Neurogenin3 (ngn3), a basic helix-loop-helix (bHLH) transcription factor, functions as a pro-endocrine factor in the developing pancreas: by itself, it is sufficient to force undifferentiated pancreatic epithelial cells to become islet cells. Because ngn3 expression determines which precursor cells will differentiate into islet cells, the signals that regulate ngn3 expression control islet cell formation. To investigate the factors that control ngn3 gene expression, we mapped the human and mouse ngn3 promoters and delineated transcriptionally active sequences within the human promoter. Surprisingly, the human ngn3 promoter drives transcription in all cell lines tested, including fibroblast cell lines. In contrast, in transgenic animals the promoter drives expression specifically in regions of ngn3 expression in the developing pancreas and gut; and the addition of distal sequences greatly enhances transgene expression. Within the distal enhancer, binding sites for several pancreatic transcription factors, including hepatocyte nuclear factor (HNF)-1 and HNF-3, form a tight cluster. HES1, an inhibitory bHLH factor activated by Notch signaling, binds to the proximal promoter and specifically blocks promoter activity. Together with previous genetic data, these results suggest a model in which the ngn3 gene is activated by the coordinated activities of several pancreatic transcription factors and inhibited by Notch signaling through HES1.
During development, the formation of the pancreas and its subsequent differentiation into the different exocrine and endocrine cell types result from the orderly activation and extinction of a large number of genes. Unique sets of nuclear transcription factors expressed in the developing and matured pancreas control these gene expression events (1,2). Among these factors, the basic helix-loop-helix (bHLH) factor neurogenin3 (ngn3) plays a central role in initiating the differentiation of the islet cells.
The pancreas forms from two groups of cells that bud from the dorsal and ventral gut endoderm at the foregut/midgut junction on embryonic days (E) 9.5–10.5 in the mouse (3). Most of the cells initially composing these buds are apparently uncommitted progenitor cells: they have the capacity to differentiate into any of the mature pancreatic cell types. The first differentiated cells to appear in the developing buds are the endocrine cells. Glucagon-expressing α-cells are the first to form, followed by insulin-producing β-cells, somatostatin-producing δ-cells, and pancreatic polypeptide-producing PP-cells. The endocrine cells initially appear at or close to the ducts, subsequently aggregate, and form islet structures late in fetal development.
In the developing pancreas, ngn3 is a marker for a population of cells that are in transit from undifferentiated epithelial progenitor cells to mature endocrine cells. Ngn3 expression peaks during the major wave of endocrine cell genesis at E15.5, is greatly diminished at birth, and is largely absent from the mature pancreas. The cells expressing ngn3 are scattered in the epithelium of the early pancreatic buds and adjacent to or in the ducts after E13. These ngn3-expressing cells are endocrine cell precursors; they coexpress some markers of differentiating endocrine cells but do not express islet hormones or markers of mature endocrine cells (4567).
Several lines of evidence suggest that ngn3 plays an essential role in determining which progenitor cells will ultimately become islet cells. Mice carrying a targeted disruption of the ngn3 gene fail to generate pancreatic endocrine cells and die 1–3 days postnatally from diabetes (6). The pancreata of the ngn3-null mice fail to express islet transcription factors isl1, pax4, pax6, and neuroD1 (6), whereas animals lacking pax6, neuroD1, nkx6.1, or nkx2.2 continue to express ngn3 in the pancreas (7), suggesting that ngn3 lies upstream of these factors and is necessary for initiating endocrine cell differentiation.
Furthermore, experiments using the Pdx1 promoter to drive early and broad expression of ngn3 in the pancreatic buds of transgenic mice demonstrate that ngn3 alone is sufficient to force precursor cells to an endocrine fate (4,7). Therefore, ngn3 is both necessary and sufficient to drive the formation of islet cells during pancreatic development, and the signals that control ngn3 expression also control islet cell formation.
Ngn3 expression appears to depend on gross positional cues that direct it to the gut and pancreatic endoderm and local signals that limit it to specific cells. The notch-signaling pathway, in a manner analogous to its role in lateral inhibition in the developing neural ectoderm, prevents ngn3 expression in all but a few scattered cells in the ducts of the developing pancreas. Disruption of notch signaling in the pancreas with a dominant-negative form of notch, or by null mutations of the genes encoding the notch ligand δ-like 1 (dll1) or the downstream mediators RBP-Jκ (4) or HES1 (8), causes premature and expanded expression of ngn3 in the pancreas and accelerated endocrine differentiation.
The positive signals that drive ngn3 expression in the pancreas are less well understood. In mice homozygous for a null mutation of the gene encoding the endoderm transcription factor hepatocyte nuclear factor (HNF)-6, ngn3 expression is reduced but not completely abolished. The mouse ngn3 promoter has at least two binding sites for HNF-6, and HNF-6 can stimulate the intact promoter (9). Other factors that activate ngn3 gene expression have not been identified.
As an initial step in understanding the regulation of ngn3 expression, we mapped and sequenced the mouse and human ngn3 gene promoters. Surprisingly, the human promoter drives transcription at a high level in all of the tested cell lines, including the fibroblast cell lines. In contrast, in transgenic mice the promoter drives β-galactosidase expression to a limited cell population in the developing pancreas and gut, and distal promoter sequences enhance this activity. The distal promoter contains a cluster of binding sites for HNF-1, HNF-3, and HNF-6, whereas the proximal promoter contains several high-affinity binding sites for the HES1 bHLH transcription factor, a negative mediator of notch signaling in the pancreas. Together, these studies provide an initial outline of the architecture of the ngn3 promoter and of the factors that regulate ngn3 expression and thereby control endocrine cell differentiation.
RESEARCH DESIGN AND METHODS
Cloning of the mouse and human ngn3 gene promoter.
A λ-phage genomic clone containing the mouse ngn3 open reading frame, clone 17/6-1-1-2 (10), was kindly provided by D. J. Anderson (California Institute of Technology, Pasadena, CA). From this phage clone, a 1-kb fragment containing sequences upstream of the open reading frame was subcloned and sequenced (Genbank accession number U76208). Human ngn3 genomic clones were obtained by screening a λ-DASH human genomic library (Stratagene) with the mouse ngn3 genomic fragment. The clone containing the longest 5′ flanking sequence, clone 14H, was subcloned, sequenced (Genbank accession number AF234829), and used for generating reporter gene plasmids.
5′ Rapid amplification of cDNA ends.
The 5′ end of the mouse ngn3 cDNA was identified by 5′-rapid amplification of cDNA end (RACE), using a modification of the protocol from the 5′-RACE System Version 2.0 (Gibco). For mouse cDNA, 2.5 pmol of specific primer JL1 (5′-ATCCTGCGGTTGGGAA-3′) was annealed to 1 μg total RNA from mouse E15.5 pancreas. Reverse transcription was carried out using SuperScript II reverse transcriptase (Gibco). After first-strand cDNA synthesis, the original mRNA template was removed by treatment with RNAse, and homopolymeric dCTP tails were then added to the 3′-end of the cDNA using terminal deoxynucleotidyl transferase. Using this product as a template, we carried out 35 cycles of polymerase chain reaction (PCR) using 5′ RACE Abridged Anchor Primer (Gibco) and JL2 (5′-TGGAAGGTGTGTGTGTGCCAG-3′) as primers. For the nested PCR, we used Abridged Universal Amplification Primer (Gibco) and JL3 (5′-GATCTAGAGACTTAGAGGTCACTGC-3′) as primers and performed 35 cycles of PCR. The PCR products were subcloned and sequenced.
Reporter gene constructs.
To generate reporter plasmids, fragments of the 5′ region of the human ngn3 gene obtained by restriction digestion were ligated upstream of the luciferase gene in the plasmid pFOXLuc1 or upstream of the thymidine kinase (TK) minimal promoter gene in the plasmid pFOXLuc1TK (11).
Cell culture and transient transfections.
βTC3-, αTC1.6-, and MPAC cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 2.5% fetal bovine serum and 15% horse serum. NIH3T3 cells were grown in DMEM medium supplemented with 10% calf serum. Cos7 cells were grown in DMEM medium with 10% fetal bovine serum with 4 mmol/l glutamine. For transient mammalian cell transfections, cells were plated in six-well tissue culture plates 24 h before transfection. For the standard reporter gene analysis, 2 μg of luciferase reporter plasmids were transfected into the cells using TransFast lipid reagent (Promega) according to the manufacturer’s instructions. For assessing the effect of the expression of HES1 on the ngn3 promoter, we cotransfected the amount of HES1 expression plasmid DNA (pcDNA3Hes1; kindly provided by R. Kageyama, Kyoto University, Kyoto, Japan ) indicated with 2 μg luciferase reporter plasmids. Cells were harvested, and luciferase assays were performed 48 h after transfection, as previously described (13). Luciferase activity was corrected for cellular protein concentration. All reporter gene analyses were performed on at least three occasions, and data are expressed as mean ± SE.
Generation of transgenic mice and detection of β-galactosidase.
We generated the plasmids pNAT6B and pNAT3B by ligating human ngn3 promoter fragments extending from −5.7 kb to 261 bp and from −2.6 kb to 261 bp upstream of the human β-globin intron and the bacterial β-galactosidase gene. Each plasmid was linearized and microinjected (1.5 ng/μl) into murine oocyte pronuclei. The injected embryos were transferred to pseudopregnant females, and the fetal pancreata with stomach and small intestine were harvested at E15.5 from the founder mice. Tissues were prefixed for 30 min at 4°C in 4% paraformaldehyde and then incubated overnight in X-gal (400 μg/ml) substrate at 37°C (−2.6 kb promoter) or room temperature (−5.7 kb promoter). Tissues were then fixed again in 4% paraformaldehyde for 30 min, paraffin embedded, and sectioned at 5 μmol/l. Genotype was determined by PCR using primers specific for the human ngn3 promoter sequence. β-galactosidase activity was assayed in six independent founder fetuses that had integrated the −2.6 kb promoter construct and in eight independent founder fetuses that had integrated the −5.7 kb promoter construct.
Immunohistochemistry was performed on paraffin-embedded sections as previously described (7). Primary antibodies were used at the following dilutions: guinea pig anti-insulin (Linco) at 1:5,000; guinea pig anti-glucagon (Linco) at 1:10,000; and rabbit anti-ngn3 (7) at 1:5,000. Biotinylated secondary antibodies (Vector) were detected with the ABC Elite immunoperoxidase system (Vector).
Preparation of proteins and elecrophoretic mobility shift assay.
HNF-3β and HNF-1α proteins were produced in vitro using SP6 and T7 TNT Quick Coupled Lysate System (Promega) using pGEM-1ratHNF-3β (generous gift from R. Costa, University of Illinois at Chicago) and pcDNA3–HNF-1α (generous gift from M. Stoffel, Rockefeller University, New York) as templates. Glutathione S-transferase (GST)-fused HES1 protein was produced in Escherichia Coli BL21–competent cells using the pGEX2T plasmid system (Promega). Nuclear extracts from αTC1.6-, βTC3-, and NIH3T3 cells were prepared following the procedure described by Sadowski and Gilman (14).
Single-stranded oligonucleotides corresponding to the sequences in the human ngn3 promoter were 5′ end-labeled with [γ-32P]ATP using T4 polynucleotide kinase. The labeled oligonucleotides were column-purified and annealed to an excess of the complementary strand. For HNF-3β and HNF-1α binding experiments, elecrophoretic mobility shift assay (EMSA) buffers and electrophoresis conditions were as previously described (11). For HES1 binding experiments conditions were the same, except that the poly(dI-dC) concentration was decreased to 15 ng/μl, and 1 μl of the in vitro reaction mixture, 2 μg of nuclear extracts, or 400 ng of GST-fused Hes1 protein was used for each binding reaction. When using antibodies, 1 μl of each antibody was incubated with the binding mix for 15 min at room temperature before gel electrophoresis. The antisera against HNF-3α, -3β, and -3γ were a generous gift from R. Costa (University of Illinois), and the HES1 antiserum was a generous gift from Y. Jan (University of California, San Francisco). The anti–HNF-1α antiserum was purchased from Santa Cruz Biotechnology.
The following oligonucleotides were used as labeled probes or competitors in EMSA reactions (top strands shown):
Mapping the human ngn3 promoter.
As an initial step in understanding the regulation of ngn3 gene expression, we identified and sequenced the mouse and human ngn3 promoters (Genbank accession numbers AF234829 and U76208) (Fig. 1). Using RNA purified from E15.5 fetal mouse pancreas, the transcription start sites of the murine ngn3 gene was determined by 5′ RACE. All 5′ RACE products identify the same start site (Fig. 1), 30 bp downstream from a putative TATAA box (data not shown). The region upstream of the start site is highly conserved in the mouse, human, and rat, with the region of highest homology in the mouse and human extending ∼300 bp upstream. A CCAAT sequence element lies at –85 bp relative to the transcription start site. Several other potential sequence elements are identified in Fig. 1.
Activity of the promoter in cell lines.
A series of progressive 5′ deletions of the ngn3 promoter each extending to +261 bp on the 3′ end were linked to the firefly luciferase gene and were tested in cell lines (Fig. 2). Serial deletions down to –502 bp do not diminish the promoter activity in vitro. Surprisingly, the promoter drives transcription to a high level in all of the tested cell lines, including the fibroblast cell lines. This high nonspecific activity appears to reside in the proximal promoter, because the shortest construct is still very active in all of the examined cell lines.
Activity of the promoter in transgenic mice.
Whereas transient transfections in cell lines may provide some indication of promoter activity, these tumor cells are not representative of the cells in the developing pancreas where ngn3 is normally expressed. Therefore, we produced mice carrying a transgene with either 5.7 or 2.6 kb of the upstream sequence from the human ngn3 gene driving the bacterial gene encoding β-galactosidase. Founder mice were harvested at E15.5, the normal peak of ngn3 expression in the fetal mouse pancreas (7).
Animals carrying the 5.7-kb construct strongly and selectively express β-galactosidase in central regions of the developing pancreas and in the gut epithelium, the same regions where ngn3 is normally expressed at this time during development (Fig. 3A). Although the level of expression is significantly lower with the 2.6 kb construct (Figs. 3B and C), the overall pattern of β-galactosidase expression is the same (data not shown).
We used immunohistochemistry to identify the cells expressing β-galactosidase in the transgenic mice carrying the 5.7 kb promoter construct (Fig. 4). The β-galactosidase–expressing cells are predominantly localized to the ducts. Most of the β-galactosidase–expressing cells do not express islet hormones, although occasional β-cells coexpress insulin and β-galactosidase.
Despite the close colocalization of β-galactosidase activity and ngn3 protein expression, specifically in the same regions of the developing pancreas and gut, there is not a perfect match. Some of the β-galactosidase–positive cells coexpress high levels of ngn3, but many do not, and many ngn3-expressing cells contain little or no β-galactosidase activity. Most likely, this discrepancy derives from differences in the timing of accumulation and degradation of the two gene products, rather than a difference in the onset and extinction of gene expression. The exact timing for initial detection of each gene product in a particular cell depends on its rate of accumulation and threshold for detection, and therefore should not be expected to be identical. In addition, some ngn3-expressing cells may randomly silence the transgene, a poorly understood phenomenon observed with many promoters in transgenic mice (15). The very brief but abundant expression of ngn3 in progenitor cells indicates that the mRNA and protein accumulate rapidly but have very short half-lives. In contrast, β-galactosidase has a fairly long half-life in mammalian cells (16) and could be expected to peak later and persist in cells after ngn3 is no longer detectable. Therefore, many of the β-galactosidase–expressing cells apparently represent a stage of islet cell differentiation that occurs after ngn3 gene production has ceased but before hormone expression has started. The large number of these cells suggests that this intermediate stage of differentiation may last longer than the initial ngn3-expressing stage.
Starting at E15.5, endogenous β-galactosidase expression can be detected at low levels along the brush border of the intestinal villi in both transgenic and nontransgenic embryos. Stronger β-galactosidase activity can also be detected in a speckled pattern that is most prominent in the small intestine of the transgenic mice but is absent in their nontransgenic littermates (Fig. 3A). Sectioning of the gut reveals that this β-galactosidase signal derives from scattered cells within the intestinal epithelium (Fig. 5). This pattern of β-galactosidase expression suggests that the ngn3 promoter is also active in a subset of progenitor cells in the developing gut; these may be progenitors for gut endocrine cells. As in the pancreas, this β-galactosidase activity partially overlaps endogenous ngn3 expression, again suggesting that the peak of β-galactosidase accumulation is delayed relative to ngn3 (Fig. 5C).
Multiple factors bind to the ngn3 promoter.
To identify nuclear factors that bind to the ngn3 promoter, we synthesized a series of oligonucleotides spanning potentially important DNA binding sites within the promoter and tested them for binding to nuclear proteins by EMSA.
Members of the HNF-3 family of winged helix transcription factors have been implicated in pancreatic development and islet function (171819202122). Based on their similarity to a consensus HNF-3 binding site (23), there are several potential HNF-3 binding sites within the 5.7-kb human ngn3 promoter. Two of the most promising sites lie at –3,687 bp and at –200 bp. We tested both binding sites by EMSA and found that both sites bind with high affinity to in vitro–produced HNF-3β (Fig. 6). Using extracts from βTC3- and αTC1.6 cells, a single major complex binds to both sites and is recognized specifically by an antiserum to HNF-3β. In addition, coexpression of HNF-3β can activate the ngn3 promoter in transiently transfected 3T3 fibroblast cells (data not shown).
The –3,687 bp HNF-3β binding site forms part of a cluster of potential DNA binding sites for known pancreatic transcription factors (Fig. 1), including potential sites for hox homeodomain–type transcription factors as well as cut homeodomain–transcription factor HNF-6 and the Pou homeodomain HNF-1 factors (24). HNF-6 binding to the ngn3 promoter has been previously demonstrated (9).
We tested an oligonucleotide spanning the potential HNF-1 binding site by EMSA and found that it can bind to in vitro–produced HNF-1α. In addition, in nuclear extracts from βTC3-cells, a major low mobility complex binds to the oligonucleotide and is recognized specifically by antiserum to HNF-1α (Fig. 7). The similar hox homeodomain–type binding site immediately downstream of the HNF-1α binding site will not bind HNF-1α (data not shown).
HES1 inhibition of the ngn3 promoter.
It has been proposed that Notch receptor signaling through the transcriptional regulator HES1 may prevent the expression of ngn3 in all but a small subset of the cells in the developing pancreas (8). To test the ability of HES1 to directly inhibit the ngn3 promoter, we expressed the HES1 cDNA from a cytomegalovirus (CMV) promoter–driven expression plasmid in 3T3 cells along with the ngn3 promoter luciferase plasmid (Fig. 8). HES1 dramatically and specifically inhibits the ngn3 promoter. Removal of 5′ sequences down to –502 bp does not significantly reduce the ability of HES1 to inhibit the promoter.
To further map sequences competent to respond to HES1 repression, plasmids were constructed with either the human ngn3 gene promoter sequence from −208 bp to 40 bp (proximal promoter) linked to the firefly luciferase gene, or the sequences from −2.6 kb to −208 bp (distal promoter) upstream of the herpes virus TK promoter linked to the firefly luciferase gene. The small proximal promoter retains most of the capacity for HES1 repression, whereas the distal sequences are repressed weakly by HES1.
Within the proximal 208 bp of the promoter, there are several potential HES1 binding sites based on the consensus binding sites for HES1 (CTNGTG) (25) and its Drosophila homologs hairy/enhancer-of-split (CGCGTC) (26,27) (Fig. 9A). We tested three oligonucleotides containing four of these sites for binding to bacterially produced HES1 protein by gel mobility shift assay. All three oligonucleotides bind HES1 and do so with greater affinity than the previously described high affinity tandem sites from the mouse HES1 gene (25) (labeled H1P in Fig. 9C). All four of these sequences are conserved in the mouse ngn3 promoter (Fig. 1).
In the present study, we identified and characterized the ngn3 gene promoter. Although 2.6 kb of the promoter is sufficient to direct expression correctly in transgenic mice, distal sequences greatly enhance this activity. The distal enhancer binds the pancreatic transcription factors HNF-6 (9), -1α, and -3β. These distal sites do not explain all of the activity of the promoter, however, as the proximal promoter has strong ubiquitous activity, which in turn is suppressed by the notch signal mediator HES1.
If ngn3 gene promoter activity is tightly restricted in vivo, then why are the same promoter constructs active in all cell lines tested? This difference could reflect differences between cells in vivo and transformed cell lines. Although the islet and ductal cell lines that were used in these experiments express low levels of ngn3 that can be detected by PCR (7), none of these cell lines are representative of the islet cell progenitors that transiently express ngn3 at high levels in vivo. Therefore, the promoter activity seen in these cell lines might be far below the activity produced by genuine ngn3-expressing cells. However, in the fibroblast cell lines ngn3 activity already approaches the activity of the potent viral Rous sarcoma virus long-terminal repeat. Because we do not observe the same indiscriminate activity in vivo, we can assume that either a potent activator present in the tested cell lines is absent from most cells in vivo, or that a strong repressor that is widely expressed in vivo is absent from the cell lines.
Alternatively, the physical state of the DNA may explain the restriction of promoter activity in vivo. In the cell lines, promoter activity was assayed by transient transfection, and therefore the reporter gene was transcribed from episomal plasmid DNA free from the constraints of chromatin structure. However, in transgenic mice the transgene integrates into the genome. Chromatin structure can have profound effects on gene activity (282930). When tightly bound to histones within the nucleosome in chromosomal DNA, the promoter may not be accessible to the general transcription factors such as SP1 and NFY that apparently drive the high nonspecific activity of the proximal promoter in plasmid DNA. Only when the local chromatin is remodeled by interactions with cell type–specific nuclear proteins could these general factors activate transcription.
The HNF-3 winged-helix transcription factors, which have at least two binding sites on the human ngn3 gene promoter, may play a critical role in opening transcriptionally silent DNA. HNF-3 actually binds with higher affinity to DNA in nucleosome core particles than to free DNA (31). HNF-3 binding can then initiate the recruitment of histone acetyltransferases and additional transcription factors and the formation of an active transcription complex including other general and cell type–specific transcription factors (313233). In this regard, it is interesting that both identified HNF-3 binding sites are adjacent to binding sites for other key transcription factors, such as SP1 and NFY in the proximal promoter and HNF-1 and HNF-6 in the distal promoter.
As initiators of transcription complex formation on a number of genes, the HNF-3 factors play an important role in patterning the early gut endoderm (34). HNF-3 factors have been implicated in the regulation of liver-specific genes (23), but they also appear to regulate the gene for the pancreatic/duodenal transcription factor PDX1 (171819), as well as endocrine-specific genes such as glucagon (21,22,35). Although HNF-3β is the predominant HNF-3 family member expressed in the pancreatic cell lines, the HNF-3 proteins that bind to the ngn3 promoter in vivo are less certain. All three HNF-3 genes are active in the pancreas, but the cell types expressing each HNF-3 subtype have not been carefully defined (36). Both HNF-3β and -3α can be detected in the fetal pancreas; and in the adult pancreas, HNF-3β can be detected in both the acinar cells and the islet cells, whereas the HNF-3α gene is active in the islets (20,37,38).
The two HNF-1 proteins, HNF-1α and -1β, are also expressed in the pancreas, but again the exact cells expressing each protein have not been defined (394041). Mice homozygous for a targeted disruption of the HNF-1α gene have smaller islets and reduced insulin secretion (42), suggesting that a decrease in ngn3 expression in these animals during development could impair islet cell formation and glucose homeostasis. However, it has been proposed that the diabetes in these animals and (by inference) the diabetes seen in humans with heterozygous mutations in HNF-1α results from defects in glucose sensing by the β-cell, rather than decreased islet mass (42).
The cut-homeodomain protein HNF-6 is expressed in the duct cells of the developing pancreas, where ngn3-expressing cells initially appear (37). Mice lacking HNF-6 have significantly reduced ngn3 expression and markedly decreased islet mass. Jacquemin et al. (9) identified two HNF-6 binding sites in the mouse ngn3 gene promoter. The proximal site is not conserved in the human promoter, but the distal site is conserved, along with at least two other potential HNF-6 binding sites in the distal promoter (Fig. 1).
Despite the importance of these endoderm transcription factors in activating the ngn3 gene, ngn3 expression is limited to a small subset of the cells expressing these factors in the developing pancreas and gut. Restriction of ngn3 expression occurs in part through the notch-signaling pathway (4,8). Notch signaling controls cell fate decisions in many different settings during development (43,44). Analogous to the limitations on ngn3-expressing cells in the ducts of the developing pancreas, lateral inhibition through notch signaling in the developing neural ectoderm limits the number of cells that can activate the default neural development program. Binding of ligand to notch receptors activates a chain of signaling events that results in the expression of the hairy/enhancer-of-split proteins, inhibitory bHLH factors that bind to N boxes in the proneuronal bHLH genes, thereby silencing these genes and preventing neurogenesis.
HES1 is the predominant mammalian hairy/enhancer-of-split homologue expressed in the developing pancreas (8). In mouse models, loss of HES1 or defects in notch signaling result in increased ngn3 expression and premature and broader endocrine differentiation in the developing pancreas (4,8). Our data demonstrate that like the proneuronal genes in Drosophila (26,27) and mash 1 in mammals (45), HES1 represses ngn3 expression by directly suppressing the ngn3 promoter.
Interestingly, HES1 targets sequences within the strong basal activation region of the proximal promoter, a region containing a high-affinity HNF-3 site that we propose plays a critical role in initiating the formation of the ngn3 gene–activating complex. In contrast to the opening of chromatin structure by HNF-3, HES1 recruits the TLE transcriptional co-repressors (46), the mammalian homologues of Drosophila groucho, which promote histone deacetylation and a closed chromatin structure (47). In this model, the precise balance of HES1 and HNF-3 effects would determine whether chromatin structure at the ngn3 promoter is open and transcriptionally active or closed and transcriptionally silent, as occurs in most of the fetal duct cells.
In summary, these studies provide an outline of the opposing signals that activate and repress ngn3 gene expression and therefore control the initiation of islet cell formation. Given the critical role that ngn3 plays in islet cell genesis, these signals may eventually prove useful in generating new islet cells for patients with diabetes. In addition, the ngn3 promoter may prove useful for targeting genes specifically to endocrine cell precursors during endocrine cell genesis.
This work is supported by grants from the Juvenile Diabetes Foundation International (to S.B.S., H.W., and R.G.M.), the Nora Eccles Treadwell Foundation, and the National Institutes of Health (DK21344 and DK55340 to M.S.G.).
We thank V. Luukela and Y. Zhang for excellent technical assistance and members of the German laboratory for helpful discussion.
R.G.M. is currently affiliated with the Department of Medicine, University of Virginia, Charlottesville, Virginia.
Address correspondence and reprint requests to Michael S. German, MD, University of California, San Francisco, HSW 1090, Box 0534, 513 Parnassus Ave., San Francisco, CA 94143. E-mail:.
Received for publication 17 December 2000 and accepted in revised form 20 February 2001. Posted on the World Wide Web at www.diabetes.org/diabetes on 11 April 2001.
bHLH, basic helix-loop-helix; CMV, cytomegalovirus; DMEM, Dulbecco’s modified Eagle’s medium; E, embryonic day; EMSA, elecrophoretic mobility shift assay; GST, glutathione S-transferase; HNF, hepatocyte nuclear factor; ngn3, neurogenin3; PCR, polymerase chain reaction; RACE, rapid amplification of cDNA end; TK, thymidine kinase.