|Home | About | Journals | Submit | Contact Us | Français|
Endocrine cells are scattered throughout the gastrointestinal mucosa from the stomach to the colon and constitute one of the largest endocrine systems in the body. Enteroendocrine cells labeled by chromogranin A comprise about 1% of all epithelial cells in the gastrointestinal tract and are surrounded by mucus, chief, and parietal cells in the stomach, and enterocyte, goblet and Paneth cells in the intestine (Figure 1). The enteroendocrine system consists of at least 15 different cell types that can be classified based on their main hormonal products and on the ultrastructure of their secretory granules . A given enteroendocrine cell secretes one or more hormone or hormone-like substance, which is released directly into the lamina propria and diffuses into the capillaries. These hormones include gastrin, histamine, serotonin, cholecystokinin (CCK), somatostatin and glucagon-like peptides (GLP1 and 2) . Although enteroendocrine cells are very scarce, they are essential regulators of digestion, gut motility, appetite, and metabolism.
The development of enteroendocrine cells is a fascinating biological problem: how is the relative proportion of the individual subtypes maintained? How are so many different cell types specified from a common precursor? And how is the endocrine compartment maintained in the gastrointestinal epithelium with its rapid and life-long turnover? However, the study of the mechanisms of enteroendocrine cell differentiation is not only an “academic exercise” but of relevance to human health and disease on multiple levels. First, deficiencies in enteroendocrine cell specification or survival contribute to human diseases, as exemplified by congenital malabsorptive diarrhea, discussed in detail below, which is due solely to deficient enteroendocrine cells. Second, several enteroendocrine cells play a role the control of glucose homeostasis and thus diabetes, and there relatedness to pancreatic beta-cells makes transdifferentiation of enteroendocrine cells an interesting possibility. Therefore, lessons learned from the study of normal gut endocrine cell development promise to be instructive to
The gastrointestinal tract is lined with a monolayer of cells that undergo continuous and rapid renewal, in contrast to the epithelia of the other endoderm-derived organs such as liver and lung that differentiate early in life and turn over slowly in adulthood [2–6]. Enteroendocrine cells found in the gastrointestinal tract express several hormones, and even transcriptional regulators, originally associated with the nervous system. Therefore, the ontology of enteroendocrine cells has long been controversial, with some scientists suggesting an endodermal, and others a neural crest origin. However, lineage tracing experiments have convincingly demonstrated that all epithelial cell types of the intestinal mucosa, including enteroendocrine cells, are differentiated from pluripotent stem cells in the crypts (or the neck region of gastric glands), and are therefore endoderm-derived [7–14]. Thus, although neurons and enteroendocrine cells express a common set of genes, they are of divergent embryological origin.
Each intestinal crypt contains 4–6 stem cells; however, the exact identity of these cells was revealed only recently. Wnt signaling had been shown to play an important role in the proliferative activity of the normal intestinal crypt [15–17]. Based on this knowledge, Clevers and colleagues identified a Wnt target gene, Lgr5/GPR49, as restricted in expression to the stem cell compartment, which is located at the very base of the crypt in the small intestine (Figure 3) . Most importantly, genetic lineage tracing studies using a Lgr5-CreERT2 transgene and Rosa26R reporter mice demonstrated conclusively that all epithelial lineages, including enteroendocrine cells, are derived from Lgr5-expressing stem cells in the intestine .
Notch is a transmembrane receptor that mediates local cell-cell communication and coordinates a signaling cascade schematized in Figure 2 . Unlike other receptors, Notch functions both at the cell surface to receive signals and in the nucleus to regulate gene expression . The key players of the Notch pathway in mammals are the ligands (Delta and Jagged), the receptor (Notch 1 through 4) and a transcription factor (RBP-Jκ) that translates the signaling process into target gene activation. Ligand binding activates proteolytic cleavage and post-translational modification of Notch, releasing the Notch intracellular domain (Notch IC) from the cell surface (; Figure 2). The Notch IC then translocates into the nucleus and associates with RPB-Jκ, a DNA binding protein , to form a complex that binds to and activates the promoters of the hairy/enhancer of split (HES) gene family . Hes1 or its relatives then repress proendocrine bHLH transcription factors such as Neurogenin 3 (Ngn3) by binding to their promoters or enhancers [21–23].
One of the major functions of Notch signaling is to mediate lateral inhibition between adjacent cells, preventing neighboring cells from adopting the same fate [24,25]. Involvement of the Notch signaling pathway in the differentiation of enteroendocrine cells was first suggested by the observation that there are never two enteroendocrine cells adjacent to each other in the gastrointestinal epithelium. This was confirmed by studies in which loss of Notch function resulted in excess enteroendocrine cells [21,26]. Notch signaling controls the bHLH transcription factor-driven differentiation program to select individual endocrine cells from expanding precursor cells [26–28]. In addition to its role in the initial decisions of epithelial lineage determination, Notch signaling is also important during later stages of enteroendocrine cell differentiation to fine-tune the number of the various enteroendocrine cell types [21,29–32].
Three proendocrine bHLH transcription factors, all targets of Notch signaling, have been shown to contribute to enteroendocrine cell differentiation through loss-of-functions studies in mice [29–33]. Math1, Neurogenin 3 (Ngn3) and NeuroD are all related to the Drosophila atonal gene, which is critical in neural differentiation in the fly [34–36]. This family of transcription factors functions in sequential order, where one factor activates another to control the initial specification of the enteroendocrine cells as well as their final differentiation.
Math1 is activated first in the immature crypts and villi of the intestinal epithelium during fetal development, although it is not detected in the stomach . Math1 is required for the initial specification of all three secretory lineages of the intestine: goblet, Paneth and enteroendocrine cells . Furthermore, it has been shown through lineage tracing studies that all secretory lineages in the intestine are derived from Math1-expressing precursor cells . Later expression of lineage-specific transcription factors such as Sox9 for Paneth cells, Klf4 for goblet cells and Ngn3/NeuroD for enteroendocrine cells is required for the proper differentiation of these secretory cells (Figure 3) [29–31,37,38]. Recently, analysis using an intestine-specific deletion of Math1 in adult mice has revealed the importance of Math1 in the balance between enterocytes and enteroendocrine cells, as intestinal precursors preferentially differentiate into enterocytes at the expense of secretory lineages in the Math1-deficient intestine (Figure 3) .
Ngn3 is first activated in the intestine as early as embryonic day 12.5 (E12.5), similar to what is observed in the developing pancreas . This expression persists into adulthood where expression is found in the intestine and glandular stomach . Ngn3 acts downstream of the transcription factor Math1, because Ngn3-deficient animals still express Math1 (Figure 3) [29,30].
The contribution of Ngn3 to enteroendocrine cell differentiation varies between intestine and stomach. In both mice and humans, all intestinal enteroendocrine cells are Ngn3-dependent, with no enteroendocrine cells left in the intestinal epithelium of mice homozygous for a null mutation in Ngn3, and few were found in humans homozygous for a NGN3 point mutation [29,40,41]. In the humans, NGN3 mutation has been reported as the cause of congenital malabsorptive diarrhea . Interestingly, the role of Ngn3 in the development of enteroendocrine cells in the stomach is more limited. Thus, while the number of glucagon-, somatostatin-, and gastrin-secreting cells is reduced in Ngn3-null mice, serotonin-, histamine-, and ghrelin-expressing cells are still present [29,30].
While loss-of-function studies suggest that normal enteroendocrine differentiation requires Ngn3 [29,30], it is less clear if Ngn3+ precursor cells contribute only to the enteroendocrine population. Several groups have developed transgenic mouse models to address this issue. In the first model, the Ngn3 promoter and 6.9 kilobases of upstream sequence were used to direct expression of the bacterial LacZ gene. In these mice, some, but not all enteroendocrine cells were marked by the transgene. A potential limitation of this model is the short Ngn3 promoter fragment employed for the transgene construction, leaving the possibility that both exclusion of essential Ngn3 enhancer elements and transgene position effects might have affected the outcome of this lineage tracing experiment.
To overcome these limitations, Andrew Leiter’s group derived mice in which the Cre-recombinase encoding cDNA was placed under the control of a Ngn3 BAC (bacterial artificial chromosome) containing 183 kilobases of the Ngn3 locus . Using this model, Leiter and colleagues demonstrated that all intestinal endocrine cells arise from Ngn3-Cre+ cells. This study, however, produced a new puzzle, as a few goblet and Paneth cells were also marked by the Ngn3-Cre transgene . These observations raise a crucial issue with genetic lineage tracing using the Cre/loxP system that is often overlooked. Because of the extraordinary specificity and activity of the site-specific DNA recombinase Cre, very low levels of expression are sufficient to effect excision of loxP-flanked targets. Thus, cells in which the Ngn3-Cre transgene is only minimally activated will be permanently marked by β-galactosidase from the recombined Rosa26R (or any other reporter) locus, even though there is yet no Ngn3 protein detectable in the same cell. In other words, the observed activation of β-galactosidase in a small percentage of goblet and Paneth cells (and the occasional entire crypt) may be due to stochastic low-level activation of the Ngn3-Cre transgene.
To better characterize the progenitors leading from the stem cells to the Ngn3+ enteroendocrine lineage, Bjerknes and Chen  have analyzed the rate of cell flux through different stages of enteroendocrine cell differentiation, finding that most if not all of the Ngn3+ cells are in the enteroendocrine lineage and Ngn3+ cells do not contribute significantly to non-endocrine lineages . The importance of Ngn3 in the enteroendocrine lineage was also demonstrated by gain-of-function experiments in which Ngn3 was over-expressed throughout the developing intestinal epithelium, resulting in increased numbers of enteroendocrine cells at the expense of goblet cells .
Genetic analysis of the third atonal-related bHLH gene, NeuroD, was the first to link Notch signaling to enteroendocrine cell differentiation . Unlike Ngn3, expression of NeuroD is restricted to a subset of enteroendocrine cells [45,46]. NeuroD has been shown to control terminal differentiation of secretin-producing cells by coordinating transcription of the secretin gene with cell cycle arrest [31,47,48]. In addition, NeuroD-deficient mice lack secretin- and cholecystokinin (CCK)-producing cells in the intestine, thus defining a subset of the enteroendocrine cell lineage . It has been suggested that NeuroD acts downstream of Ngn3 and is regulated by Ngn3, which was confirmed by the absence of NeuroD expression in the Ngn3-deficient mice (Figure 3) [29,49].
Hes1, which is transcriptionally controlled by Notch signaling, encodes a bHLH transcriptional repressor [22,50] and antagonizes the activity of the aforementioned bHLH transcriptional activators [51–54]. Hes1 is critical for maintaining endocrine precursor pools in the undifferentiated state. In Hes1-deficient mice several bHLH factors are up-regulated, resulting in precocious differentiation of enteroendocrine subtypes in the stomach and intestine . In these mice, excess numbers of glucagon-, CCK/gastrin-, somatostatin- and GIP- cells are found in the gastric epithelium during development . In addition, increased numbers of cells positive for serotonin, CCK, proglucagon, somatostatin, and GIP are found in the intestine . There more goblet cells and fewer enterocytes in Hes1-deficient mice, suggesting that the secretory lineage is augmented at the expense of enterocytes in the intestine (Figure 3) .
Transcription factors other than bHLH factors have also been shown to play critical roles during enteroendocrine cell differentiation. As mentioned above, concomitant with the premature differentiation of enteroendocrine cells in the Hes1-deficient animals , the Pax4, Pax6, Nkx2.2, Isl-1 genes are activated in the gut . Instead of controlling the global differentiation of enteroendocrine cells like the bHLH genes, these factors are likely to play important roles in fine-tuning cell specification decisions within the enteroendocrine population.
The zinc-finger containing transcription factor Gfi1 is expressed in scattered cells throughout the developing intestine epithelium. Once the intestine matures, Gfi1 is found mainly in the crypts but also in some villi in the small intestine . Double immunostaining analysis showed that Gfi1 expression is restricted to enteroendocrine lineages . In addition, Gfi1 is expressed within the crypt in Math1-dependent progenitors, and Gfi1 expression is affected in Math1-deficient embryos, suggesting that Gfi1 acts downstream of Math1 in the differentiation program (Figure 3B) .
Another zinc-finger protein, Insm1 (also termed IA-1), is found in scattered cells of the intestinal epithelium, many of which also co-express NeuroD1 and chromogranin A . In Insm1 null mice, substance P or neurotensin expressing endocrine cells are absent and serotonin, CCK or PYY expressing cells are present at a reduced number, while enteroendocrine precursors cells are generated normally . This genetic analysis reveals that Insm1 controls differentiation of specific enteroendocrine lineages in the intestine and acts downstream from Notch and Ngn3 after the initial specification of enteroendocrine precursors.
As mentioned briefly above, many homeobox genes including Isl-1, Pdx1, Nkx6.1 and Nkx2.2 have been shown to be involved in enteroendocrine cell differentiation. Isl-1 transactivates somatostatin expression in islet tumor cells in vitro  and its expression is restricted to G/D precursors (gastrin/somatostatin co-expressing cells) and a subpopulation of somatostatin cells in the stomach . This implies that Isl-1 could be involved in the asymmetric division of G/D precursors . Unfortunately, thus far examination of the requirement of Isl-1 in enteroendocrine differentiation has not been possible due to the early embryonic lethality of the Isl-1 null mice .
Pdx1 is expressed in the pancreatic buds at the onset of pancreatic organogenesis . Deletion of Pdx1 leads to an arrest of pancreatic differentiation and altered duodenal morphology [61–64]. Most relevant here is the fact that the Pdx1-deficient duodenum is also devoid of gastrin cells . Interestingly, Pdx1-deficient mice have an increased number of serotonin cells in the distal stomach . Pdx1 is clearly essential for normal gastrin cell development in the gastric epithelium; however, other factors must also contribute to this lineage since not all gastrin cells are Pdx1 positive and a few gastrin cells still remain in the Pdx1-deficient mice .
Two paired box (Pax) genes, Pax4 and Pax6, are expressed throughout the gastrointestinal tract during development. Mice deficient for either Pax4 or Pax6 fail to properly elaborate pancreatic endocrine cells [66,67]. Similar to their role in the pancreas, the two genes also control important stages of enteroendocrine cell differentiation in the gut.
Deletion of Pax4 affects enteroendocrine cell development only in the rostral gastrointestinal tract. In the stomach of Pax4 deficient mice, the number of somatostatin and serotonin-expressing cells is significantly reduced, while gastrin-expressing cells are not affected. In the duodenum, serotonin-, secretin-, CCK-, GIP-, and PYY- expressing cell differentiation is severely impaired . Interestingly, there are no significant differences in enteroendocrine cell numbers in the ileum and colon between the Pax4 deficient mice and the controls .
Pax6 expression is also detected in the developing gastrointestinal epithelium. In Pax6 deficient mice, there is a severely diminished elaboration of somatostatin- and gastrin-expressing cells in the distal stomach; however, serotonin-expressing cells are not affected. In addition, the number of GIP-expressing cells is reduced in the duodenum of Pax6 deficient mice .
Nkx6.1 expression is detected in the gastric mucosa, but not in the duodenum . The majority of cells expressing Nkx6.1 are serotonin-expressing cells, with a smaller population expressing gastrin . In Pdx1-deficient mice, Nkx6.1 expression is absent, indicating Pdx1 acts upstream of Nkx6.1 in the enteroendocrine cell program .
Nkx2.2 expression is first detected in the proximal forestomach and then in scattered cells throughout the small intestine, and expression persists through birth . Many intestinal hormone-producing enteroendocrine populations are reduced in Nkx2.2-deficient mice, including serotonin-, CCK-, GIP-, and gastrin-expressing cells . Interestingly, the number of ghrelin-expressing cells is significantly increased in the Nkx2.2 deficient mouse . Further analysis suggested that ghrelin-expressing cells are specified at the expense of other endocrine cell lineages in the intestine while the total number of enteroendocrine cells remains unchanged .
Nkx6.3 expression is detected exclusively within the epithelium of the antrum in developing, newborn and adult mice [72,73]. Phenotypic analysis using Nkx6.3 null mice showed a selective and severe reduction in gastrin-secreting G cells and an increase in the number of somatostatin-expressing D cells. Interestingly, Nkx6.3 null mice express normal levels of transcription factors (Pdx1, Pax6 and Ngn3) that are required for the development of gastrin-secreting G cells [29,30,65,68,73]. Conversely, normal expression of Nkx6.3 was detected in the Ngn3 null animals . These data suggest that Nkx6.3 functions independently of other transcription factors in G cell differentiation.
Many unanswered questions remain regarding how enteroendocrine cells are specified in the gastrointestinal tract, and if it might be possible to control their relative contribution with therapeutic benefit. Unlike endocrine cells in the pancreas, which cluster together to form islets, enteroendocrine cells are scattered as individual cells throughout the gastrointestinal mucosa. Very little is known about what directs the region-specific expression of hormones in the gut. Although the functions of several key transcription factors during enteroendocrine cell differentiation have been uncovered using genetic means, exactly how these factors direct and control the gene expression in the multitude of endocrine cell types remains unresolved. It is also not known whether and to what extent these transcription factors control hormone expression in the adult. However, this limitation will be overcome in the near future through the use of inducible gene ablation with the Cre/LoxP system, which permits gene deletion at all stages of ontogeny. Because of the continuous renewal of the gastrointestinal epithelium throughout adult life, it might be possible one day to manipulate the differentiation of these cells for the treatment of metabolic and gastrointestinal diseases.
The authors thank Dr. Joshua R. Friedman for critical reading of the manuscript and Johanna Murray for providing the immunostaining images. Related work in KHK’s laboratory was supported by NIH grants R01-DK053839 and R01-DK055342, and in CLM’s laboratory by NIH grant R01-DK078606.
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.