|Home | About | Journals | Submit | Contact Us | Français|
Epithelial tubular organs are essential for life in higher organisms and include the pancreas and other secretory organs that function as biological factories for the synthesis and delivery of secreted enzymes, hormones and nutrients essential for tissue homeostasis and viability. The lungs, which are necessary for gas exchange, vocalization and maintaining blood pH, are organized as highly branched tubular epithelia. Tubular organs include arteries, veins and lymphatics, high-speed passageways for delivery and uptake of nutrients, liquids, gases and immune cells. The kidneys and components of the reproductive system are also epithelial tubes. Both the heart and central nervous system of many vertebrates begin as epithelial tubes. Thus, it is not surprising that defects in tube formation and maintenance underlie many human diseases. Accordingly, a thorough understanding how tubes form and are maintained is essential to developing better diagnostics and therapeutics. Among the best-characterized tubular organs are the Drosophila salivary gland and trachea, organs whose relative simplicity have allowed for in depth analysis of gene function, yielding key mechanistic insight into tube initiation, remodeling and maintenance. Here, we review our current understanding of salivary gland and trachea formation – highlighting recent discoveries into how these organs attain their final form and function.
Model organisms, such as Drosophila, have proven to be excellent for identifying the underlying molecules and mechanisms of many aspects of development and disease. Understanding the cell biology of epithelial tube formation has been no exception. As with all multicellular organisms, Drosophila contain a variety of tubular organs, many of which form during the first several hours of development, including the salivary gland, trachea, Malpighian tubules and others. Many of the developmental mechanisms that Drosophila employs during tube formation are conserved from worms to humans (Andrew and Ewald, 2010). Therefore, a more complete understanding of the cell biology of epithelial tube formation is relevant across many fields.
Progress toward unraveling the molecular and cellular details of Drosophila tube formation is facilitated by the unsurpassed genetic tools available for this organism. Mechanisms exist for easily creating loss-of-function and gain-of-function mutations as well as for affecting gene function in a tissue- or cell-type specific manner (Brand and Perrimon, 1993; Venken and Bellen, 2005; Maggert et al., 2008). Tools exist for genetically marking entire organs, parts of organs or even individual cells. Protein tags can be directed to any membrane or organelle within the cell, facilitating studies of cell rearrangement, cell shape change and cell migration. Live imaging of tubulogenesis is possible due to the relatively translucent Drosophila embryo and the timeline of tube formation. Tubes that lie close to the surface can easily be visualized in real time with standard confocal imaging (Ribeiro et al., 2002; Chihara et al., 2003; Kato et al., 2004; Ribeiro et al., 2004) and problems resulting from challenging deep-tissue imaging have been overcome using two-photon confocal microscopy (Cheshire et al., 2008).
The simplest of Drosophila tubes is the secretory portion of the salivary gland (SG) (Bradley et al., 2003). The SG secretory tubes form from two primordia of ~100 polarized epithelial cells found in the most posterior region of the head and flanking the ventral midline (Figure 1). Neither cell death nor cell division occurs once the SG primordia have been specified. Thus, only changes in cell size, shape, adhesion and position transform the two plates – or placodes - of polarized epithelial cells on the embryo surface into simple un-branched cigar-shaped epithelial tubes positioned deep within the embryo. These elongated epithelial tubes connect to the larval mouth through the salivary ducts, which arise from epithelial cells positioned immediately ventral to the secretory primordia (Kuo et al., 1996; Jones et al., 1998; Haberman et al., 2003; Kerman et al., 2008).
The trachea (TR) is a somewhat more complicated branched tubular network that facilitates oxygen exchange through simple diffusion of air from the outside to every cell in the animal (Manning and Krasnow, 1993). Despite its ultimate relative complexity, tracheal tube formation utilizes many of the same cellular mechanisms required to form the salivary gland. Tracheal tubes arise from 20 separate primordia, which form ten placodes on each side of the embryo along the lateral surfaces of the thorax and abdomen (Figure 1). The ~40 cells within each tracheal placode undergo one final round of cell division during invagination, resulting in approximately 80 cells per tracheal segment, also known as a tracheal metamere. Once internalized, subsets of cells within each metamere migrate in stereotypical directions to form distinct branches. Specialized tracheal cells – fusion cells – from adjacent and contralateral segments ultimately fuse to form tubular connections between different metameres giving rise to the final interconnected tubular tracheal network. Subcellular tubes extend from the terminal cells at or near the ends of each branch to contact target tissues throughout the animal.
Also among the better-characterized tubular organs in Drosophila are the four elongated Malpighian tubules (MT) of the Drosophila excretory system (Jung et al., 2005; Denholm and Skaer, 2009; Beyenbach et al., 2010). MT primordia evaginate from the hindgut to form rudimentary MT tubes that elongate using a variety of mechanisms, including cell division, cell enlargement, cell shape change, cell rearrangement and cell recruitment (Skaer, 1996; Denholm et al., 2003; Campbell et al., 2009; Campbell et al., 2010). Not surprisingly, several of the signaling pathways and downstream effector genes required for SG and tracheal development also function in MT development, supporting the idea of common molecular programs underlying tube morphogenesis (Blake et al., 1998; Blake et al., 1999; Jack and Myette, 1999; Liu et al., 1999). Other, less-well characterized Drosophila tubular organs include the embryonic hindgut, an excellent system for characterizing genes affecting left-right polarity, the wing veins, which form as the dorsal and ventral surfaces of the wing contact and seal, providing an excellent system for discovering novel integrin signaling pathway components, the dorsal appendages of the mature ovary, the dorsal vessel and others. The reader is referred to the following reviews for more information regarding our current understanding of how the MT and other Drosophila tubes form and attain their final geometries (Bier, 2000; Lengyel and Iwaki, 2002; De Celis and Diaz-Benjumea, 2003; Crozatier et al., 2004; Blair, 2007; Tao and Schulz, 2007; Berg, 2008; Denholm and Skaer, 2009; Beyenbach et al., 2010).
Global patterning genes determine the location and number of cells that form the salivary gland (Figure 1; Panzer et al., 1992; Andrew et al., 1994; Kuo et al., 1996; Henderson et al., 1999; Henderson and Andrew, 2000). Salivary gland (SG) formation requires the homeotic gene Sex combs reduced (Scr), which is initially expressed in a limited anterior-posterior domain of the embryo known as parasegment two (PS2) (Panzer et al., 1992; Andrew et al., 1994). Scr works with two more globally expressed cofactors, Extradentical (Exd) and Homothorax (Hth), to induce expression of a set of early-expressed SG transcription factors (Henderson and Andrew, 2000). In the absence of Scr (zygotic loss), exd (maternal and zygotic loss) and hth (zygotic loss), SGs fail to form, and when Scr is expressed globally, SGs form in two additional segments (PS0 and PS1). Scr fails to induce SG fates in more posterior segments because of two negatively-acting factors; Teashirt (Tsh) – a zinc finger protein – prevents SG formation in PS3-13, whereas Abdominal A (AbdA) – another homeotic protein – prevents SG formation in PS14 (Andrew et al., 1994). Spatial limits on SG formation are also provided by dorsal-ventral patterning genes (Panzer et al., 1992; Henderson et al., 1999). Dpp signaling (TGF-β pathway) provides the dorsal limit on SG formation, whereas ventral midline activation of EGF signaling distinguishes the salivary gland duct primordium from the more lateral secretory primordia (Kuo et al., 1996; Henderson et al., 1999; Haberman et al., 2003).
Expression of the genes that specify the SG disappears shortly after morphogenesis begins (Henderson and Andrew, 2000). However, expression of many early SG transcription factors initially induced by Scr/Exd/Hth continues through larval life. The continued expression of early transcription factor genes is mediated both by auto- and cross-regulation, with the Drosophila FoxA transcription factor Fork head (Fkh) playing a major role (Figure 2; Zhou et al., 2001; Chandrasekaran and Beckendorf, 2003; Abrams and Andrew, 2005; Abrams et al., 2006; Maruyama et al., 2011). Fkh and the other early expressed transcription factors also orchestrate SG morphogenesis and the specialization of the SG as a secretory organ (Myat and Andrew, 2000a; Myat and Andrew, 2000b; Myat and Andrew, 2002; Abrams and Andrew, 2005; Abrams et al., 2006; Fox et al., 2010).
Within four hours, the SG primordia change geometry from two mono-layered plates of tightly adherent epithelial cells on the embryo surface to two elongated, fully internalized secretory tubes (Figure 3). The first step is a thickening of the SG primordia into SG placodes, a process mediated by the epithelial cells changing shape from cuboidal to columnar. As primordia, the SG cells are in the same orientation as all surface ectoderm; their apical surfaces face out toward the extraembryonic membranes and their basal surfaces face in, contacting the underlying mesoderm. Shortly after the placodes form, cells in a dorsal-posterior position of the primordia undergo another shape change – apical constriction, a process whereby the apical domain constricts to create pyramidal shaped cells driving tube internalization (Myat and Andrew, 2000a). The cells that invaginate first form the most distal portion of the mature SG tubes. Shortly after the first cells are internalized, cells in a dorsal anterior domain also undergo apical constriction and push inward creating a slight anterior bulge in the ingressing tubes. Finally, the remaining SG cells internalize, most likely through a wrapping-type mechanism, wherein the primordia fold inward to form a trough-like structure that eventually seals along the two edges to form a tube (Chung and Andrew, 2008).
As the SG cells continue to internalize, the tubes are pushed further into the embryo. Once the distal cells of the SG tubes contact a layer of dorsally-positioned mesodermal cells, the visceral mesoderm, the cells turn posteriorly and actively migrate along this tissue to arrive at their final position in the embryo, with the elongated tubes aligned along the anterior-posterior body axis (Bradley et al., 2001; Bradley et al., 2003; Chung and Andrew, 2008). The journey of the SGs to their final position involves contact with multiple distinct tissues with each tissue providing either guidance cues and/or functioning as a substrate for directed migration (Bradley et al., 2001; Bradley et al., 2003; Kolesnikov and Beckendorf, 2005; Vining et al., 2005; Harris and Beckendorf, 2007; Chung and Andrew, 2008).
Invagination of the SG primordia involves nuclei moving to the basal side of the cell and the apical domain constricting, a mechanism utilized in many other tissues including the mammalian optic cup (Hilfer, 1983) and the neural tubes of many vertebrates (Schoenwolf and Smith, 1990). Loss of fkh results in a complete failure of apical constriction and SG internalization; fkh loss does not, however, affect basal nuclear movement, indicating that these processes can be uncoupled (Myat and Andrew, 2000b). Knowing how Fkh is linked to apical constriction awaits the identification and characterization of the relevant transcriptional targets, several candidates of which are currently under study (Maruyama and Andrew, unpubl.). Nonetheless, studies have implicated the small GTPase Rho, its regulators and its downstream effectors in the process. Mutations in 18-wheeler, which encodes a Toll-like receptor, folded gastrulation (fog), which encodes a G-protein coupled receptor ligand, RhoGEF2, as well as two Rho-GAPs, lead to either delays and/or partial failure of SG internalization (Nikolaidou and Barrett, 2004; Kolesnikov and Beckendorf, 2007). Studies of Rho mutants using both loss-of-function alleles as well as SG-specific expression of a Rho dominant-negative suggest two mechanisms for Rho action (Xu et al., 2008): (1) Rho affecting levels of the apical membrane protein Crumbs (Crb), which has been implicated in cytoskeletal reorganization during apical constriction in the invaginating trachea (Letizia et al., 2011) and (2) Rho acting through Rho kinase (ROK) to affect apical constriction, presumably through its affects on myosin contractility. The finding that most SG cells eventually internalize in all the above-mentioned mutant backgrounds, whereas the SGs completely fail to invaginate in fkh mutants, argues for the existence of additional Fkh-dependent events driving this process.
Once internalized, the SG elongates along its proximal-distal axis. Loss-of-function mutations in the transcription factors encoded by huckebein (hkb) or ribbon (rib) have profound effects on this process. hkb mutant SGs nearly completely fail to elongate, giving rise to short puck-shaped SGs with very limited apical membrane domains (Myat and Andrew, 2000b; Myat and Andrew, 2002). rib mutant SGs elongate more slowly and ultimately achieve only 60% the lumenal length of WT glands (Bradley and Andrew, 2001; Cheshire et al., 2008; Kerman et al., 2008).
Hkb drives tube elongation in part by activating expression of klarsicht (klar), which encodes a putative dynein-associated protein known to mediate directed movement of organelles, including lipid droplets and nuclei (Welte et al., 1998; Mosley-Bishop et al., 1999). In the SG, Klar mediates the polarized delivery of vesicles to the apical membrane where they contribute to apical membrane growth and tube elongation. Too little apical membrane expansion, as occurs in both hkb and klar loss-of-function mutants, results in shortened tubes, whereas too much, as occurs with either loss of hairy (a negative regulator of hkb expression) or with Gal4-driven overexpression of hkb or klar, results in either branched or bulbous SG lumena (Myat and Andrew, 2002). Although Hkb affects SG transcription of crb, Hkb through Klar also increases Crb protein levels. crb encodes a transmembrane protein that localizes to a domain just apical to the adherens junctions (the subapical region) in all ectodermally-derived epithelia and functions as an apical determinant (Tepass et al., 1990; Wodarz et al., 1993); the loss of crb results in a severe loss of SG cells, with the remaining cells forming epithelial cysts with small central lumena (Wodarz et al., 1995). Thus, Hkb mediates tube elongation through two downstream targets that function to increase the apical membrane domain: Klar, which mediates apical targeting of membrane vesicles, and Crb, which mediates apical membrane expansion through somewhat less well understood mechanisms (Myat and Andrew, 2002).
The Rib transcription factor, which is required for tube elongation in the SG, the trachea and the Malpighian tubules, also controls events at the apical surface (Bradley and Andrew, 2001; Cheshire et al., 2008; Kerman et al., 2008). Wild-type Rib increases both crb mRNA and protein levels and affects the phosphorylation state of Moesin. Active phosphorylated Moesin is thought to bind apical membrane components (e.g. Crb), cross-linking them to the underlying actin cytoskeleton, thereby increasing the stiffness of the apical membrane (Medina et al., 2002; Karagiosis and Ready, 2004; Kunda et al., 2008). Loss of rib results in reduced crb/Crb, increased active Moesin, a depletion of the Rab11-positive apical vesicle membrane population and an increase in apical microvillar structure in both the SG and the branch of the trachea that elongates by the same mechanisms as the SG (Kerman et al., 2008). The molecular changes observed in rib mutant tubes increase apical stiffness, consistent with both the slower and limited apical expansion observed with live imaging (Cheshire et al., 2008).
Regulated turnover of the adherens junction protein E-cadherin through endocytosis is also key to SG elongation as revealed by experiments disrupting the small GTPase Rac (Pirraglia et al., 2006; Pirraglia et al., 2010). Blocking Rac activity with either combined loss-of-function mutations in multiple Racs or by SG-specific expression of a dominant-negative form of Rac results in reduced levels of Dynamin-mediated endocytosis, increased levels of E-cadherin, and reduced tube elongation. Expression of a constitutively-active form of Rac significantly decreases E-cadherin levels, resulting in the dispersal (and death) of SG cells. Moreover, related studies in the trachea revealed that tightly regulated E-cadherin turnover mediated by Rac and Src is critical to tube elongation in this tissue as well (Uemura et al., 1996; Chihara et al., 2003; Shindo et al., 2008). Altogether, these studies suggest that E-cadherin levels (and consequent assembly and disassembly of adherens junctions) must be tightly regulated to allow for increased flexibility during tube remodeling but to also provide sufficient cell adhesion for both the SG and trachea to elongate as intact cohesive structures.
Regulation of E-cadherin turnover is also important for other aspects of SG lumenal dimensions based on loss-of-function and overexpression studies of Pak1 (Pirraglia et al., 2010). Pak1 belongs to a family of serine-threonine kinases that bind and are activated by the Rho family GTPases, Rac and Cdc42 (Arias-Romero and Chernoff, 2008). Loss of Pak1 leads to defects in the polarized Rab5- and Dynamin-dependent endocytosis of E-Cadherin required for increasing the elongation ratio of the apical surface (proximal-distal dimensions relative to orthogonal dimensions), resulting in expanded SG lumena. Interestingly, SG expression of activated Pak1, a myristoylated form that constitutively localizes to membranes, results in the formation of multiple ectopic lumens. The overall phenotypes and epistasis analysis suggest that in the SG, Pak1 functions downstream of Cdc42 rather than Rac, explaining the apparent disparities in the phenotypic consequences of misregulation of Rac and Pak1. Since both Rac and Cdc42/Pak1 affect E-cadherin turnover, understanding the spatial and temporal dynamics of their activation will be key to understanding the differential consequences of loss and gain of activity on SG (and tracheal) morphology.
The final correct position of the SG depends on its active migration along a number of tissues that provide either traction or signals for directed movement. The internalizing SG tubes follow an approximately dorsal trajectory as they elongate. Once SG cells contact the overlying visceral mesoderm (VM), the glands reorient and begin to migrate posteriorly, initially in direct contact with cells that form the inner layer of gut musculature (the circular visceral muscle or cVM) (Bradley et al., 2003). Direct contact between the dorsal side of the SG and cVM is disrupted during mid-embryogenesis as the longitudinal visceral muscle (lVM) precursors migrate anteriorly along the cVM between the cVM and SG (Vining et al., 2005). The SG then directly contacts the lVM, with contacts between the distal SG tubes and the gastric caecae (the tubular appendages that develop in the most anterior domains of the gut) persisting through the end of embryogenesis. Whereas the dorsal portion of the SG makes direct contact with the VM during migration, the ventro-lateral portion of the SG makes direct contact with alternating populations of somatic musculature (SM) and fat body (FB) precursors (Vining et al., 2005). Finally, the ventromedial portion of the gland is in close proximity to the developing neural tube, with proximal portions of the SG making direct contact with the CNS in late embryos. Analysis of SG migration in embryos in which the development of the cVM, lVM, SM and FB is disrupted suggests that each of these tissues contributes to SG positioning (Vining et al., 2005).
Posterior migration of the SG absolutely requires integrin signaling (Bradley et al., 2003). In embryos mutant for αPS1, an alpha integrin subunit expressed in the early SG (inflated [if]), or mutant for αPS2, an alpha integrin subunit expressed in mesoderm (multiple edematous wing [mew]), or mutant for α1,2 laminin, a laminin expressed in the VM (wing blister [wb]), the SGs invaginate, the distal end of the tube contacts the VM but the tubes completely fail to migrate. As more SG cells internalize and the tube continues to elongate, it buckles and often folds in half. A similar failure of the SGs to migrate is observed with maternal and zygotic loss of the only β integrin subunit expressed in embryos. These findings suggest that integrins expressed in the substrate tissue (the VM) bind laminin in the ECM, which is also a substrate for the integrins expressed in the migrating tissue (the SG). As the WT SG migrates along the VM, it extends multiple dynamic lamellipodia, which are not visible in the SGs of the integrin subunit mutants. Such structures are reminiscent of the leading edges of migrating cells in a tissue culture dish and bring up the question of what features of single cell migration are conserved in a tissue such as the SG, which migrates as intact polarized epithelia.
Whereas integrin signaling is essential for migration, this pathway is unlikely to provide directional signals for migration. Such signals come from tissues surrounding the SG, including the yolk, FB, SM, VM and neural tube. Indeed, genetic studies support the existence of both yolk-dependent and FB-dependent migration cues, although the exact signals coming from these tissues await discovery (Bradley et al., 2003; Vining et al., 2005). A combination of attractant and repellent signals coming from the neural tube and VM (Wnts, Netrin and Slit) are critical for keeping posterior SG migration on a course parallel to both tissues (Kolesnikov and Beckendorf, 2005; Harris and Beckendorf, 2007). PDGF/VEGF signaling is also required for keeping SG migration on track, since loss-of-function mutations in the receptor as well as loss -of-function mutations in any of the three ligands or their ectopic expression result in SG migration defects (Harris and Beckendorf, 2007). Altogether, these studies support a role for multiple tissues providing distinct migratory cues at different developmental stages. How a tissue that migrates as a highly polarized epithelium integrates these signals to coordinate the movement of its approximately 100 constituent cells remains to be elucidated.
The major function of the Drosophila SG is secretion. The CrebA transcription factor, which is expressed early and continuously in this tissue, as well as to somewhat lower levels in other secretory tissues, is required for high-level secretory capacity (Andrew et al., 1997; Abrams and Andrew, 2005; Fox et al., 2010). Loss of CrebA diminishes expression of secretory pathway genes to levels observed in surrounding non-secretory tissues and expression of CrebA in new cell types is sufficient to elevate secretory gene expression. Moreover, loss of CrebA results in reduced SG lumen size and content as well as a reduction in the size and number of apically-localized secretory vesicles (Fox et al., 2010). CrebA directly binds a conserved consensus site in the enhancer regions of genes encoding the protein components of the secretory machinery to elevate expression above the basal levels required in all cells. This appears to be an ancient role for this family of proteins based on the demonstration that expression of the active forms of the human proteins most similar to CrebA (known as Creb3L1 and Creb3L2) are also capable of activating expression of secretory pathway genes in heterologous cell types, including Drosophila embryos and HeLa cells (Fox et al., 2010). CrebA also appears to enhance expression of tissue-specific secretory cargo genes although the expression of the SG cargo genes also depends on the other SG transcription factors Fkh and Sage (Fox et al., 2010; Maruyama et al., 2011)
Among tubular organs, the Drosophila trachea has emerged as the premier system for understanding the molecular and cellular underpinnings of tube formation (Metzger and Krasnow, 1999; Ebner et al., 2002; Petit et al., 2002; Rosin and Shilo, 2002; Ribeiro et al., 2003; Uv et al., 2003; Cabernard et al., 2004; Myat, 2005; Uv and Samakovlis, 2005; Kerman et al., 2006; Swanson and Beitel, 2006; Casanova, 2007; Affolter and Caussinus, 2008; Andrew and Ewald, 2010; Schottenfeld et al., 2010). The genes specifying trachea have been discovered, the early regulators controlling major aspects of tubulogenesis, including invagination and migration, are known, as are the molecules that specify the size and types of cellular tubes comprising this organ.
The identification and cloning of the earliest-expressed tracheal genes, trachealess (trh), ventral veinless (vvl; also known as drifter [dfr]) and knirps (kni) (Anderson et al., 1995; de Celis et al., 1995; Isaac and Andrew, 1996; Wilk et al., 1996; Chen et al., 1998), provided important probes for learning how the tracheal primordia are specified. These discoveries identified several genes that impose limits on trachea formation (Figure 4). Early expression of the spalt major (salm) gene “bookends” trachea formation, limiting tracheal formation to the posterior thoracic and abdominal regions (regions corresponding to parasegments 4 – 13; Kuhnlein and Schuh, 1996; Boube et al., 2000; Zelzer and Shilo, 2000). Wingless-signaling, Dpp-signaling and (likely) EGF-signaling limit trachea formation to a subset of cells within each thoracic and abdominal segment (de Celis et al., 1995; Isaac and Andrew, 1996; Wilk et al., 1996). Characterization of the enhancers has revealed that JAK/STAT signaling through Stat92E directly induces expression of the two earliest-expressed tracheal genes, trh and vvl (Brown et al., 2001; Chen et al., 2002; Li et al., 2003; Sotillos et al., 2010), thus acting as the positive signaling pathway inducing trachea formation. Interestingly, expression of the spatially limited component of JAK/STAT signaling, unpaired (upd), which encodes the secreted ligand, persists in the trachea up through embryonic stage 13 in a Trh-dependent manner (Sotillos et al., 2010). This finding may explain why there are somewhat fewer tracheal primordia in trh mutant embryos (Chung et al., 2011). Early Trh-dependent upd expression could be required to recruit additional cells to a tracheal fate in a non-cell autonomous manner.
Based on the coordinate expression of the known early tracheal transcription factor genes, as well as the finding that the initial expression of each of these genes is independent of the function of the others, it has been suggested that, as in the SG, no single gene controls development of this organ. Instead, it is thought that the coordinate activities of a few early transcription factor genes, such as trh, vvl and kni, mediate tracheal development (Boube et al., 2000; Zelzer and Shilo, 2000). These early transcription factor genes were proposed to work together to regulate expression of key downstream target genes, such as breathless (btl), which encodes an FGF-receptor essential for tracheal migration, and to independently activate expression of other tracheal genes, such as rhomboid (rho) and heartbroken (hbr; also known as downstream of FGF [dof]) (Michelson et al., 1998; Vincent et al., 1998; Imam et al., 1999). However, recent studies suggest that Trh plays the major role in maintaining and implementing a tracheal cell fate; Trh is required not only to maintain its own expression but is also required for the expression of every other tested tracheal gene (~100 genes were tested in this recent study; Chung et al., 2011). On the other hand, Vvl affects only an estimated 25 – 30% of tracheal genes. Kni is entirely dependent on Trh for its tracheal expression and has only one known transcriptional target gene in the trachea, salm, which is expressed in a subset of tracheal branches. Therefore, Trh is the major regulator of tracheal development, consistent both with a complete failure of the trachea to undergo any aspects of morphogenesis in trh null embryos (Isaac and Andrew, 1996; Llimargas, 1999) and the formation of additional tracheal metameres observed with global expression of trh (Wilk et al., 1996).
As with the SG, cells within the tracheal placodes invaginate through an apical constriction mechanism to form internalized tracheal sacs (Isaac and Andrew, 1996; Llimargas, 1999). Invagination of the primordia requires Trh (Isaac and Andrew, 1996; Wilk et al., 1996); in trh mutants, the tracheal primordia completely fail to accumulate apical actin and undergo apical constriction, remaining on the embryo surface with the surrounding ectoderm (Isaac and Andrew, 1996; Llimargas, 1999). Trh controls invagination in part through transcriptional activation of the rhomboid (rho) gene (Boube et al., 2000; Zelzer and Shilo, 2000), which encodes a protein essential for processing the EGF ligand (Lee et al., 2001). rho is the spatially limited component for EGF signal activation since the ligand, receptor and downstream effectors are expressed relatively ubiquitously. In rho mutants, and in embryos mutant for other components of the EGF signaling pathway, many tracheal cells fail to invaginate (Wappner et al., 1997; Llimargas, 1999; Bradley et al., 2001; Nishimura et al., 2007). The partial invagination of the tracheal primordia in these mutants may indicate some residual function from maternal stores of EGF components. Alternatively, additional pathways may function downstream of Trh to mediate internalization. Trh appears to be a more general regulator of tube formation, as at least two other Drosophila tissues express and require this gene for tube invagination (Isaac and Andrew, 1996). As tracheal cells invaginate, their final mitotic division is oriented in a direction that facilitates internalization of the primordia (Nishimura et al., 2007).
The final round of mitotic division that occurs during tracheal invagination results in approximately 80 cells per metamere. Once the tracheal cells have invaginated to form the tracheal sac, subsets of cells within each tracheal metamere begin to migrate in stereotypical directions (Figure 5). Subsets of cells will migrate dorsally to populate the dorsal branch (DB), whereas other cells will migrate ventrally to populate either the lateral trunk (LT) or ganglionic branch (GB). Still other cells migrate internally to populate the visceral branch (VB) and a final group of cells will migrate along the anterior-posterior axis of the embryo to populate the major artery of the trachea, known as the dorsal trunk (DT). Cells that remain near the site of invagination form the transverse connective (TC), the stalk connecting all of the tracheal branches within each metamere. Migration of all tracheal cells absolutely depends on the fibroblast growth factor (FGF) receptor tyrosine kinase-signaling pathway. Initially, all tracheal cells express btl, which encodes one of the two known Drosophila FGF receptors (Glazer and Shilo, 1991; Klambt et al., 1992). btl expression in the trachea requires Trh and Vvl (Dfr/Vvl), the two earliest expressed tracheal transcription factors (Anderson et al., 1995; Anderson et al., 1996; Ohshiro et al., 2002). btl expression is maintained in migrating tracheal branches through activation of the Btl- and Dpp- signaling pathways (Ohshiro et al., 2002; Myat et al., 2005). The tracheal branches migrate toward the source of the Branchless (Bnl) FGF ligand, which is expressed in the nearby target tissues (Sutherland et al., 1996). The loss of Btl, Bnl or other downstream signaling components results in the tracheal branches completely failing to migrate and remaining as internalized sacs (Klambt et al., 1992; Younossi-Hartenstein and Hartenstein, 1993; Reichman-Fried et al., 1994; Reichman-Fried and Shilo, 1995; Lee et al., 1996; Sutherland et al., 1996; Michelson et al., 1998; Vincent et al., 1998; Imam et al., 1999). In embryos expressing an activated form of the Btl receptor or in embryos where the Bnl ligand is expressed ubiquitously, ectopic tracheal branching is observed (Lee et al., 1996; Sutherland et al., 1996). Interestingly, FGF (Btl)-signaling is also required at later stages of larval development for tracheal branch outgrowth in response to hypoxia (Jarecki et al., 1999).
Migration of specific tracheal branches along distinct trajectories has been shown to require integrin signaling, EGF signaling, and at least one of the signaling pathways known to function in axonal pathfinding and SG migration, the Slit/Robo signaling cascade (Boube et al., 2000; Englund et al., 2002; Gallio et al., 2004; Kolesnikov and Beckendorf, 2005). The requirement for integrin signaling has been most cleanly demonstrated in the migration of the visceral branch (VB) of the trachea along the visceral mesoderm (VM). In embryos mutants for the α-integrin subunit αPS1, which is expressed in the VB of the trachea, or embryos mutant for the α-integrin subunit αPS2, which is expressed in the mesoderm, the VB of the trachea reaches the VM, but fails to migrate along it, much like the SG (Boube et al., 2000; Bradley et al., 2003). Similar defects in VB migration are observed with mutations in the more broadly expressed β-integrin subunit gene, although many aspects of embryonic development are also abnormal in these animals. EGF signaling has been shown to function to prevent mid-line crossing of the tracheal ganglionic branch (GB) (Gallio et al., 2004). Rho3, which is expressed in a specific midline neuronal cell type, known as the VUM neurons, is thought to activate ligand release, which is sensed by the migrating GB. Tracheal specific expression of either dominant-negative EGF receptor (EGFR) or Ras leads to midline crossing whereas tracheal specific expression of activated EGFR induces premature turning away from the midline. Migration of several tracheal branches, including the DB, VB, DT and GB are affected by mutations in the Slit/Robo signaling cascade (Englund et al., 2002; Lundstrom et al., 2004). Although in both the Drosophila and mammalian nervous system Slit/Robo signaling primarily functions in axonal repulsion, in the trachea (and SG), this signaling pathway functions in both attraction and repulsion, with the Robo receptor primarily mediating repulsion and the Robo2 receptor mediating attraction. Slit acts as an attractant for both the DB and VB, whereas Slit acts as a repellent for the GB.
The distinction between specific tracheal branches and the types of tubes they form (see next section) depends on the signaling pathways that are activated in a given group of cells. For example, cells in the primordia that are exposed to dorsal and ventral domains of Dpp signal as well as continued FGF signaling will maintain expression of two early tracheal transcription factors, Knirps (Kni) and Knirps-related (Knrl), and will migrate either dorsally to populate the DB or ventrally to populate the LT and GB (Affolter et al., 1994b; Vincent et al., 1997; Chen et al., 1998; Myat, 2005). Kni/Knrl expression in the cells that form the DB shuts off expression of Salm, a transcription factor expressed and required in the cells that form the DT (Franch-Marro and Casanova, 2002). Salm expression in the DT also requires Wg, which is expressed in ectodermal cells in close proximity to this branch (Chihara and Hayashi, 2000; Llimargas, 2000). In embryos in which Wg signaling is blocked, the DT cells lose expression of Salm and migrate internally with the VB. Maintained Salm expression thus ultimately correlates with formation of the DT and, interestingly, with the types of tubes the DT forms (Kuhnlein and Schuh, 1996; Kuhnlein et al., 1997; Ribeiro et al., 2004). Repression of Salm expression correlates with the formation of other branches and with the formation of a different type of tube.
In the fully formed trachea, four different types of tube geometries are evident, types I-IV (Samakovlis et al., 1996). Type I tubes are multicellular tubes in which the apical surfaces of two or more cells surround a central common lumen, with the cells connected to one another through intercellular junctions. Type II tubes comprise a linear array of single cells whose apical surfaces completely surround a central lumen. Cells within type II tubes connect with themselves through intracellular (or autocellular) junctions and with their proximal and distal neighbors through intercellular junctions. Type III tubes are donut-shaped tubes formed by the cells that connect adjacent or contralateral tracheal metameres. The apical surfacess of these cells span the inner wall of the donut with intercellular junctions connecting to adjacent cells on both edges of the inner wall. Finally, type IV tubes are the subcellular tubes that form within the terminal cells as long cytoplasmic extensions, known as tracheoles (Guillemin et al., 1996). These elongated subcellular tubes contact target tissues throughout the animal for gas exchange.
All of the primary branches of the trachea begin as multicellular type I tubes. As the trachea develop, the different types of tubes begin to emerge depending on the position of the cells within each branch of the trachea (Samakovlis et al., 1996). Cells that form the stalk of the trachea can remain as either multicellular type I tubes, such as those forming the bulk of the DT and large portions of the TC, or they can form multicellular type II tubes through cell rearrangement (Jazwinska et al., 2003; Ribeiro et al., 2004). The stalks of the DB, LT, GB and VB are type II tubes that form from the type I tubes of the primary branches. The formation of type II tubes from type I tubes involves cell rearrangements wherein one cell within a pair of cells that surround a shared portion of lumen reaches around the lumen at one end to contact itself. At the same time, the other cell reaches around the lumen at the other end and contacts itself. The autocellular junctions that initiate at the sites of self-contact then zipper up as the two cells slide past one another. This process stops once the distal end of the proximal cell contacts the proximal end of the distal cell through only a persistent intercellular junction. Two proteins found in the apical extracellular matrix are required to prevent the zippering process from continuing to a point where the two cells would separate and disrupt tracheal continuity, Piopio (Pio) and Dumpy (Dp) (Jazwinska et al., 2003). Stalk cell intercalation is driven by the migration of tip cells, which generate tensile forces on the entire branch through attachment of the proximal portions of the branch to the remainder of the tracheal system (Caussinus et al., 2008).
The Salm transcription factor is required to prevent type I tubes from undergoing the cell rearrangements required to form type II tubes; pan-tracheal expression of Salm results in all stalk cells forming type I tubes, whereas the loss of salm function results in stalk cells forming type II tubes (Ribeiro et al., 2004). Recent work suggests that Salm prevents stalk cell intercalation in type I tubes, at least in part, by upregulating Rab11-dependent recycling of E-cadherin to the apical membrane (Shaye et al., 2008), perhaps relieving the tension required for stalk cell intercalation by allowing apical membrane expansion. Interestingly, the failure to achieve full elongation of the SG and of the type I tracheal tubes in rib mutants has also been linked to a diminished population of Rab11-positive vesicles (Kerman et al., 2008).
Cells at or near the ends of each primary tracheal branch form either fusion cells or terminal cells. Fusion cells connect the individual tracheal metameres to each other. There are five fusion cells per metamere, one at the end of the DB, which connects with the contralateral DB fusion cell on the other side of the embryo, and one each at the anterior and posterior ends of the DT and LT, which contact the fusion cells of the DT and LT in the adjacent anterior and posterior tracheal metameres. Terminal cells form subcellular type IV tubes, which are the blind-ended tracheoles that directly contact target tissues for gas exchange. Cells near the ends of the dorsal and ventral tracheal branches, which receive the highest levels of Dpp, Wg and Btl signaling, express high levels of Delta and become fusion cells (Ikeya and Hayashi, 1999; Llimargas, 1999; Steneberg et al., 1999; Chihara and Hayashi, 2000). Delta, one of the two known Notch ligands in Drosophila, signals to the adjacent tracheal cells to block them from also becoming fusion cells, thus limiting the number of fusion cells within each tracheal branch. In conditional Notch mutants, more fusion cells form at the expense of stalk cells. Since mutations disrupting Notch signaling also lead to an increase in the number of terminal cells, Notch may also function in other cell fate decisions within the trachea (Llimargas, 1999; Steneberg et al., 1999). In addition to Delta, fusion cells also express the Escargot transcription factor, which activates local expression of E-cadherin, one of several proteins required to form the donut-shaped fusion cell tubes (Tanaka-Matakatsu et al., 1996; Uemura et al., 1996). Once fusion cells contact one another, the basal accumulation of E-cadherin recruits a number of cytoskeletal proteins and initiates formation of an actin rich-structure between the apical surfaces of the contacting fusion cells (Tanaka-Matakatsu et al., 1996; Uemura et al., 1996; Lee and Kolodziej, 2002a; Lee and Kolodziej, 2002b; Lee et al., 2003; Tanaka et al., 2004; Jiang and Crews, 2007; Kakihara et al., 2008). This structure serves a scaffolding function for the assembly of the apical membrane surface that eventually connects the lumens between neighboring tracheal metameres.
The cells near the ends of the primary branches of the Drosophila trachea that do not become fusion cells but that receive high levels of Bnl signal become terminal cells, which contain the subcellular type IV tubes or tracheoles. Nineteen of the approximately 80 cells in each metamere become terminal cells, with varying numbers within each branch. Three genes have been implicated in terminal branching during embryogenesis. sprouty, which encodes a cytoplasmic factor that is activated in response to high-level Btl-signaling at the tips of primary branches, functions to antagonize the pathway in neighboring cells to limit the number of stalk cells that take on a terminal cell fate (Hacohen et al., 1998). pruned encodes a transcription factor required for the elaboration of cytoplasmic extensions and whose activity must be controlled to prevent ectopic projections and to limit the invasion of tracheoles into territories normally supplied by other tracheal cells (Affolter et al., 1994a; Guillemin et al., 1996). Finally, a nuclear lamin has also been implicated in the formation of cytoplasmic extensions during terminal cell differentiation, presumably also acting through effects on gene expression (Guillemin et al, 2001). During late embryonic stages, apical sprouts contiguous with the existing apical domains of terminal cells extend in the direction of cytoplasmic growth, creating elongated subcellular lumens that are contiguous with the remainder of the tracheal lumen (Gervais and Casanova, 2010). Lumenal outgrowth in terminal cells appears mechanistically similar to apical membrane expansion in the SG, with the polarized delivery of vesicles to the minus ends of microtubules, which are enriched at the apical domain. Organization of the microtubule network in elongating terminal cells is thought to occur through FGF-signaling dependent assembly of actin rich structures near the basal ends of the elongating terminal cells (Gervais and Casanova, 2010). Maintenance of the terminal branches and their luminal organization requires integrin signaling (Levi et al., 2006).
Tube size control is tightly linked to branch identity and, surprisingly, is not affected by the number of individual tracheal cells; instead, tube size is mediated by the coordinate control of the apical (lumenal) surface of tracheal cells (Beitel and Krasnow, 2000). The cloning and characterization of many of the genes regulating tracheal tube size revealed that several encode components of septate junctions (SJs; junctional structures functionally and molecularly similar to vertebrate tight junctions) (Behr et al., 2003; Llimargas et al., 2004; Wu et al., 2007; Bachmann et al., 2008; Hijazi et al., 2009). For example, both megatrachea and sinuous (two genes required to restrict tube length) encode proteins that localize to septate junctions and are closely related to vertebrate claudins (Behr et al., 2003; Wu and Beitel, 2004). Claudins are highly divergent multi-span transmembrane proteins that through homo- and heterophilic interactions provide paracellular barrier function to tight junctions, which prevents the diffusion of solutes and water between epithelial cells (Tsukita and Furuse, 2000,2002; Tsukita et al., 2001). This barrier function of claudins is shared with the related fly proteins. Importantly, mutations in a number of proteins encoding previously known components of SJs were subsequently shown to affect both tracheal tube length and to affect barrier function (Paul et al., 2003). Based on the characterization of the other major class of genes affecting tube length, it appears that SJs affect tube length because they are required for the polarized secretion of components regulating the apical extracellular matrix (ECM). This other class of genes affecting tracheal tube size encodes the enzymes required for the synthesis and modification of the apical ECM. Specifically, synthesis and secretion of a transient chitinous matrix is required to achieve uniform tube diameter throughout the different branches of the trachea (Araujo et al., 2005; Devine et al., 2005; Tonning et al., 2005; Moussian et al., 2006; Tonning et al., 2006). The subsequent modifications of this chitinous matrix by enzymes whose secretion depends on intact SJs serve to limit overall tube length (Luschnig et al., 2006; Wang et al., 2006). The requirement for the synthesis and secretion of an intact apical ECM is also shared by the salivary gland; mutations reducing levels of secretion as well as mutations affecting the modification of apically-secreted proteins result in irregular salivary gland lumena with multiple bulges and constrictions (Seshaiah et al., 2001; Abrams et al., 2006). Unlike with the trachea, chitin does not appear to be a major constituent of the apical ECM of the salivary gland.
The characterization of serrano (sano), a novel apically-enriched cytosolic protein expressed in several embryonic tubular organs, has implicated the planar cell polarity (PCP) genes in tube size regulation (Chung et al., 2009). The highly conserved PCP pathway controls cell polarity in the plane of the epithelia, orthogonal to the apical-basal axis. Like Sano, PCP components are apically enriched, partially overlapping the adherens junctions. In the precursors to adult epithelial structures, many PCP components are also asymmetrically localized to either the proximal or distal side of adult precursor cells, although the same asymmetric localization of these components in the trachea has not been reported. sano loss results in over-elongated trachea, whereas its overexpression results in shortened trachea (and shortened SGs). sano overexpression during the larval or pupal stages leads to the mislocalization of several core PCP proteins as well as characteristic PCP phenotypes in multiple adult epithelial tissues. As with sano, the loss or overexpression of multiple PCP genes causes defects in tracheal tube length and Sano directly binds Disheveled, a core PCP component. Interestingly, both the tube length defects and the adult PCP phenotypes caused by Sano overexpression are linked to a decrease in apical domain size, suggesting a role for Sano and the PCP components in regulating membrane turnover and/or the linkage of membrane to the underlying apical cytoskeleton. Supporting a role for Sano/PCP components in apical membrane turnover is the finding that loss of wurst, which encodes a highly conserved transmembrane protein essential for clathrin-mediated endocytosis, leads to the same tracheal tube length defects as loss of sano (Behr et al., 2007).
Among the final steps in maturation is the clearing of the matrix- and liquid-filled tracheal lumen to allow gas to fill the tubular network. The identification and characterization of several gas-filling mutants reveals that following the lumenal deposition of the secreted chitinous proteins required for tube expansion, clathrin-mediated endocytosis rapidly clears tracheal solids from the lumen, followed shortly by the clearance of lumenal liquids (Behr et al., 2007; Tsarouhas et al., 2007). Careful examination of mutants compromised for secretion or endocytosis indicates that the processes are interdependent: high amplitude lumenal secretion is required for the subsequent endocytic wave, which is, in turn, required for liquid clearance (Tsarouhas et al., 2007). Based on genetic interaction studies (Behr et al., 2007), epithelial Na+ channels may be important cargo for endocytosis, since these proteins are involved in liquid clearance in both the Drosophila trachea (Liu et al., 2003) and mammalian neonatal lung (Hummler et al., 1996).
Both the SG and trachea have proven to be extremely useful systems for understanding the basic mechanics of how epithelial tubes initially form and achieve their final architecture. A comparison of what has been learned about both systems over the past two decades has revealed many key similarities as well as differences in how these tubes are made. Both the salivary gland and trachea begin as polarized epithelial placodes that form tubes by an invagination mechanism involving apical constriction. Although two very different transcription factors appear to drive this process – Fkh for the salivary gland, Trh for the trachea – the apical membrane protein Crb has been implicated in the invagination of both tissues (Myat et al., 2000a; Isaac and Andrew, 1996; Wilk et al., 1996; Xu et al., 2008; Letizia et al., 2011). The geometry of SG invagination differs somewhat from that of the trachea in that SG invagination begins with a small number of concentrically arranged cells, whereas tracheal invagination begins with two rows of cells flanking the midline of each tracheal placode. Consequently, the salivary gland invaginates to form a single elongating tube, whereas the tracheae invaginate to form internalized sacs that are somewhat elongated along their dorsal-ventral axes. The initial geometry established during invagination is likely to impact final tube morphology (Myat and Andrew, 2002). Both the SG and trachea remain fully polarized intact epithelia throughout morphogenesis. Although the SG cells stop dividing by the onset of SG-specific gene expression, tracheal cells one additional mitotic division a little later during invagination. Once internalized, however, both tissues grow by increases in cell size rather than increases in cell number. Both tubes elongate and actively migrate to arrive at their final positions. However, unlike some of the branches of the trachea, which undergo significant remodeling of cellular junctions to form other tube types, the salivary gland remains a type I multicellular tube. Nonetheless, regulated turnover of apical junction proteins such as E-cadherin is required in the SG (Pirraglia et al., 2006; 2010) as well as in the type I tubes of the trachea, which do not undergo significant remodeling (Uemura et al., 1996; Chihara et al., 2003;Shindo et al., 2008). Both of these tube types also share a requirement for the transcription factor Rib and its downstream targets for full tube elongation (Bradley et al., 2001; Kerman et al., 2008). Although many of the signaling pathways that provide directional cues for migration await discovery, at least on signaling pathway has already been implicated in the migration of both tissues (Kolesnikov et al., 2005; Englund et al., 2002; Gallio et al., 2004). Finally, although it is clear that establishing and maintaining a uniform patent lumen requires the secretion and modification of apical ECM, the composition of that matrix appears quite different in the two tissues.
Many of the molecules and mechanisms that participate in the formation of the Drosophila SG and trachea are shared in the formation of tubular organs in vertebrates. For example, the neural tube, lung and liver of many vertebrates form through an apical constriction mechanism related to the cell shape changes required to internalize the SG and tracheal primordia (Colas and Schoenwolf, 2001). FGF signaling has been implicated in branching morphogenesis of numerous tissues, including the mammalian lung, salivary glands, lacrimal glands, mammary glands, prostate glands and kidneys (Kato and Sekine, 1999; Lebeche et al., 1999; Makarenkova et al., 2000; Warburton et al., 2000; Costantini and Shakya, 2006; Patel et al., 2006; Sternlicht, 2006; Thomson and Marker, 2006). Similarly, Notch signaling appears to play roles in cell fate choices at multiple stages of vertebrate blood vessel formation, initially in the choice between arteriole versus venous vessel formation, and, later in the choice between forming tip versus stalk cells in angiogenic sprouts (for review, see (Torres-Vazquez et al., 2003; Hofmann and Iruela-Arispe, 2007). Likewise, the Slit-Robo pathway has been implicated in signaling blood vessel outgrowth during tumor angiogenesis (Liu and Herlyn, 2003).
Although much has been revealed regarding the specification, morphogenesis and physiological specialization of each of the tube types discussed in this review, many gaps remain in our understanding of the molecular and cellular underpinnings of tube development. In the Drosophila systems, the challenges include the identification and characterization of the genes that mediate invagination. Although several effectors are known, they either fail to fully account for the events of invagination and/or they are not yet linked to the transcription factors that provide temporal and spatial regulation to the process. Clearly additional signaling pathways that guide migration of the SGs and trachea to their final destinations await discovery. Learning how complex guidance information provided by multiple independent sources is processed to coordinate movement of entire tissues promises to be an exciting area of research. We still need to learn what gives each tubular organ its unique mechanical characteristics so that it can withstand the forces of movement and have the right flexibility for function. It will also be interesting to learn if formation of the adult Drosophila tissues, which occurs during metamorphosis, simply recapitulates the events of embryonic tube morphogenesis or if new programs are implemented. Learning more about salivary gland physiology could yield new and effective strategies for combating vector borne diseases such as malaria, Lyme's disease and Dengue fever. Finally, for tissue engineering using stem cell technologies, it will be necessary to eventually understand the regulation of mammalian tubular organ development to the same level as has been attained for Drosophila tissues.
We thank the following members of the Andrew lab for their comments and careful reading of the manuscript: Rebecca M. Fox, Caitlin D. Hanlon and SeYeon Chung. We also thank two anonymous reviewers for their thoughtful suggestions for improving this paper. We gratefully acknowledge Flybase, DGRC, DHSB, and the Bloomington Stock Center for data and reagents they have provided over the years to support our own work on Drosophila tube formation. We also thank our many colleagues studying Drosophila tube formation for the comprehensive and elegant work they have published over the past two decades and we apologize to our colleagues whose work we were unable to include.