Many of our essential organs, such as the endocrine glands, lungs, and vasculature, are composed of epithelial and endothelial tubes. Tubes perform several physiological functions, including the transport of secretory products, elimination of waste products, absorption of nutrients and water, and exchange of gases. Failure of tube formation often leads to organ failure, as evidenced by several disorders, such as polycystic kidney disease and spina bifida. Thus, understanding the basic mechanisms by which tubes form during embryonic development could lead to novel methods for the diagnosis and treatment of such disorders.
Rapid advances in the study of Drosophila tube formation are made possible by the great variety of approaches being applied. The Drosophila model is attractive because if the organism's relatively small genome size, easy husbandry, and a plethora of sophisticated genetic tricks. These tricks include various methods by which mutations affecting the tissue of interest can be generated in large-scale screens and the relative ease with which corresponding wild-type genes can be cloned due to the availability of the complete genomic sequence provided by the Berkeley Drosophila Genome Project and Celera genomics. Furthermore, transposable P-elements can be used to manipulate individual genes or large genomic regions, as well as for gene expression studies (Ish-Horowicz and Pinchin, 1987; O'Kane and Gehring, 1987; Schneuwly et al., 1987; Wilson et al., 1989; Brand and Perrimon, 1993). Although the use of reverse genetics to study gene function was limited previously to model organisms other than Drosophila, it is now possible to generate gene knockouts using a P-element mediated strategy designed by Rong and Golic (2001). Lastly, P-elements can be used as vectors for inducing expression of double-stranded RNA for the specific silencing of target gene expression (Kennerdell and Carthew, 2000). Thus, the multitude of genetic techniques combined with the ability to visualize tube formation in vivo by live microscopy has allowed the Drosophila researcher to identify genes involved in tube formation and understand the cellular process they regulate.
This review discusses the cellular events by which several different types of tubular organs are formed in the Drosophila embryo and the genes involved in these events. Because recent progress has been made in studies of the hindgut, Malpighian tubules, proventriculus, salivary gland, and trachea, this review will be limited to these tubular organs (Fig. 1). Because this review will focus on tube morphogenesis and not on earlier patterning events, the reader is advised to refer to recent reviews of the hindgut (Lengyel and Iwaki, 2002), Malpighian tubule (Cagan, 2003), salivary gland (Bradley et al., 2001), and trachea (Lubarsky and Krasnow, 2003), which provide a more comprehensive discussion.
Tubular organs form by several different types of morphogenic movements such as, invagination, directed migration, and convergent extension (Fig. 2). The organs discussed in this review are largely derived from the ectodermal layer of the embryo. Thus, primordial cells that initially reside at the surface of the embryo must be brought into the interior of the embryo. This process occurs mostly through invagination, whereby cells within a monolayer change shape and fold inward, forming a tubular structure. Invagination occurs not only during the formation of Drosophila organs, but also in more complex organisms, such as during formation of the avian and mammalian neural tube (reviewed by Schoenwolf and Smith, 1990), formation of the avian lens and optic cup (Hilfer, 1983), and the avian otic vesicle (Alvarez and Navascues, 1990).
During invagination, epithelial cells change shape from columnar to pyramidal (Fig. 2A). This cell shape change is accompanied by the constriction of the apical domains and basal migration of nuclei. Basal migration of nuclei is not dependent on apical constriction, because in salivary gland cells of Drosophila embryos mutant for the transcription factor Fork head, nuclei migrate basally yet apices fail to constrict (Myat and Andrew, 2000b). An apically localized actin cytoskeleton is thought to be involved in apical constriction. Supporting evidence comes from studies in Xenopus, for which expression of the actin binding protein Shroom (Haigo et al., 2003) is sufficient to induce apical constriction during Xenopus neurulation by recruiting apically localized actin, as well as from studies of Drosophila gastrulation where folded gastrulation and concertina are thought to induce and coordinate apical constriction most likely by regulating the actin cytoskeleton through RhoGEF2 (Hacker and Perrimon, 1988; Barrett et al., 1997). Although cell shape change accompanies invagination, it is not known whether it is the cause or the effect of invagination. Other mechanisms, such as changes in cell–cell adhesion and the apical extracellular matrix may also be involved in driving invagination. Changes in cadherin-mediated adhesion have been observed during invagination of mesodermal cells in Drosophila embryos, supporting the cell–cell adhesion model for invagination (Oda et al., 1998). Furthermore, in gastrulating sea urchin embryos, greater hydration of a newly deposited apical extracellular matrix is thought to drive invagination (Lane et al., 1993). Thus, the mechanism by which cells invaginate may differ from one tissue to another. Alternatively, several mechanisms may be coordinated to result in the invagination of a particular tissue.
Once primordial cells have been internalized and a rudimentary tube is formed, additional morphogenic movements, such as directed cell migration and convergent extension, occur to give rise to the final shape and size of the tube. A critical requirement for cell migration is that the cell must polarize its protrusions, whether they are broad, fan-like lamellipodia, or spike-like filopodia, in the direction of migration (Fig. 2B; reviewed by Ridley et al., 2003). This cellular polarity is achieved by the polarized growth of actin at the leading edge of the migratory cell. Precise regulation of integrin-mediated adhesion is also important for cell migration. The continual assembly and disassembly of adhesion sites at the front and back of the cell, respectively, provide the migrating cell with propulsive force and traction. The Rho family of small GTPases, which include, Rac, Rho, and Cdc42, as well as their interacting proteins, such as protein kinases and lipid-modifying enzymes, are key regulators of actin and adhesion site remodeling during cell migration (reviewed by Van Aelst and Symons, 2002). In some developmental contexts, cells do not migrate alone and instead migrate attached to neighboring cells. For example, during formation of the Drosophila proventriculus, salivary gland, and tracheal tubes, cells migrate as a group. Thus, cell migration during tube morphogenesis has the added complexity of coordinating adhesion to neighboring cells simultaneously with the other requirements characteristic of single migratory cells.
In convergent extension (CE), cells intercalate between one another to narrow and extend the tissue in a particular direction (Fig. 2C). CE is a common morphogenic movement known to occur in such diverse species as ascidians, teleost fish, birds, and mammals (reviewed by Keller, 2002). CE is involved in shaping tissues and organs, such as elongation of the vertebrate neural tube and of the Drosophila hindgut. Studies using live imaging suggest a cell traction–cell substrate model where polarized protrusions adhere to and exert traction on adjacent cells, thereby pulling them between one another (Shih and Keller, 1992). Recent studies on germ-band elongation in the Drosophila embryo identify an important role for Myosin II in cell intercalation whereby the polarized recruitment of Myosin II allows remodeling of cell–cell junctions (Bertet et al., 2004). It is possible that, in tissues undergoing convergent extension, formation of polarized protrusions as well as polarized remodeling of cell–cell junctions are both necessary for cells to intercalate. In addition to the morphogenic movements discussed above, cell division can also contribute to tube formation, as in the case of the Drosophila Malpighian tubules, where precise regulation of cell proliferation is critical for the tubes to reach their correct length.
The alimentary track of the Drosophila embryo consists of the foregut, midgut, and hindgut. In this review, I will focus on morphogenesis of the foregut and hindgut tubes, which originate from the ectoderm to from single-layered tubes that become ensheathed with visceral mesoderm (Skaer, 1993; Campos-Ortega and Hartenstein, 1997). Both the hindgut and the foregut tubes are formed by invagination of its primordial cells, with the foregut originating from cells of the stomodeum that reside in the anterior region of the embryo and the hindgut originating from cells of the proctodeum in the posterior region.
Hindgut formation begins with the invagination of its primordial cells to form a globular epithelial sac (Fig. 3; Campos-Ortega and Hartenstein, 1997). The globular sac is then transformed into an elongated tube through convergent extension, increases in cell size by DNA endoreplication and increases in tubular surface area due to cell shape changes (reviewed by Lengyel and Iwaki, 2002). Cell division also plays an important role in hindgut elongation with the number of hindgut cells almost tripling during this process (Harbecke and Janning, 1989; Campos-Ortega and Hartenstein, 1997; Iwaki et al., 2001). The elongated hindgut is organized into three morphologically distinct domains: small intestine, large intestine, and rectum. The different domains of the hindgut differ in cell size; cells of the large intestine are larger than those of the small intestine and rectum, due to a 1.7-fold endoreplication of their DNA (Smith and Orr-Weaver, 1991; Fuss et al., 2001). Cells of the small intestine are smaller and more cuboidal (Iwaki et al., 2001).
Hindgut patterning and cell rearrangement during tube elongation is controlled by a transcriptional hierarchy consisting of the Drumstick (Drm), Lines (Lin), and Bowl transcription factors (Fig. 3). drm and bowl encode small zinc-finger proteins of the Drosophila Odd-skipped family. drm is expressed specifically in the anterior region of the hindgut that will give rise to the small intestine, whereas both bowl and lin are expressed throughout the hindgut (Johansen et al., 2003). Genetic analysis and gene expression studies revealed that bowl acts downstream of lin and lin acts downstream of drm (Johansen et al., 2003). The normal role of Drm is to antagonize Lin such that the small intestine fate is established in the more anterior regions of the hindgut tube (Fig. 3). Furthermore, Lin repression by Drm is transcriptional and direct and involves the first C2H2 finger of Drm (Green et al., 2002). In addition to regulating hindgut patterning and morphogenesis, the drm-lin-bowl transcriptional hierarchy is also involved in patterning of the proventriculus during foregut development, which suggests a general role for these transcription factors in development of the gut tubes.
drm, lin, and bowl regulate hindgut cell rearrangement by controlling the expression of Unpaired (Upd), the ligand for the JAK/STAT signaling pathway. Upd is expressed specifically in the hindgut cells that will form the small intestine, whereas the JAK/STAT receptor (encoded by domeless), JAK (encoded by hop), and STAT (encoded by Stat92E) are expressed more broadly in the hindgut epithelium (Iwaki et al., 2001; Johansen et al., 2003). In embryos mutant for upd or other JAK/STAT signaling pathway components, the hindgut is shorter and wider due to failed cell rearrangement (Johansen et al., 2003). Thus, Upd acts cell nonautonomously in the large intestine to control cell rearrangement by activating the JAK/STAT signaling pathway in the large intestine. In addition to regulating cell rearrangements during Drosophila hindgut elongation, the JAK/STAT signaling pathway is also known to regulate similar types of cellular movements in vertebrate embryos. For example, JAK1 kinases are required for zebrafish gastrulation in a process called epiboly, where cells from a deeper layer move up and intercalate between more superficial layers, resulting in the thinning and spreading of the tissue (Conway et al., 1997).
The Malpighian tubules (MT) form by the evagination of cells from the junction between the hindgut and the midgut to form four small buds (Fig. 4). During the bud evagination step, cells are continually recruited into these buds from the proctodeum in a step involving the walrus gene, which has not been cloned (Liu et al., 1999). Although very little is known about the cellular changes associated with evagination of MT cells, it may proceed in a manner similar to proventriculus evagination, which is thought to be driven by cell shape changes involving localized apical constrictions (see below). The four buds that are initially cylindrical extend to become crescent shaped by the migration of cells toward the tip of the tube and/or cell rearrangement. The circumference of these primitive tubes is approximately eight cells (Janning et al., 1986; Skaer and Arias, 1992). The MT then undergoes dramatic elongation to form tubes with only two cells in its circumference (Skaer, 1993).
The zinc-finger type transcription factor, Kruppel, controls the bud evagination step, partly through regulation of a homeobox containing transcription factor, encoded by the cut gene (Harbecke and Janning, 1989; Gaul and Weigel, 1991; Jack et al., 1991; Liu and Jack, 1992). Subsequent to bud evagination, extension of the cylindrical buds into crescent-shaped tubes requires ribbon (rib), which encodes a BTB/POZ-type nuclear protein (Jack and Myette, 1997; Bradley and Andrew, 2001; Shim et al., 2001) and faint sausage (fas), which encodes an extracellular, immunoglobulin-like protein (Lekven, 1998; Liu et al., 1999). Both rib and fas are also required for formation of the salivary glands with rib being additionally required for tracheal tube formation (Myat and Andrew, 2000a Bradley and Andrew, 2001; Shim et al., 2001). Because rib directed mediates cell migration in both the salivary gland and trachea, it is possible that, in the MT, rib also plays a migratory role in transforming the cylindrical-shaped buds into crescent-shaped ones. MT formation also requires activity of the hedgehog and wingless signaling pathways, which allow the ureters and the MT to reach their normal lengths by mediating cell movements and cell proliferation, respectively (Skaer and Arias, 1992; Harbecke and Lengyel, 1995; reviewed by Ainsworth et al., 2000).
As in hindgut tube elongation, cell division also plays an important role in elongation of the MTs. Mutations in genes required for cell division, such as faint little ball, barren, and three rows, prevent tube elongation (Skaer, 1989; Baumann and Skaer, 1993; Jack and Myette, 1997; Kerber et al., 1998; Liu et al., 1999). Cell division in the MT is mediated by the tip cell, which resides at the distal tip of each MT and stimulates cell proliferation in neighboring cells through a process requiring epidermal growth factor (EGF) signaling (Hoch et al., 1994). Of interest, MT formation also involves a mesenchymal to epithelial transition (MET). MT primordial cells give rise to principal cells, which line the tubules and control the flow of hydrogen, sodium, and potassium ions (O'Donnell et al., 1998). In addition, stellate cells, which control the flow of other ions, such as chloride, originate from a group of posterior mesodermal cells and undergo a mesenchymal to epithelial transition before becoming integrated into the MT epithelium (Denholm et al., 2003). This mesenchymal–epithelial transition parallels vertebrate kidney development, where nephrogenic mesenchymal cells undergo MET to form the glomerulus and nephron (reviewed by Shah et al., 2004). Furthermore, the recruitment of stellate cells into MT requires hibris, an ortholog of vertebrate NEPHRIN, which is localized to the foot processes of glomerular podocytes and is mutated in two types of congenital nephritic syndrome (Artero et al., 2001; Saleem et al., 2002; Denholm et al., 2003). Thus, the Drosophila Malpighian tubules and the vertebrate kidney appear to share certain cellular and molecular requirements even though they are widely diverged organs.
Foregut Tube and Proventriculus
Like the hindgut, the embryonic foregut is a single-layered epithelial tube formed from the ectoderm. In its fully developed stage, the foregut consists of the atrium, pharynx, esophagus, and proventriculus. The foregut forms from the anterior region of the blastoderm embryo by the invagination of cells of the stomodeum. In addition to the cellular events associated with invagination, cell division also contributes to the inward movement of the foregut tube (Skaer, 1993; Campos-Ortega and Hartenstein, 1997). The most posterior part of the internalized foregut tube gives rise to the proventriculus, which forms at the junction with the anterior region of the midgut (Fig. 5). It is a multilayered organ that serves as a valve to regulate the passage of food into the midgut (Strasburger, 1932). Proventriculus development begins when cells at the posterior boundary of the foregut tube evaginate to form a ball-like structure. This evagination event occurs through cell shape changes initiated by local constriction of apical membranes. Subsequent constrictions at the boundary of the ectodermal and endodermal cells result in the formation of the keyhole structure (Fig. 5). Cells from the anterior portion of the ectodermal keyhole region, which is devoid of mesoderm, migrate inward into the endodermal keyhole region to form a heart-shaped structure (Fuss et al., 2004). These migrating cells are characterized by a stretched morphology with long cytoplasmic extensions (Pankratz and Hoch, 1995).
Mutational analyses identified several genes involved in cell migration during proventriculus formation. The drm-lin-bowl transcriptional hierarchy, which patterns the hindgut and mediates cell rearrangement during hindgut elongation, is again involved in proventriculus patterning and morphogenesis. In the proventriculus, Lin normally represses keyhole function (Fig. 3) and repression of Lin by Drm and Bowl is required to form the keyhole structure (Johansen et al., 2003). As in the hindgut, the drm-lin-bowl transcriptional hierarchy regulates expression of Upd, the ligand for the JAK/STAT signaling pathway, in the region of the foregut that will become the posterior part of the keyhole (Johansen et al., 2002). In contrast, other members of the JAK/STAT pathway have a broader expression pattern. This expression pattern parallels that in the hindgut epithelium where Upd is expressed specifically in cells of the small intestine and other JAK/STAT members are expressed throughout the hindgut epithelium (Iwaki et al., 2001; Johansen et al., 2002, 2003). In JAK/STAT pathway mutants, the anterior boundary cells of the proventriculus fail to migrate inward into the endoderm layer (Josten et al., 2004). The endodermal cell layer of the keyhole is also malformed in mutants for domeless, which encodes Upd receptor. The JAK/STAT pathway is also required for cell shape changes and migration of several cell types, including Drosophila tracheal cells and border cells of the Drosophila ovary (reviewed by Hou et al., 2002; Li et al., 2003), which indicates a general requirement for the JAK/STAT signaling pathway in morphogenesis.
Cell migration in the proventriculus also requires the localized activity of the Notch signaling pathway in the anterior and posterior boundary cells (Fuss et al., 2004). In Notch signaling mutants, such as Delta, Notch, Fringe, and Su(H), there is a delay in the inward migration of ectodermal cells of the keyhole region into the endodermal cell layer (Fuss et al., 2004). In addition, collapse of the endodermal proventriculus rim suggests a defect in the posterior boundary cells where the Notch receptor is expressed (Fuss et al., 2004). Notch-dependent signaling is mediated through the cytoskeletal linker protein Short Stop (Shot), a member of the spectraplakin superfamily (Roper et al., 2002). It is possible that Notch signaling mediated by Shot regulates organization of the actin cytoskeleton in a process also involving the small GTPase, Cdc42 (Fuss et al., 2004). Shot, in turn, is required for the proper localization of the Notch receptor in the boundary cells.
Fork head, which encodes a transcription factor with an HNF3/fork head DNA-binding domain is expressed in several tubular organs of epithelial origin, including the entire foregut region, the salivary gland, and the hindgut (Weigel et al., 1989a, b; Weigel and Jackle, 1990). fkh mutants display a variety of gut defects; the proventriculus and esophagus of the foregut fail to form, the hindgut fails to elongate, and the midgut and salivary glands disintegrate (Weigel et al., 1989a, b; Weigel and Jackle, 1990). Studies support a model in which fkh establishes signaling centers that define sites where specific morphogenic events are to occur by activating the hedgehog, wingless, and decapentaplegic (dpp) signaling pathways in the gut primordia. In embryos mutant for hh or a temperature-sensitive allele of wg, the keyhole structure forms; however, the proventricular cells fail to migrate inward (Pankratz and Hoch, 1995). The effects of wg and hh appear to be due to defects in morphogenic movements instead of cell proliferation, because proventricular cells that fail to migrate inward remain clustered on top of the proventriculus in both hh and wgts mutant embryos (Pankratz and Hoch, 1995). In contrast to wg and hh signaling, dpp signaling suppresses cell movements in the more anterior regions of the foregut, thus allowing only the keyhole region to undergo the morphogenic events that form the proventriculus (Pankratz and Hoch, 1995). Although fkh is required for cell migration in the proventriculus, fkh in the hindgut is required for cell rearrangements during tube elongation; in fkh mutant embryos the hindgut fails to elongate and instead remains as a short, straight tube (Weigel et al., 1989a). Thus, Fkh's ability to regulate different types of morphogenic movements may be determined by the temporal and spatial expression of target genes in the tissue of interest.
The inward migration of proventricular cells to form the cardiac valve also requires the integrin family of cell surface adhesion receptors, which mediate cell–cell and cell–substratum adhesions. Integrins exist as α β heterodimers with a β subunit associating with several different α subunits to form functional receptors (Hynes, 2002). In the developing proventriculus, the α2 subunit is expressed in the ectodermal layer of the keyhole region, the α1 subunit in the foregut ectoderm, and the β subunit throughout most of the foregut epithelium (Pankratz and Hoch, 1995). In embryos mutant for the β integrin subunit, encoded by the myospheroid (mys) gene (Leptin et al., 1989; MacKrell et al., 1998), migration of proventricular cells fail to occur and the cells remain clustered at the top of the esophagus, a phenotype similar to that of hh and wg temperature-sensitive mutants (Pankratz and Hoch, 1995).
Cell migration during proventriculus development also requires the activity of gap junction proteins which form intercellular channels that allow cells to exchange ions and small molecules (Goodenough et al., 1996). In embryos mutant for kropf, which encodes Innexin 2, a member of the innexin multigene family of gap junction channel proteins, the keyhole structure fails to form and subsequent migration of proventriculus cells fails to occur (Phelan and Starich, 2001; Bauer et al., 2002). Innexin 2 expression is regulated by the Wingless signaling pathway (Bauer et al., 2002). The innexin gap junction proteins are structurally and functionally analogous to the connexins, which encode gap junction channel proteins in vertebrates (Willecke et al., 1991). Integrins and innexins are also required for formation of other tubular organs. Mutations in the β-integrin subunit also affect salivary gland posterior migration (Bradley et al., 2003) and kropf mutants were recently shown to severely affect the size of the hindgut, Malpighian tubules, salivary glands, and trachea (Bauer et al., 2004). The studies described above have identified a wide range of genes involved in proventriculus cell migration. However, much still remains to be learned about the how the activities of these genes are coordinated spatially and temporally to result in the inward migration of cells into the endodermal region of the keyhole.
SALIVARY GLAND AND DUCT
The salivary gland consists of a pair of elongated secretory tubes that are connected to the larval mouth through the salivary duct tubes. Cells of the salivary gland synthesize and secrete protein during embryogenesis and throughout larval development, such as the salivary gland glue proteins, which allow the larva to adhere to solid substrates before pupariation. The tubes of the salivary duct system consist of two individual ducts that connect to the secretory tubes and a central common duct that connects the individual ducts to the larval mouth (reviewed by Bradley et al., 2001). The duct tubes transport the secretory products of the gland to the larval mouth. In addition to salivary duct and gland cells, the imaginal ring cells, which are the precursors to the adult salivary gland, form a ring of cells between the proximal ends of the secretory tubes and the distal ends of the individual duct tubes.
Like the hindgut and foregut, salivary gland formation begins with invagination of its primordial cells. The cellular and molecular events associated with salivary gland invagination are beginning to be understood in detail. During salivary gland invagination, columnar epithelial cells, which initially comprise two plates, or placodes, at the ventral surface of the embryo, change shape from columnar to pyramidal in a process involving constriction of the apical domains and basal migration of nuclei (Figs. 2, 6; Myat and Andrew, 2000a). Salivary gland cell shape change occurs in a regulated and sequential manner, starting with cells in the dorsal–posterior part of the placode, then dorsal–anterior, then ventral–anterior and, finally, the ventral–posterior domain (Myat and Andrew, 2000a). The order of invagination determines the final shape of the organ; in embryos mutant for the transcription factor Huckebein (Hkb) or the Ig-like cell-adhesion protein Faint sausage (Fas), the order of invagination is abrogated and abnormally shaped glands are formed (Myat and Andrew, 2000a). fas has also been shown to be required for specification and morphogenesis of the Drosophila heart tube (Haag et al., 1999).
The transcription factor Fork head (Fkh), which is required for formation of the hindgut, foregut, MT, and proventriculus, plays an important role in early aspects of salivary gland morphogenesis. Fkh is required for apical constriction of salivary gland cells. In fkh mutant salivary gland cells, apical domains do not constrict, even though nuclei migrate basally (Myat and Andrew, 2000b). fkh salivary gland cells do not change shape from columnar to pyramidal and fail to invaginate, most likely as a result of failure to constrict apically. In addition to its role in cell shape change, fkh is also required to promote secretory cell survival by preventing apoptosis (Myat and Andrew, 2000b). fkh mediates cell survival partly through senseless, which encodes a zinc-finger transcription factor (Chandrasekaran and Beckendorf, 2003). Downstream target genes of Fkh involved in salivary gland cell shape change are currently not known. In addition to its roles in early salivary gland development, fkh is also required during larval stages for the expression of the glue genes, which allow the larva to adhere to a solid substrate (Lehmann and Korge, 1996; Mach, 1996; Roth, 1999). Thus, fkh appears to function repeatedly throughout Drosophila embryo and larval development to regulate the expression of a variety of target genes involved in such diverse processes as apical constriction and cell survival. fkh's roles in promoting cell survival may not be limited to salivary gland cells. In fkh mutant embryos, the proventriculus and Malpighian tubules, among other structures, are missing (Jurgens and Weigel, 1988). The midgut epithelium of fkh mutant embryos degenerates and is populated by macrophages, suggesting that the absence of fkh function in the midgut may promote cell death, as in the salivary gland (Weigel et al., 1989). Interestingly, the fkh homolog in Caenorhabditis elegans, daf-16, acts in the insulin-receptor signaling pathway to lengthen life span (Lin et al., 1997).
The shape and size of the salivary gland can be regulated by the extent of apical membrane generated during invagination. In mutants where either too little or too much apical membrane is produced, the shape of the gland and the lumen are profoundly altered (Myat and Andrew, 2000a, 2002). In embryos mutant for huckebein (hkb), short, dome-shaped glands instead of elongated tubes are formed (Fig. 7). This change in tube morphology is due to insufficient growth of apical membrane during invagination (Myat and Andrew, 2002). Hkb generates apical membrane in salivary gland cells by positively regulating two targets, crumbs (crb), which encodes an apical membrane protein required for apical membrane specification and growth (Tepass et al., 1990; Wodarz et al., 1995; Tepass, 1996; Izaddoost et al., 2002; Pellikka et al., 2002), and klarsicht (klar), which encodes a putative regulator of dynein ATPase proposed to promote minus-end directed microtubule-dependent organelle transport (Fischer-Vize and Mosley, 1994; Welte et al., 1998; Mosley-Bishop et al., 1999). Hkb, in turn, is negatively regulated by the transcription factor, Hairy. In hairy mutants, prolonged expression of hkb is followed by increased apical membrane growth, which leads to the formation of either branched glands or glands with expanded lumina (Fig. 7). Thus, the transient and dynamic pattern of hkb and its target genes is critical for forming salivary glands of the correct size and shape.
After all the salivary gland cells have invaginated, the distal tip of the salivary gland tube comes in contact with the overlying tissue known as the visceral mesoderm (VM; Fig. 6). The distal most cells of the salivary gland turn and migrate posteriorly while remaining connected to the rest of the gland. Posterior migration of the salivary gland requires ribbon (rib), which encodes a nuclear BTB/POZ protein, (Bradley and Andrew, 2001; Bradley et al., 2003). In rib mutants, the salivary gland stalls at or near the point where the distal-most cells normally contact the VM and the gland fails to turn and migrate posteriorly (Bradley and Andrew, 2001). In the trachea, Rib allows branch elongation by controlling growth of the apical membrane (Shim et al., 2001). It is currently not known whether Rib also regulates salivary gland migration by a similar mechanism. BTB/POZ domain containing proteins may play general roles in the morphogenesis of tubular organs. For example, GZFI, a GDNF-inducible protein in the developing kidney with a BTB/POZ domain is highly expressed in the branching ureteric bud and collecting ducts of the mouse metanephric kidney (Fukuda et al., 2003). Furthermore, inhibition of its function interferes with ureteric bud branching in organ culture studies.
The VM is thought to provide both a suitable substrate for salivary gland migration as well as to act as a physical barrier to prevent movement in alternative directions. In mutations that affect either the migration or the specification of VM cells, such as those that affect components of the fibroblast growth factor (FGF) or Dpp signaling pathways, the VM is broken up and the salivary gland migrates aberrantly (Beiman et al., 1996; Gisselbrecht et al., 1996; Shishido et al., 1997; Michelson et al., 1998; Imam et al., 1999; Bradley et al., 2003). The salivary glands also fail to migrate posteriorly, even though they reach the turning point in embryos mutant for the integrin α-subunits PS1 and PS2 (Bradley et al., 2003). PS1 and PS2 integrins are also required for tracheal cell migration (Boube et al., 2001), demonstrating yet another parallel between salivary gland and tracheal migration. These studies collectively demonstrate that posterior migration of the salivary gland requires ribbon, integrins, and an intact VM.
Although development of the salivary glands has been extensively studied, little is known about salivary duct formation. The duct forms from two to three rows of the ventral-most cells of the salivary placode. After invagination of the salivary gland cells is complete, cells in the posterior half of the duct primordium invaginate to form the individual ducts followed by cells in the anterior half, which form the common duct (Jones et al., 1998). Duct morphogenesis involves coordinated cell rearrangements, similar to the cellular movements that result in elongation of the hindgut tube (Jones et al., 1998; Keller, 2002). Invagination of the duct cells requires the basic helix-loop-helix–PAS family transcription factor trachealess trh, which interestingly, also regulates tracheal invagination (Isaac and Andrew, 1996). A target gene of TRH in duct cells is eyegone (eyg), which is expressed in the individual duct and proximal secretory cells, as well as the cells that of the imaginal ring (Jones et al., 1998). Functional studies of eyg using embryos carrying a small deficiency that removes an estimated 20 genes, or embryos carrying the deficiency in trans to an inversion that maps near the eyg gene suggests a role for eyg in distinguishing the individual ducts from the common duct (Kuo et al., 1996) and in specification of the imaginal ring cells, a process also requiring Serrate (Ser), a ligand for the Notch receptor (Haberman et al., 2003).
Duct tubes of wild-type larvae have actin-rich rings along the apical surface facing the lumen (Haberman et al., 2003). These structures are absent in larvae mutant for the Notch ligand Serrate (Haberman et al., 2003). These apical rings may be precursors to the functional equivalent of tracheal taenidia, which are a series of ridges composed of extracellular matrix, thought to provide strength and flexibility. In support of the existence of taenidia-like structures in the salivary duct is the observation that many proteins encoding ZP domains, which form part of the apically secreted cuticle in the wing and tracheal epithelia, are also expressed in the salivary duct (Roch et al., 2002). Thus, several parallels exist between the duct tubes and tracheal tubes; both organs require trh for their invagination and both types of tubes may be strengthened by taenidia-like structures and/or an apical extracellular matrix deposited by ZP domain containing proteins.
In contrast to the other tubular organs discussed here, which consist of elongated, unbranched tubes, the trachea is an interconnected network of branched epithelial tubes responsible for transporting oxygen and other gases throughout the embryo. The entire network is formed from 10 placodes, or plates of approximately 90 ectodermal epithelial cells on each side of the embryo, which invaginate into the underlying mesoderm to form elongated sacs (Fig. 8A,B). Invagination of the tracheal cells requires the transcription factor Trachealess (Trh) and EGF and Wg signaling (Isaac and Andrew, 1996; Wilk et al., 1996; Llimargas and Casanova, 1999; Llimargas, 2000). The invaginated tracheal cells then migrate out to form six primary branches. Some of the primary branches, such as the anterior and posterior dorsal trunk and the visceral branch grow along the anterior–posterior axis, whereas the dorsal, lateral, and ganglionic branches grow along the dorsal–ventral axis. Subsequently, the dorsal trunk branches and the lateral branches on each side of the embryo fuse to form an interconnected tracheal network. After primary branching is complete, the secondary and tertiary branches form. Extension and fusion of some branches generates the highly branched interconnected tracheal tree present at the end of embryogenesis. The FGF signaling pathway through the FGF receptor Breathless (Btl) is one of the key signaling pathways that regulate tracheal branching morphogenesis. The FGF-like ligand Branchless (Bnl) is expressed in nontracheal cells surrounding the invaginating tracheal cells in a pattern that presages the migratory paths of the six primary branches. In response to Bnl, the invaginated tracheal cells expressing the Btl receptor and an FGF-specific downstream signaling protein encoded by heartbroken (also known as downstream of FGF or stumps) begin to migrate toward the Bnl source, thus forming the six primary branches (Michelson et al., 1998; Vincent et al., 1998; Imam et al., 1999). In embryos mutant for bnl, btl, or dof, tracheal cells fail to migrate to form the six primary branches (Michelson et al., 1998; Vincent et al., 1998; Imam et al., 1999). Furthermore, ectopic expression of bnl is sufficient to induce tracheal cell migration to novel sites (Sutherland et al., 1996). In vivo confocal microscopy revealed that Bnl acts as a chemoattractant to induce the formation of filopodial extensions exclusively at the tips of migrating tracheal cells (Fig. 8C,D; Ribeiro et al., 2002; Sato and Kornberg, 2002). The motile force generated by these tip cells is thought to pull along the rest of the cells of the growing branch, thus elongating the tube.
The Bnl/Btl signaling pathway establishes the general pattern of tracheal branching; however, additional signaling pathways, such as the Decapentaplegic (Dpp) and Wingless/WNT (Wg) pathways, control the migration of specific branches. Dpp regulates the migration of those branches that grow in the dorsal–ventral direction, such as the dorsal, ganglionic, and lateral trunk branches (Wappner et al., 1997; Vincent et al., 1998; Llimargas and Casanova, 1999), whereas Wg signaling is required for dorsal trunk formation, which grows in the anterior–posterior direction (Chihara and Hayashi, 2000; Llimargas, 2000). Wg signaling is also required for earlier invagination events and for establishment of the fusion cell fate. In the absence of Dpp signaling, cells of the dorsal branch initiate migration by forming filopodial extensions; however, dorsal branch migration fails as the bud-like outgrowths are subsequently resorbed into the dorsal trunk (Ribeiro et al., 2002). Thus, for productive outgrowth of the dorsal branch, both the Bnl/Btl and Dpp signaling pathways are required, with Bnl/Btl mediating filopodial extension and Dpp mediating additional cellular events, such as cell shape changes or cell adhesions specific to the dorsal branch.
As in salivary gland migration, tracheal cell migration also requires interactions with surrounding tissues. The visceral branch, which transports oxygen to the gut, contacts the visceral mesoderm during its migratory path. Drosophila integrins, αPS1 and αPS2 are specifically expressed on the surface of cells of the visceral branch and the visceral mesoderm, respectively. Interaction of these integrins with the surrounding extracellular matrix promotes migration of the visceral branch over the visceral mesoderm (Boube et al., 2001). Tracheal cells also rely on surrounding neural tissue for guidance during migration. For example, a single terminal cell of the ganglionic branch migrates through the ventral nerve cord following a tortuous path (Englund et al., 2002). This migratory path is controlled by the guidance protein Slit acting through Robo receptors and their effector Vilse, a Rac/Cdc42 GAP that guides the migration of both tracheal and neuronal cells (Englund et al., 2002; Lundstrom et al., 2004).
The tracheal system is composed of three types of tubes, distinguished by the number of cells that comprise each type of tube and the type of adherens junctions that seal the tube. Multicellular tubes, such as the dorsal trunk, consist of two or three cells surrounding a central lumen. Single-celled tubes, such as the dorsal branch at the end of embryogenesis, consist of a single cell that wraps around to form autocellular adherens junctions with itself. Lastly, subcellular tubes, such as those found in the terminal branches, are also single-celled but have an intracellular lumen (Samakovlis et al., 1996). The mechanism by which single-celled tubes with autocellular junctions are formed was largely unknown until recently, when Jazwinska and colleagues (2003) demonstrated a role for apical matrix proteins in this process. A family of genes encoding proteins containing a zona pellucida (ZP) domain is expressed in numerous epithelial tissues of the Drosophila embryo (Roch et al., 2002). Two of these ZP genes, piopio (pio) and dumpy (dp), are required for the conversion of multicellular tracheal tubes into single-celled tubes with autocellular junctions. During this remodeling process, cells that were once arranged side by side rearrange their positions by intercalating to adopt an end-to-end position. The dorsal movement of the cells toward the source of the Bnl chemoattractant is one possible force that elongates the cells and breaks the initial alignment between them. As the cells intercalate, the intercellular AJs of adjacent cells are progressively replaced by autocellular AJs. To ensure that the lumen remains connected, the conversion of intercellular AJs into autocellular ones needs to be stopped once the cells reach an end-to-end position. In embryos mutant for pio or dp, tracheal cells of the dorsal, lateral, and visceral branches fail to rearrange properly. As a result, hollow cysts form instead of fine tracheal tubes (Jazwinska et al., 2003). Pio and Dp are thought to form an apical extracellular matrix at the luminal side, which prevents the complete conversion of intercellular AJ into autocellular AJ, thereby restricting the reduction of the lumen. Such regulation is necessary to maintain luminal connectivity and epithelial integrity. The levels of E-cadherin during the cell rearrangements that convert intercellular AJs into autocellular ones also requires the small GTPase Rac (Chihara et al., 2003). The requirement for adhesion junction remodeling during tracheal cell rearrangement parallels that during convergent extension of the Drosophila germ-band (Bertet et al., 2004). In addition to pio and dp, 18 additional genes encoding ZP proteins are found in Drosophila (Jazwinska and Affolter, 2004). Many of these ZP genes are expressed in tubular organs of epithelial origin, such as the trachea, foregut, hindgut, Malpighian tubules, and salivary glands. However, the role of such ZP genes in the morphogenesis of tubes other than the trachea remains to be investigated.
After all the primary branches have formed, the lumen of tracheal tubes between adjacent metameres fuse in a process called anastomosis. The primary branches consist of tubes where the lumen is closed at the tips of the branches. Tube fusion occurs when specialized cells known as fusion cells located at the tips of the branches mediate branch fusion and give rise to a continuous tubular structure. Fusion of the dorsal trunk requires a specialized mesodermal cell called the bridge cell, which provides the substratum over which fusion can occur (Wolf and Schuh, 2000). In addition, E-cadherin, encoded by shotgun, and Armadillo, the Drosophila homolog of β-catenin, form a track together with F-actin and Shot, a cytoskeletal-linker protein most likely to guide the formation of new apical surfaces between the two fusion cells (Uemura et al., 1996; Beitel and Kransnow, 2000). The small GTPase RhoA is also required for anastomosis; gain-of-function RhoA mutants have a similar tracheal fusion phenotype to that of shot loss-of-function mutants. Thus, RhoA appears to negatively regulate assembly of the luminal track, perhaps through its regulation of the apical cytoskeleton or the transport of luminal contents (Lee and Kolodziej, 2002).
After primary branch formation, tracheal tubes undergo a tube expansion phase where their dramatic growth allows them to expand up to 40 times their original size. Although all branches of the tracheal system are interconnected, each branch has a characteristic size and shape and its growth is regulated independently. Recent studies identified two mechanisms by which tracheal tube expansion occurs. In the first mechanism, tracheal tube expansion is controlled at the apical surface of individual cells which faces the lumen. Morphometric measurements by Beitel and Krasnow (Beitel and Kransnow, 2000) demonstrated that during dilation of the dorsal trunk, dramatic growth of the apical surface was accompanied by very little change in the basal surface. The apical surface mechanism requires the activity of the Grainy head (Grh) transcription factor, which functions to limit apical membrane growth and restrict tracheal branch elongation in response to Bnl/Btl signaling (Beitel and Kransnow, 2000; Hemphala et al., 2002). Thus Bnl/Btl signaling is thought to first promote tracheal branching and tube elongation and then inhibit further luminal growth through activation of Grh. It is currently not know how Grh affects apical membrane growth in tracheal cells. Regulation of branch elongation by apical membrane growth appears to also require the putative transcription factor Ribbon, which likely functions to integrate multiple signaling pathways to affect the cytoskeleton (Bradley and Andrew, 2001; Shim et al., 2001). An apical surface mechanism is also used in the salivary gland to regulate tube size and shape; however, the molecular requirements appear to differ from those in the trachea (Myat and Andrew, 2002).
A second mechanism for tracheal tube size control involves septate junction organization. The invertebrate pleated septate junctions form a continuous barrier around cells to prevent the free diffusion of water and solutes (reviewed by Tepass et al., 2001). They are morphologically and molecularly distinct from the vertebrate tight junctions and, thus, are analogous rather than homologous. From mutagenesis screens, Beitel and colleagues identified a group of tube expansion mutants that affect the length and diameter of tracheal tubes without affecting early aspects of tracheal development (Beitel and Kransnow, 2000; Behr et al., 2003; Paul et al., 2003; Llimargas et al., 2004; Wu et al., 2004). Many tracheal branches, most noticeably the dorsal trunk of mutants, such as megatrachea (mega) and cystic, are more convoluted compared with wild-type, and the lumen is abnormally expanded and/or constricted (Fig. 9). Several of the genes corresponding to the tube expansion mutations encode components of septate junctions. For example, sinuous and mega encode claudin-like proteins, which are required for SJ organization, tracheal tube size control, and for paracellular barrier function (Behr et al., 2003; Wu et al., 2004); however, paracellular permeability and tube size expansion appear to operate by separate mechanisms because mutants affecting one function do not necessarily affect the other. One mechanism by which the SJs control tracheal tube size is by affecting cell shape because mutations in the SJ proteins, mega and lachesin (lac), which encode a cell surface protein mediating homophilic cell adhesion (Behr et al., 2003; Llimargas et al., 2004), cause tracheal cells to adopt an irregularly stretched morphology.
Tubulogenesis is a complex and fascinating process that appears to be regulated at multiple levels. Studies on development of different tubular organs in Drosophila embryos are beginning to elucidate the cellular and molecular mechanisms by which tubes of different sizes and shapes are formed. Although the many embryonic tubes of Drosophila differ morphologically, they share several key characteristics in their formation, not only cellularly, such as invagination of primordial cells and cell rearrangements during tube elongation, but also molecularly. Such similarities are evidenced by the requirement for Lin, Bowl, Drm, and JAK/STAT signaling in the hindgut and foregut tubes; in the requirement for Trh for the invagination of both the tracheal and salivary duct precursors; and in the requirement for Ribbon in the formation of the trachea, salivary gland, and Malpighian tubules (Table 1). Thus, it appears that a set of morphogenic regulators are used repeatedly to control the downstream events that sculpt a two-dimensional tissue into different types of three-dimensional structures. Such regulators appear to have a high degree of flexibility in terms of the cellular movements they regulate. For example, Fkh is required for cell rearrangements in hindgut elongation and also for apical constriction in salivary gland invagination. Similarly, the JAK/STAT signaling pathway regulates cell rearrangement in the hindgut and cell migration in the proventriculus. In the development of most tubular organs of ectodermal origin, epithelial invagination is the method of choice by which primordial cells of ectodermal origin are internalized. Subsequent cell migration and cell rearrangement events, which are sometimes coupled with cell proliferation, result in a tube of a unique size and shape. It is an extraordinary feat of nature to generate the wide diversity of tubular structures from only a handful of morphogenic movements. It is also remarkable that significant parallels exist between the formation of tubular organs in Drosophila and more complex organisms, such as those in the Drosophila Malpighian tubules and the vertebrate kidney. Thus, studies of tubular organs in Drosophila will continue to make great contributions to our understanding of developmental processes not only in the Drosophila embryo but also in vertebrate organisms as well as our understanding of certain diseased states. Whereas great strides have been made in the study of tube morphogenesis in Drosophila embryos, there is still much work ahead, as effector proteins need to be identified and the coordinated activities of transcription factors, signaling proteins, and effector proteins elucidated. The near future will indeed be an exciting time for the tube enthusiast, with studies in the Drosophila embryo spearheading the endeavor.
Table 1. Common Players in Morphogenesis of Tubular Organs
Gap junction proteins
Fork head, Drm, Lin, Bowl
Fork head, Ribbon
Fork head, Drm, Lin, Bowl
JAK/STAT/Upd, Notch/Delta/Ser, Hh, Wg, Dpp
Integrin α1/α2/β, Shot
Salivary gland and duct
Fork head, Ribbon, Trh
Htl, Hbr, Dpp Notch/Del/Ser
Integrin α1/α2/β, Faint sausage
Hh, Htl, Hbr, Wg, Dpp
Integrin α1/α2/β, Shot
I thank Markus Affolter, Greg Beitel, Bernhard Fuss, Frank Josten, and Marc Neumann for kindly providing figures and Tanya Gnanaraj for help with references. I thank Deborah Andrew, Greg Beitel, Pamela Bradley, Adam Haberman, Markus Schober, Melissa Vining, and members of my lab for their valuable criticisms and discussions about tube formation.