|Home | About | Journals | Submit | Contact Us | Français|
Notch signaling inhibits differentiation of endocrine cells in the pancreas and intestine. In a number of cases, the observed inhibition occurred with Notch activation in multipotential cells, prior to the initiation of endocrine differentiation. It has not been established how direct activation of Notch in endocrine precursor cells affects their subsequent cell fate. Using conditional activation of Notch in cells expressing Neurogenin3 or NeuroD1, we examined the effects of Notch in both organs, on cell fate of early endocrine precursors and maturing endocrine-restricted cells respectively. Notch did not preclude the differentiation of a limited number of endocrine cells in either organ when activated in Ngn3+ precursor cells. In addition, in the pancreas most Ngn3+ cells adopted a duct but not acinar cell fate; whereas in intestinal Ngn3+ cells, Notch favored enterocyte and goblet cell fates, while selecting against endocrine and Paneth cell differentiation. A small fraction of NeuroD1+ cells in the pancreas retain plasticity to respond to Notch, giving rise to intraislet ductules as well as cells with no detectable pancreatic lineage markers that appear to have limited ultrastructural features of both endocrine and duct cells. These results suggest that Notch directly regulates cell fate decisions in multipotential early endocrine precursor cells. Some maturing endocrine-restricted NeuroD1+ cells in the pancreas switch to the duct lineage in response to Notch, indicating previously unappreciated plasticity at such a late stage of endocrine differentiation.
Endocrine cells in the pancreas and intestine differentiate from multipotential epithelial cells derived from the early gut endoderm. In the pancreas, relatively undifferentiated epithelial cells give rise to the duct, acinar, and endocrine lineages prior to birth. At least five different endocrine cell types form the islets of Langerhans including insulin producing β cells, as well as α, , PP, and ε cells that produce glucagon, somatostatin, PP, and ghrelin (Habener et al., 2005; Oliver-Krasinski and Stoffers, 2008). In contrast, enteroendocrine cells that express one or more of 12 hormones continuously differentiate from precursors throughout postnatal life.
Endocrine differentiation in both the pancreas and intestine is regulated by the temporal expression of basic helix loop helix (bHLH) transcription factors to sequentially restrict subsequent differentiation to specific lineages. Expression of the bHLH transcription factor, Neurogenin 3 (Ngn3) initiates endocrine differentiation following specification of the pancreatic epithelium by the homeodomain protein, Pdx1 early in pancreagenesis (Gu et al., 2002).
The absence of pancreatic endocrine cells in Ngn3−/− mice suggests that Ngn3 is required for their specification (Gradwohl et al., 2000). Lineage analysis of the descendants of Ngn3+ cells showed that all endocrine cells in the pancreas arose from Ngn3+ cells, indicating that the effects of Ngn3 were cell autonomous. However, lineage tracing also revealed that small numbers of acinar and duct cells arose from Ngn3 expressing cells, suggesting that Ngn3+ cells were not restricted to an endocrine cell fate (Gu et al., 2003; Schonhoff et al., 2004).
NeuroD1, another bHLH protein, was initially described as an activator of the insulin gene (Naya et al., 1995). NeuroD1 knockout mice develop severe diabetes with reduced numbers of β cells (Naya et al., 1997). The absence of NeuroD1 in Ngn3 null mice indicates that NeuroD1 is downstream of Ngn3 (Gradwohl et al., 2000). Ngn3 and the homeodomain protein, NKX2.2, (Anderson et al., 2009; Huang et al., 2000) directly activate NeuroD1 transcription, suggesting that NeuroD1 is expressed at a later stage of islet differentiation.
In the intestine, the three secretory lineages, enteroendocrine, Paneth, and goblet cells require the bHLH protein, Atoh1 for differentiation (Shroyer et al., 2005; Yang et al., 2001). Presumably, Ngn3 initiates endocrine differentiation as enteroendocrine precursor cells segregate from a common secretory progenitor cell. As in the pancreas, intestinal enteroendocrine cells are absent from Ngn3 null mice although some endocrine cells in the stomach differentiate in the absence of Ngn3 expression (Jenny et al., 2002; Lee et al., 2002). Secretin and cholecystokinin cells fail to develop in neuroD1 null mice whereas other enteroendocrine cell types are present (Naya et al., 1997). NeuroD1 is expressed in nearly all enteroendocrine cell types where its may have a role in inhibiting cell proliferation as cells mature (Mutoh et al., 1998; Ratineau et al., 2002).
A number of studies suggest that Notch signaling inhibits endocrine differentiation in both the pancreas and the intestine. Many of the effects of Notch result from its inhibition of bHLH proteins that activate cellular differentiation programs. Thus Ngn3 and NeuroD1 are potential targets of Notch in the pancreas and the intestine. Notch signaling increases expression of its transcriptional effector protein, Hes1, in the pancreas (Jarriault et al., 1998), inhibiting expression and/or transcriptional activity of bHLH proteins (Davis and Turner, 2001). The Ngn3 gene has multiple Hes1 binding sites and is repressed by Hes1, suggesting it may be a target of Notch (Lee et al., 2001).
Activation of Notch at a relatively early stage of pancreatic development in cells expressing Pdx1 resulted in failure to develop both endocrine and acinar cells with the remaining pancreatic cells trapped in a relatively undifferentiated state (Hald et al., 2003; Murtaugh et al., 2003). Likewise, activation of Notch in cells expressing either villin or fatty acid binding protein showed a significant reduction in enteroendocrine and goblet cells (Fre et al., 2005; Stanger et al., 2005). The timing and context of Notch activation is a major determinant of its effects on organogenesis and differentiation. The studies described above involved the effects of Notch activation in broad populations of multipotential cells in each organ that give rise to several cell lineages in addition to endocrine cells. The observed effects of Notch in these studies were likely the result of its activation prior to the initiation of endocrine differentiation. The inhibitory role of Notch was further suggested by excessive endocrine differentiation seen with widespread disruption of Notch function in the pancreas (Apelqvist et al., 1999; Jensen et al., 2000) and in the fetal intestine (Jensen et al., 2000).
The effects of Notch activation in endocrine precursor cells in the pancreas have been examined in limited detail. The results were interpreted to suggest complete inhibition of endocrine differentiation in early endocrine precursors with less pronounced effects in maturing islet cells (Greenwood et al., 2007; Murtaugh et al., 2003). The effects of Notch activation in enteroendocrine precursor cells have not been described.
In the present study, we have examined the effect of conditionally activating Notch signaling in early and late endocrine precursor cells expressing Ngn3 or NeuroD1 respectively in the pancreas and intestine. Unexpectedly, activation of Notch in Ngn3+ cells did not prevent the initiation of endocrine differentiation in the pancreas or the intestine. However a number of islet cell types failed to mature in the pancreas. Many cells switched to a duct cell fate in pancreas and enterocyte and goblet cell fate in intestine. Surprisingly, some NeuroD1+ cells in the pancreas respond to Notch and differentiate into duct cells, whereas maturing enteroendocrine precursors were relatively unaffected by Notch.
Sequences encoding Cre recombinase that included a nuclear localization signal and simian virus 40 polyadenylation sequences were introduced into a murine BAC clone, RPCI-23188B11 (Invitrogen) containing the NeuroD1 locus, by homologous recombination in E. Coli with a linear DNA fragment amplified from the suicide vector pKD4 as described previously (Cotta-De-Almeida et al., 2003; Datsenko and Wanner, 2000; Schonhoff et al., 2004). The Cre encoding sequences were inserted into the translation initiation ATG of the NeuroD gene. A Ngn3-CreERT2 transgene was generated by inserting CreERT2 sequences into the same Ngn3 BAC described previously (Schonhoff et al., 2004).
Transgenic mice were generated by pronuclear injection of the purified circular NeuroD1-Cre BAC or Ngn3-CreERT2 BAC DNA into the pronuclei of fertilized oocytes of B6XB6D2F1 mice. Founders were generated and were identified by genotyping with primers specific for the Cre transgene. We generated 2 NeuroD1-Cre and 3 Ngn3-CreERT2 pedigrees. For each line, the pedigrees showed nearly identical pattern of Cre expression. The Ngn3-Cre mice have been described earlier (Schonhoff et al., 2004). The PCR primers used for generating BAC NeuroD1-Cre transgene were: Sense- TGCTTGCCTCTCTCCCTGTTCAATACAGGAAGTGGAAACATGCCCAAGAAGA AGAGGAA and antisense-GGCTCGCCCATCAGCCCGCTCTCGCTGTATGATTTGGTCATCCTCCTTAGTTCCTATTCCGA.
For lineage tracing experiments, NeuroD1-Cre transgenic mice were crossed with ROSA26-LacZ (B6.129-Gt26Sortm) or ROSA26-EYFP (B6.129X1-Gt (ROSA) 26Sortm1(EYFP) Cos) indicator mice (Soriano, 1999). To determine if Ngn3 and NeuroD1 expressing cells are susceptible to Notch, Ngn3-Cre (Schonhoff et al., 2004), NeuroD1-Cre and Ngn3-creERT2 mice were crossed with ROSA26Notch/Notch knock-in mice that contain the intracellular domain of murine Notch1 and a bicistronic nuclear EGFP reporter downstream of a floxed stop sequence (Murtaugh et al., 2003). Cre activity was induced in Ngn3-creERT2;ROSANotch/+ mice by treating 6 weeks old mice with 2 mg tamoxifen (Sigma, T5648) dissolved in corn oil with one intraperitoneal injection per day for 5 days. Tissues were harvested and analyzed 7 days later. The Institutional Animal Care and Use Committee approved all vertebrate animal studies in accordance with NIH regulations.
Tissues for X-Gal staining were performed as previously described (Schonhoff et al., 2004). Fixed tissue samples were processed for either frozen or paraffin sections using standard methods.
The following primary antibodies were used for immunohistochemical and immunofluorescence analysis: Goat anti-NeuroD1 (1:500, Santa Cruz), Goat anti-EGFP (1:100, Abcam), Rabbit anti-GFP (1:10,000, Invitrogen), Guinea pig anti-Insulin (1: 5000, Sorin Biomedica), Guinea pig anti-PYY (1: 5000, a gift from G. Aponte, University of California, Berkley), Rabbit anti-ChromograninA (1:1000, Immunostar), Rabbit anti-Somatostatin (1:8000, a gift from R. Lechan, Tufts Medical Center), Rabbit anti-Glucagon (1: 3000, a gift from M. Appel, University of Massachusetts), Rabbit anti-Pancreatic Polypeptide (1: 5000, a gift from Chance, Eli Lilly), Rabbit anti-β-Gal (1:600, Cappel), Wisteria Fluoribunda Agglutinin (WFA, 1: 600, Sigma), Biotinylated Dolichos Biflorus Agglutinin (DBA, 1:1000, Vector Laboratories), Mouse anti-Neurogenin3 (1:2000, Beta cell Biology Consortium), Cytokeratin 19 (1:50, Developmental Studies Hybridoma Bank), Rabbit anti-Pdx1 ( 1:5000, Invitrogen), Rabbit anti-Muc2 antibody (1:400, a gift from Dr. S. Krasinski, Harvard Medical School), Rabbit anti-LPH (1:1,000, K. Yeh, Louisiana State University), Rabbit anti-Secretin (1:400, a gift from Dr. W. Chey, University of Rochester), Rabbit anti-CCK (1:1,000, a gift from Dr. W. Chey, University of Rochester). Intestinal epithelial Alkaline Phosphatase activity was detected using an Alkaline Phosphatase Substrate Kit (Vector Lab.)
For immunofluorescence analysis, Alexa Fluor 488 or 594 conjugated secondary antibodies were used at a dilution of 1:800. Tyramide Signal Amplification (TSA) kit (Molecular probes) was used for co-localization studies with two primary antibodies raised in the same species (Brouns et al., 2002). For immunohistochemical analysis, biotin conjugated secondary antibodies were used at a dilution of 1: 400 (Jackson ImmunoResearch) and were amplified with Vectastain ABC and DAB kits (Vector Laboratories) using standard protocols. Images were obtained with Nikon Eclipse E600 microscope. For morphometric analyses of cells expressing EYFP or EGFP, multiple sections separated by at least 25 μ from at least 3 mice were examined for each genotype. Results were normalized to the total area of pancreatic or islet tissue on the slides examined. Type II two-tailed or type III one-tailed Student t-test was used to analyze grouped data sets. P<0.05 was considered significant. Confocal images were obtained with a Zeiss LSM 510 Meta microscope. Digital deconvolution was performed on confocal images to confirm colocalization of multiple fluorescent labels in the same cell in multiple focal planes using the EPR program that was developed by the imaging group at the University of Massachusetts Medical School.
Pancreata from adult transgenic and age-matched control mice were fixed in 2.5% glutaraldehyde or in a mixture of 2.5% glutaraldehyde and 2% paraformaldehyde solution in 0.1 M cacodylate buffer pH 7.4. Part of the samples was post-fixed in 1% osmium tetroxide in 0.1 M phosphate buffer, pH 7.4 and embedded in Epon-Araldite. Areas of interest were trimmed, sectioned with a LKB Type 8802 ultramicrotome (60–100 nm), collected on uncoated nickel 200–300 mesh grids, desiccated, counterstained with uranyl acetate and Reynold’s lead citrate and observed in a Philips EM208S transmission electron microscope (Philips, The Netherlands). Indirect immunogold labeling studies for GFP were performed on osmified samples with the rabbit polyclonal serum (1:100, Invitrogen, USA), using 10 nm gold-labeled anti rabbit immunoglobulins (BB International, UK) as described (Rindi et al., 2002).
To assess the effect of Notch signaling in early endocrine progenitor cells of the pancreas, we conditionally activated Notch signaling in Ngn3+ cells by crossing ROSANotch/Notch mice (Murtaugh et al., 2003) with mice expressing Cre recombinase under control of a BAC containing the Ngn3 locus (Schonhoff et al., 2004). Excision of stop sequences by Cre enabled expression of the intracellular domain of Notch1 (NICD) as well as EGFP, which facilitated identification of cells expressing NICD. Ngn3-Cre;ROSANotch/+ mice were born at the expected Mendelian frequency and were indistinguishable from their wild type littermates at birth. Thereafter, Ngn3-Cre;ROSANotch/+ mice lost weight, became hyperglycemic, and died by postnatal day 3.
Islets failed to develop in pancreata of Ngn3-Cre;ROSANotch/+ mice, unlike control ROSANotch/+ mice (Fig. 1A, B). However we frequently observed individual chromogranin A (CGA) stained cells that expressed EGFP, indicating that Notch signaling in Ngn3+ cells did not preclude endocrine differentiation (Fig. 1B) as previously suggested (Murtaugh et al., 2003). Further characterization of the EGFP+ endocrine cells revealed that peptide YY (PYY) was the most frequently expressed hormone (Fig. 1C). A smaller number of cells expressed glucagon, with some coexpressing PYY (Fig. 1D, arrow in the inset points to one cell coexpressing PYY/Glucagon). However, we were unable to identify insulin-, somatostatin- or PP-expressing cells.
The neonatal lethality of Ngn3-Cre;ROSANotch/+ mice limited our studies to two-day-old mice. In order to examine the effects of Notch activation in early endocrine precursor cells of adult mice, we generated transgenic mice that expressed a tamoxifen inducible Cre (CreERT2) under control of the same Ngn3 BAC used for the Ngn3-Cre transgene. Following Tamoxifen induction in 8 week-old mice, we observed approximately 0–4 cells that expressed EYFP or EGFP per islet from both Ngn3-CreERT2;ROSAEYFP and Ngn3-CreERT2;ROSANotch/+ mice respectively (Fig. 1E, F). Morphometric analysis indicated that the frequency of labeling was not was not statistically different with Notch activation (p=0.43, type II, 2-tailed Student t-test). The relatively low frequency of labeling most likely is the result of the very small number of Ngn3+ cells normally present in the adult pancreas as well as the mosaic recombination characteristic of inducible Cre enzymes. Approximately 90% of Ngn3+ cells in adult mice differentiate into insulin-expressing cells (Fig. 1E, F) with nearly all staining for chromogranin A (not shown) in both Ngn3-CreERT2;ROSAEYFP and Ngn3-CreERT2;ROSANotch/+ mice. In contrast to the reduced endocrine differentiation in fetal pancreas described above with the noninducible Cre, our observations suggest that Ngn3+ cells in adult mice are no longer responsive to Notch.
Ngn3+ cells give rise to occasional duct cells appearing as individual or pairs of cells in the normal neonatal pancreas (Fig. 2A–C). In contrast, the majority of EGFP+ cells in neonatal Ngn3-Cre;ROSANotch/+ mice stained for dolichos biflorus agglutinin (DBA), a lectin expressed on pancreatic duct cells (Fig. 2D). We identified both small clusters as well as sheets of DBA+ cells with Notch activation. EGFP-expressing cells also stained for Wisteria Floribunda Agglutinin (WFA) and Cytokeratin 19, further identifying them as duct cells (Suppl. Fig. 1, top panel). In many cases, endocrine cells were in close proximity to duct cells but we were unable to identify any cells expressing both duct and endocrine markers (Fig. 2E, F). The absence of hormone colocalizing with duct lineage markers suggested that many Ngn3+ cells switched to the duct lineage as a result of Notch activation.
The increased frequency of duct cells arising from Ngn3+ cells following Notch activation was far greater than what we previously described in Ngn3-Cre;ROSA26 mice (Schonhoff et al., 2004). To determine whether the apparent increase in duct cells represented an increase in the absolute number versus a relative increase resulting from the loss of endocrine cells we compared the number of duct cells (standardized per unit area) arising from normal Ngn3+ cells in Ngn3-Cre;ROSAEYFP mice to the those arising from Ngn3+ cells in Ngn3-Cre;ROSANotch/+ mice. We found that 9.1% of EYFP+ cells in 2-day-old Ngn3-Cre;ROSAEYFP mice stained for DBA. The total number of cells arising from Ngn3+ cells in Ngn3-Cre;ROSANotch/+ mice was comparable to Ngn3-Cre;ROSAEYFP mice (Fig. 2G, p>0.064). With Notch activation, we observed a large increase in the fraction that was fated to the duct lineage (Fig. 2G. p<0.00003), suggesting that Notch led to an absolute reduction in endocrine cells with a corresponding increase in the number of duct cells. The duct cells identified in Ngn3-Cre;ROSANotch/+ mice only occasionally stained for cell proliferation markers (not shown), suggesting that they did not arise from expansion of an otherwise minor cell population. Although we previously showed that very small numbers of acinar cells in adult mice arose from Ngn3+ cells (Schonhoff et al., 2004), we were unable to identify acinar cells arising from Notch activation in Ngn3+ cells, consistent with Notch inhibition of acinar differentiation (Kopinke et al., 2012).
To determine whether cells expressing NeuroD1, a downstream target of Ngn3, retain the capacity to respond to Notch, we generated transgenic mice that express Cre recombinase under control of the NeuroD1 gene in order to activate Notch in NeuroD1+ cells. Cre was inserted at the translational initiation site of a NeuroD1 bacterial artificial chromosome (BAC) containing 118 KB of 5′ and 82 KB of 3′ flanking sequence (Cotta-De-Almeida et al., 2003; Datsenko and Wanner, 2000; Yu et al., 2000) (Fig. 3A).
We initially crossed NeuroD1-Cre mice to ROSA26-LacZ (R26R) indicator mice to identify all cells that arose from NeuroD1+ cells by recombination based cell lineage analysis (Soriano, 1999; Zambrowicz et al., 1997). Cre expression appeared to be identical for each of two independent pedigrees with expression of β-galactosidase in all NeuroD1+ cells in the pancreas, indicating that Cre expression was restricted to the appropriate cells, recapitulating the expression of the endogenous gene (Fig. 3B). Staining for β-galactosidase activity was restricted to islets with no staining seen in either duct or acinar cells (Fig. 3C, left panel). Whereas Ngn3+ cells give rise to small numbers of non-endocrine cells (Schonhoff et al., 2004), NeuroD1-expressing cells were restricted to an endocrine cell fate.
Further analysis of NeuroD1-Cre;R26R mice revealed that all insulin-expressing β cells stained for β-galactosidase, indicating that they arose from NeuroD1 expressing cells (Fig. 3D). To determine whether non-β cells arose from cells that expressed NeuroD1, we stained sections for β-galactosidase either with glucagon, somatostatin, or PP (Fig. 3E). In order to resolve the non-β cells from surrounding β cells in slightly offset focal planes, we combined confocal microscopy with deconvolution analysis. Approximately 15% of non-β endocrine cells stained for β-galactosidase, indicating that most (~95%) of NeuroD1+ cells are committed to a β-cell fate as we suggested previously (Itkin-Ansari et al., 2005).
We examined NeuroD1-Cre;ROSAEYFP mice at e16.5 to determine whether NeuroD+ cells in the fetal pancreas were multipotential. EYFP expression was restricted to chromogranin A expressing endocrine cells at e16.5 (Fig. 3C, right panel). Likewise, we were unable to identify β-galactosidase activity in either duct or acinar cells of NeuroD1-Cre;R26R mice (Fig. 3C, middle panel), indicating that NeuroD1+ cells were endocrine restricted in both the fetal and adult pancreas, unlike Ngn3+ cells that gave rise to increased number of nonendocrine cells prior to birth (Cras-Meneur et al., 2009; Magenheim et al., 2011; Schonhoff et al., 2004). These observations suggest that expression of NeuroD1 represents a distinct, later stage of endocrine differentiation where cells are committed to an endocrine cell fate.
We next examined the effect of Notch activation in NeuroD1+ cells in the islets to determine whether cells at this later stage of differentiation still respond to Notch signaling. In contrast to Ngn3-Cre;ROSANotch/+ mice, NeuroD1-Cre;ROSANotch/+ mice were healthy, and indistinguishable from their normal littermates. Most islets of NeuroD1-Cre;ROSANotch/+ mice appeared histologically indistinguishable from islets in normal mice in contrast to the absence of islet morphogenesis when Notch was activated at an earlier stage of endocrine differentiation in Ngn3+ cells
Unlike the uniform staining for insulin throughout the central islet core of normal mice (Fig. 3D), some cells in the center of islets in NeuroD1-Cre;ROSANotch/+ mice did not stain for insulin (Fig. 4A). Instead we observed occasional cells that expressed glucagon, somatostatin (Fig. 4B, C), and PP (not shown) in the islet core rather than exclusively at the islet periphery as seen in normal islets (Fig. 3E). This change in distribution of hormone expression was present in most islets of mice ranging from 2 days to over one year of age. Most insulin, glucagon, somatostatin, and PP cells in the islet core expressed EGFP (Fig. 4A–C), suggesting that Notch signaling does not prevent endocrine differentiation of NeuroD1+ cells although it may influence the localization of different cell types in the islet core. EGFP expressing cells in most islets continued to express NeuroD (Fig. 4D–F), suggesting that NeuroD may not be a direct target of Notch signaling in contrast to Ngn3,.
Overall there was little change in the total number of non-insulin expressing endocrine cells. The Ki67 labeling index for glucagon-expressing cells was not significantly increased in NeuroD1-Cre;ROSANotch/+ mice, suggesting that the effects of Notch activation did not result from expansion of the α cell lineage. Glucagon expressing cells in the islet center of NeuroD1-Cre;ROSANotch/+ mice did not express MafA (not shown), a marker of mature β cells (Nishimura et al., 2006). None of the glucagon-, somatostatin-, or PP expressing cells in the islet center coexpressed insulin, further suggesting that they did not retain properties of mature β cells (Suppl. Fig. 2). NeuroD1-Cre;ROSANotch/+ mice were normoglycemic with normal glucose tolerance (not shown), suggesting that Notch activation had little impact on β cell function and did not impair glucose homeostasis.
The effects of Notch activation on endocrine differentiation in NeuroD1+ cells appeared to be relatively mild compared to the absence of islet morphogenesis seen with Notch activation in early endocrine precursor cells expressing Ngn3 (Fig. 1B). Whereas 90% of islets in the pancreas of NeuroD1-Cre;ROSANotch/+ appeared normal, we observed a group of EGFP-expressing cells in ~10% islets that did not express Chromogranin A (Fig. 4G, inset, arrows), suggesting that Notch activation may alter the cell fate of NeuroD1+ cells normally destined to become endocrine cells.
To determine if NeuroD1+ cells switched to alternate nonendocrine cell fates in response to Notch, we examined EGFP-expressing cells for expression of markers for duct and acinar cells. Some of the EGFP+ cells appeared to be individual duct cells or in small intraislet ductules (Fig. 4G, H, L). With Notch activation, approximately 7± 3% of EGFP+ cells expressed DBA, whereas we never identified DBA expression in cells arising from NeuroD in NeuroD1-Cre;ROSA26 mice (p<0.02, type III, one-tailed Student t-test). Unlike EGFP-expressing endocrine cells (Fig. 4D–F) these EGFP labeled duct cells no longer expressed NeuroD1 (Fig. 4H), suggesting that Notch either directly or indirectly suppressed NeuroD1 expression in cells that switched to a duct cell fate. We were unable to detect Ngn3 in the adult pancreas of NeuroD1-Cre;ROSANotch/+ or normal mice (Suppl. Fig. 3), consistent with the difficulties encountered by others in detecting Ngn3 expression with available antibodies in the adult pancreas (Wang et al., 2009). In addition, we were unable to identify amylase expression in EGFP+ cells, suggesting that they did not switch to the acinar lineage (Suppl. Fig. 4). We observed EGFP expressing cells in approximately 10% of ducts of NeuroD1-Cre;ROSANotch/+ mice at all ages from P2 to >1 year. Some ducts were entirely comprised of EGFP+ cells (Fig. 4G) whereas others contained one to several EGFP+ cells (Fig. 4H, left panel). Approximately 1–2% of islets were associated with large dilated cystic structures lined with cuboidal EGFP+ cells (Fig. 4I) similar to structures reported previously with Notch activation at very early stages of pancreagenesis in Pdx1+ cells (Hald et al., 2003; Kopinke et al., 2011; Murtaugh et al., 2003). We also identified large EGFP-expressing ducts associated with islets as well as small intraislet ductules within a large adjacent islet (Fig. 4J, K).
We further characterized ducts arising from NeuroD1+ endocrine precursor cells in NeuroD1-Cre;ROSANotch/+ mice for expression of the homeodomain protein, Pdx1, which is expressed in immature duct cells. Pdx1 expression in ducts decreases during gestation and is not seen in mature ducts (Gu et al., 2002; Wescott et al., 2009). In 2-day-old animals, Pdx1 expression occurred mostly in developing islets (Suppl. Fig. 5). The absence of Pdx1 in most duct cells following Notch activation suggests that these cells are relatively mature, although we identified rare Pdx1+ cells that stained for both EGFP and the duct marker, WFA (Suppl. Fig. 5, arrow).
Our results suggest that a fraction of maturing endocrine cells that express NeuroD1 retain the capacity to respond to Notch signaling with resultant minor changes in the intraislet distribution of endocrine cells. Less frequently, NeuroD1+ cells switch to an alternate duct cell fate but never become acinar cells (Suppl. Fig. 4). To ensure we had identified all potential cell types that arose from activation of Notch in NeuroD+ cells, we examined EGFP+ cells in sections from NeuroD1-Cre;ROSANotch/+ mice for both WFA, and chromogranin A expression (Fig. 5A). As anticipated, we identified EGFP+ endocrine cells that stained for chromogranin A, (Fig. 5B–D, large arrows), and EGFP+ cells staining for the ductal marker, WFA (Fig. 5B–D, small arrows). In addition we identified a subset of EGFP+ cells that did not express either WFA or chromogranin A (Fig. 5B–D, arrowheads).
We examined pancreata of NeuroD1-Cre;ROSANotch/+ mice by electron microscopy to further characterize other cell types present. We identified a group of four cells in what appeared to be an intraislet ductule. These cells had relatively featureless cytoplasm and surrounded a lumen lined by microvilli and occasional cilia, both characteristics of duct cells (Fig. 6A [a–c]). The cells in this intraislet ductule also display desmosomes (Fig. 6A–b), indicating their epithelial origin, and are clearly distinguished from adjacent granule filled endocrine cells (Fig. 6A–a, right side). Unexpectedly, the lumen forming cells have rare cytoplasmic electron dense granules with outer membranes that are characteristic of endocrine secretory granules (Fig. 6A [b, d] arrowheads).
The cells described above that formed an intraislet ductule represent a potentially novel cell type with features of both ductal and endocrine differentiation. We identified two individual, relatively featureless islet cells that appeared very similar to the intraislet ductules just described and were surrounded by cells filled with cytoplasmic electron dense secretory granules characteristic of β cells. These cells had relatively clear cytoplasm much like the intraislet duct cells and formed tight junctions with neighboring cells, indicating that they were epithelial in origin (Fig. 6B–a). Other features characteristic of duct cells included two centrioles in one cell and a long microvillus in the other (Fig. 6B–b). In the same two cells, we also identified rare cytoplasmic electron dense granules with either coarse or punctate cores, characteristic of endocrine secretory granules (Fig. 6B [b–e]). Thus the clear cells also show features of both duct (microvilli) and endocrine (secretory granules) differentiation in a single cell and are ultrastructurally similar to the ductule cells shown earlier (Fig. 6A). These appear to be relatively immature cells with relatively limited structural features of duct and endocrine cells. As a result these cells may not be sufficiently differentiated to express markers of lineage specific differentiation for ducts or endocrine cells using standard staining techniques. Immunogold labeling for EGFP showed nuclear gold particles over these clear cells, indicating that they were expressing Notch-IC (Fig. 6B–f).
Ngn3 and NeuroD1 are also involved in the differentiation of intestinal endocrine cells, suggesting that regulation of enteroendocrine cell fate may share common mechanisms with the pancreas. We examined the fate of enteroendocrine precursors cells following activation of Notch in either Ngn3-Cre;ROSANotch/+ or NeuroD1-Cre;ROSANotch/+ mice. We identified many EGFP+ cells in the neonatal small intestine of Ngn3-Cre;ROSANotch/+ mice (Fig. 7B). However, EGFP+ cells only occasionally expressed chromogranin A compared to Ngn3-Cre;ROSAEYFP mice, suggesting that as in the pancreas, Notch limited but did not prevent endocrine differentiation (Fig. 7A, B). EGFP+ cells appeared as both individual cells as well as clusters of several columnar epithelial cells (Fig. 7B). The neonatal lethality of Ngn3-Cre;ROSANotch/+ mice limited our studies to two-day-old mice, prior to the maturation of the enteroendocrine cell population in normal mice. To examine the effects of Notch activation in early endocrine precursor cells of adult mice, we examined the cell fate of intestinal Ngn3+ cells using the Ngn3-CreERT2 mice to conditionally activate Notch. As expected, we observed mosaic recombination in Tamoxifen treated mice, with a fraction of enteroendocrine cells expressing EYFP when Ngn3-CreERT2 mice were crossed to ROSAEYFP indicator mice (suppl. Fig. 6).
Adult Ngn3-CreERT2; ROSANotch/+ mice treated with Tamoxifen appeared healthy. With the exception of a single CGA+EGFP+ cell (Fig. 7C, inset), we were unable to identify chromogranin A expression in several hundred other EGFP+ cells. We observed EGFP expressing cells in adult Ngn3-CreERT2;ROSANotch/+ mice as individual cells or as groups of columnar cells. The columnar cell clusters did not stain for Ki67, indicating that they were not an actively expanding population (Fig. 7D).
Further examination of EGFP+ cells that did not express chromogranin A in the intestine of Ngn3-Cre;ROSANotch/+ mice showed intense apical border staining for lactase (LPH), a brush border enzyme expressed at high levels in neonatal enterocytes (Fig. 8A), suggesting that these endocrine precursors adopted an enterocyte cell fate following Notch activation. In adult intestine of tamoxifen treated Ngn3creERT2; ROSANotch/+ mice, we further identified that those groups of EGFP+ columnar cells showed apical expression of alkaline phosphatase, suggesting they switched to the enterocyte lineage (Fig. 8B), consistent with our findings in neonatal intestine.
We frequently identified EGFP+ cells that stained for the goblet cell marker, Muc2 (Fig. 8C, D) Since up to 15% of duodenal goblet cells arise from Ngn3 expressing cells in normal mice (Schonhoff et al., 2004), it was difficult to determine if the EGFP expressing goblet cells were normal descendants of Ngn3+ cells or endocrine precursor cells that adopted an alternate goblet cell fate in response to Notch. We examined the frequency of goblet cells arising from Ngn3+ cells with and without Notch activation in the colon, where Ngn3 expressing cells normally contribute to 1.6% of the goblet cell population (Schonhoff et al., 2004). In Tamoxifen treated Ngn3-CreERT2;ROSAEYFP mice, we identified one Muc2+ EYFP+ cell out of 54 EYFP expressing cells, a percentage consistent with our previous lineage tracing studies with Ngn3-Cre mice. In contrast, 58% of EGFP-expressing cells coexpressed Muc2 in Ngn3-CreERT2;ROSANotch/+ mice after Notch activation. The substantially increased frequency of goblet cells arising from Ngn3 expressing cells suggests that they switched to an alternate goblet cell fate as a result of Notch activation rather than from the inherent low frequency that normally arise from Ngn3+ cells. These observations also suggest that the Ngn3+ cells were already committed to secretory lineage differentiation and that at this stage, Notch favors goblet cell differentiation in contrast to the inhibition of this lineage with Notch activation in multipotential cells (Fre et al., 2005; Stanger et al., 2005).
We previously showed that a significant fraction of Paneth cells arose from Ngn3+ precursor cells (Schonhoff et al., 2004; Wang et al., 2007). Since Paneth cells do not appear until several weeks after birth, we activated Notch in Ngn3+ cells of adult mice using the inducible Cre line to circumvent neonatal lethality seen with the noninducible Ngn3-Cre transgene. We crossed ROSAEYFP mice with Ngn3-CreERT2 mice to determine the relative fraction of Ngn3+ cells that become lysozyme staining Paneth cells using the conditional transgene. Nearly 40% of EYFP+ cells in the ileum and 9% in the duodenum were lysozyme stained Paneth cells (Fig. 8E, G). In contrast, no EGFP+ cells in Ngn3-CreERT2;ROSANotch/+ mice expressed lysozyme, indicating that Notch activation strongly selected against Paneth cell differentiation, a major alternate cell fate of Ngn3+ cells (Fig. 8F, G). Thus at this stage of differentiation, Notch selects against endocrine and Paneth cell fates in favor of enterocyte and goblet cell differentiation.
In order to determine the effects of Notch activation in maturing enteroendocrine cells expressing NeuroD1, we first analyzed the normal cell fate of NeuroD1+ cells in the intestine by crossing NeuroD1-Cre mice to either ROSAEYFP or ROSA26 (lacZ) indicator mice. Nearly all cells arising from NeuroD1-expressing cells appeared to be endocrine restricted as most EYFP+ cells stained for chromogranin A (Fig. 9A, arrow, inset). In contrast to Ngn3-Cre mice, NeuroD1-Cre mice did not label either goblet or Paneth cells with β-galactosidase expression when crossed to ROSA26 mice (Fig. 9B, C). Most chromogranin A expressing enteroendocrine cells in NeuroD1-Cre;ROSAEYFP mice expressed EYFP, indicating that they arose from cells that expressed or continue to express NeuroD. The enteroendocrine cell populations of NeuroD1-Cre;ROSANotch/+ mice appeared identical to wild type mice, including expression of secretin and cholecystokinin, two NeuroD1 dependent hormone genes, suggesting that Notch did not directly affect NeuroD1 (Fig. 9D–F). Our observations suggest that in the intestine, once maturing cells express NeuroD1, they are committed to the enteroendocrine lineage and are no longer responsive to Notch.
Specification and differentiation of endocrine cells of the pancreas and the intestine involves the stepwise restriction from multipotent progenitor cells. This stepwise restriction is controlled by the sequential expression of transcription factors that direct differentiation of each emerging cell type. In many systems, the effects of Notch depend on timing of the Notch signal, with Notch acting reiteratively in a stepwise manner at different developmental stages to regulate multiple cell fate choices and outcomes during the course of differentiation. In the nervous system, Notch regulates the cell fate of neural stem cells between neural or glial precursors. As these precursors further differentiate, Notch inhibits differentiation of neuronal precursors to neurons as well as regulating differentiation of glial precursors to astrocytes versus oligodendrocyctes (Grandbarbe et al., 2003). In the immune system, Notch signaling favors differentiation of early lymphoid progenitors into T-cell progenitors, inhibiting generation of B-cell precursors. Notch appears to regulate T cell precursor subtype segregation at later stages of T cell development (Anderson et al., 2001).
Prior Notch gain of function and loss of function studies have concluded that Notch inhibits endocrine differentiation in both the pancreas and the intestine. In most cases, conditional Notch activation was initiated in multipotential cells expressing either Pdx1 in the pancreas or Villin in the intestine, which give rise to multiple epithelial lineages in the pancreas and intestine respectively (Fre et al., 2005; Hald et al., 2003; Murtaugh et al., 2003; Stanger et al., 2005). Thus the observed inhibition of endocrine cell fate may have resulted from effects on multipotential progenitor cells rather than a direct effect on endocrine precursor cells.
Likewise, Notch loss of function studies showed excessive endocrine differentiation with widespread disruption of Notch pathway components in the pancreas (Apelqvist et al., 1999; Jensen et al., 2000). The effects of loss of Notch function on enteroendocrine cells are less clear. Hes1 null mice showed increased enteroendocrine cell differentiation (Jensen et al., 2000). In one study, treatment with a gamma secretase inhibitor had no apparent affect on endocrine cell number (van Es et al., 2005), while another showed increased endocrine differentiation (VanDussen et al., 2012). Most prior studies used generalized Notch loss of function or examined the effects of altered Notch activity in multipotential cells. As a result, it is unclear whether Notch directly regulates endocrine cell fate as opposed to an indirect effect of Notch on an earlier multipotential progenitor. More recently, conditional deletion of Presenilins in Ngn3-expressing cells resulted in their switching to an acinar cell fate (Cras-Meneur et al., 2009). However, deletion of Presenilins did not alter the eventual endocrine cell fate of more mature Pax6+ cells, indicating that the effects of Notch loss of function is dependent on the developmental stage, as we have shown with Notch activation.
The present work indicates that Notch activation in early endocrine precursor cells limits but does not preclude endocrine differentiation as was suggested by earlier studies (Fre et al., 2005; Hald et al., 2003; Murtaugh et al., 2003; Stanger et al., 2005). The identification of significant numbers of PYY- and glucagon-expressing endocrine cells in the pancreas when Notch was activated in Ngn3+ cells with the absence of cells expressing insulin, somatostatin, PP, and ghrelin may indicate endocrine differentiation was limited to cells expressing two hormones that appear at the earliest stages of pancreatic endocrine differentiation at e9.5 and do not contribute to β cells that appear later (Schonhoff et al., 2005; Upchurch et al., 1994). As a result, Notch activation may prevent the expansion of endocrine cells that gives rise to islets as well as β and cells later in gestation. We believe that the differences between our study and earlier work suggesting that Notch completely blocked endocrine differentiation in Ngn3+ cells (Murtaugh et al., 2003) can be explained by our use of an Ngn3-Cre transgene where Cre expression was driven by a much larger fragment of the Ngn3 gene with resultant expression in all Ngn3+ cells. In addition, the transgene used previously to activate Notch resulted in unexplained lethality at e13.5, at an early stage of endocrine differentiation that precluded analysis of the cell fate of Ngn3+ cells (Murtaugh et al., 2003).
In the present study, Notch activation by a large Ngn3-Cre transgene enabled mice to survive after birth, allowing us to study the effects of Notch on the cell fate of Ngn3+ cells. The observed switching of Ngn3+ cells to the duct lineage, is consistent with our previous finding (Schonhoff et al., 2004), as well as recent findings by others (Magenheim et al., 2011), that Ngn3+ cells give rise to duct cells in the developing pancreas. Several studies have suggested that the level of Ngn3 expression in endocrine precursor cells plays an important role in their cell fate with higher levels of Ngn3 protein favoring endocrine differentiation and lower levels favoring the duct lineage (Magenheim et al., 2011; Wang et al., 2010). In the absence of Ngn3 protein, increased numbers of cells expressing the Ngn3 gene in the fetal pancreas became duct cells, resulting in thickened ducts with reduced branching (Magenheim et al., 2011). Reduced expression of Ngn3 protein when Notch is active may have a similar role for Ngn3+ cells switching from an endocrine to a duct cell fate. Although we previously showed that rare acinar cells arise from Ngn3-expressing cells by lineage tracing that in normal adult mice (Schonhoff et al., 2004), we never observed switching to the acinar lineage with Notch activation.
The effects of Notch on maturing pancreatic endocrine cells have not been extensively studied. Surprisingly, we observed that a subpopulation of NeuroD1+ cells in the pancreas retain sufficient plasticity to respond to Notch to become duct cells at a relatively late stage of endocrine differentiation. Several findings reported here suggest that NeuroD1+ cells in the pancreas are not directly shunted to the duct lineage in response to Notch as was proposed for Pax4-expressing cells (Greenwood et al., 2007). Lineage tracing established that NeuroD1 expressing cells are endocrine-restricted, never giving rise to duct cells as available alternate cell fate. In addition, we found that duct cells arising from NeuroD1+ cells in response to Notch no longer expressed NeuroD1. The evidence for direct shunting of Pax4+ cells to the duct cells was not conclusive since lineage tracing of Pax4-expressing cells failed to identify any nonendocrine cell fates as would be predicted if cells were shunted as suggested.
Multipotential precursor cells have been proposed to reside in the tips of embryonic pancreatic epithelium that differentiate into acinar cells or migrate away from the tips to become either duct or endocrine cells (Zhou et al., 2007). Cell lineage tracing of Hes1 expressing cells suggest that Notch strongly selects against acinar differentiation in both embryonic tips as well as in the adult pancreas (Kopinke et al., 2012; Kopinke et al., 2011). A number of studies have since proposed that the trunks of the fetal pancreatic ducts contain a bipotential progenitor cell that may give rise to the duct or endocrine cell lineage (Kopinke et al., 2012; Kopinke et al., 2011; Magenheim et al., 2011). The most direct evidence for the existence of a bipotential progenitor, has been described recently, indicating that Ngn3+ cells give rise to endocrine or duct cells and have lost the competence to become acinar cells at an early stage of pancreagenesis (Beucher et al., 2012). The primitive cell we identified with limited ultrastructural features of both the duct and endocrine cells with Notch activation in NeuroD1+ cells (Fig. 6), could be related to such a bipotential cell. We speculate that this novel cell may arise from NeuroD1+ cells by reverting to an immature bipotential cell (endocrine, duct restricted) since NeuroD1+ cells never directly give rise to duct cells unlike Ngn3+ cells.
As was the case in the pancreas, the effects of Notch activation in early enteroendocrine precursors limits but does not completely block endocrine differentiation in contrast to the complete absence of enteroendocrine cells with activation of Notch in villin-expressing cells (Fre et al., 2005). Villin is expressed in all intestinal lineages as well as their multipotential precursors. The absence of enteroendocrine cells is likely the result of Notch activation prior to the initiation of endocrine differentiation rather than a direct effect on enteroendocrine precursors.
A recent study showed that expression of the Notch ligand Dll1 in intestinal secretory cell lineages was coordinated with their exit from the cell cycle. Conditional deletion of Dll1 in multipotential cells of the intestine resulted in increased numbers of goblet and enteroendocrine cells, suggesting that lateral inhibition of adjacent cells limited secretory lineage differentiation of neighboring cells, which continue to divide with expansion of the absorptive lineage (Stamataki et al., 2011). In the present work, Notch was specifically activated in Ngn3+ enteroendocrine precursors, making direct comparison to the study described above difficult. While we observed that enteroendocrine precursor cells switch to an enterocytic cell fate in response to Notch, they do not appear to reenter the cell cycle, remaining post-mitotic.
Although we previously found that about 15% of goblet cells in the proximal small intestine arose from Ngn3+ cells (Schonhoff et al., 2004), our findings that Notch induced some enteroendocrine precursor cells to become goblet cells was unexpected. A number of earlier studies showed that increased Notch activity increased Hes1 expression resulting in a profound reduction of goblet cell numbers (Fre et al., 2005; Stanger et al., 2005) whereas inhibition of Notch signaling resulted in a significant expansion of this lineage (Pellegrinet et al., 2011; Stamataki et al., 2011; van Es et al., 2005; VanDussen et al., 2012). Another study reported a modest (40%) increase in goblet cells with conditional activation in Notch in multipotential cells and suggested that this occurred only in postmitotic cells (Zecchini et al., 2005). The unexpected switching to the goblet lineage with over a 30-fold increase in colonic goblet cells may result from Notch activation at a later stage of differentiation than prior studies which targeted multiple epithelial populations in the intestine. Since most Ngn3+ cells are post mitotic, our results may reflect the effects of Notch on nondividing cells (Bjerknes and Cheng, 2006; Stamataki et al., 2011).
We previously showed by lineage tracing that up to 45% of Paneth cells in the proximal small intestine arise from Ngn3-expressing cells (Schonhoff et al., 2004). Surprisingly, Notch appears to completely inhibit differentiation of those Paneth cells that arise from Ngn3+ cells. The role of Notch in Paneth cell differentiation has received relatively little attention, which may be in part due to the failure of mice to survive to the age of 4 weeks after birth when Paneth cells first appear when Notch pathway is perturbed in multipotential cells. A potential role for Notch in Paneth cell differentiation was first suggested from increased numbers of “intermediate cells” seen in mice with deletion of the tumor suppressor gene, Lkb1, from the intestine (Shorning et al., 2009). More recently, disruption of the Notch pathway by conditional deletion of RBPJκ in the intestine resulted in increased numbers of Paneth cells, consistent with our observations that Notch inhibits Paneth cell differentiation (Kim and Shivdasani, 2011).
In contrast to the pancreas, Notch activation in maturing enteroendocrine cells expressing NeuroD1 had no effect on intestinal differentiation, further indicating that timing and context are the major determinants of the response to Notch. Despite the much higher turnover rate of the intestinal epithelium compared to the pancreas, differentiating enteroendocrine cells do not retain the plasticity to respond to Notch as we observed for pancreatic endocrine cells.
In conclusion, these studies reveal unanticipated and previously uncharacterized differences in the response of early and late endocrine precursor cells in the pancreas and the intestine to active Notch signaling. Notch activation in early endocrine precursor cells in the pancreas and intestine limits endocrine differentiation with many cells switching to alternate non-endocrine cell fates including duct cells in the pancreas as well as to enterocytes and goblet cells in the intestine. Surprisingly, some late pancreatic precursor cells retain to capacity to change their cell fate in response to Notch that may involve a novel mechanism of reversion to an immature bipotential cell, whereas maturing enteroendocrine cells have lost the competence to respond to Notch.
This work was supported in part by NIH grants DK43673, DK67166, and DK90000 to A.B.L., a grant from the Caring for Carcinoid Foundation, the Tufts-GRASP Digestive Disease Center, P30DK34928 and T32-DK007542, core resources of the University of Massachusetts Diabetes Endocrinology Research Center grant DK032520, the University of Massachusetts Digital Light Microscopy Core, and the Transgenic Core Facility at Tufts Medical Center. The authors thank Douglas Melton for generously providing ROSANotch/+ mice. The authors also thank Dr. T. D’Adda and Mrs. E. Corradini for skillful technical assistance with electron microscopy. The cytokeratin 19 (TROMA III) antibody developed by Rolf Kemler was obtained from Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the university of Iowa, Department of Biology, Iowa City, IA 52242.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.