectopic expression of Ngn3 in pancreatic ductal cells in vitro induces the expression of many of the genes that mark the development of pancreatic islet cells [5, 6].

Despite numerous reports describing its role, our understanding of the regulatory networks surrounding Ngn3 remains limited. It has been reported that the expression of Ngn3 is directly regulated by HNF6 [7], Hes1 [8], and Ngn3 itself [9]. In addition, Ngn3 has been shown to directly regulate the expression of NeuroD1 [10], Pax4 [9], and Nkx2.2 [11]. Recent microarray studies have implicated a number of additional candidate targets of Ngn3 in the developing pancreas [12] and in pancreatic ductal cells [5]. These studies were performed on different microarray platforms and were limited to approximately 12,000 or 18,000 genes, respectively. In addition, the increase in Ngn3 expression in the cells studied in these experiments was relatively low: 10-fold (independent analysis of microarray data) and 12.5-fold, respectively.

To expand our understanding of the molecular mechanisms regulated by Ngn3 during development, we have generated a stable mESC line with inducible uniform expression of Ngn3 and used this line to construct and identify an Ngn3-regulated expression profile for approximately 39,000 genes. In addition, we have validated a new role for Ngn3 in regulating Notch signaling with implications for islet cell development. These studies also suggest a possible method by which hormone-producing cells might be more efficiently generated in vitro from ESCs.

Materials and Methods

Generation of Inducible Ngn3 ESC Line

Murine Ngn3 cDNA was isolated by reverse transcription-polymerase chain reaction (RT-PCR) amplification from differentiated D3 embryoid body (EB) [13] mRNA. First-strand cDNA synthesis was performed with the AMV cDNA Synthesis Kit (Roche Applied Science, Indianapolis, http://www.roche-applied-science.com) following the manufacturer's recommended protocol with 1 μg of total RNA. Ngn3 amplification was performed using the forward primer (5′-CCCACGCGTGCCACCATGGCGCCTCATCCCTTGGA-3′), which incorporates a kozak sequence immediately upstream of the Ngn3 start codon, and the reverse primer (5′-CCCTCTAGATCAATGGTGATGGTGATGGTGCAAGAA GTCTGAGAACACCA-3′), which incorporates a six histidine (6XHis) coding sequence prior to and in frame with the Ngn3 stop codon. PCR was conducted with AccuPrime DNA polymerase (Roche Applied Science), 1× reaction buffer, 0.5 μl of first-strand cDNA, 300 μM each primer, and cycling conditions as follows: 95°C for 3 minutes; 40 cycles of 95°C for 30 seconds, 58°C for 45 seconds, and 68°C for 1 minute; and 68°C for 5 minutes. The PCR product was purified and cloned into the pCR2.1 vector by TA cloning as recommended by the supplier (Invitrogen) and sequenced at the University of Wisconsin Biotechnology Center. The MluI/XbaI Ngn3 cDNA-containing fragment was subcloned into the corresponding restriction sites of the plox vector (gift from M. Kyba and G. Daley) to generate a plox-Ngn3–6XHis construct. This construct was coelectroporated [14] with pSalk-Cre [15] into the Ainv15 ESC line (gift from M. Kyba and G. Daley and available at American Type Culture Collection, Manassas, VA, http://www.atcc.org). Eight stable site-specific integrants (named Ngn3-C3, -C4, -C5, -C6, -D3, -D4, -D5, and -D6) were obtained (with 350 μg/ml G418 [Invitrogen, Carlsbad, CA, http://www.invitrogen.com] selection), expanded, and screened by PCR using Qiagen DNeasy (Qiagen, Hilden, Germany, http://www1.qiagen.com) purified genomic DNA, and LoxinF and LoxinR primers with the sequences, 5′-CTAGATCTCGAAGGATCTGGAG-3′ and 5′-ATACTTTCTCGGCAGGAGCA-3′, respectively. PCR was performed using AmpliTaq DNA polymerase (Invitrogen), 1× PCR II reaction buffer, 1.5 mM MgCl2, 200 μM dNTP, 200 nM each primer, 500 ng of genomic DNA, and cycling conditions as follows: 94°C for 5 minutes; 30 cycles of 94°C for 1 minute, 58°C for 1 minute, and 72°C 1 minute; and 72°C for 7 minutes. Single-copy integration of Ngn3 cDNA was confirmed using quantitative real-time RT-PCR (QPCR) with genomic DNA and gene expression assays (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) for Ngn3 (Assay ID Mm00437606_s1) and β3 tubulin (Mm00727586_s1), which produce an amplicon from a single exon. The comparative threshold method for relative quantitation was used [16] so that the ΔCT was derived by subtracting the threshold cycle (CT) of Ngn3 from that of β3 tubulin and the ΔΔCT was derived by subtracting the ΔCT for the inducible Ngn3 cell line sample from the ΔCT for the parental Ainv15 cell line. Fold change was then determined by 2−ΔΔCT.

Cell Culture and Induction of Ngn3 Expression

Induction of the Ngn3 transgene in the eight cell lines was determined using QPCR for Ngn3 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Mm99999915_g1) in ESCs maintained in LIF and treated daily for 3 days with 1 μg/ml doxycycline (Dox) (BD Biosciences, San Diego, http://www.bdbiosciences.com) relative to untreated cells. The Ngn3-D6 cell line was selected for further analyses because its transgene mRNA levels in response to Dox were superior to the other lines; and given the nature of the site-specific recombination used to develop the cell lines, all eight are safely assumed to be genetically identical [15]. Induction was also evaluated using immunofluorescent staining [13] of induced and uninduced cells with rabbit anti-mouse Ngn3 antibody (number 7979, bleed 2 at 1:2,000; a gift from M. German) and an anti-rabbit 568 Alexa Fluor-conjugated secondary antibody (Invitrogen) at 1:2,000. The pluripotency of the Ngn3 ESC line was evaluated by injecting approximately 1 × 106 undifferentiated ESCs into the hind limb of nonobese diabetic-severe combined immunodeficiency (NOD/SCID) mice (Jackson Laboratory, Bar Harbor, ME, http://www.jax.org) and subsequent hematoxylin-and-eosin staining of histological sections of teratomas harvested 6 weeks after injection.

Undifferentiated Ngn3-D6 and Ainv15 ESCs (Ainv15 ES0) were cultured in gelatinized (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) 100-mm tissue culture plates (BD Biosciences) on a layer of γ-irradiated neomycin-resistant mouse embryonic fibroblasts (Millipore, Billerica, MA, http://www.millipore.com) in the presence of Dulbecco's modified Eagle's medium (DMEM)-High Glucose (Invitrogen) with 15% fetal calf serum (HyClone, Logan, UT, http://www.hyclone.com), 100 U of penicillin/ streptomycin (Invitrogen), 2 mM l-glutamine (Invitrogen), 5.5 × 10−2 M β-mercaptoethanol (Invitrogen), 0.11 mM nonessential amino acids (Invitrogen), 350 μg/ml G418 (Invitrogen), and 1,000 U/ml leukemia inhibitory factor (LIF) (Chemicon International, Temecula, CA, http://www.chemicon.com) at 37°C and 5% CO2. ESCs were treated with (ON) 1 μg/ml Dox (Invitrogen) from day 0 to 3, along with removal of LIF and feeder cells (Ngn3 ES3 ON). Removal of LIF during induction of Ngn3 may be critical to making Ngn3-specific observations considering the recent findings that LIF may promote increased expression of Ngn3 through Janus tyrosine kinase/signal transducer and activator of transcription (JAK/STAT) signaling [17]. In addition, LIF and feeder cell removal should provide an environment more amenable to differentiation and provide a more biologically relevant context in which to induce Ngn3 expression. In parallel, EB formation was initiated by removing the undifferentiated ESCs with 1.5 ml of 2 mM EDTA (Invitrogen) and 2% chicken serum (Invitrogen) for 15 minutes at 37°C. Three million cells were transferred to a siliconized (Sigma-Aldrich) 60-mm Petri dish (BD Biosciences) with the same media and conditions as above, but without LIF or feeder cells. Ngn3 or Ainv15 EBs were renewed every 24 hours with fresh media. Ngn3 EBs were also treated daily with (ON) or without (OFF) 1 μg/ml Dox from either EB day 0–3 (Ngn3 EB3 ON or OFF) or EB day 7–10 (Ngn3 EB10 ON or OFF). The 72 hours of exposure to Dox was chosen based on the fact that Ngn3 expression is transient within the developing pancreas and therefore represents a more biologically relevant amount of expression in vitro.

Analysis of Germ Layer Gene Expression Markers During EB Formation

To identify the most appropriate timing and context for Ngn3 induction, and with the goal of recapitulating embryonic pancreas development in vivo, expression of markers of the three embryonic germ layers was evaluated by QPCR at different time points during EB formation. Ngn3 EBs were grown as described above and were harvested for RNA by resuspension and homogenization in 1 ml of TRIzol (Invitrogen). RNA was isolated by following the manufacturer's recommended protocol (Invitrogen), and was further purified using the RNeasy Mini Kit RNA cleanup protocol (Qiagen) with on-column DNase I treatment (Qiagen). QPCR was performed as described above using the primer/probe sets from Applied Biosystems for β3-tubulin, brachyury (T), goosecoid, sox17, pdx1, and Ngn3.

Microarray Data Acquisition

Cells treated with and without Dox at various stages (Ngn3 ES3 ON, EB3 ON and OFF, EB10 ON and OFF cells, and parental Ainv15 ES0, EB3, and EB10) were harvested for RNA isolation as described above. Afterward, the RNA was quantified and qualified using UV spectroscopy and agarose gel electrophoresis, respectively. Ten micrograms of RNA were converted to double-stranded cDNA using the Roche Microarray cDNA Synthesis Kit and purified using the Roche Microarray Target Purification Kit as recommended by the supplier (Roche Applied Science). The cDNA was converted to biotinylated cRNA using the GeneChip IVT Labeling Kit as recommended by the supplier (Affymetrix, Santa Clara, CA, http://www.affymetrix.com). The biotin-cRNA was purified using the RNeasy Mini Kit RNA cleanup protocol and analyzed by UV spectroscopy. The purified biotin-cRNA was fragmented and hybridized to the GeneChip Mouse Genome 430 2.0 Array as recommended (Affymetrix), and the GeneChip was washed and scanned using the GeneChip Fluidics Station 400 and the GeneArray Scanner (Agilent Technologies, Palo Alto, CA, http://www.agilent.com), respectively, all at the Biomedical Genomics Center of the University of Minnesota. All data have been deposited in the Gene Expression Omnibus (GEO) (National Center for Biotechnology Information [NCBI], Bethesda, MD, http://www.ncbi.nlm.nih.gov/geo) and are accessible through GEO series accession number GSE3653.

Microarray Data Analyses

ArrayAssist (Stratagene, La Jolla, CA, http://www.stratagene.com) was used for probe-level data analysis. GCRMA (GeneChip Robust Multiarray Average) was used for signal normalization [18]. Ainv15 ES0 data were defined as the baseline to generate differential expression values for all hybridizations. A multiclass analysis of variance was performed to generate p values for every probe set. To evaluate transcriptome similarities among treatment groups, differentially expressed genes with expression of at least 2 and not greater than −2 and p values of not greater than .001 were selected from each group and combined for hierarchical clustering using a correlation similarity metric, an average clustering method, and by clustering both rows and columns. Also, probe set lists of Ngn3-induced differentially expressed genes were produced for each differentiation context, ES3, EB3, and EB10 cells. Using the three probe set lists derived by individually comparing Dox-treated ES3, EB3, and EB10 cells with Ainv15 ES0 cells, a Venn diagram was constructed to generate a subset list of Ngn3-regulated genes that were common among the three groups. The intersecting probe sets representing “context-independent” differentially expressed genes were systematically evaluated for significant enrichment of canonical signaling and metabolic pathways using MetaCore (GeneGo, Inc., St. Joseph, MI, http://www.genego.com). Context-dependent probe sets were also identified and included in the supplemental Tables file (http://www.ncbi.nlm.nih.gov/geo).

Microarray QPCR Validation

Five micrograms of the RNA isolated from the same Ngn3 EB10-induced and uninduced cells, as in the microarray analyses described above, and two additional biological replicates were used in a separate cDNA synthesis reaction using the Roche AMV cDNA Synthesis Kit as recommended for five reactions (Roche Applied Science). The cDNA was used to confirm the induction of Ngn3 prior to the microarray analyses using the comparative threshold method as described above. In addition, this material was used as a template for QPCR confirmation of the fold change values observed by microarray analyses for a subset of significantly regulated genes. All primers were supplied as gene expression assays (Applied Biosystems) and included pancreas transcription factor 1 a (Ptf1a) (Mm00479622_m1), Hes5 (Mm00439311_g1), Notch1 (Mm00435245_m1), Dll1 (Mm00432841_m1), Brachyury (T) (Mm00436877_m1), NeuroD1 (Mm01280117_m1), Isl1 (Mm00627860_m1), FoxA2 (Mm00839704_mH), peptide YY (Mm00435350_m1), Onecut1 (Mm00839394_m1), and single minded 1 (Sim1) (Mm00441390_m1).

Notch Signaling Activity

Ngn3 D6 ESCs were grown as EBs for 7 days (as previously described) before being treated with both 1 μg/ml Dox and 1 μM γ-secretase inhibitor (γ-SI) XVIII (Compound E) (Calbiochem, San Diego, http://www.emdbiosciences.com), Dox alone, Compound E alone, or none of the above treatments for 72 hours. RNA from the resulting four groups (n = 3) of Ngn3 EB10 cells was harvested in TRIzol, purified with Qiagen RNeasy columns, converted to cDNA, and used for QPCR analysis as described above using the gene expression assays for Ngn3, Sim1, Ptf1a, and GAPDH. A two-tailed Student's t test was performed to determine whether a significant difference existed between mean ΔΔCT values for Ngn3-uninduced and -induced (Compound E-treated relative to untreated) samples. A one-tailed paired Student's t test was performed to determine whether islet hormone expression (insulin1, glucagon, and somatostatin) was significantly different in γ-SI treated compared with untreated EB10 cells. QPCR was performed for insulin1 (forward primer GACCCACAAGTGGAACAACTG, reverse primer AACGCCAAGGTCTGAAGGT, probe CTGGGAGGAAGCCCCGGG), glucagon (Mm00801712_m1), somatostatin (Mm00436671_m1), and GAPDH using gene expression assays as described above.

Sim1 Analysis

Mouse insulin promoter-green fluorescent protein transgenic mice [19] (Gift from M. Hara) were bred and sacrificed for embryo removal and dissection 15 days after fertilization. Successful removal of pancreas tissue was confirmed by observing green fluorescence using UV light and a dissecting microscope (Leica Microsystems GmbH, Wetzlar, Germany, http://www.leica.com). Tissue was homogenized in TRIzol using a 1-minute pulse with a disposable tip homogenizer. QPCR was performed, as described above, to quantify Sim1 and Ngn3 levels in separately pooled RNA from five to nine dissected pancreata (n = 3), one or two brains (n = 3), and one whole embryo (n = 3). A two-tailed Student's t test was performed to determine whether a significant difference existed between mean ΔΔCT values. QPCR was also performed on EB7–10 cells, with (n = 4) or without (n = 4) Dox treatment, and for Ngn3, Sim1, NeuroD1, Nkx2.2 (Mm00839794_m1), Pax4 (Mm01159036_m1), Ptf1a, and GAPDH genes using the ΔΔCT method as described above.

Islet Hormone Production

Ngn3 ESCs were differentiated into EBs as described above for 7 days and then plated on gelatinized coverslips in a 24-well plate in 10% fetal calf serum DMEM, 1% penicillin/streptomycin, and l-glutamine. After 7 days of growth, tissue cultures were either untreated or induced for Ngn3 for 3 days as described above. In the subsequent 3 days, induced and uninduced cultures were treated or untreated with the γ-SI as described above. Coverslips were then harvested for immunofluorescent staining and QPCR analysis immediately after the treatments, 13 days after plating (EB7 + 13), or until a total of 28 days after plating (EB7 + 28). These time points were chosen based on previous analysis of optimum hormone transcriptional levels (data not shown) in ESCs described elsewhere [13]. QPCR was performed for insulin1, glucagon, somatostatin, and GAPDH using gene expression assays as described above. Antibody staining was performed as previously described [13] using mouse anti-glucagon (K79bB10; Sigma-Aldrich) and rabbit anti-somatostatin (number 20067; ImmunoStar, Inc., Hudson, WI, http://www.immunostar.com). Isotype controls were mouse immunoglobulin Gκ (M-7894; Sigma-Aldrich) and normal rabbit serum (R9133; Sigma-Aldrich). Secondary antibodies were anti-mouse 488 Alexa Fluor-conjugated antibody (1:2,000) and anti-rabbit 568 Alexa Fluor-conjugated antibody (1:4,000).


Ngn3-Inducible mESC Line

We generated an ESC line capable of Dox-inducible, uniform expression of Ngn3 using the system described in Kyba et al. [15]. Site-specific integration of Ngn3 cDNA into the HPRT1 locus of the Ainv15 ESC line was confirmed using a genome-specific primer and a vector-specific primer that results in a 500-bp PCR product in successfully integrated clones (Fig. 1A). Single-copy integration of murine Ngn3 cDNA was confirmed by QPCR of genomic DNA, in which approximately 1.5-fold Ngn3 was detected in the Ngn3 ESC line relative to the Ainv15 parental cell line (Fig. 1B). In addition, the Ngn3 D6 ESC line reproducibly expressed more-than-200-fold Ngn3 transcript (Fig. 1C) and increased Ngn3 protein (Fig. 1D) when treated with 1 μg/ml Dox for 72 hours. Finally, histological analysis confirmed that, when injected into the hind limb of immunodeficient mice, the Ngn3 ESC line was capable of forming tissues representative of all three germ layers (Fig. 1E).

Figure Figure 1..

Generation of a Ngn3-inducible murine ESC line. (A): The D6 Ngn3 ESC line has stable site-specific integration of murine Ngn3 cDNA based on polymerase chain reaction analysis of genomic DNA. The most intense band within the molecular weight marker (M) represents 500 bp, which is the expected amplicon size of a successfully integrated transgene. A 500-bp amplicon is observed with the D6 Ngn3 ESC line and in a previously developed transgenic tet-eGFP cell line (+) [15]. Neither the parental Ainv15 ESC line (A) nor the no-template control (−) shows amplification. (B): The D6 Ngn3 ESC line has a single-copy integration of Ngn3 cDNA based on quantitative real-time reverse transcription-polymerase chain reaction (QPCR) analysis of genomic DNA. Ngn3 (C) transcription and (D) translation is induced, relative to no doxycycline (−), by the addition of 1 μg/ml doxycycline (+) into the culture medium of D6 Ngn3 ESCs based on QPCR and immunofluorescent staining of an ESC colony, respectively. (E): Histological analysis of teratomas formed 6 weeks after hindlimb injection of 1 × 106 cells demonstrated that the D6 Ngn3 ESC line is capable of forming tissues representative of all three germ layers: (i) gut mucosal-like cells (endoderm), (ii) keratinizing squamous epithelium (ectoderm), and (iii) bone (mesoderm).

Endoderm and Pancreatic Developmental Context

To identify the most relevant, pancreas-oriented context in which to induce Ngn3 expression, we evaluated a set of genes associated with endoderm (Sox17 and FoxA2) and pancreatic endoderm and endocrine development (Pdx1 and Ngn3). Based on the previously demonstrated similarities between EBs and the gastrulating embryo [20], these marker genes were characterized for changes during EB formation (Fig. 2). Markers of undifferentiated ESCs (Oct4), mesoderm (brachyury and goosecoid), and ectoderm (β3-tubulin) were assessed in parallel. Transcription of Oct4 decreased approximately 28-fold after 11 days of EB formation. In addition, β3-tubulin, brachyury, and goosecoid transcripts also decreased over the course of EB formation. In contrast, Sox17, FoxA2, Pdx1, and Ngn3 mRNA increased during mid-late EB formation, suggesting that endoderm development is favored during EB formation in our Ngn3-inducible ESC line. In addition, the sequence in which peak transcriptional levels occur during EB formation is consistent with the sequence of gene expression during in vivo embryonic development of the pancreas. For example, peak transcript levels of Sox17 occurring at EB7 precedes the Pdx1 peak at EB9, which in turn precedes the peak in Ngn3 transcription at EB11.

Figure Figure 2..

Endoderm development is favored during EB formation in Ngn3 ESCs. Quantitative real-time reverse transcription-polymerase chain reaction analysis of a panel of marker genes representing pancreatic endocrine (Pdx1 and Ngn3), endoderm (Sox17 and FoxA2), mesoderm (brachyury and goosecoid), ectoderm (β3-tubulin), and undifferentiated (Oct4) cells at 3, 5, 7, 9, and 11 days of EB formation and relative to undifferentiated cells (EB0). Abbreviations: EB, embryoid body; ESC, embryonic stem cell; Ngn3, neurogenin 3.

Ngn3 Regulation of Gene Expression

The results of identifying the most relevant pancreas developmental context led to microarray analysis of gene expression after induction of Ngn3 from EB7 to EB10 (Ngn3 EB10 ON). Subsets of genes identified as significantly regulated by Ngn3 induction and relevant to pancreatic islet cell development were validated using QPCR (Table 1). For example, NeuroD1, a previously identified direct target gene of Ngn3 [10], was found to be significantly upregulated in Ngn3-induced EBs. In addition to known targets, potentially novel gene expression relationships were uncovered by the microarray analysis. Of note, Ngn3 induction led to the apparent upregulation of transcription of Sim1, which is involved in the differentiation of neuroendocrine cells of the hypothalamic-pituitary axis [21]. The pancreas-specific gene Ptf1a, which is expressed in early pancreatic precursors of exocrine, endocrine, and duct lineages of the pancreas [22], was also differentially upregulated in Ngn3-expressing EBs compared with uninduced EBs. Finally, we found that the mesoderm marker gene brachyury (T) was one of the most significantly downregulated genes in the analysis.

Table Table 1.. Validation of microarray data by QPCR (fold change in Ngn3 EB10 ON relative to OFF)
original image

To evaluate the consequence of Ngn3 expression in a more homogenous and pluripotent cellular context, a microarray analysis was also performed after induction of Ngn3 in undifferentiated ESCs either in conjunction with EB formation for 3 days (Ngn3 EB3 ON) or with removal of feeder cells and LIF for 3 days (Ngn3 ES3 ON). These data sets along with the appropriate controls (publicly available through the NCBI GEO Web site http://www.ncbi.nlm.nih.gov/geo; accession number GSE3653) were clustered based on genome-wide gene expression in each treatment group (Fig. 3A). This led to the identification of similarities within the transcriptome of Ngn3-induced ESCs and uninduced EBs, indicating that Ngn3 can initiate a gene marker expression profile that is similar to EB formation profiles.

Figure Figure 3..

Hierarchical clustering and identification of context independent Ngn3-regulated genes. (A): The expression levels of every probe set on the Mouse 430 2.0 GeneChip were evaluated relative to undifferentiated ESCs and used to determine the hierarchical relationships for each Ngn3 ESC treatment group. (B): Genes with significant differential expression in Ngn3-expressing cells (ON) compared with uninduced cells (OFF), under three differentiation contexts (ES3, EB3, or EB10), were identified (see Supplementary Tables at http://www.ncbi.nlm.nih.gov/geo; accession number GSE3653 for gene lists) and cross-referenced using a Venn diagram to identify context-independent probe sets at the intersection. The 757 genes at the intersection were evaluated for significant enrichment of canonical signaling and metabolic pathways using MetaCore software (Table 2). Abbreviations: EB, embryoid body; ESC, embryonic stem cell; Ngn3, neurogenin 3.

Venn diagram analysis was performed on probe sets representing genes with significant differential expression relative to undifferentiated parental ESCs (Ainv15 ESC) from each of three Ngn3 induction contexts (ES3, EB3, and EB10; Fig. 3B). At the intersection of this Venn diagram were 757 context-independent probe sets, which were subsequently evaluated for significant enrichment of signaling and metabolic pathways as well as network processes (Table 2). Context-dependent genes were also identified and included in the supplemental Table. For example, the 3,755 genes identified in the bottom center circle represent those genes regulated in a more “endoderm/pancreas”-enriched context of late EB formation, although other contexts or cell types may be present under these conditions.

Table Table 2.. Results of identifying significant (p < .05) signaling and metabolic pathways or network processes using systematic gene ontology analysis of 757 context-independent, neurogenin 3–regulated genes
original image

Ngn3 Regulation of Notch Signaling

A gene ontology analysis of the 757 Ngn3-regulated genes identified by microarray analysis demonstrated that Notch signaling is the most significant signaling pathway regulated by Ngn3 at the transcriptional level (Table 2). Notch signaling genes upregulated by Ngn3 included the Dll1 and Dll3 ligands; the Notch1 receptor; the presenillin1 and nicastrin γ-secretase subunits that process the Notch intracellular domain (NotchICD); the CBFA2T1h transcription factor that is activated by interaction with the NotchICD; and the Notch transcriptional target genes Hes5, Hes6, and Hey1 (Table 1 and data not shown).

This led us to the hypothesis that Ngn3 may regulate the activity level of Notch signaling. To test this idea, a γ-SI (Compound E) was used to evaluate the relative sensitivity (to Notch signaling inhibition) of the Notch signaling target gene, Hes5, in induced and uninduced Ngn3 EB10 cells. As previously shown, Ngn3 transcript levels were confirmed to be induced by an average of 380-fold relative to uninduced cells with the addition of 1 μg/ml Dox for 72 hours. We then determined that Hes5 gene expression was significantly more sensitive to Compound E-mediated Notch signaling inhibition in the Ngn3-induced cells compared with the uninduced cells (p < .0001) (Fig. 4A), whereas Sim1 and Ngn3 mRNA levels were not. Interestingly, we also observed that Ptf1a gene expression was significantly more sensitive to Notch signaling inhibition in Ngn3-expressing cells compared with uninduced cells (p < .05).

Figure Figure 4..

Ngn3-induced ESCs are more sensitive to Notch signaling inhibition than uninduced ESCs, and Sim1 behaves as a pancreas-relevant and direct target of Ngn3. (A): Quantitative real-time reverse transcription-polymerase chain reaction (QPCR) was performed to determine the expression levels of Hes5, Ptf1a, Sim1, and Ngn3 in uninduced and Ngn3-induced Ngn3 ESCs after treatment with 72 hours of 1 μM γ-secretase inhibitor Compound E († relative to untreated cells). Mean Ngn3 induction was determined to be 380-fold by comparing induced with uninduced Ngn3 levels in Compound E–treated and untreated cells (data not shown). Significance was determined using a Student's t test (∗∗∗ p < .0001; ∗ p < .05). (B): QPCR analysis was also performed to evaluate Sim1 and Ngn3 transcript levels (relative to the entire E15.5 embryo) in the developing embryonic day 15.5 mouse fetal pancreas and brain (∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .0001). (C): In addition, a timecourse analysis was performed to evaluate Ngn3 target gene response kinetics. The Sim1 response is similar to the known direct target genes, NeuroD1, Nkx2.2, and Pax4, whereas the indirect target gene Ptf1a shows a delayed response. Fold change values are relative to uninduced EB7 cells and were measured in 24-hour increments from EB7 to EB10. Abbreviations: EB, embryoid body; Ngn3, neurogenin 3; Ptf1a, pancreas transcription factor 1 a; Sim1, single minded 1.

Sim1 Regulation

As a result of determining that Sim1 upregulation in Ngn3-induced EB10 cells was independent of Notch signaling activity (Fig. 4A), and because Sim1 expression is lost in the developing neural tube of Ngn3-deficient mice [21, 23], we began to characterize its expression in the developing pancreas. We used QPCR to quantify the relative levels of Sim1 and Ngn3 transcripts in embryonic day 15.5 (E15.5) pancreas, brain, and whole embryo (Fig. 4B). Sim1 mRNA levels were significantly greater in the pancreas (p < .01) and brain (p < .05) relative to the whole embryo. As expected, Ngn3 mRNA was significantly enriched in the pancreas compared with the brain and whole embryo (p < .0001). Interestingly, the levels of Ngn3 found in the developing E13.5 pancreas relative to GAPDH (ΔCT = 2.61 ± 0.35) were similar to, but slightly lower than, the levels found in the induced Ngn3 EB10 cells relative to GAPDH (ΔCT = −0.86 ± 0.23). We also investigated the kinetics of Ngn3 target gene upregulation after Ngn3 induction (Fig. 4C). Sim1 levels increased in an immediate manner, similar to that seen for the known Ngn3 direct target genes, NeuroD1, Nkx2.2, and Pax4. In contrast, the levels of Ptf1a were increased in a delayed manner, supporting the observation that this gene is regulated indirectly through Ngn3 regulation of Notch signaling (Fig. 4A).

Ngn3 Overexpression Increases Endocrine Hormone Expression in ESCs

To test the hypothesis that overexpression of Ngn3 in ESCs promotes pancreatic islet endocrine differentiation, we studied the insulin, glucagon, and somatostatin gene expression in differentiated Ngn3 ESCs. We also tested the effect of Notch signaling inhibition by treatment with γ-SI on islet hormone expression. When γ-SI was added to EB7 cells at the same time as Ngn3 induction, a significant increase in the levels of insulin (p = .026) and somatostatin (p = .001), but not glucagon (p = .199), was observed at EB10 (Fig. 5A). To compare hormone expression levels observed in Ngn3-inducible ESCs with the spontaneous levels in cells treated as previously reported [13], Ngn3 ESCs were also differentiated as EBs for 7 days and then plated in adherent cultures for either 13 or 28 days. Cells were treated after plating with Dox for 3 days starting at day 7 after plating, γ-SI for a 3-day period starting at day 10, or both and then analyzed by QPCR and immunofluorescence microscopy. Whereas γ-SI alone had no effect on promoting endocrine hormone gene expression over that seen in untreated cells, Dox treatment and induction of Ngn3 enhance the transcription of insulin 1, glucagon, and (at later stages) somatostatin and increase the detection of glucagon and somatostatin protein (Fig. 5B, 5C). The sequential presence of γ-SI simultaneously after uniform Ngn3 overexpression had no significant impact on the expression of endocrine hormones induced by Ngn3.

Figure Figure 5..

Ngn3 induction in ESCs increases islet hormone expression. (A): Quantitative real-time reverse transcription–polymerase chain reaction analysis was performed on Ngn3 ESCs differentiated as EBs for 7 days, induced with Dox, treated with or without γ-SI for 3 days, and evaluated for hormone transcript levels at EB10. (B): Ngn3 ESCs were also differentiated for 13 or 28 days post-EB7 formation (EB7 + 13 or EB7 + 28) to evaluate hormone transcript levels (relative to uninduced γ-SI untreated cells grown to the same time point). A Student's t test was performed to evaluate significance for data in both (A) and (B) (∗ p value < .05; ∗∗ p value of < .001). (C): In addition, immunofluorescent staining was performed on γ-SI alone, Dox alone, and γ-SI- and Dox-treated cultures at EB7 + 13 and EB7 + 28 time points for glucagon and somatostatin. Abbreviations: Dox, doxycycline; EB, embryoid body; γ-SI, γ-secretase inhibitor; Ngn3, neurogenin 3.


The data presented here implicate a number of molecular mechanisms regulated by Ngn3. First, we found that the pancreas-specific transcription factor Ptf1a was upregulated by 95-fold in Ngn3-induced EB10 cells (Table 1). This observation supports the recent findings reported by Schonhoff et al. [24], which suggest that Ngn3 is expressed in a small number of pancreatic acinar cells. Lin et al. [25] also suggested that Ptf1a may play a role in the development of certain endocrine cells of the developing zebrafish pancreas. The observation that Ngn3 induction of Ptf1a was significantly reduced upon Notch signaling inhibition further suggests that the Ngn3-Ptf1a relationship may be indirect by acting through Ngn3-mediated activation of Notch signaling components. Therefore, a role for Ngn3 in the regulation of Ptf1a in pancreatic exocrine and endocrine cell development may be of further interest.

We also confirmed that Sim1 transcript levels were significantly enriched in the developing pancreas (Fig. 4B). This is consistent with the observations that Sim1 expression is enriched in Ngn3-GFP+ cells of the developing E13.5 pancreas (independent analysis of data in [12]) and appears to be lost in the ventral neural tube of Ngn3-deficient E13.5 mice [23]. We also identified two putative Ngn3 binding sites in the Sim1 promoter that are identical to the experimentally validated sites in the NeuroD1 [10], Pax4 [9], and Nkx2.2 [11] promoters (data not shown). Although Sim1 was previously identified as a potential target of Notch signaling in Drosophila [26], our functional data in mouse ESCs support the concept that Sim1 upregulation by Ngn3 is independent of Notch signaling activity (Fig. 4A). Moreover, we have also shown that Sim1 temporally responds to Ngn3 induction in a manner similar to the response observed for known direct target genes of Ngn3, NeuroD1, Nkx2.2, and Pax4 (Fig. 4C). Together, the cis-regulatory relationship and functional information suggest that Sim1 may be a direct target of Ngn3 in the developing pancreas, where both transcripts are highly enriched. Because Sim1 is known to play an important role in the differentiation of neuroendocrine cells of the hypothalamic-pituitary axis [21], it will be interesting to further characterize the potential relationship between Ngn3 and Sim1 in the developing pancreas.

Despite the existing knowledge that Notch signaling is important for islet cell development, the mechanism by which Notch signaling is more active within Ngn3-expressing cells is unknown. Studies to date have not distinguished between transcriptional control and other control mechanisms. The data presented here suggest a mechanism by which Notch signaling components are regulated by Ngn3 at the transcriptional level. This observation is supported by previous reports that found that Dll1 and -4, presenillin1 and -2, Notch2, and Hes6 were upregulated in pancreatic ductal cells infected with a viral Ngn3 expression construct (5 and 6). Upregulation of many of these genes and other Notch signaling pathway genes by Ngn3 was also observed in vitro in our Ngn3-inducible mESC line (Tables 1, 2).

It has been shown that cells having premature expression of Ngn3 in the developing pancreas are sensitive to genetic downregulation of Notch signaling components [2]. In addition, Murtaugh et al. [27] have demonstrated that Ngn3-expressing cells in the developing pancreas must escape Notch signaling in order to differentiate. Consistent with these in vivo observations, we have demonstrated, in vitro, that Notch signaling is more sensitive to γ-secretase inhibition in EB10 cells induced to express Ngn3 (Fig. 4A). Simultaneous Notch signaling inhibition and Ngn3 activation resulted in significantly increased pancreatic hormone transcripts, including insulin and somatostatin, indicating that this artificial system may recapitulate some important molecular mechanisms that occur during the development of islet hormone-producing cells. The Ngn3 ESC line, therefore, represents a useful experimental tool to better understand the role of Ngn3 in development and will be of particular use to further investigate the nature of promoter regulation using combined chromatin immunoprecipitation and microarray technologies.


The authors indicate no potential conflicts of interest.


We thank the anonymous reviewers for helpful comments and suggestions to strengthen this manuscript. We also thank Jacques Michaud and Gérard Gradwohl for helpful discussions regarding Sim1 and Ngn3, respectively, and Archana Deshpande for technical assistance with microarray data ac-quisition. This work was supported by the University of Wisconsin-Madison Department of Surgery National Institutes of Health (NIH) postdoctoral training grant (T32 AI52037) (to N.R.T.) and by grants from the NIH National Institute of Diabetes and Digestive and Kidney Diseases Beta Cell Biology Consortium (U19-DK061244) and Juvenile Diabetes Research Foundation (1-2004-145) (to J.S.O.). N.T. and R.V. contributed equally to this work.