Vertebrate Intestinal Endoderm Development

Abstract

The endoderm gives rise to the lining of the esophagus, stomach and intestines, as well as associated organs. To generate a functional intestine, a series of highly orchestrated developmental processes must occur. In this review, we attempt to cover major events during intestinal development from gastrulation to birth, including endoderm formation, gut tube growth and patterning, intestinal morphogenesis, epithelial reorganization, villus emergence as well as proliferation and cytodifferentiation. Our discussion includes morphological and anatomical changes during intestinal development as well as molecular mechanisms regulating these processes.

Introduction

The intestine is a highly organized, complex organ that serves many important functions, including digestion and nutrient absorption, endocrine function, and immunity. The mature (endoderm-derived) intestinal epithelium is comprised of the stereotypic crypt-villus unit and contains absorptive (enterocytes) and secretory (goblet, enteroendocrine, Paneth, tuft) cell types, as well as resident intestinal stem cells and rapidly dividing progenitors. Adjacent to the intestinal epithelium are a complex assortment of (non-endoderm derived) cells comprising the lamina propria, submucosa, and muscular layers. In order to reach its mature form, the intestine must go through several complex developmental stages.

In this review, we focus on vertebrate intestine development with an emphasis on mammalian development. The authors recognize the importance of invertebrate model organisms and their contribution to the overall understanding of gut development, however, in an attempt to cover intestine development from gastrulation to the mature organ, invertebrate development is beyond the scope of this review. Excellent reviews on invertebrate gut development may be found elsewhere (Wessel and Wikramanayake, 1999; Lengyel and Iwaki, 2002; Maduro and Rothman, 2002; McGhee, 2007; Hou).

We have organized this review to cover intestine development along a developmental timeline starting with gastrulation and endoderm specification and ending with the mature organ shortly after birth. Along the way, we will cover the topics of gut tube formation, hindgut patterning, intestinal epithelial reorganization and villus emergence as well as proliferation and cytodifferentiation of the embryonic intestine.

Endoderm Specification during Gastrulation

This section is meant only to be a brief review of endoderm specification during gastrulation. Many excellent reviews exist which are dedicated to gastrulation and endoderm specification (For review, see: (Wells and Melton, 1999; Zernicka-Goetz, 2002; Tam et al., 2003; Technau and Scholz, 2003; Rossant, 2004; Tam et al., 2006; Spence and Wells, 2007; Tam and Loebel, 2007; Zorn and Wells, 2007; Rossant and Tam, 2009; Zorn and Wells, 2009).

The process of gastrulation gives rise to three primary germ layers including ectoderm, mesoderm and endoderm. Experimental evidence from fish, frogs and mice indicate that mesendoderm progenitor cells give rise to either endoderm or mesoderm as gastrulation proceeds (Lawson and Pedersen, 1987; Lawson et al., 1991; Rodaway et al., 1999; Kimelman and Griffin, 2000; Rodaway and Patient, 2001; Spence and Wells, 2007; Zorn and Wells, 2007; Zorn and Wells, 2009). First demonstrated in Xenopus, The TGF-β superfamily member Nodal is required for mesoderm and endoderm specification in all vertebrates (Green and Smith, 1990; Clements et al., 1999; Aoki et al., 2002; Ben-Haim et al., 2006; Hagos and Dougan, 2007). Exposure to Nodal signaling as the mesendoderm progenitor cells move through the streak determines if a progenitor cell becomes endoderm or mesoderm, with high levels of Nodal specifying endoderm and lower levels promoting mesoderm (Tremblay et al., 2000; Lowe et al., 2001; Vincent et al., 2003; Weng and Stemple, 2003; Tada et al., 2005; Shen, 2007). Activin, another TGF-β superfamily member, is able to mimic Nodal activity by binding to the same receptors (Smith et al., 1990; Thomsen et al., 1990; Gamer and Wright, 1995; Henry et al., 1996; Gray et al., 2003).This universal requirement for Activin/Nodal signaling across species has been utilized to direct differentiation of human and mouse embryonic stem cells (hESCs and mESCs, respectively) into endoderm (Kubo et al., 2004; D'Amour et al., 2005; Tada et al., 2005; Spence and Wells, 2007).

Nodal signaling promotes a complex transcriptional network, which acts to separate the endodermal and mesodermal lineages as well as give regional identity to the newly formed endodermal layer (Lewis and Tam, 2006; Zorn and Wells, 2009).

Morphogenetic movements of the endoderm during gastrulation

In mouse, birds and humans gastrulation starts in the pluripotent epiblast layer with the formation of the primitive streak. In the mouse, cells migrate through the primitive streak displacing and intercalating with the underlying visceral endoderm (VE) (which gives rise to mainly extraembryonic structures) and forms the endodermal layer of the embryo (Lawson and Pedersen, 1987; Tam and Beddington, 1992) The first cells to emerge through the primitive streak migrate anteriorly and give rise to the anterior definitive endoderm (ADE) (Lawson and Pedersen, 1987; Lawson and Schoenwolf, 2003; Kimura et al., 2006; Tam et al., 2007; Franklin et al., 2008). The mid- and hindgut endoderm emerges at the mid-streak stage of gastrulation, with all definitive endoderm (DE) formation completed prior to somite formation (Tam et al., 2007; Franklin et al., 2008; Zorn and Wells, 2009). Recently, it has been demonstrated that definitive endoderm that migrates through the primitive streak does not entirely displace visceral endoderm as previously thought (Kwon et al., 2008). Instead, the newly specified DE intercalates and mixes with existing underlying VE such that in the gut tube at E8.75, the VE accounts for ~10% of foregut endoderm, ~15% of midgut and ~35% of hindgut endoderm (Kwon et al., 2008). These genetic studies are consistent with cell labeling studies, which suggested that the VE might contribute to the gut tube (Tam and Beddington, 1992). This recent evidence for intercalation confounds genetic loss-of function studies for endoderm development because a complete failure DE formation via gastrulation may be masked by the contribution of VE to the DE layer after gastrulation. This problem may be especially true in the hindgut, where up to 35% of the DE after gastrulation consists of VE.

Gut tubulogenesis and elongation

Once the primary germ layers are established, the endoderm undergoes a complex series of changes to give rise to a gut tube. Gut tube morphogenesis differs dramatically between organisms. For example, in zebrafish and Xenopus, the endoderm forms from a “rod” of cells which will give rise to the gut tube (Warga and Nusslein-Volhard, 1999; Chalmers and Slack, 2000; Horne-Badovinac et al., 2001), whereas in chick, mouse, and human, the endoderm exists as a sheet of simple epithelium covering the mesoderm. The endodermal sheet undergoes a series of morphogenetic movements, which will ultimately give rise to the gut tube (Figure 1). Fate-mapping studies have shown that specific regions of the naïve endoderm contribute to specific domains of the gut tube, and in general, anterior endoderm cells contribute to the anterior intestinal portal (AIP) whereas posterior endoderm cells contribute to the caudal (posterior) intestinal portal (CIP) (Lawson et al., 1986) (Lawson and Pedersen, 1987; Tam et al., 2004; Franklin et al., 2008). Furthermore, fate-mapping studies have shown that individual organs, such as the liver, can arise from multiple early endoderm domains (Rosenquist, 1971; Warga and Nusslein-Volhard, 1999; Chalmers and Slack, 2000; Tremblay and Zaret, 2005).

Figure 1. Gut tube formation.

A. Schematic of the morphogenetic movements giving rise to the gut tube from post-gastrulation (E7.25) through gut tube formation (E9.5) in the mouse embryo. At E7.25, after gastrulation, the embryo is organized in a cup shape with the endoderm on the outer-most surface (yellow) and the mesoderm and ectoderm on the inside of the cup (black). At E7.75, pits that will give rise to the AIP and CIP are evident (black arrows). Note that the AIP forms several hours prior to the CIP, but for ease of illustration they are shown together. By E8.0 the AIP and CIP are clearly visible. Between E8.0 and E9.0, the lip of both the portals moves toward the center of the embryo, and the lateral endoderm continues to fold ventrally. Between E9.0 and E9.5 the endoderm has finished folding and exists as a tube surrounded by mesenchyme.

B. Ventral view of the endoderm folding. At E7.5, shortly after gastrulation has ended, the endoderm is schematically represented as a relatively naive flat sheet. By E7.75, the endoderm begins to fold along the longitudinal axis, much like a sheet of paper being rolled into a tube. At the same time, the intestinal portals begin to form (black arrows point to invaginations where the portals will form). Between E8.0 and E8.5, the ends of the endodermal “tube” are formed and represent the anterior and caudal intestinal portals. Simultaneously, the endoderm continues to fold ventrally along the longitudinal axis. Since the endoderm is covered by mesoderm, as the endoderm forms a tube, mesoderm is visible. By E9.0–E9.5 the endodermal tube is closed such that the endoderm is entirely surrounded by mesoderm. The dashed line denotes a cross section through the closed tube, where the endoderm is surrounded by the mesoderm.

C. FoxA2creER;R26R embryos treated with Tamoxifen via maternal oral gavage (0.12mg/g) at E6.5. Tamoxifen activation of FoxA2creER causes extensive recombination of the R26R allele, leading to β-galactosidase expression in the endoderm and notochord. Embryos were collected and stained for β-galactosidase activity with X-Gal (shown in blue) at E8.5 and E9.5. Anterior is denoted by the red arrow and posterior is denoted by the blue arrow. The middle panel demonstrates the CIP at E8.5, with the arrow pointing to the lip of the CIP.

In chick and mouse, one of the first steps in gut tube closure is the formation of anterior and posterior pits in the endoderm, which are the beginning of the AIP and CIP respectively (Figure 1A,B; 2A). As the pits invaginate further, they form pockets. The intestinal portals expand rostrally (AIP) and caudally (CIP) while the openings of both the AIP and CIP move towards each other. The lateral endoderm of the midgut folds ventrally and meets together to form a closed gut tube as the embryo is turning at e9.0 (Figure 1; 2B,C) (Lewis and Tam, 2006).

Figure 2. Gut tube histology.

A–B. At left, a schematic drawing of the embryo shows the level of the histological sections shown at right. The endoderm in colored yellow in the schematic and pseudocolored yellow in the histological sections.

A. Transverse sections through the caudal region of an E7.75 embryo demonstrating the open endoderm mid-embryo (section 1) which is starting to form a CIP (section 2 and 3).

B. Transverse sections through the caudal region of an E8.5 embryo showing the open gut tube toward the middle of the embryo, which closes forming the CIP at the posterior of the embryo.

C. Changes in the gut tube epithelium between E9.0 and E11.5. The gut tube endoderm condenses between E9.0 and E9.5 to give rise to a pseudostratified epithelium. Between E9.5 and E10.5 the gut tube epithelium circumference increases giving rise to a bigger lumen and an increased epithelial surface area.

All images are the same magnification (40X). Endoderm is pseudocolored yellow in all histological sections.

Little is known about how intestinal portal formation and gut tubulogenesis is controlled. Extensive cross talk and inductive cues between the mesoderm and endoderm have been described (Schultheiss et al., 1995; Wells and Melton, 1999; Cleaver and Krieg, 2001; Deutsch et al., 2001; Rossi et al., 2001; Withington et al., 2001; Couly et al., 2002; David et al., 2002; Zaret, 2002; Sugi and Markwald, 2003; Serls et al., 2005; Wandzioch and Zaret, 2009). Despite this extensive body of work, it does not appear that mesodermal movements directly guide gut tube morphogenesis. That is, while the mesoderm is undoubtedly providing information to the endodermal layer, it does not appear as though the cellular movements of the mesoderm are directly tethered to endodermal movements since cell labeling experiments show that labeled adjacent mesoderm and endoderm can end up at different locations (Tremblay and Zaret, 2005).

Transcription factors and signaling pathways that disrupt gut tube formation have been identified. In most cases, however, disruption of gut tubulogenesis is a secondary consequence of disrupting endoderm specification or maintenance. For example, disruption of Fox factors (FoxA2, FoxH1), Gata factors, Sox17, Mixl1 or Smad signaling all lead to defects in gut tube morphogenesis which are secondary to endoderm development and specification defects (Soudais et al., 1995; Hudson et al., 1997; Narita et al., 1997; Bossard and Zaret, 1998; Dufort et al., 1998; Clements and Woodland, 2000; Tremblay et al., 2000; Hart et al., 2002; Kanai-Azuma et al., 2002; Liu et al., 2004; Tam et al., 2007; Hoodless et al., 2001). New evidence suggests that endoderm which will give rise to anterior or posterior endoderm may be specified differently during gastrulation. Analysis of FoxA2 and FoxH1 null embryos with novel markers for foregut, midgut and hindgut showed that mid and posterior endoderm is specified in these embryos while anterior endoderm fails to form (McKnight et al., 2010). As mentioned above, however, this study did not perform lineage-tracing experiments to rule out the possibility that the mid- and posterior endoderm seen in these mutant embryos was contributed by the VE.

Recently, it has been demonstrated that regulation of convergent extension movements control gut tube formation and elongation in mouse (Garcia-Garcia et al., 2008; Wen et al., 2010). Chato, a KRAB zinc-finger protein, regulates body axis elongation in all three germ layers through a process independent of non-canonical Wnt signaling. Chato was demonstrated to control convergent extension in the endoderm, and in embryos lacking Chato the endoderm failed to elongate and undergo gut tube closure (Garcia-Garcia et al., 2008). In a second study, Dact1 is involved in gut tube morphogenesis by regulating non-canonical Wnt-dependent planar cell polarity (PCP) signaling. Dact1 null mice had severely impaired posterior development and hindgut defects, including a failure of the hindgut endoderm to form a CIP and failure of the endoderm to fold ventrally at e8.25. By e10.5 Dact1 null mice fail to form a cloaca and have no obvious hindgut (Wen et al., 2010). While Dact1 regulates non-canonical Wnt signaling, canonical Wnt signaling is also important for caudal development as a lack of Wnt3a, Wnt5a, LRP6 or Tcf1/Lef1 all lead to posterior developmental defects that are mainly mesodermal in origin (Takada et al., 1994; Yamaguchi et al., 1999; Pinson et al., 2000). TCF1/4 double null mice have a lack of caudal structures, however, unlike other disruptions in canonical Wnt signaling, the primary defect in these mice is in the endoderm. At E8.5 TCF1/4 double null mice fail to form a CIP and lose expression of the hindgut endoderm markers such as FoxA1 and Sox17. By E9.5 embryos have an open midgut and completely lack hindgut structures concomitant with a loss of posterior endodermal Shh expression (Gregorieff et al., 2004). Lastly, studies in Xenopus have shown that generation of a lumen in the primitive gut tube, as well as gut tube elongation require Rho, ROCK and Myosin II for proper cellular movement and tissue rearrangement (Reed et al., 2009). Taken together, these studies demonstrate that both Wnt-dependant and Wnt-independent mechanisms control gut tube morphogenesis, and both canonical and non-canonical Wnt signaling are required for development of the posterior gut tube. Overall, our understanding of the dynamics and molecular regulation of gut tubulogenesis is still in its infancy and a more in-depth understanding of this process will require additional work.

Gut tube patterning

Nodal signaling, which is required for mesendoderm specification during gastrulation, plays a second role in specifying anterior identity during anterior-posterior (A-P) patterning (Osada and Wright, 1999; Schier and Shen, 2000; Brennan et al., 2001; Hoodless et al., 2001; Yamamoto et al., 2001; Robertson et al., 2003; Vincent et al., 2003; Duboc et al., 2004; Lu and Robertson, 2004; Yamamoto et al., 2004), such that by the end of gastrulation, the endoderm is patterned into molecularly distinct A-P domains as demonstrated by anterior Hhex, FoxA2 and Sox2 expression and posterior Cdx expression (Cdx1, 2, 4) (For review, see (Dufort et al., 1998; Martinez Barbera et al., 2000; Chawengsaksophak et al., 2004; Kinkel et al., 2008; Sherwood et al., 2009; Zorn and Wells, 2009).

After the initial A-P patterning during gastrulation, the endoderm continues to receive patterning and inductive signals, mostly from the mesoderm, such that by e8.5 distinct organ-specific molecular domains are already in place (Sherwood et al., 2009). These inductive and patterning steps occur simultaneously with tissue rearrangements and gut tube formation, making this time during embryogenesis extremely dynamic. The anterior gut tube will give rise to the esophagus, lungs, thyroid, liver, pancreas and biliary system, where as the midgut will give rise to the stomach and the small intestine and the hindgut will give rise to the large intestines as well as the lining of the genito-urinary system (Wells and Melton, 1999; Zorn and Wells, 2007; Seifert et al., 2008; McLin et al., 2009; Spence et al., 2009; Zorn and Wells, 2009).

Signaling pathways that play a role in A-P patterning of the endoderm include FGF, Wnt, BMP and retinoic acid signaling, and are reviewed elsewhere (Summarized in Figure 4A). (Wells and Melton, 1999; Wells and Melton, 2000; Tiso et al., 2002; Kumar et al., 2003; Grapin-Botton, 2005; Dessimoz et al., 2006; Lewis and Tam, 2006; Marikawa, 2006; Tam et al., 2006; McLin et al., 2007; Pan et al., 2007; Zorn and Wells, 2007; Goessling et al., 2008; Wills et al., 2008; Bayha et al., 2009; McLin et al., 2009; Zorn and Wells, 2009). An understanding of developmental signaling pathways that pattern the endoderm in the embryo has been used to inform embryonic stem cell (ESCs) differentiation into specific endodermal lineages (D'Amour et al., 2006; Cai et al., 2007; Kroon et al., 2008; Basma et al., 2009; Maehr et al., 2009; Brolen et al., 2010; Mfopou et al., 2010). Additionally, how multiple signaling inputs are integrated to determine positional A-P identity in the embryo is also starting to become clearer through the use of ESCs because of fine control over growth factor timing and dosage that in vitro studies allow. It has recently been demonstrated that a specific dose of FGF4 likely patterns hESC derived endoderm such that it is able to increase Pdx1 expression induced by RA signaling. In addition, RA and FGF4 appear to work synergistically to induce posterior Cdx2 expression (Johannesson et al., 2009). Moreover, it has recently been reported that varying concentrations of FGF2 can pattern hESC-derived endoderm from anterior (low doses) to posterior (high doses)(Ameri et al., 2010). Our work has recently shown that FGF and Wnt signaling synergize to efficiently push hESC derived endoderm into Cdx2+ hindgut epithelium, which gives rise to hindgut spheroids that are equivalent to early embryonic (E8.5) mouse hindgut. These spheroids can then be grown in intestine specific growth conditions in vitro and go through a series of developmental steps to give rise first to fetal gut-like tissue and then to tissue reminiscent of adult intestine (Spence JR, 2010). It is becoming evident that in vitro differentiation of different tissue lineages is able to greatly benefit from closely mimicing embryonic developmental cues, and that including very early embryonic patterning steps in differentiation protocols is an important first step in this process.

Figure 4. Schematic of mouse intestinal development from gastrulation to birth.

A. The grey box contains a schematic of an E8.0 mouse embryo with the endoderm in yellow. Signals involved in anterior-posterior patterning are sent from the mesoderm (Meso, white cells) to the endoderm (Endo, yellow cells). Signaling pathways important for A-P patterning are summarized, and include fine regulation of Bmp, Wnt, Fgf and RA signaling along the axis.

B–F. The grey boxes contain a schematic of a transverse section through the proximal intestine (duodenum) at different developmental time points. The endoderm-derived intestinal epithelium is shown in yellow. A close up of the intestinal epithelium and underlying mesoderm-derived mesenchyme is shown under the grey box. Key molecular events regulating intestine development are shown based on the developmental time during which evidence has been presented for each event.

B. Pseudostratified intestinal endoderm at E10.5.

C. At E14.5, intestinal morphogenesis is underway. The intestinal epithelium becomes stratified and the formation of Ezrin+ secondary lumina is apparent (Ezrin expression is shown in red).

D. By E15, secondary lumina have fused with the luminal surface and villus emergence is evident. In addition, condensed mesenchyme expressing higher levels of BMPs and PDGFR-α underlying nascent villi are present. Major signaling pathways (described in detail in the text) involved in epithelial-mesenchymal crosstalk responsible for modulating villus emergence are depicted.

E. At E16.5 villi and intervillus regions are evident. Sox9 expression is restricted to the proliferative intervillus region (green cells). Recent evidence has shown that β-catenin activity is present in the villus epithelium, but is excluded from the proliferative intervillus region.

F. By late E18.5, β-catenin activity has transitioned and is present in the proliferative Sox9+ intervillus villus, and is absent from the villi. A schematic of the pathways involved in mesenchymal-epithelial crosstalk at E18.5 that regulate both mesenchymal and epithelial proliferation is shown.

While it is clear that secreted factors signaling from the mesoderm to the endoderm are crucial in A-P patterning, these signaling events must ultimately induce an endodermal epithelium intrinsic program that will specify hindgut and intestinal identity. For example, xenografting studies show that the hindgut endoderm is only partially able to be reprogrammed along the A-P axis by the mesoderm just prior to cytodifferentiation (Duluc et al., 1994; Grapin-Botton and Melton, 2000) indicating that the intrinsic identity of the epithelium becomes fixed at a certain point during development. Hox factors play an important role in patterning the mesoderm and neurectoderm (McGinnis and Krumlauf, 1992; Krumlauf, 1994; Deschamps et al., 1999), however, mutations in Hox factors have only minor effects on the gut endoderm. (Manley and Capecchi, 1995; Boulet and Capecchi, 1996; Aubin et al., 1997; Warot et al., 1997; Zacchetti et al., 2007). It has recently been shown that a member of the ParaHox gene cluster, Cdx2, is perhaps the most critical intrinsic factor for hindgut and intestinal specification and patterning (Gao et al., 2009; Grainger et al., 2010), whereas a loss of Cdx1 or Cdx4 does not have an intestinal phenotype (Subramanian et al., 1995; van Nes et al., 2006). Cdx2 null mice die at E3.5, and therefore assessment of Cdx2 function at later stages requires a conditional loss-of-function approach. Conditional loss of Cdx2 in the endoderm around E9.5 results in the loss of intestinal identity. It becomes apparent by E12.5 that the gut epithelium is replaced by esophageal epithelium. In these mice, the posterior Hox code was only transiently affected such that only 5/13 Hox genes examined at e12.5 had reduced expression levels, which eventually recovered by e14.5. Furthermore other region-specific markers, such as Pdx1 (pancreas and duodenum) and Barx1 (stomach) remained unchanged. This indicates that while the endoderm has a posterior-to-anterior transformation in Cdx2 loss-of-function mice, other aspects of the A-P pattern remain intact (Gao et al., 2009). Conditional loss of Cdx2 at later developmental time points, around e13.5, led to a transformation of the small intestine to a stomach-like identity (Grainger et al., 2010). It is clear from these studies that Cdx2 is initially required for establishing and maintaining posterior identity and at later stages, it is required for maintaining A-P identity.

It is likely that a combination of signaling molecules responsible for A-P patterning integrate to drive endoderm specific Cdx2 activity, since the Cdx family is regulated in many different contexts by Wnt, Fgf, Bmp and RA signaling (Houle et al., 2000; Allan et al., 2001; Ikeya and Takada, 2001; Beland et al., 2004; Beland and Lohnes, 2005; Keenan et al., 2006; Pilon et al., 2006; Joo et al., 2010). It is also important to note that different signaling pathways will likely have different effects on Cdx2 expression and activity during different developmental time-points. For example, double-null mutations in the Wnt downstream effectors TCF1 and TCF4 have early posterior defects, including a loss of the caudal-most hindgut at E8.5. At E13.5 these TCF1/4 -mutant mice had a homeotic transformation such that the duodenum had reduced Cdx2 expression and expressed the stomach marker, Sox2 (Gregorieff et al., 2004). This phenotype is strikingly similar to loss of Cdx2 at E13.5 (Grainger et al., 2010) and indicates that Wnt signaling may positively regulate Cdx2 expression during early development. However, a second recent study shows that de-repression of Wnt signaling by loss of the Wnt antagonist Pinin around e13.5 also leads to downregulation of Cdx2 (Joo et al., 2010). Collectively, these studies show that there will be complex temporal, spatial and context specific roles for patterning factors in specifying hindgut identity.

Early intestinal development

Little is known about changes occurring in the intestinal epithelium after the hindgut epithelium is fully formed, but before villus emergence begins. In mouse, after the gut tube is fully formed around e9.0, the simple epithelium condenses and becomes a pseudostratified epithelium by e9.5 (Figure 2C). Between e9.5 and e13.5, the intestine lengthens with the growing embryo and the circumference of the gut mesenchyme and epithelium increases along with the size of the lumen (Figure 2C and 3) (Lepourcelet et al., 2005; Cervantes et al., 2009). Wnt5a, working through non-canonical Wnt signaling, has been implicated in the control of gut elongation and regulation of epithelial architecture. Wnt5a null mice have an 80% reduction in small intestine length and a 63% reduction in large intestine length by E18.5, but differences were apparent by E14.5. Wnt5a mutant mice also had disrupted apical-basal polarity of the intestinal epithelium (Cervantes et al., 2009). Similarly, mice lacking proteins of the Secreted Frizzled Related Protein (Sfrp) family (Sfrp1−/−;Sfrp2−/− or Sfrp1−/−;Sfrp2−/−;Sfrp5+/−), which physically interact with and inhibit Wnt5a, also had shortened intestines by E13.5. Furthermore, apical-basal polarity of the epithelium and core components of the PCP pathway were disrupted in these mice, consistent with a defect in non-canonical Wnt signaling (Matsuyama et al., 2009).

Figure 3. Intestinal epithelial reorganization.

Intestinal development at E12.5 (A, A’), E14.5 (B, B’), E16.5 (C, C’) and E18.5 (D, D’). A–D shows the Cdx2 positive epithelium (Red) and Vimentin positive mesenchyme. A’–D’ shows the E-cadherin positive epithelium. Secondary lumina can be seen at E14.5 (Boxed region is magnified in the inset; L indicates primary lumen, s indicates secondary lumen) (B’) as epithelial reorganization begins. By E16.5 there are clear intervillus regions and villi (C, C’).

In addition to Wnt signaling, Fgf signaling is also important during early intestinal development. Fgf9, Fgf10 and FgfR2b are all required for proper cecum development. Fgf9 is expressed in the intestinal epithelium and signals to the mesenchyme, driving mesenchymal proliferation and lengthening of the intestine (Geske et al., 2008). Furthermore, Fgf9 signals from the epithelium are required for mesenchymal Fgf10 expression. Fgf10 then signals back to the epithelium through FgfR2b, driving proliferation, which is required for cecal budding and growth (Burns et al., 2004; Fairbanks et al., 2004; Sala et al., 2006; Zhang et al., 2006).

During this time in intestinal development, in addition to an increase in length and circumference, there are dramatic changes in the intestine at the transcriptional level. Data mining approaches have been employed and proven useful to look at the entire transcriptome, as well as transcription factor specific regulation across developmental timepoints (Lepourcelet et al., 2005; Choi et al., 2006). Based on the limited understanding of early intestinal development, and the lack of literature available for intestinal development at this stage, it is clear that additional research is needed in this area.

Epithelial reorganization and villus emergence

Starting at about E14 in the mouse (E17 in the rat, 9 weeks in humans), the stratified epithelium of the midgut and hindgut endoderm begins to undergo extensive reorganization to form the simple columnar epithelium which covers the lumenal surface of the intestines (Figures 3 and 4). This reorganization has been carefully examined by light and electron microscopic analysis of the developing intestine in animal models and human fetuses. (Johnson, 1910; Patten, 1948; Grand et al., 1976; Trier and Moxey, 1979). Prior to these morphological changes, the endoderm transitions from a tightly packed simple epithelium with nuclei at several levels within the apicobasal axis, called pseudostratified epithelium, to a stratified epithelium in which the apical (lumenal) cells are connected by junctional complexes, but the basally located cells are more loosely associated (Figure 3A,B; 4B,C) (Dunn, 1967). By E14 in the mouse, a wave of rostral-to-caudal (proximal-to-distal) epithelial reorganization initiates formation of the columnar intestinal epithelium. This reorganization begins with secondary lumina (also called intra-epithelial cavities) forming in the deeper basal layers of the stratified epithelium, with nascent junctional complexes in cells surrounding these secondary lumina (Figure 3B; 4C) (Mathan et al., 1976; Matsumoto et al., 2002). Over the next day, extension of these junctional complexes to adjacent cells enlarge these secondary lumens, and they begin to fuse with the primary lumen (Figure 4D). Mitosis of cells lining the secondary lumina may also contribute to lumina expansion (Toyota et al., 1989). At the same time that the intestinal endoderm begins to reorganize into a simple columnar epithelium, the mesenchyme begins to invaginate into the epithelium to form nascent villi (Figure 3C; 4D,E). Like epithelial reorganization, villus emergence progresses in a rostral-to-caudal wave and is evident by E15 in the mouse, by E19 in the rat, and 9 weeks in human (Figure 4D,E) (Mathan et al., 1976; Madara et al., 1981).

Development of the large intestine (colon) follows a similar pattern as the small intestine, with some important distinctions. As in the small intestine, the early single-layer hindgut endoderm develops into a multilayered epithelium surrounded by mesenchyme and an outer mesothelium (mouse E13–E15, rat E16–18, human 9 weeks) (Polak-Charcon et al., 1980; Colony et al., 1989). Interestingly, the fetal colon develops villus-like structures before adopting the deeper crypts and flat intercrypt table that constitutes the mature colonic epithelium (Patten, 1948; Helander, 1973). These villus-like structures first appear as longitudinal ridges in the multilayered epithelium, which begin to develop secondary lumina in a process of epithelial reorganization similar to the small intestine (E16 in mouse, E19 in rat, 10 weeks in humans) (Helander, 1973; Bell and Williams, 1982). At the same time, the mesenchyme begins to invaginate into the epithelial ridges, forming longitudinal epithelial folds. These folds resolve into villi which are covered with stratified epithelium that has initiated cytodifferentiation, evident with the emergence of goblet cells (11–12 weeks in humans) (Lev and Orlic, 1974). These larger primary villi subsequently split in a process mediated by extensive epithelial and mesenchymal rearrangements. Crypt-shaped structures initially form as secondary lumina within the basal layers of the stratified intervillus epithelium, and extend as a single layered epithelium to become continuous with the lumen. At the same time, mesenchymal cells extend into the epithelial layer adjacent to the nascent single layered epithelium, thus dividing the primary villi into smaller secondary villi, each with a core of mesenchymal lamina propria covered by a simple columnar epithelium. (E17–18 in mice, 12–13 weeks in humans) (Bell and Williams, 1982).

Although the morphological changes in the fetal gut have been well described in several animal models, less is known about the mechanisms that initiate epithelial reorganization and villus emergence. Apicobasal polarity is required for epithelial reorganization, likely supported by the planar cell polarity pathway (Matsuyama et al., 2009). For example, the apical membrane organizing protein Ezrin is required for normal polarization of the epithelium from E14–E15 (Figure 4C,D). At E14, Ezrin is concentrated at both the lumenal surface as well as secondary lumina undergoing epithelial reorganization. At this stage, the stratified intestinal endoderm of Ezrin deficient mice appears grossly normal, but by E15 some cells fail to properly polarize, ultimately resulting in villus fusion and gross disorganization of the intestinal mucosa (Saotome et al., 2004).

Epithelial-Mesenchymal crosstalk during villus emergence

Extensive crosstalk between the developing endoderm and underlying mesenchyme are critical for normal development of the intestines. This concept was first established by tissue reassortment experiments in which the endoderm and mesenchyme from different regions, stages, and/or species of the developing gastrointestinal tract were recombined and allowed to develop (Lebenthal, 1989). These studies demonstrate that both permissive and instructive reciprocal interactions between the intestinal endoderm and mesoderm are required for normal morphogenesis and differentiation. For example, in a more recent study, xenografts of stratified endoderm from proximal jejunum combined with mesenchyme from proximal colon (ePJ/mPC) gave villi, jejunal-specific absorptive enterocytes expressing the digestive enzyme sucrase-isomaltase, and endocrine cells expressing jejunal-type hormones (CCK), but did not support production of small-intestine specific Paneth cells. The converse xenograft (ePC/mPJ) also gave villi, sucrase-isomaltase expressing enterocytes, and in this case the xenografts contained Paneth cells, but endocrine cells expressed distal-type hormones (PYY, GLP-1) (Ratineau et al., 2003). This example demonstrates the complex nature of the reciprocal epithelial-mesenchymal interactions in the developing intestines.

Further evidence supporting the role of epithelial-mesenchymal crosstalk comes from mice bearing mutations in Epimorphin, a syntaxin protein involved in targeting secretory vesicles to the plasma membrane of mesenchymal cells (Hirai et al., 1992). Loss of epimorphin inhibits epithelial morphogenesis and enhances epithelial proliferation, likely by interfering with multiple intercellular communication pathways including BMP and Hedgehog (Fritsch et al., 2002; Shaker et al., 2010). Several other intercellular signaling pathways, including TGF-β, PDGF, FGF, WNT, and EGF, have been implicated in epithelial-mesenchymal crosstalk during intestinal epithelial reorganization and villus formation.

Hedgehog signals from the developing epithelium regulate maturation of the underlying mesenchyme (reviewed in (van den Brink, 2007). Both Shh and Ihh are expressed throughout the pseudostratified epithelium of the early intestinal endoderm. Shh and Ihh continue to be expressed in the single-layer columnar epithelium with Shh becoming restricted to the villus base during villus emergence and eventually becoming extinguished, whereas Ihh expression is retained by the differentiated epithelium (Kolterud et al., 2009). Shh mutant mice display villus overgrowth, whereas Ihh mutant mice show reduced epithelial proliferation and have smaller, fewer villi, suggesting opposing effects of Shh and Ihh (Ramalho-Santos et al., 2000)). Interestingly, inhibition of all intestinal Hedgehog signals, by deletion of both Shh and Ihh in the early endoderm, antibody injection, or expression of the Hedgehog antagonist Hhip, causes defective intestinal mesenchymal development, leading to severely disrupted villus formation and epithelial maturation; moderate Hedgehog inhibition leads to impaired crypt-villus axis formation with ectopic and branched crypts proliferating on the villi (Wang et al., 2002; Madison et al., 2005; Mao et al., 2010). These results suggest that Hedgehogs may be one of the primary and most important signals during the phase of villus emergence (Figure 4D,F).

In the subepithelial mesenchyme, Hedgehog signals are interpreted by Gli transcription factors Gli2 and Gli3, which directly activate transcription factors of the Forkhead Box winged-helix superfamily (Madison et al., 2009). Mutation of the Gli target gene Foxl1/Fkh6 causes a delay in epithelial reorganization and villus emergence, leading to hyperproliferation and abnormal crypt branching in the adult (Kaestner et al., 1997). Additional Hedgehog targets FoxF1 and FoxF2 cooperate to support mesenchymal expansion, differentiation, and maintenance. Loss of these genes causes defects in the mesenchyme leading to disintegration of the tissue shortly after villus formation (Figure 4F) (Ormestad et al., 2006). In addition to cell-autonomous mesenchymal effects of Gli and Fox proteins, these transcription factors regulate secreted morphogens, such as Wnts and BMPs, that signal back to the developing endoderm.

BMPs, in particular BMP2, 4, and 7, play important roles in reciprocal epithelial-mesenchymal signaling downstream of the Hedgehog pathway in the developing intestine (Figure 4D,F) (Roberts et al., 1995; Madison et al., 2005). BMPs are primarily expressed in the subepithelial mesenchyme, especially in the mesenchymal condensations that underlie nascent villi (Karlsson et al., 2000). Perturbation of BMP signaling in the developing chick hindgut caused abnormal morphogenesis and differentiation of endoderm, mesoderm, and ectoderm derivatives in the gut (De Santa Barbara et al., 2005). Mutation of the BMP receptor BMPR1a causes aberrant hyperproliferation and ectopic crypt formation on the villi, leading to polyp formation in a disease termed Juvenile Polyposis (Howe et al., 2001; He et al., 2004). However, it is clear that loss of BMP signaling in nonepithelial cells is critical for this process, as epithelial-specific deletion of BMPR1a causes only hyperplasia but not polyp formation (Auclair et al., 2007; Shroyer and Wong, 2007). The hypothesis that BMP signaling is critical for development of the crypt-villus axis was further supported by experiments in which epithelial expression of the BMP2/4/7 antagonist Noggin caused abnormal villus formation in transgenic lines (Batts et al., 2006). In these mice, epithelial reorganization occurs but subepithelial mesenchymal condensation is less robust, leading to fewer but larger villi formed by birth. Later, BMP inhibition consistently causes extensive disorganization with ectopic proliferating crypts on the villi, leading to polyp formation in the small and large intestines (Haramis et al., 2004; Madison et al., 2005; Batts et al., 2006). These defects are associated with increased Wnt, PDGF and Hedgehog activities, suggesting multiple levels of feedback regulating epithelial-mesenchymal interactions (Haramis et al., 2004; Madison et al., 2005; Batts et al., 2006).

PDGF epithelial to mesenchymal signaling functions in parallel with the Hedgehog pathway. PDGF-A and its receptor, PDGFR-α, are expressed in the endoderm and mesenchyme respectively, prior to villus emergence (Figure 4D). PDGF-A becomes progressively more restricted to the intervillus epithelium and subsequently to the crypts, whereas PDGFR-α expression remains scattered throughout the subepithelial mesenchyme, with particularly strong expression in mesenchymal clusters beneath nascent villi which retain strong expression and remain at the growing tip of the villus core as the villi elongate. Genetic ablation of this epithelial-to-mesenchymal signaling system causes abnormally folded small intestinal villi and disrupted colonic mucosal architecture, likely due to early differentiation of mesenchymal smooth muscle cells. (Karlsson et al., 2000). However, unlike Hedgehog and BMP pathway manipulations, in PDGF mutants proliferation remains appropriately restricted to the intervillus epithelium and crypts.

EGF is another potent morphogen that acts on the developing intestine. Deletion of the EGF receptor, EGFR, caused delayed epithelial reorganization and villus formation, leading to reduced proliferation, villus blunting and tissue disintegration (Miettinen et al., 1995). Interestingly, this phenotype was highly variable in different inbred strains of mice (Sibilia and Wagner, 1995; Threadgill et al., 1995). Postnatally, EGF is considered a potent mitogen for the intestine. (Barnard et al., 1995) However, EGF has been reported to have varying effects on the intestine of different species at various developmental stages. For example, EGF delivered in utero or in organ culture enhances intestinal epithelial maturation in mice (Calvert et al., 1982; Beaulieu et al., 1985). In contrast, EGF decreased proliferation and maturation of human fetal intestinal organ cultures (Menard et al., 1988; Menard et al., 1990). EGF treatment of intestine explanted prior to villus emergence enhanced villus formation and expression of cytodifferentiation markers for enterocytes and goblet cells, along with enhanced epithelial proliferation. Conversely, inhibition of EGFR blocked mesenchymal growth and development, as well as reducing epithelial proliferation and goblet cell formation (Duh et al., 2000). Thus, EGF is important in the early development of the intestine, but has varying effects depending on the developmental stage, and proximodistal region, and species examined.

Various additional growth factors, including HGF, IGFs, KGF/FGF7, and TGF-β have poorly defined roles in intestinal development, but are expressed in the intestinal mucosa along with their receptors and have protective or reparative effects in various contexts (Goke and Podolsky, 1996; MacDonald, 1999; Howarth, 2003; Xian, 2003; Ido et al., 2005). Much work remains to be done to identify all of the important intercellular communication pathways involved in epithelial reorganization and villus formation.

Transcriptional control of villus emergence

Several transcription factors have been implicated in intestinal villus formation. Elf3 is an ETS transcription factor that cooperates with the Crif1 transcriptional coactivator to regulate morphogenesis and epithelial differentiation of the intestine (Ng et al., 2002; Kwon et al., 2009). Both Elf3- and Crif1-deficient mice develop fewer, abnormal villi with disorganized lamina propria mesenchyme and defective epithelial cells, likely arising at the time of villus formation. Both mutant lines also show reduced expression of the Tgf-β type II receptor, Tgf-βRII. In an elegant study, re-expression of Tgf-βRII in the epithelium rescued all intestinal phenotypes associated with Elf3 deficiency (Flentjar et al., 2007). These results suggest that Elf3/Crif1 plays an important role during villus emergence that is mediated by Tgf-β signaling. However, the specific ligands and cells of origin of this morphogenetic signal remain to be identified.

HNF4α is another transcription factor expressed throughout the intestinal epithelium. Deletion of HNF4α in the early embryonic endoderm causes a severe colonic phenotype including loss of crypt-villus architecture, defects in epithelial and mesenchymal maturation, and reduced epithelial proliferation (Garrison et al., 2006). In contrast, deletion after villus formation does not perturb intestinal development but instead leads to abnormal homeostasis in the adult crypt progenitors (Cattin et al., 2009; Darsigny et al., 2009).

Mutation of Nkx2.3, a mesenchymal transcription factor, causes a significant reduction in mesenchymal cells, reduced epithelial proliferation, and delayed villus formation, with about half of the mice dying prior to weaning. Mice that survive recover from the neonatal paucity of intestinal villi and instead show abnormal architecture of the small and large intestines including branched villi, thickened mucosa, and epithelial hyperproliferation (Pabst et al., 1999).

In addition to the specific transcription factors discussed here, global chromatin remodeling has also been implicated at this stage of intestinal development. A dominant negative mutation of the p300 histone acetyltransferase (HAT) causes delayed villus emergence associated with a failure of subepithelial mesenchymal condensation and weak BMP4 expression at the sight of presumptive villus formation (Shikama et al., 2003). Villi eventually emerge in these p300 mutant animals, along with differentiated epithelium, but continued proliferation of the villus core mesenchyme suggests a continual requirement for histone acetylation in the developing intestine, especially in the mesenchymal compartment. Interestingly, a similar mutation in the HAT protein CBP does not affect prenatal development of the intestine.

Histone deacetylases (HDACs) catalyze the reverse modification as HATs. Examination of HDAC 1 and 2 showed strong expression throughout the pseudostratified intestinal endoderm which became restricted to weaker, villus-specific expression once villi had formed (Tou et al., 2004). In intestinal explants, overexpression of HDACs blocked epithelial differentiation, whereas treatment with HDAC inhibitors caused premature villus formation and epithelial differentiation, associated with increased histone acetylation on genes expressed in differentiated epithelial cells (Tou et al., 2004). Thus, dynamic modification of chromatin is likely to be an important component regulating development of the intestine.

Proliferation and cytodifferentiation of the embryonic intestinal epithelium

Proliferation occurs throughout the stratified midgut and hindgut endoderm prior to villus emergence. No evidence of functional cytodifferentiation is present at this time, although the epithelial cells at the apical surface form a functional barrier to passive diffusion of macromolecules (reviewed in (Grand et al., 1976; Trier and Moxey, 1979)). As the pseudostratified endodermal epithelium reorganizes to form a simple columnar epithelium and villi begin to emerge, proliferation increases in the epithelium at the bases of emerging villi (Figure 4E). Later, proliferation becomes restricted exclusively to the intervillus epithelium, and then to the crypts of Lieberkühn.

Distinct epithelial cell types can be identified by morphological and molecular markers during villus emergence. Each lineage in the intestinal epithelium has a distinct function. In the embryo, four main cell types can be identified: columnar cells that bear apical microvilli, termed enterocytes in the small intestine and colonocytes in the large intestine; mucous-producing goblet cells; caveolated or tuft cells; and a diversity of hormone-producing enteroendocrine cells. Antimicrobial peptide secreting Paneth cells appear in the small intestine once crypts develop, which occurs after villus emergence in human fetuses and postnatally in rodents. Enterocytes and colonocytes are collectively termed absorptive cells to denote their primary role in absorbing nutrients and electrolytes. Goblet, enteroendocrine, and Paneth cells are collectively termed secretory cells to reflect the abundant secretory granules which typify these cells. Tuft cells, a relatively rare and understudied component of the intestinal epithelium, is also classified as a secretory cell based on recent lineage studies (Gerbe and Jay, personal communication). In addition, highly specialized enterocytes termed M cells develop to overlie the lymphoid follicles of maturing Peyer’s patches. All of these epithelial cell types are derived from multipotent stem cells in the adult intestine, but whether distinct stem cells exist in the embryo is unclear. Although it is beyond the scope of this review to detail all of the factors involved in proliferation and cytodifferentiation in the intestine, two major pathways that influence this process, Wnt/β-catenin and Notch, will be discussed here. For further information on intestine epithelial stem cells and cytodifferentiation see recent reviews by (Barker et al., 2008; Garrison et al., 2009; van der Flier and Clevers, 2009)

The Wnt/β-catenin signaling pathway is essential for supporting intestinal epithelial stem cell maintenance, proliferation, and differentiation in the adult intestine, but its role in the prenatal intestine is less clear. Canonical Wnt signaling blocks APC-mediated degradation of cytoplasmic β-catenin, leading to β-catenin accumulation and translocation to the nucleus (reviewed in (Gregorieff et al., 2005)). In the nucleus, β-catenin associates with DNA binding proteins of the Tcf/Lef family, and recruits coactivators to initiate transcription of target genes. Components of the Wnt/β-catenin signaling pathway are dynamically expressed in endodermal and mesodermal layers of the developing intestines, but β-catenin transcriptional activity is first detected in the intestinal epithelium after villus emergence (Kim et al., 2007; Garcia et al., 2009; Joo et al., 2010). Kim et al report that β-catenin activity is restricted to the postmitotic cells of the villus epithelium, where it continues to be expressed until birth, at which point β-catenin activity redistributes to the proliferative intervillus epithelium (Figure 4E,F). Conflicting reports on the localization of β-catenin activity in embryonic intestine suggest some controversy in the field.

Evidence supporting a role for the Wnt/β-catenin pathway in fetal intestines comes from mice in which Tcf4 is disrupted: in these mutant animals, villus emergence occurs but epithelial proliferation in the small intestine is subsequently halted, and some aspects of cytodifferentiation are abnormal (Korinek et al., 1998). This result was confirmed in Tcf4 mutant zebrafish, which also show loss of proliferation in the middle and distal intestines (Muncan et al., 2007). Similarly, in the mature intestine inhibition of Wnt/β-catenin causes acute loss of proliferation, depletion of the progenitor compartment, and a block in differentiation of goblet, enteroendocrine, and Paneth cells (Pinto et al., 2003; Ireland et al., 2004; Kuhnert et al., 2004). The similarity in these phenotypes led to the conclusion that β-catenin transcriptional activity drives proliferation in both the embryonic and adult intestine. Tcf4 is expressed in the intervillus epithelium of the embryonic intestine, whereas Tcf3 is expressed in the embryonic villus epithelium (Kim et al., 2007). In the adult, Tcf4 is the main Tcf/Lef factor and is expressed throughout the intestinal epithelium, and β-catenin is active in crypt progenitors (Gregorieff et al., 2005; Davies et al., 2008). Thus, based on the report that β-catenin is activated only in the embryonic villus epithelium, β-catenin may utilize different Tcf binding partners to activate distinct fetal or postnatal transcriptional programs, (Figure 4E,F) (Kim et al., 2007). These results may also suggest β-catenin independent functions for Tcf4 in the embryonic intestine. Conversely, β-catenin may partner with other nuclear proteins such as Sox factors to execute Tcf-independent transcriptional programs (Sinner et al., 2004; Sinner et al., 2007).

Premature activation of β-catenin throughout the intestinal endoderm causes disruption of villus emergence (Kim et al., 2007). However, this result may arise from a homeotic transformation of mutant tissue to non-intestinal endoderm, since Cdx2 was lost in cells with active β-catenin. A previous study in which hyperactive β-catenin was expressed in intestinal epithelium that was already appropriately patterned showed no defects in morphogenesis but instead caused apoptosis of the mutant cells (Wong et al., 2002). The concept that β-catenin is normally repressed in the developing intestinal endoderm is further supported by recent studies of Pinin, a nuclear speckle associated protein that physically binds to and represses transcriptional activity of nuclear β-catenin. Deletion of Pinin in the intestinal endoderm impaired villus formation and cytodifferentiation, associated with premature activation of β-catenin in the intervillus epithelium and misregualtion of Cdx1 and 2 (Joo et al., 2010). Together, these studies demonstrate the key role of β-catenin activity in intestinal development, but emphasize that clarifying the role of Wnt/β-catenin in the perinatal intestinal epithelium is an important area of future investigation.

The Notch signaling pathway is essential for regulating proliferation and cytodifferentiation in the developing intestine. Notch proteins are transmembrane receptors that are bound by ligands of the Delta/Serrate/Lag-2 family, which are expressed by adjacent cells. Once bound, Notch receptors are cleaved to release a cytoplasmic transcriptional activation domain (Notch Intracellular Domain; NICD). Overexpression of NICD in the emerging duodenum blocked proliferation and inhibited villus growth (Stanger et al., 2005). In contrast, after villus emergence in the intestine NICD blocked formation of secretory cells (goblet and enteroendocrine cells) and expanded proliferation (Fre et al., 2005). Conversely, inhibition of Notch activity enhanced secretory cell production and reduced proliferation (Milano et al., 2004; van Es et al., 2005b). Notch1 and Notch2 constitute the essential Notch receptors in the mature intestine (Riccio et al., 2008). The specific ligands essential for signaling in the intestine have yet to be identified.

In the nucleus, NICD binds to CSL/RBP-J proteins to activate transcription of target genes, the best known being Hes1. Studies of Hes1-deficient mice demonstrated that Hes1 promotes absorptive enterocyte differentiation and represses formation of secretory cells (Jensen et al., 2000). Subsequent studies showed that Hes1 represses Atoh1/Math1, which is required for secretory cell formation (Yang et al., 2001; Shroyer et al., 2007). A model emerged in which a balance between Hes1 and Atoh1 expression, controlled by Notch activity, determines absorptive vs. secretory cell differentiation. This model is further supported by studies in other model organisms: inhibition of Notch in zebrafish and Drosophila gut caused a change in epithelial cell fate and dysregulated cell turnover (Crosnier et al., 2005; Micchelli and Perrimon, 2006; Ohlstein and Spradling, 2006). Recent evidence suggests that Atoh1 is both necessary (Kazanjian et al., 2010; van Es et al., 2010) and sufficient (Vandussen and Samuelson, 2010) for epithelial progenitors to adopt a secretory cell fate. Thus, Notch and Atoh1 are critical gatekeepers of intestinal epithelial cell fate.

The Notch and Wnt/β-catenin pathways interact at several levels to regulate intestinal epithelial cell fate (reviewed in (Nakamura et al., 2007)). Crosstalk between these pathways was first suggested by studies in mice with global inhibition of intestinal Wnt activity (Pinto et al., 2003), which blocked epithelial proliferation and also phenocopied the loss of secretory lineages observed in Atoh1-mutant mice. Recent evidence from adult intestine suggests Wnt/β-catenin can enhance Notch signaling by upregulating receptor and ligand expression, and can also directly activate key Notch targets including Hes1 (Peignon G, 2010). In addition, Atoh1 can be targeted for degradation by the APC/axin complex when Wnt signals are present, and Atoh1 has also been reported to be a direct target of β-catenin transcriptional activation (Tsuchiya et al., 2007). On the other hand, active Notch signaling can enhance GSK3β-mediated β-catenin degradation (Koo et al., 2009). Thus, the Notch and Wnt/β-catenin pathways crossregulate one another via multiple complex mechanisms. Furthermore, both Notch and Wnt/β-catenin can be coordinately regulated, for example by the reactive oxygen species (ROS)-producing oxidase Nox1(Coant et al., 2010). Additional work is needed to define the precise mechanisms that mediate context-dependent crosstalk between these two important pathways.

Within the secretory lineage, several genes are known to function downstream of Atoh1. Gfi1 is a zinc-finger transcriptional repressor that is an important regulator of the fate of Atoh1-expressing secretory progenitors. Gfi1 mutants divert goblet and Paneth cell progenitors towards the enteroendocrine fate (Shroyer et al., 2005). Neurogenin3 is a transcription factor in the same basic helix-loop-helix family as Atoh1, which it is dependent upon for expression in the intestine. Neurogenin3 is required for formation of all enteroendocrine cells (Jenny et al., 2002). Conversely, overexpression of Neurogenin3 in the developing intestine can divert secretory progenitors towards the enteroendocrine fate at the expense of goblet cells (Lopez-Diaz et al., 2007). Recent evidence suggests that Gfi1 exerts its effects by repressing Neurogenin3 (Bjerknes and Cheng, 2010). Thus, a balance between Gfi1 and Neurogenin3 may control subtype allocation within the secretory lineage. After commitment to the enteroendocrine lineage, several transcription factors (e.g., NeuroD1, Pax6, Nkx2.2, Insm1) regulate differentiation and maintenance of specific hormone-producing cells (reviewed in (Schonhoff et al., 2004).

Several additional transcription factors function downstream of Wnt/β-catenin and Notch/Atoh1 to control intestinal epithelial cytodifferentiation and proliferation. SPDEF is a target of both the Wnt/β-catenin and Notch/Atoh1 pathways, and directs terminal differentiation and maturation of goblet cells (Gregorieff et al., 2009; Noah et al., 2010). Sox9 is a transcriptional target of β-catenin which can feed back to inhibit its activity, and regulates epithelial proliferation and formation of Paneth and goblet cells (Bastide et al., 2007; Mori-Akiyama et al., 2007). Klf5 and Klf4 are respectively expressed in the proliferating and differentiated cells of the intestinal epithelium. Klf5 is thought to stimulate whereas Klf4 inhibits proliferation by interactions with the Wnt/β-catenin pathway (McConnell et al., 2007; Flandez et al., 2008; McConnell et al., 2009). Klf4 is also a target of the Notch pathway where it regulates colonic goblet cell differentiation (Katz et al., 2002; Zheng et al., 2009).

Crypt development and postnatal changes to the intestinal epithelium

Crypt development initiates by anchorage of precursor stem cells to the intervillus epithelium, and subsequent remodeling of the tissue surrounding these cells to form flask-shaped crypts by upward migration of the crypt-villus junction, rather than downward penetration of the developing crypt (Calvert and Pothier, 1990). Crypts contain multiple stem cells but are monoclonal, that is, derived from a single stem cell, in adult intestines (Ponder et al., 1985; Griffiths et al., 1988). In mice, crypts become monoclonal during the first 2 weeks of postnatal development. In utero, progenitor cells mix extensively, with no clonal grouping of cells (Shiojiri and Mori, 2003). One week postnatally, mixed crypts are observed, but at 2 weeks most crypts are monoclonal (Schmidt et al., 1988). The mechanisms that produce this clonality have not been elucidated, but likely derive both from the rapid proliferation of the intestinal epithelium, and from the mechanism of crypt expansion by fission (see below). In the perinatal intestinal epithelium, proliferation of putative stem cells occurs at a rapid rate to maintain growth of the organ, resulting in patches of cells that are derived from a single earlier progenitor. As crypts develop, they are clonally derived within the patch, but are polyclonal at the border between two patches. Subsequently, new crypts arise by fission of existing crypts, with stem cells from the parent crypt partitioning randomly between the daughter crypts. As new crypts arise from rapid expansion of resident stem cells and subsequent fission, these polyclonal crypts become monoclonal. This clonality is maintained throughout adulthood, and can serve to fix genetic and epigenetic changes geographically within the intestines (Fuller et al., 1990; Endo et al., 1995; Novelli et al., 1996) For example, mutations can arise in crypts and these mutant crypts can expand by fission to create patches of clonally-derived cells with identical mutations (Greaves et al., 2006; Gutierrez-Gonzalez et al., 2009). Epigenetic marks are similarly stable and can be used to study stem cell dynamics in the intestines (Yatabe et al., 2001).

Less is known about pathways that control crypt development. As discussed above, the Hedgehog and BMP pathways regulate the location of developing crypts in the intervillus epithelium. BMP signals likely restrict stem cells to the emerging crypts, as phospho-SMAD 1/5/8 is observed primarily on the villus epithelium (Haramis et al., 2004). This restriction may be related to inhibition of the Wnt/β-catenin pathway via the PTEN/PI3K/AKT pathway (Tian et al., 2005). Wnt/β-catenin activity is likely to be required for emergence and maintenance intestinal stem cells, but direct evidence that it is required for crypt development is lacking. Wnt/β-catenin is required for expression of EphB2 and EphB3 in the intervillus epithelium, and for restriction of the EphB ligand Ephrin-B1 to the postmitotic villus epithelial cells. These EphB/EphrinB proteins contribute to crypt formation by restricting proliferating cells to the intervillus epithelium, and later for localization of Paneth cells to the crypt base (Batlle et al., 2002). The canonical β-catenin transcriptional target c-Myc also likely contributes to crypt emergence (Bettess et al., 2005), although conflicting data suggests that c-Myc may be required for stem cell maintenance rather than crypt formation (Muncan et al., 2006). Finally, several β-catenin target genes have been identified as molecular markers of crypt base columnar (CBC) stem cells. One such β-catenin target, Ascl2, is of special interest because regulation of its activity controls stem cell numbers: increased Ascl2 expression expanded proliferation and caused formation of crypt-like structures on the villi, whereas deletion of Ascl2 blocked renewal of CBCs and loss of mutant crypts (van der Flier and Clevers, 2009). Determining the mechanisms that control crypt development, and distinguishing these from those that regulate intestinal stem cells, is an important goal for the future.

Paneth cells begin to differentiate in the small intestinal epithelium coincident with the development of crypts. As with crypt development, the Wnt/β-catenin pathway is essential for maturation of Paneth cells. Mutation of the Wnt receptor Frizzled5 causes mislocalization of Paneth cells similar to EphB mutation, and blocks their final maturation (van Es et al., 2005a). Conversely, restriction of β-catenin activity in the late prenatal intestine via Lgr5-mediated feedback inhibition prevents premature emergence of Paneth cells (Garcia et al., 2009). Furthermore, β-catenin targets genes such as Sox9 and SPDEF are known to regulate Paneth cell differentiation and maturation.

Thus, Wnt/β-catenin activity appears to be centrally important to coordinate formation of crypts, stem cells, and Paneth cells. An excellent example of this coordinate regulation is seen in mice bearing mutations in FGFR3, which have defects in crypt formation, stem cell maintenance, Paneth cell maturation, and reduced expression and activity of β-catenin (Vidrich et al., 2009). However, these three properties of β-catenin can be separated, for example other genes that regulate Paneth cell maturation, such as PPAR-β, are not regulated by Wnt/β-catenin and do not show defects in crypt formation or stem cell maintenance (Varnat et al., 2006). Likewise, β-catenin is critical to regulate active CBC stem cells, but evidence is emerging for a slowly-cycling quiescent/reserve stem cell population that may not be directly regulated by β-catenin (reviewed in (Montgomery and Breault, 2008; Li and Clevers, 2010). Much work remains to fully understand how the simultaneous formation of crypts, stem cells, and Paneth cells is accomplished.

A final major step during development of the mammalian intestine occurs during the suckling to weaning transition (Pacha, 2000). During this period, the animal must change its nutritional absorptive capacity from a primarily lipid-based diet (milk) to a more complex carbohydrate-based mix (solid food). This transition is regulated by both cell-intrinsic and external cues, such as a surge in glucocorticoid levels. Precocious development of the adult digestive enzyme repertoire can be induced by systemic administration of steroids (Moog, 1953; Henning, 1981). However, the regional expression of appropriate transporters (i.e., jejunal expression of lactase; ileal expression of ASBT) appears to be primarily encoded by the intestine. Transcription factors of the GATA family are important for regulating this regional identity: mutation of Gata4 causes expression of ileal-specific digestive enzymes in the jejunum (Bosse et al., 2006; Battle et al., 2008). However, not all aspects of jejunal identity are lost in Gata4 mutant mice. Thus, much remains to be learned about the mechanisms that control regional identity in the intestine, and what controls intestinal digestive and absorptive capacity.

Perspectives

A daunting number of key developmental steps must be properly coordinated in order for the mature intestine to develop. Headway has been made over many decades in understanding how critical processes at each stage of development are regulated and our understanding of the cellular and molecular aspects of intestinal development has expanded significantly in recent years. However, several areas of intestinal development are understudied and unclear. Gut tubulogenesis, the process of making a gut tube from a sheet of endoderm, is only recently starting to be understood at the molecular level. It is currently unknown what controls the initial formation of the anterior and caudal pits which give rise to the AIP and CIP. Furthermore, endoderm specific deletion of genes controlling cellular movements leading to gut tube defects has not been performed. Thus, whether reported gut tube morphogenic defects are a consequence of endoderm-specific gene perturbations is unknown. While a lot of work has focused on mesenchymal-epithelial interactions in the foregut at this early stage of development, less work has been done studying crosstalk in the hindgut.

Another major area of intestine development that remains unclear is that of villus emergence. Many signaling pathways are clearly involved in this process, but it is unknown how they are all intimately tied together to drive villus formation. Additionally, transcriptional regulators that control the signaling pathways and drive villus emergence are lacking. During this process, it is also unclear how cells are designated to become an intervillus progenitor cell or a cell in the villus.

While there is some understanding of how the intervillus progenitor cells are regulated, existing literature and dogma surrounding regulation of adult intestinal stem cells (ISCs) make this an area for further investigation. For example, it is well established in the adult that Wnt/β-catenin signaling is required for regulation and maintenance of ISCs. Furthermore, the downstream effector of Wnt signaling, TCF4, is required for maintenance of the intervillus progenitor region in the embryo. In contrast to this, recent studies suggest that active canonical Wnt/β-catenin signaling is absent from the intervillus region until postnatal life in the mouse. Thus, while active Wnt/β-catenin signaling is required for adult ISC maintenance, it appears that there is a repressive role for Wnt signaling in maintenance of the intervillus progenitor zone during development. In addition to clarifying a role for Wnt/β-catenin in regulating intervillus progenitors during development, identification of additional regulators of progenitor establishment and maintenance are critical to further our understanding of intestine development. Many questions remain surrounding the transition from an embryonic progenitor to an adult ISC. That is, it is unknown if all intervillus progenitors are capable of becoming an ISC or if a small pool of cells is set-aside during development, which will give rise to ISCs in the crypt.

Lastly, one of the final hurdles is to translate the mechanisms controlling intestinal development into directing differentiation of pluripotent stem cells (PSCs; induced and/or embryonic) into intestinal tissue. It seems that this lofty goal will require a synthesis of the information reviewed here to piggyback on work that has led to differentiation of other endodermal organs such as liver and pancreas from PSCs.