One shall become two: Separation of the esophagus and trachea from the common foregut tube

Abstract

The alimentary and respiratory organ systems arise from a common endodermal origin, the anterior foregut tube. Formation of the esophagus from the dorsal region and the trachea from the ventral region of the foregut primordium occurs via a poorly understood compartmentalization process. Disruption of this process can result in severe birth defects, such as esophageal atresia and tracheoesphageal fistula (EA/TEF), in which the lumina of the trachea and esophagus remain connected. Here we summarize the signaling networks known to be necessary for regulating dorso-ventral patterning within the common foregut tube and cellular behaviors that may occur during normal foregut compartmentalization. We propose that dorso-ventral patterning serves to establish a lateral region of the foregut tube that is capable of undergoing specialized cellular rearrangements, culminating in compartmentalization. We review established as well as new rodent models that may be useful in addressing this hypothesis. Finally, we discuss new experimental models that could help elucidate the mechanism behind foregut compartmentalization. An integrated approach to future foregut morphogenesis research will allow for a better understanding of this complex process.

INTRODUCTION

During gestation, two endodermal tubes, the trachea and the esophagus, must be formed correctly for newborn infants to be able to breathe and eat. These two tubes arise during mid-gestation from a single developmental intermediate, the foregut endodermal tube. A complex set of morphogenetic events must occur for the proper separation of the esophagus and trachea. Over the last decade, our understanding of the genes necessary for proper foregut compartmentalization has advanced, but the basic cellular behaviors underlying this event are still unknown. In this review, we focus first on our current understanding of the morphogenesis of the foregut and the birth defects that can arise if foregut development goes awry. Next, we discuss regulation of dorsal-ventral patterning, midline cellular behaviors during compartmentalization, and finally, recent experimental tools that will allow further investigation into the cell biology of foregut compartmentalization.

OVERVIEW OF FORGUT COMPARTMENTALIZATION

The anterior foregut tube is initially derived from the embryonic endoderm, an epithelial sheet that folds ventrally to form a single endodermal tube around embryonic day 8 (E8.0) in the mouse. This tube is sandwiched between the neural tube (dorsal) and the developing heart loop (ventral). Once the tube is formed, the most dorsal cells resolve from the endodermal tube to establish the notochord (Aoto et al., 2009; Franklin et al., 2008; Tremblay and Zaret, 2005). The process of notochord resolution in mouse, which occurs between E8.25 and E9.5 (days of embryonic development since fertilization), has been described in detail (Jurand 1974). During development, the notochord acts as a signaling center that secretes signals important for the proper patterning of the foregut, neural tube and surrounding tissues (Chamberlain et al., 2008).

After notochord resolution the single common foregut tube compartmentalizes into the trachea and esophagus by an undefined set of cellular behaviors. The compartmentalization process begins around E9.5 with the formation of lung buds from the ventral foregut endoderm (FGE) at the level of the 6^th pharyngeal arch (Cardoso & Lü 2006). Starting at the site of lung bud formation, the common foregut primordium will then separate into two tubes, the esophagus on the dorsal side and the trachea on the ventral side (Fig 1A–C) (Zaw-Tun 1982; O’Rahilly & Muller 1984). Compartmentalization appears to occur at the dorso-vental midpoint of the lateral walls of the common foregut tube (Fig 1B″). For the purpose of this review we refer to the nexus of the dorsal and ventral regions of the lateral wall as the dorso-ventral midline (Fig 1B′). The compartmentalized region forms in a caudal to rostral direction extending to the future larynx. Once the esophageal and tracheal tubes are established, they differentiate into structurally and functionally distinct organs. By the time of birth, the esophageal epithelium consists of stratified squamous cells and the esophageal mesenchyme develops into concentric rings of smooth muscle (Yu et al. 2005), aiding in the passage of ingested materials to the stomach. On the other hand, the tracheal epithelium differentiates into a pseudostratified epithelium, and the mesenchyme develops C-shaped cartilage rings ventrally and the trachealis muscle dorsally (Perl et al. 2002; McAteer 1984), providing a conduit for air exchange with the lungs. Without proper patterning and differentiation, the esophagus and trachea will not function correctly upon birth. Much work has focused on the differentiation of the respiratory and digestive systems after compartmentalization but many questions remain as to how separation occurs.

Figure 1. The foregut compartmentalization between E10 and E11.5.

A–C) The anterior foregut compartmentalizes into the esophagus (marked in pink) and trachea (marked in green) between embryonic day 10.0 (28 somites) and E11.5. Prior to compartmentalization the foregut tube is a single cell layer thick, shown in panel A′. During compartmentalization the ventral foregut endoderm becomes pseudostratified (panel B′). The area we refer to as the dorso-ventral midline (ie. the lateral midpoint of the foregut wall) is marked with a square (panel B′). The endoderm then comes together at the dorso-ventral midline (panel B″) prior to resolving into two separate tubes (panel C′). It is important to remember that this process is occurring in 3-dimensions and that cellular rearrangements are occurring throughout the endoderm.

MODELS OF COMPARTMENTALIZATION

Three different models have been proposed to explain the division of the common foregut tube into the esophagus and trachea: 1) the outgrowth model (O’Rahilly and Muller, 1984; Zaw-Tun, 1982), 2) the watershed model (Sasaki et al., 2001a) and 3) the septation model (Fig 2) (Qi and Beasley, 2000; Sutliff and Hutchins, 1994) (Fig 2). The outgrowth model hypothesizes that the trachea grows out of the common foregut tube like an elongating bud (Fig 2, see arrow), while the common foregut tube per se differentiates into the esophagus (Zaw-Tun 1982; O’Rahilly & Muller 1984). This model is not supported by experimental results reported in the literature. For example, increased levels of proliferation specifically in the region of the ventral primordium at the point where compartmentalization begins are necessary correlates of this mechanism, but have not been detected (Ioannides et al. 2010, our unpublished observation). Additionally, this explanation is unlikely because the single undivided foregut tube contains two domains which express either respiratory or esophageal markers (Fig 1A) (Aubin et al. 1997).

Figure 2. Models of foregut compartmentalization.

Prior to compartmentalization the E10.0 undivided foregut has a dorsal (pink) and ventral (green) domain. There are three main models of foregut compartmentalization: Outgrowth model, Watershed model and the Septation model. The compartmentalization event starts at the level of the lung buds (marked with asterisk). The future location of this level of tissue is marked with asterisks in the models showing compartmentalization occurring. In the outgrowth and watershed models, arrows showing the direction of growth mark regions of outgrowth/proliferation. In the septation model an arrow demonstrates where the septation event is moving rostrally along the undivided foregut tube.

Alternatively, the watershed model suggests that both the trachea and esophagus are elongating while separated by a mesenchymal septum that blocks elongation of the dorso-ventral midline of the lateral wall (Fig 2) (Sasaki et al. 2001a). The watershed model proposes that the mesenchyme and epithelium are both active in separating the common foregut tube. This requires the existence of increased proliferation in the dorsal and ventral regions compared to the midline of the lateral walls (Fig 2, see arrows). As noted earlier, increased levels of proliferation in these regions have not been reported. Additionally, this model depends on the presence of a mesenchymal septum. It is possible that a mesenchymal septum does exist and moves rostrally as the compartmentalization event occurs. However, this hypothesis remains to be tested.

Both the watershed model and the outgrowth model suggest that the common undivided foregut tube does not change in length over the course of compartmentalization (Fig 2). To test this hypothesis, Ionnadies, Copp and colleagues (Ioannides et al. 2010) measured the length of the common foregut tube before and after compartmentalization. They found that, counter to predictions from these models, the common foregut tube actually decreases in length (Fig 2). This evidence suggests that the outgrowth and watershed models are incorrect, because these potential mechanisms of compartmentalization would result in no change in the length of the uncompartmentalized foregut tube.

Finally, the septation model, is based on the concept that “lateral edges” (Fig 1B″) occur at the dorso-ventral midline of the foregut tube (Fig 2) (Keith and Spicer, 1906; Qi and Beasley, 2000). These ridges are thought to be regions of the epithelium that thicken at the dorso-ventral midline, make contact across the lumen and fuse. The point of contact then moves rostrally to separate the common foregut tube into the esophagus and trachea. In this model, the epithelium actively fuses to separate the common foregut tube into two tubes (Fig 1C′, Fig 2). The septation model has been widely accepted in the field as the model most consistent with experimental evidence (Qi & Beasley 2000). However, based on three-dimensional reconstructions of wild type mouse foreguts and scanning electron microscopy of wild type chicken foreguts, there is little evidence that overt lateral ridges or a septum exist (Sasaki et al. 2001a; Metzger et al. 2011). Interestingly, Metzger et al. described the region of the epithelium where the trachea and esophagus are separating as a “saddle”. While there is no evidence of distinct lateral ridges or an epithelium septum, it is possible that there is an epithelial “saddle” which develops when the lung buds form and moves rostrally to separate the two tubes. This type of separation event would require apical constriction and collective migration of epithelial cells through the “saddle” region as it moves.

Without live imaging of the active compartmentalization event, it is unlikely that any model will be proven correct. The morphogenetic mechanism likely involves a combination of mesenchymal and epithelial cell movements and rearrangements that occur at the D/V midline, separating the common primordium into two tubes. Until there is a much deeper understanding of the cellular behaviors occurring during compartmentalization, the exact mechanism underlying foregut compartmentalization will remain a mystery.

HUMAN FOREGUT DEFECTS

Defects in the development of the respiratory and digestive systems are especially detrimental to survival. Consequently, the mortality rate of infants born with foregut abnormalities was 100% until the 1940s (Choudhury et al. 1999). However, with new surgical interventions the mortality rate is currently below 10%. In humans a range of foregut phenotypes can occur depending on the region of the foregut that is affected. Most common is the presence of an esophageal atresia (EA) with or without a tracheo-esophageal fistula (TEF). EA/TEF is characterized by an esophagus ending in a blind pouch and a fistula connecting the stomach to the trachea, usually around the level of the lungs. EA’s occur in about 1 in 3,500 live births and are classified into four types, labeled A–D (Torfs et al. 1995; Gross 1957). The most common of these is Type C (Fig 3B), which is the presence of both an EA and a TEF. These EA/TEFs can be corrected surgically, however, treatment requires multiple surgeries and typically results in a diminished quality of life for the patient (Kovesi & Rubin 2004).

Figure 3. Disruptions in foregut compartmentalization can result in a variety of defects of the esophagus and trachea.

A) The foregut normally separates into an esophagus (in pink) and a trachea with lungs budding off (in green). B) Esophageal atresia with tracheoesophageal fistula (EA/TEF) involves the esophagus ending in a blind pouch, and an open passage (fistula) connecting the trachea to the stomach. The fistula often contains both esophageal and tracheal characteristics (marked in yellow). This type of defect is found in several genetic mouse models and a pharmacological model (Adriamycin). While Shh mutants do have EA/TEF, their phenotype is often more complex and may be caused by different underlying mechanisms. C) Another common foregut defect is tracheal agenesis, in which the ventral foregut does not form a trachea but differentiates into an esophagus instead. This defect appears to be caused by a loss of ventral patterning, marked by Nkx2.1 expression. D) A more severe form of tracheal agenesis can occur when in addition to a loss of ventral fate, lung agenesis occurs as well. This defect has been found in mouse mutants where ventral WNT signaling has been completely removed. E) A final potential class of foregut defect is when compartmentalization appears to have stalled and the single primordium does not separate into two tubes. In these mutants dorso-ventral patterning remains intact but separation does not occur. While Ephrin-B2 and RAR mutants seem likely to have normal dorso-ventral patterning, the expression of Sox2 and Nkx2.1 have not been examined in these models.

A better understanding of the underlying causes of foregut defects may enable development of earlier and better treatments. To complicate matters, 48% of the time human foregut defects occur along with a host of other developmental abnormalities, making it difficult to determine the underlying cause of the phenotype (Holder et al. 1964). The most common of these are congenital heart defects and defects in other endodermal organ systems. For example, often patients diagnosed with VACTER/VACTERL syndrome, will have EA/TEFs along with either vertebral, anal, cardiac or renal defects. Many of these defects result from improper morphogenesis of an endodermal tissue, suggesting that there is a common developmental mechanism between the organ systems. VACTER/VACTERL syndrome is tentatively linked to changes in multiple genes including HOXD13, ZIC3, PTEN and FOX genes (Shaw-Smith, 2010). However, any role of these genes in the development of EA/TEF has yet to be determined. Studying common features of this group of defects instead of focusing on one organ system at a time may lead to more information about the development of EA/TEF and VACTER/VACTERL syndrome 

Interestingly, several patients with EA/TEF have deletions spanning the chromosomal region which contains NOGGIN (Puusepp et al., 2009). This finding in conjunction with Noggin mouse mutant studies suggested that mutations in NOGGIN could cause human foregut defects. However, when the coding region of NOGGIN was examined for point mutations in 50 patients with EA/TEF, only one patient had a mutation within the coding region, and this resulted in a predicted conservative amino acid change (Murphy et al., 2012). Such data suggest that mutations in the coding region of NOGGIN are not major factors in the molecular etiology of human EA/TEF. To determine if NOGGIN is involved human EA/TEF, the regulatory region of the gene will have to be examined further.

Mutations in other genes, including MYCN (van Bokhoven et al., 2005) and SOX2 (Williamson et al. 2006), are linked to the development of EA/TEF in human patients independently of VACTER/VATERL defects. While the links between these genes and patient pathology is not well studied or understood, Sox2 mutations in mice suggest that the loss of Sox2 causes foregut defects by disrupting dorsal patterning (Que et al., 2007). Using experimental mouse models to further investigate the roles of these genes may help us gain a better understanding of the morphogenetic mechanisms underlying this spectrum of defects.

RODENT MODELS OF FOREGUT COMPARTMENTALIZATION DEFECTS

Currently, our knowledge of foregut compartmentalization comes primarily from genetic manipulation in mouse targeted gene mutation models (Table 1). These mutants have been used to probe the molecular genetics and morphogenetic regulation of foregut compartmentalization, as well as the later differentiation of the tracheal and esophageal tubes. These genetic mouse models have also helped elucidate the underlying causes of murine foregut phenotypes (Table 1), and therefore are helpful in thinking about the pathogenesis of the corresponding human malformations. The two main etiologies that have been shown to result in foregut defects are notochord resolution defects (Fausett et al., 2014; Gillick et al., 2003; Mortell et al., 2004; Possoegel et al., 1999; Qi and Beasley, 1999) and dorsal-ventral patterning (D/V patterning) defects (Minoo et al. 1999; Que et al. 2007).

Table 1.

Table of genetic mouse models with foregut defects and the causative mechanism

Mouse Gene	Malformation	D/V patterning	Mechanism
Shh−/−	EA/TEF	established	Unknown
Gli2−/−;Gli3+/−	No E, T or lungs	unknown	Unknown
Nkx2.1−/−	TEF, small lungs	Loss of ventral Nkx2.1	Loss of D/V
Noggin−/−	EA/TEF	established	Notochord resolution
Sox2GFP/cond	EA/TEF	Moderate loss of sox2	Loss of D/V
EphrinB2LacZ/LacZ	TEF	unknown	Unknown
Foxf1+/−	TEF, narrow E, small lungs	unknown	Unknown
BMP4cond	TEF; tracheal agenesis	Loss of ventral Nkx2.1	Loss of D/V
Adriamycin Treated	EA/TEF	established	Notochord resolution
β-catenincond	TEF; tracheal agenesis	Loss of ventral Nkx2.1	Loss of D/V
β-cateninact	TEF	present	Unknown
Wnt2/2bcond	TEF; tracheal agenesis	Loss of ventral Nkx2.1	Loss of D/V
RAR mutants	TEF; tracheal agenesis	unknown	Unknown
Barx1−/−	TEF	present	Unknown

Improper notochord resolution can cause esophageal atresia and trachoesophageal fistulas

When defects in notochord resolution occur, the foregut typically displays EA/TEF (Fig 3B). This phenotype is thought to result from excess dorsal foregut endodermal tissue remaining attached to the notochord during resolution (Li et al. 2007; Fausett et al. 2014). The improper resolution event leaves insufficient dorsal foregut endoderm (dFGE) behind to form the esophagus during compartmentalization. EA/TEF has been found in mutant mice lacking Noggin, a BMP signaling inhibitor (Fausett et al., 2014; Li et al., 2007). The phenotype found in Noggin^−/− mice closely resembles the Type C type of human EA/TEF (Fausett et al., 2014; Li et al., 2007). Noggin mutants have notochord abnormities caused by defective notochord resolution prior to foregut compartmentalization.

Another model of EA/TEF is the Adriamycin rodent model. Adriamycin is an anthracycline antibiotic and chemotherapeutic. VACTERL-like phenotypes occur in a fraction of offspring when Adriamycin is injected into pregnant wild-type mice and rats, as long as the injection occurs prior to foregut compartmentalization (Fig 3B) (Ioannides et al., 2010; Qi and Beasley, 1999). This system has been extensively studied as a model for VACTERL defects. A recent study suggests that the notochord defects detected in embryos from Adriamycin treated dams were due to a delayed down-regulation of adhesion markers during notochord resolution, resulting in a loss of dorsal foregut endoderm (Hajduk et al., 2012). This defect in notochord resolution closely mirrors the defect found in Noggin mutants with EA/TEF (Fausett et al., 2014).

Proper dorso-ventral patterning is necessary for foregut compartmentalization

The second type of defect that leads to a failure of compartmentalization is loss of either dorsal or ventral patterning. The result is esophageal agenesis (Fig 3C), where the esophagus does not form when dorsal patterning is lost, or tracheal agenesis, where trachea does not form when ventral patterning is lost. This defect is hypothesized to be caused by a lack of differentiation of either the esophagus or trachea (Que et al. 2007; Minoo et al. 1999). Multiple mouse models (Fig 3C–D) exhibit this type of defect and we will discuss them in more detail below, in the section on dorso-ventral patterning.

While both of these classes of defect are medically relevant, neither of them appears to be the result of a disruption in the actual process of compartmentalization. As a result we currently know little about what drives separation of the foregut tube endoderm or what cellular behaviors are involved.

Undetermined causes of foregut compartmentalization defects

While the two previously described etiologies are well studied in mouse mutants, there is a third class of foregut defects that is predicted to exist, but has not yet been characterized. These are mouse mutant phenotypes in which the common foregut does not compartmentalize yet retains normal dorso-ventral patterning without notochord abnormalities (Fig 3E). There are currently three published mouse models that may have this phenotype but they have not been investigated thoroughly. These models are the ephrin-B2^LacZ/LacZ mouse (Dravis et al. 2004), the Barx1^−/− mouse (Woo et al. 2011) and mice with loss of retinoic acid receptor signaling (Fig 3E) (Mendelsohn et al. 1994; Luo et al. 1996; Desai et al. 2006). These models reflect the exciting possibility that their defects may stem from disruptions in the process of compartmentalization itself. Further investigation of these mice may provide an experimental system to uncover the mechanisms of the active compartmentalization process, as well as a new pathogenesis of compartmentalization defects.

DORSO-VENTRAL PATTERNING DURING FOREGUT DEVELOPMENT

A significant body of research exists on the necessity of dorso-ventral (D/V) patterning within the foregut endoderm. The dorsal foregut endoderm (dFGE) is marked by expression of the HMG-domain transcription regulator Sox2 (Que et al. 2007), whereas the ventral FGE is marked by Nkx2.1 (Minoo et al. 1999), another transcription regulator (Fig 4). Data suggest that both of these factors must be expressed in their proper domains for compartmentalization to occur (Minoo et al., 1999; Que et al., 2007).

Figure 4. A signaling network of specific genes is necessary to establish proper Nkx2.1 and Sox2 expression, leading in turn to either respiratory fate or esophageal fate.

The expression of Sox2 and Nkx2.1 are regulated by a complex set of mesenchymal (blue) and endodermal (dorsal marked by pink and ventral marked by green) signaling events involving BMP, Eph-ephrin, FGF and WNT signaling. When any steps in this pathway are disrupted, a loss of dorso-ventral patterning can occur, leading to foregut compartmentalization defects.

In the ventral domain, Nkx2.1 regulates multiple respiratory specific genes and is necessary for differentiation of trachea and lung specific cell types (Bohinski et al. 1994; Ray et al. 1996). In Nkx2.1 null mice, the foregut endoderm is dorsalized and Sox2 is expressed throughout the entire foregut tube (Fig 3C) (Minoo et al. 1999). Nkx2.1^−/− foreguts do not compartmentalize and resemble a human defect called tracheal agenesis, where the differentiated foregut is mainly esophageal in character with circumferential smooth muscle rings surrounding the foregut tube. In tracheal agenesis the foregut is a single tube that connects the oral cavity with the lungs and stomach.

While Nkx2.1 expression is necessary to establish the ventral foregut domain, Sox2 is necessary to establish the dorsal foregut domain. Until recently, little was known about the role of Sox2 in foregut compartmentalization because a complete deletion of Sox2 in mice results in death prior to gastrulation. However, a hypomorphic allele of Sox2 shows that a reduction in Sox2 levels does result in an EA/TEF phenotype 60% of the time (Fig 3B) (Que et al. 2007). Furthermore, conditional removal of Sox2 in the ventral endoderm results in EA/TEF in 10% of embryos and a shortened trachea in 60% of embryos (Que et al. 2007). While both Nkx2.1 and Sox2 are required for the development of an esophagus and trachea, it is unknown if they play a role in the compartmentalization event itself or if they are simply necessary to pattern the tissue domains to set up the cellular machinery of compartmentalization.

Multiple developmental pathways regulate dorso-ventral patterning

Multiple signaling pathways are responsible for the proper localization of Nkx2.1 and Sox2 within the foregut endoderm (Fig 4). Both Nkx2.1 and Sox2 must be properly localized for proper foregut development, so understanding the signaling network that regulates them could shed light on the early stages of compartmentalization.

The ventral establishment of Nkx2.1 depends on both BMP and WNT signaling from the surrounding mesenchyme and vFGE (Fig 4) (Goss et al. 2009; Harris-Johnson et al. 2009; Domyan et al. 2011; Li et al. 2008). A ligand for the BMP pathway, BMP4, is present in the ventral mesenchyme and signals via BMP receptors 1A and 1B (Bmpr1a;b), present in the ventral endoderm (Fig 4) (Domyan & Sun 2011; Rodriguez et al. 2010; Li et al. 2008). In the absence of mesenchymal BMP4 or endodermal Bmpr1a;b, Nkx2.1 protein expression is lost and tracheal agenesis occurs (Li et al. 2008; Domyan et al. 2011). When ventral Nkx2.1 expression is lost, dorsal Sox2 expression expands into the ventral domain resulting in an esophageal fate (Fig 3C). Interestingly, in mice with Bmpr1a;b conditionally removed from the ventral endoderm, a reduction in Sox2 can rescue both Nkx2.1 expression and the tracheal agenesis phenotype (Domyan et al. 2011). These data suggest that BMP signaling is necessary for repressing ventral Sox2 expression, not for inducing ventral Nkx2.1 expression (Fig 4).

Defining the role of BMP signaling in the foregut is complicated by the fact that secreted BMP agonists are present in and around the foregut endoderm. Noggin, a secreted BMP agonist, is expressed in the dorsal foregut and the notochord (Fig 4), suppressing BMP signaling in the dorsal foregut region (Stottmann et al. 2001); reducing the genetic dosage of BMP4 (by half) (Que et al., 2007) or of BMP7 (entirely) (Li et al., 2007) rescues the Noggin mutant compartmentalization defect completely. Repression of dorsal BMP signaling by Noggin allows Sox2 to be expressed in the dorsal region (Fig 4). In Noggin^−/− mice, the foregut has normal dorso-ventral patterning, suggesting that inhibition of BMP signaling is not necessary for establishment or maintenance of D/V patterning (Fausett et al. 2014). Instead, research suggests that Noggin is necessary for proper resolution of the notochord from the foregut endoderm prior to compartmentalization (Li et al. 2007; Fausett et al. 2014).

While BMP signaling is important in repressing ventral Sox2, WNT signaling is necessary for establishing ventral Nkx2.1 (Fig 4). The WNT ligands important in foregut compartmentalization are Wnt2 and Wnt2b, which are both found in the ventral foregut mesenchyme (Fig 4) (Goss et al. 2009). When both Wnt2/2b are removed from the ventral mesenchyme, endodermal Nkx2.1 is not established and respiratory agenesis occurs (Fig 3D) (Goss et al. 2009). In addition, when β-Catenin (also known as Ctnnb1), encoding the downstream target of canonical WNT signaling, is conditionally removed from the foregut endoderm, the result is a phenocopy of the Wnt2/2b mutant mouse (Fig 3C) (Harris-Johnson et al. 2009). WNT signaling is also sufficient to establish Nkx2.1 expression in certain contexts. Conditional activation of β-Catenin in the ventral foregut and anterior stomach endoderm results in ectopic Nkx2.1 expression in the stomach (Fig 3E) (Harris-Johnson et al. 2009; Goss et al. 2009). However, Domyan and colleagues (Domyan et al. 2011) activated β-Catenin in the endoderm of mice without endodermal Bmpr1a;b and found that activated β-Catenin is not sufficient to rescue Nkx2.1 expression when BMP signaling is not present (Fig 4). Therefore, BMP signaling is not acting downstream of β-Catenin in the ventral foregut. Furthermore, when β-Catenin was conditionally removed from the ventral foregut endoderm there was no change in BMP signaling levels (Domyan et al. 2011). These two sets of data support a model in which BMP signaling is functioning independently of WNT signaling in establishing ventral foregut identity (Fig 4).

While the establishment of the ventral endoderm identity has been well studied, it is less clear how Sox2 expression in the dorsal FGE is established and regulated. It appears that Sox2 is the default state of the early foregut endoderm, with all cells expressing Sox2 at E8.5. Although Sox2 seems to be turned on throughout the epithelium by default, multiple pathways repress Sox2 in the ventral FGE (Fig 4). This repression results in a gradient of Sox2 expression with Sox2-Hi cells located in the dorsal FGE and Sox2-Lo cells found in part of the ventral FGE. Sox2 repression and Nkx2.1 activation, results in the establishment of a very clear Sox2-Nkx2.1 pattern by E9.5. As was previously discussed, active endodermal BMP signaling represses Sox2 in the ventral FGE (Domyan et al. 2011).

In addition to BMP signaling, fibroblast growth factor (FGF) signaling is active in the ventral foregut mesenchyme. For example, the FGF ligand Fgf10 is present in the ventral foregut mesenchyme (Fig 4) (Min et al. 1998). When it is genetically ablated, the foregut separates normally, but lung development is impaired (Min et al. 1998). Even though loss of Fgf10 does not result in compartmentalization defects, in foreguts cultured ex vivo, FGF10 is able to inhibit Sox2 expression (Que et al. 2007). This suggests that Fgf10 may play a role in repressing ventral Sox2 expression.

An especially interesting aspect of Sox2/Nkx2.1 regulation is the reciprocal inhibition that occurs between Sox2 and Nkx2.1 (Fig 4). In stem cells, Sox2 has been shown to bind the promoter region of Nkx2.1 and inhibit its transcription (Boyer et al. 2005). Furthermore, genetic manipulations have suggested that Sox2 and Nkx2.1 mutually repress each other to maintain dorso-ventral patterning (Que et al. 2007; Minoo et al. 1999; Harris-Johnson et al. 2009; Goss et al. 2009; Domyan et al. 2011). This mutual repression results in the most dorsal and most ventral endoderm expressing only Sox2 or Nkx2.1. Intriguingly, at the dorso-ventral midline region, where compartmentalization occurs, there are cells that express both transcription factors at lower levels (Fig 1B′). The importance of this double positive region of cells has not been investigated, but we speculate they may play a role during compartmentalization.

The role of Sox2 and Nkx2.1 in establishing a differentiation program of either esophageal or tracheal fate is well understood. However, if their role were as simple as controlling differentiation then embryos which have both dorsal Sox2 and ventral Nkx2.1 expression, such as crosses with a ventral activation of β-Catenin (Goss et al. 2009; Harris-Johnson et al. 2009), should in theory compartmentalize normally. It is likely that the boundary where Sox2 and Nkx2.1 meet is necessary for maintaining dorso-ventral midline gene expression of important positional cues, which in turn regulate specific cellular behaviors. However, very few genes are known to be expressed specifically in the midline region, either in the epithelium or in the mesenchyme. More work will need to be done to determine if the dorso-ventral boundary is playing an active role during foregut compartmentalization.

NEW GENETIC MODELS OF FOREGUT COMPARTMENTALIZATION DEFECTS

Reverse Eph/ephrin signaling during foregut development

A recent study has shown that Eph-ephrin signaling appears to be active specifically at the D/V midline of the foregut (Fig 4) (Dravis & Henkemeyer 2011). Eph-ephrin signaling is a complex receptor tyrosine kinase signaling pathway that is often active in the development of boundaries during morphogenesis (Himanen et al. 1998). This signaling pathway has historically been studied largely in the development of the brain and nervous system, but recently has been implicated in multiple compartmentalization events, including uro-genital compartmentalization, palate shelf closure, body-wall closure and foregut compartmentalization (Dravis et al. 2004; Dravis & Henkemeyer 2011). The Eph-ephrin pathway is comprised of 13 members, which are separated into two classes: an A-subclass and a B-subclass (Toth et al. 2001; Himanen et al. 1998). These two classes are distinguished by the structure of their ligands. The A-subclass ligands are extracellular and attached to the cell membrane by a glycosylphophatidylinositol anchor. The B-subclass ligands span the membrane and have a cytoplasmic tail, giving the ligand the ability to reverse signal to their own cell when attached to a receptor (Holland et al. 1996; Brückner et al. 1997). For a more in-depth review of Eph-ephrin signaling biology see Kullander and Klein 2002 (Kullander and Klein, 2002).

The various compartmentalization defects all occurred when ephrin-B2 reverse signaling was specifically disrupted by replacing the cytoplasmic signaling domain with a LacZ domain (Dravis et al. 2004). As a result, foregut defects occurred in ephrin-B2^LacZ/LacZ embryos at an incidence rate of 47% of null embryos (Dravis & Henkemeyer 2011). Ephrin-B2 binds to B-subclass ligands (EphB1-4, EphB6), but only EphB2 and EphB3 are present in the foregut at the time of compartmentalization (Kretovich unpublished observation). The foregut expression pattern of EphB3 is especially interesting because it is found specifically at the foregut midline prior to and during compartmentalization (Dravis & Henkemeyer 2011). Because a loss of reverse signaling results in a loss of foregut compartmentalization, the localization of the active ligand is crucial to determining the role of Eph-ephrin signaling in the foregut. Active ephrin-B ligand (ephrin-B1-3) is found throughout the foregut epithelium and the mesenchyme at the D/V midline via staining with a phospho-antibody (Dravis & Henkemeyer 2011). These data together suggest that Eph-ephrin signaling is active at the midline region during compartmentalization and is at least partially necessary for compartmentalization. Furthermore, Eph-ephrin signaling is an interesting candidate to regulate key cell behaviors at the dorso-ventral midline because in other developmental contexts it regulates processes such as boundary formation, guided migration and cytoskeletal dynamics, which could be important for foregut compartmentalization (Kullander & Klein 2002). Unfortunately, it is unknown whether D/V patterning is established in ephrin-B2 mutants. If so then this would be a very attractive model to directly examine the cellular behaviors involved in foregut compartmentalization.

Barx1 signaling during foregut development

Another recently published mouse model with foregut compartmentalization defects is the Barx1 null mouse (Woo et al. 2011). Embryos lacking Barx1, an indirect WNT inhibitor, appear to have almost normal dorso-ventral patterning, based on Sox2 and Nkx2.1 expression, yet the foregut fails to compartmentalize (Woo et al. 2011). During stomach and foregut development, Barx1 is found in the mesenchyme underlying the endoderm (Kim et al. 2005). It may be required in this context for the expression of secreted WNT agonists, secreted frizzled-related proteins (sFRPs) 1 and 2, which block WNT signaling within the overlying endoderm (Woo et al. 2011; Kim et al. 2005). In the foregut, Barx1 is found specifically in the dorsal mesenchyme (Fig 4) and Barx1^−/− embryos appear to have a similar TEF foregut defect to embryos with conditionally activated β-Catenin (Fig 3C) (Harris-Johnson et al. 2009; Goss et al. 2009). While Barx1 is not expressed specifically at the D/V midline, it is one of the only known models that display almost normal dorso-ventral patterning and TEF. Because Barx1 mutants appear to retain expression of Sox2 and Nkx2.1, they may be a good model to investigate the cellular behaviors necessary for the actual compartmentalization event, rather than the preceding patterning. Further examination of the cellular mechanism disrupted in Barx1^−/− mice is necessary to gain a better understanding of the role of Barx1 during foregut compartmentalization.

CELLULAR BEHAVIORS DURING COMPARTMENTALIZATION

Programmed cell death

Programmed cell death has long been thought to be necessary for foregut compartmentalization. Excess cell death is present at the point of compartmentalization, and reduced levels of cell death occur in Adriamycin treated mice with EA/TEF (Orford et al. 2001; Qi & Beasley 2000; Williams et al. 2000; Zhou et al. 1999; Ioannides et al. 2010). To determine the exact role of cell death during compartmentalization, Ioannides et al. 2010 (Ioannides et al. 2010) quantified the percentage of dying cells present during the process of foregut compartmentalization. In fact, there is more cell death present (up to 35% of epithelial cells undergoing PCD) at the D/V boundary than the dorsal or ventral domains of the foregut in the common foregut tube at the point of compartmentalization (Ioannides et al. 2010). However, the foregut develops normally in Apaf1^−/− mice, which lack programmed cell death (Ioannides et al. 2010). If the foregut epithelium were undergoing a fusion event, as suggested by the septation model, it would be expected that excess cell death would be occurring at the point of fusion. The cells would need to eliminated, but would not be required for the dorso-ventral midline fusion event itself. This suggests that the excess apoptosis occurring at the point of compartmentalization is likely a byproduct of the separation event, and is not necessary for compartmentalization to occur.

Proliferation

Both the outgrowth model and the watershed model hypothesize that the foregut epithelium is growing rapidly to form the esophagus and trachea (O’Rahilly & Muller 1984; Zaw-Tun 1982). Higher levels of proliferation found specifically in the growing trachea or esophagus would support the outgrowth and/or watershed models, depending on the localization of the dividing cells. However, no research has shown a presence of excess proliferation in the dorsal or ventral regions of the epithelium, or at the D/V midline (Ioannides et al., 2010). While many organs do bud and grow out from the foregut endoderm during development (Slack 1995; Fagman et al. 2006; Tremblay 2011), this does not seem to be the case with the development of the trachea and esophagus.

Cellular rearrangements during foregut development

Many rearrangements of epithelial cells including polarity, actin dynamics, and differential adhesion play a substantial role in embryonic development. These behaviors can cause epithelial mesenchymal plasticity (EMP) in epithelia undergoing a collective migration or major rearrangement where normal epithelial connections hinder cell movements. After the cells in an EMP state rearrange, they can return to their polarized, adhered epithelial state to reestablish the epithelium. A good example is mammary gland morphogenesis, where epithelial polarity is disrupted to allow for necessary cellular rearrangements (Ewald et al. 2008). The mammary gland cells partially downregulate their cell-cell adhesion as shown by a decrease in tight junctions and β-Catenin localization (Ewald et al. 2008). However, the epithelium does not become truly mesenchymal, as it retains basement membrane and luminal markers.

In the case of the septation model of foregut compartmentalization, cells in the foregut “saddle” region could be entering an EMP-like state at the point of dorso-ventral midline fusion. During compartmentalization, the dorso-ventral midline of the foregut epithelium has increased WT1 expression (Que et al., 2007), a transcription factor involved in EMT(Miller-Hodges and Hohenstein, 2012), and decreased p63 expression, an inhibitor of EMT(Lindsay et al., 2011; Que et al., 2007). These data suggest that EMP may be occurring in the foregut dorso-ventral midline epithelium during compartmentalization. If the hypothesis of a “saddle” (Metzger et al. 2011) moving rostrally to separate the common foregut is correct, than the epithelium would decrease cell adhesion allowing cells to move through the “saddle” region, before reforming into the esophagus and trachea. To determine if these events are occurring, high resolution imaging of the foregut compartmentalization event would be necessary to detect relatively small decreases in adhesion markers.

Additionally, changes in epithelial actin dynamics may play a role in compartmentalization, similar to their role in neural tube closure and lung budding (Kim et al. 2013; Brouns et al. 2000). The lateral epithelia of the neural tube develops actin rich filopodia that reach out across the luminal space and may help bring the two sides together (Brouns et al. 2000). These filopodia structures are not usually made by normal epithelia, but are commonly found on migrating cells. We have found that actin-rich cellular projections, filopodia-like structures, do appear to be present at the point of compartmentalization (Fig 5A–C). In addition to filopodia, actin could be involved in increased apical constriction along the dorsal and ventral sides of the foregut, forcing the epithelium together at the D/V midline. This constriction event would be a more complex version of the apical constriction that occurs during lung budding (Kim et al. 2013). It is plausible that actin constriction is necessary to push the epithelium together at the dorso-ventral midline, where a subsequent epithelial fusion event occurs. However, if this is the case, there should be lateral ridges present when examined by SEM (Metzger et al. 2011). Further work will need to be done to determine if major actin rearrangements, such as constriction or filopodia, are important for foregut compartmentalization.

Figure 5. Actin-rich protrusions are present at the point of foregut compartmentalization.

At E10.5 (32s) the foregut is in the process of compartmentalizing and has apical actin present throughout the epithelium, as visualized by Phalloidin staining. A) The epithelium has a layer of apical actin in the uncompartmentalized foregut but does not have obvious actin protrusions. B) Where the epithelium is about to undergo the compartmentalization event, a large amount of apical actin is present and when examined at a higher magnification (Panel D), actin-rich protrusions are present at the D/V midline (white arrow). C) After compartmentalization apical actin is still present in the esophagus and trachea. Scale Bar=.1 mm

NEW EXPERIMENTAL SYSTEMS TO STUDY FOREGUT COMPARTMENTALIZATION

Traditionally, approaches of genetic ablations and visualization using antibodies have proved to be unsuccessful at determining the cellular behaviors occurring during foregut compartmentalization. Recently, computer modeling of lung bud formation predicted that apical constriction is sufficient to initialize lung budding in chicken development (Kim et al. 2013). This work is an example of how computer modeling could be used to answer many of the questions that still exist about foregut compartmentalization. As previously discussed, there are multiple cellular behaviors whose necessity during compartmentalization is not easily addressed experimentally. During foregut compartmentalization, there is a large amount of apical actin and potential constriction events occurring but without live imaging or modeling we cannot assess the role of apical constriction. Furthermore, the mesenchyme is likely playing an active role by pushing on the epithelium to shape the foregut. Without a way to model this behavior, it is almost impossible to assess the morphogenetic movements with the genetic tools in the field. To take full advantage of the computer modeling system, it will be necessary to use organ culture to collect temporal data on developing foreguts. Fortunately, the mouse foregut is very amenable to culturing ex vivo. The uncompartmentalized foregut will develop normally in culture for up to 72 hours and separate into an esophagus and trachea (our unpublished observations). In addition, it is possible to label relevant tissue domains with GFP by using genetically modified mice. Such labeling should allow the tracking of movements of cells over time. In the future, utilizing computer modeling, in conjunction with ex vivo culture, will aid in the advancement in our understanding of the cellular behaviors occurring during foregut compartmentalization.

Another embryological model system that should be useful for studying foregut compartmentalization is chicken (Metzger et al. 2011). This is often overlooked because there are numerous genetic mouse models with foregut defects. Yet we have found that the chicken foregut develops in a very similar manner to that of the mouse, starting as a single endodermal tube and separating into an esophagus and trachea (Fig 6A–B). Many of the same signaling pathway genes involved in early mouse foregut development are conserved and similarly expressed, including components and targets of FGF, BMP, and Shh signaling (Morrisey & Hogan 2010; Metzger & Krasnow 1999; Affolter et al. 2003). In addition, it is likely that similar physical mechanisms are working to shape the developing epithelium (Morrisey & Hogan 2010; Davies 2002). To examine the normal cellular behaviors occurring during foregut development, the chicken may be a superior model to the mouse. For example, in ovo manipulations of chicken embryos are substantially more facile, either by electroporation (Fig 6C) or treatment with pharmaceuticals, than live manipulations of mouse embryos. The chicken model allows for more directed alterations of wild type behaviors by injecting plasmids or drugs straight into the foregut tube and allowing the embryo to develop in ovo (Fig 6C).

Figure 6. Chicken as a model system for studying foregut development.

A) At stage 20 in chicken the foregut starts as a single tube (FG) (A′) with lung buds (LB) (A″) starting to develop. B) By stage 25 the chicken foregut has compartmentalized (B′) into an esophagus (E) and trachea (T). Caudal to the separated esophagus and trachea, (B″) the trachea ends in two lung buds. C–D) The chicken is a system that is amenable to electroporation of plasmids directly into the foregut lumen of an embryo in ova prior to foregut compartmentalization. The embryo can then develop in ova until after foregut compartmentalization. The electroporation can be targeted to either the dorsal or ventral side of the foregut, visualized by inclusion of a marker such as GFP. C) GFP targeted to the ventral domain of the foregut. The electroporation occurred prior to compartmentalization and the foregut was dissected after compartmentalization had occurred (stage 25). ((A–C. Scale Bar= .1mm) D–E) Treatment of chicken embryos in ova with Adriamycin at stage 11 results in caudal regression (white arrows compare posterior structures), a loss of the posterior structures of the embryo. Caudal regression is seen in Noggin mutants and Adriamycin treated embryos. (D–E. Scale Bar = 1mm)

Furthermore, the chicken has been published as a model of Adriamycin induced foregut defects (Naito et al., 2009). This was accomplished by injecting Adriamycin into a developing egg prior to compartmentalization and allowing the chicken to develop. In our own hands this approach resulted in caudal regression, a loss of posterior structures, which is often associated with VACTERL syndrome (Fig 6D–E). However, in our hands, such embryos do not display EA/TEF, which suggests this may not be the most sensitive phenotype to Adriamycin exposure. Nonetheless, this example establishes a method for potentially altering many different signaling pathways and cellular behaviors by drug treatment in ova. To do these same experiments in mouse, the foregut must by cultured ex vivo leading to confounding factors from the loss of surrounding structures and signals. More research will need to be done to confirm that chicken foreguts have a D/V patterning network similar to that in mice. Once that point is established, electroporation can be used to disrupt dorso-ventral specific genes identified from mouse studies to examine which cellular behaviors are important. Furthermore, live imaging of foregut compartmentalization could be accomplished in chicken embryos. The uncompartmentalized foregut endoderm can be labeled using DiI and the labeled cells tracked through compartmentalization. By tracking the cells from multiple regions of the foregut, we could determine what types of movements are occurring during compartmentalization.

To summarize, using computer modeling to analyze the validity of certain cellular movements would help narrow the focus to cellular behaviors that are predicted to occur. Furthermore, by using techniques available in the chicken system, data about cellular movements during foregut compartmentalization could be collected and modeled.

FUTURE DIRECTIONS

Rodent genetic and pharmacologic models have made major contributions to our understanding of important genetic effectors and processes required for foregut compartmentalization. The necessity of proper notochord formation and the regulation of dorso-ventral patterning markers, Sox2 and Nkx2.1, are now clearly established, and several genes involved in these processes are known. However, a big unanswered question remains: How does the foregut compartmentalize? The separation of the common foregut tube into the esophagus and trachea clearly involves a complex set of signaling networks that regulate patterning and the cellular rearrangements in both the epithelium and the mesenchyme. Specifically we need to know the extent to which cells are moving or rearranging at the point of fusion, and if dorso-ventral patterning regulates these behaviors. To accomplish this, the foregut epithelium of mouse mutant models with and without changes in dorso-ventral patterning needs to be examined carefully for changes in cell shape, actin localization and polarity in relation to the site of compartmentalization. Furthermore, using a drug treatment approach to disrupt key cellular behaviors, such as cytoskeletal rearrangement, could shed light on the mechanics behind foregut compartmentalization. This is important not just for understanding foregut compartmentalization but also for increasing our understanding of other complex epithelial morphogenesis events. Moving forward, the field needs to use a multifaceted and multidisciplinary approach by taking advantage of existing genetic models in combination with computer modeling, organ culture and basic embryological and cell biological techniques. With increased experimental power we will be able to better define the mechanism by which the foregut compartmentalizes and the role of certain cellular behaviors in the process.