|Home | About | Journals | Submit | Contact Us | Français|
Epithelial tubes are a fundamental tissue across the metazoan phyla and provide an essential functional component of many of the major organs. Recent work in flies and mammals has begun to elucidate the cellular mechanisms driving the formation, elongation and branching morphogenesis of epithelial tubes during development. Both forward and reverse genetic techniques have begun to identify critical molecular regulators for these processes and have revealed the conserved role of key pathways in regulating the growth and elaboration of tubular networks. In this review, we discuss the developmental programs driving the formation of branched epithelial networks, with specific emphasis on the trachea and salivary gland of Drosophila melanogaster and the mammalian lung, mammary gland, kidney, and salivary gland. We both highlight similarities in the development of these organs and attempt to identify tissue and organism specific strategies. Finally, we briefly consider how our understanding of the regulation of proliferation, apico-basal polarity, and epithelial motility during branching morphogenesis can be applied to understand the pathologic dysregulation of these same processes during metastatic cancer progression.
Much progress has been made recently in delineating the cellular and molecular mechanisms driving tube morphogenesis. In this review, we discuss the major outstanding questions regarding the formation, elongation and elaboration of epithelial tubular networks. We pay particular attention to emerging common themes and differences between tube formation processes in Drosophila and mammals. We focus on those experiments that have revealed the underlying cellular and molecular mechanisms utilized during tube formation and branching morphogenesis. Finally, we consider whether present data argue for common cellular or molecular programs for tube morphogenesis or whether each organ utilizes a specific spatial and temporal configuration of more fundamental cellular and molecular mechanisms. It is our hope that this review will stimulate future work focused on resolving tube morphogenesis into a series of molecularly regulated changes in the discrete properties and behaviors of individual cells.
Tubular organization is a common feature of many developing tissues. Tubes can represent a transient phase of organ development. The vertebrate neural tube initially forms as a simple columnar epithelium, but ultimately gives rise to the complex architecture of the brain and spinal cord. Tubular organization is quite useful and can serve many important physiologic roles, including: control and delivery of gases, nutrients, waste and hormones, compartmentalization of organ function, and barrier function between the organism and its environment. The respiratory, circulatory, and secretory organs are all built of networks of interconnected tubes. Tubular epithelial organs arise from each of the germ layers, ectoderm (e.g. mammary gland), mesoderm (e.g. kidney), and endoderm (e.g. liver). Many of the same cellular and molecular mechanisms are utilized during the formation, elongation and elaboration of endothelial tubes, a topic that is reviewed in detail in this issue and elsewhere (Risau and Flamme, 1995; Beck and D’Amore, 1997; Carmeliet, 2000; Ellertsdottir et al., 2009). In this review, we focus specifically on epithelial tubular organs.
Epithelial tubes are only one of the constituent tissues of an organ, and are themselves surrounded by a complex mixture of supporting cells and extracellular matrix (ECM), collectively referred to as a stroma or mesenchyme. Mature epithelia are distinguished by a few key features, including close cell-cell contact and adhesion, strong apico-basal polarity, specialized intercellular junctions connecting neighboring cells, and a basally located basement membrane or basal lamina (Figure 1A). During the formation of an epithelial tube, each of these features can be independently modulated. Also, during morphogenesis and in the very early embryo, the organization of the epithelium of a given organ can vary significantly from its mature quiescent organization. Furthermore, there are some characteristic interspecies differences in epithelial organization (Knust and Bossinger, 2002). Chief among these differences is the organization and location of the barrier junctional complex. In mammals, tight or occluding junctions are composed of claudins and sit apical to the adherens junction (Figure 1A). In Drosophila melanogaster, the septate junction is also composed of claudins, but sits basal to the adherens junction. Both tight and septate junctions are responsible for controlling paracellular permeability (Knust and Bossinger, 2002).
As we consider the different model systems for studying tubular epithelia, it is worth noting that epithelia can have very different, but organ-specific, differentiated architecture and organization. Classically, the epithelia are subdivided into types, based first on the number of cell layers (Figure 1B). Single layered epithelia are termed simple and multilayered epithelia are termed stratified. Epithelia in which the nuclei occupy more than one layer, but all cells have direct contact with the basement membrane, are considered pseudo-stratified. In cases where the epithelium has several organizations in close juxtaposition, it is collectively considered to be a transitional epithelium. This basic classification scheme can be further refined to indicate cell shapes within the epithelium, whereby cells with similar height to width ratios are termed cuboidal, those with greater height than width are columnar, and those with greater width than height are squamous. Many of the secretory epithelia are further enveloped in a layer of cells expressing muscle markers such as alpha-smooth muscle actin. If these cells are between the epithelial cells and the basement membrane, they are termed myoepithelial cells, and if they are outside the basement membrane, they are termed myofibroblasts (Figure 1C). For any given tissue, there is a characteristic adult differentiated epithelial type, but the organization can be drastically different in early development or during developmental remodeling. For example, the mammary gland luminal epithelium lines and defines the luminal space and has a simple epithelial organization, but transiently reorganizes into a stratified epithelium during embryonic and postnatal branching morphogenesis (Hinck and Silberstein, 2005; Ewald et al., 2008;). We might expect that different strategies are employed to build different types and architectures of tubular networks.
In contrast to epithelia, mesenchymal tissues are much looser and are characterized by a predominance of single cells dispersed through extensive ECM. Integrin based cell-matrix adhesions are relatively more common than cadherin based intercellular adhesions. These distinctions are not absolute however, as epithelial tissues can undergo an epithelial to mesenchymal transition (EMT) and give rise to dispersed single cells. Prominent physiologic examples of EMT include the vertebrate neural crest, the ingressing cells of the chick epiblast during gastrulation, and the sclerotome and associated derivatives of the ventral portion of the somite. The converse process of mesenchymal to epithelial transition (MET) can also occur, most notably in the condensation of mesodermally-derived mesenchymal cells in the formation of the vertebrate kidney. At present, it is unclear whether epithelium and mesenchyme should be considered to be two alternative organizations or two opposing poles that define a continuum of organization (O’Brien et al., 2002). As we begin to understand the cellular processes underlying epithelial morphogenesis, we need to pay close attention to the extent to which epithelial organization is maintained throughout the process. Perhaps the most critical aspect of this organization is the degree of apico-basal polarity. It may be easier and more reliable to monitor and dissect the regulation of apico-basal polarity specifically rather than epithelial organization more broadly.
As we try to understand tube morphogenesis, we should start with: what needs to be accomplished by the tissue? Then, how do the cells do it? Then, what are the molecules that mediate/direct/specify the cell behaviors? Finally, which cellular and molecular mechanisms are general across tissues or across phyla? There are several distinct subprocesses that must occur for an undifferentiated early embryonic tissue to become a mature differentiated tubular network. The precise order of these events can vary dramatically between tissues, but each of these subprocesses must occur at least once.
First, the cells of an epithelial tube must be specified and made molecularly distinct from the surrounding cells that will give rise to other tissue types. This epithelial identity then needs to be maintained, though systems vary greatly in the extent to which it is continuously maintained or iteratively de- and re-differentiated. It is also possible for tubular morphogenesis to include a full or partial EMT followed by reestablishment of full epithelial identity. Once the future epithelial tissue is specified, a lumen needs to be built, as a tube is defined by the existence of a specialized luminal space (Figure 2A). As with epithelial identity, different systems may vary in the extent to which the lumen is permanently maintained once established. Tubes then need to elongate and elaborate (Figure 2B-E). Three conceptually distinct mechanisms exist for tube elaboration. The end of the tube can split, with further elongation in both directions, a process known as bifurcation. New tubes can initiate from sites distant from either tube end, a process referred to as side branching. The tube can separate and exchange cell-cell contacts for cell matrix contacts without significant elongation, a process referred to as clefting. Additionally, both the diameter and length of the tube can be regulated and the final tissue and cell type specific differentiation program needs to be completed (Figure 2F). It is important to distinguish to what extent these subprocesses of tube specification, lumen formation, elongation, and elaboration occur as a temporally distinct series of events or whether one or more of these subprocesses occurs simultaneously.
A critical starting point in understanding the elongation and elaboration of tubular networks is to define the simplest functional unit of morphogenesis. To what extent is the process of epithelial branching morphogenesis meaningfully autonomous to the epithelium? What is the minimum set of cell types and extracellular matrix that is sufficient to build the form and differentiated function of a given epithelial tube? In the mammary gland, there is clear coordination between epithelial cells and stromal cells (Van Nguyen and Pollard, 2002; Ingman et al., 2006), as well as coordinate motility in different epithelial cell types such as the luminal and myoepithelial cells (Ewald et al., 2008). In the mouse lung, it is also clear that important patterning information is exchanged between the epithelial and mesenchymal compartments, with the lung vasculature playing an important signaling role (Warburton et al., 2000). It is important to determine whether epithelial tube morphogenesis can be fully captured based on an understanding of the actions of the epithelial cells alone, or whether it is necessary to understand the epithelium in its normal context of extracellular matrix and stromal cells (Nelson and Bissell, 2006). It is clear in the mammalian kidney (Constantini, 2006; Dressler, 2006; Nigam and Shah, 2009), lung (Metzger et al., 2008), salivary gland (Patel et al., 2006) and mammary gland (Lu and Werb, 2008) that there is a close juxtaposition of many stromal cell types basal to the epithelium during branching. There are also changes in the stromal composition and differentiation during branching. In the mammary gland, loss of specific stromal cell populations has a significant effect on ductal elongation (Ingman et al., 2006). Further work is required in each of these systems to determine the extent to which epithelial-stromal communication is a refining mechanism or a central source of patterning information.
In an ideal world, we would have a single model system that developed and elaborated tubular networks through conserved mechanisms and that was ideally suited for genetics, cell biology and imaging. Instead, we have a number of different model systems with different strengths and limitations. Some reagents and techniques work best in specific systems. Simplified cell culture models with immortalized cells have provided critical insights into the molecular mechanisms of lumen formation and apico-basal polarity (Debnath and Brugge, 2005; O’Brien et al., 2002). These models have been more limited for modeling the 3D complexity of organ architecture, but have provided mechanistic insights relevant to in vivo development.
Recently, techniques have been developed in multiple labs to model the development of epithelial organs in a more organotypic fashion, with successful protocols established for mammary development (Simian et al., 2001; Wiseman et al., 2003; Meyer et al., 2004; Fata et al., 2007, Ewald et al., 2008), submandibular gland (Steinberg et al., 2005), and lung (Liu et al., 2004). These cultures are alternately referred to as organotypic, organoid, or mesenchyme-free epithelial cultures. All share conceptually similar isolation protocols in which the epithelial compartment of the gland is separated from the stromal compartment and then cultured in 3D gels of ECM proteins. It is worth noting that these preparations are rarely completely mesenchyme free, as there are tightly attached stromal cells that can readily be carried along with the epithelial fragments. These cultures are modular in that they allow arbitrary recombinations of epithelium, stroma, and ECM. This experimental flexibility allows the role of complex factors to be tested in an organotypic context. For example, it is difficult in vivo to eliminate whole classes of stromal cells or stromal proteins without systemic effects elsewhere in the animal, whereas it is comparatively easy in 3D culture. It is also valuable that similar techniques can be used with primary human epithelium (Yu et al., 2007) and so insights from model systems can be readily and directly tested in human disease states.
It is also possible to culture the entire epithelial organ (Sakai and Onodera, 2008), most notably in salivary gland (Larsen et al., 2003), kidney (Srinivas et al., 1999), lung (Bellusci et al., 1997), and mammary gland ( Topper et al., 1975; Gallo-Hendrikx et al., 2001). Since the epithelium is cultured in its normal stromal environment, development is often even more similar to normal in vivo development. These cultures are a major advantage when the molecules being tested are broadly required in the early embryo (e.g. E-cadherin or fibronectin). Importantly, organ cultures make mammalian organ development observable. Branching morphogenesis in mammalian tissues occurs inside an embryo inside the mother (salivary, lung, etc) or inside an adolescent mouse over weeks (mammary gland). A major advantage to externally developing embryos is that organ development can often be directly imaged in the intact organism. Drosophila in particular has been a rich experimental system for understanding tube morphogenesis as the genetics are well developed and live imaging in vivo is practical due to the clarity of the embryo and the relatively close proximity of relevant tissues to the embryo surfaces (Ribeiro et al., 2002; Kato et al., 2004; Caussinus et al., 2008; Cheshire et al., 2008). Similarly, in vivo imaging of organ development in zebrafish is beginning to provide additional insights into the mechanisms of tube morphogenesis, most notably during kidney development (Vasilyev et al., 2009). The availability of these various model systems increases the importance of the validation of in vitro results in vivo and also of the cross-validation of results from one model system to others and to human development and disease.
The morphogenesis of epithelial tubes is intrinsically a question of the morphogenesis of a tissue, not a single cell. This requires both coordination among the cells of the epithelium and between the epithelial compartment and surrounding stromal cells. We seek to resolve complex tissue transformations (e.g. mammary or tracheal branching morphogenesis) into a series of discrete changes in the properties and behaviors of the individual constituent cells of that tissue. We expect that the apparently complex tissue behavior of branching morphogenesis in different systems can be explained as unique configurations of simpler, conserved subprocesses. Essentially the problem is: what happens when and how is it regulated?
Imaging provides the foundational description of the cellular and molecular events that occur during branching morphogenesis. Live cell imaging is particularly important as it enables accurate reconstruction of the sequence of these events. In highly stereotyped or geometrically simple tissues, a series of fixed samples can provide a useful starting point, but there is no substitute for watching the same cells over time. When you are additionally interpreting the phenotypes of molecular perturbations with less than 100% penetrance or some degree of mosaicism, the problem is compounded further. Even a single high-resolution movie can unambiguously resolve the sequence of cellular processes involved. Recent improvements in computer hardware, software and microscope automation enable simultaneous collection of dozens to hundreds of time-lapse movies. This multiplicity enables researchers to reconstruct most or all of the different normal developmental trajectories. Coupled with recent advances in molecular imaging, it is now a reasonable goal to image branching morphogenesis in each of the major models systems, with distinct channels of information for the dynamics of actin, microtubules, junctional complexes and various signaling molecules.
With an ever increasing understanding of the spatial and temporal dynamics of cells driving tube morphogenesis, attention naturally turns to identifying the critical molecular regulators for these cell behaviors. Forward and reverse genetic approaches have been and will continue to be used to isolate and identify molecular regulators governing these changes in cell behaviors and properties. Mammalian and Drosophila studies have diverged significantly in their technical approaches to these questions, with forward genetic screens dominating in Drosophila and reverse genetic approaches, chiefly gene deletion by homologous recombination, providing the mainstay of murine genetic approaches to branching morphogenesis. The challenge today is to integrate the combined insights of imaging and genetics to achieve a cellular resolution understanding of tube morphogenesis and its regulation. There is a critical role for mammalian systems with close physiologic similarity to humans and for simpler model systems with unique experimental advantages.
Tube architecture can vary enormously; the basic structure, nonetheless, is a highly polarized epithelium surrounding a central luminal space. The (free) apical surfaces of the epithelial cells face the lumen, the basal surfaces contact other tissues or a basement membrane, and the lateral surfaces connect adjacent epithelial cells through specialized junctions. The junctions serve to anchor the cells, provide barrier function and segregate the different membrane components – both proteins and lipids – into their appropriate compartments (Figure 1; Martin-Belmonte and Rodriguez-Fraticelli, 2009). Thus, at its core, tube formation simply requires the assembly of a polarized epithelium around a central lumen (Figure 2). This cellular organization is achieved in some tissues by remodeling an already polarized epithelium to surround a luminal space and in others by polarizing the precursors and creating luminal space de novo (Bryant and Mostov, 2008; Chung and Andrew, 2008).
In tubular organs that form from an already polarized primordium, two closely-related processes have been described: wrapping or budding (Figure 3A,B; Lubarsky and Krasnow, 2003). In both processes, the tube primordium folds inward to create a trough or cup-like structure. The shape of the nascent structure depends on the geometry of the primordium with respect to the surrounding epithelia, and the number and position of cells that are undergoing cell shape changes at any given time. With wrapping, the final tube is parallel to the plane of epithelial origin and is eventually completely separated from the surrounding non-tube epithelia by the fusion of epithelial folds at the edges of the tube primordium. With budding, the tube forms by the extension of an invagination or evagination of cells orthogonal to the plane of the epithelia. The elongated bud often remains contiguous with the epithelium of origin. With both wrapping and budding, the integrity and polarity of the epithelial primordium is maintained throughout the process of initial tube formation.
Most of the anterior portion of the neural tubes of birds and mammals, and the entire neural tube of amphibians form by wrapping during a process called primary neurulation. The neural tube primordium arises from an already polarized epithelium. This primordium is initially created early in embryogenesis in birds and mammals when the inner cell mass delaminates into the epiblast and hypoblast. Cells that remain in the epiblast and that do not move inward during gastrulation form the ectoderm, with the neural tube precursors positioned along the axis of the dorsal midline. The neuronal precursors are polarized with their apical (free) surfaces facing out and their basal surfaces contacting the underlying mesoderm. Primary neurulation begins with the elongation of neuronal precursors along their apical-basal axes to form the columnar cells of the neural plate. Elongation is a microtubule-driven process referred to in some embryology textbooks as “palisading” (Schoenwolf, 1983; Schoenwolf and Powers, 1987). Microtubule assembly in this system is controlled by γ-tubulin (Gunawardane et al., 2000; Meads and Schroer, 1995), which localizes to an apical domain within epithelial cells through the action of the Shroom family of PDZ domain and actin-binding proteins (Lee et al., 2009; Lee et al., 2007). Shroom proteins have been shown to regulate both apical-basal elongation and apical constriction (Haigo et al., 2003; Hildebrand and Soriano, 1999; Hildebrand, 2005) (Lee et al., 2007; Lee et al., 2009; Sawyer et al., 2009). Subsequently, the inward bending or folding of the neural plate along the midline and into the embryo brings the lateral edges together to seal off the tube. Wrapping is driven in part by contraction of the apical cell surfaces, an actinomyosin-driven process, and in part by the basal positioning of nuclei, a nuclear position that in the neural tube is linked to cell division (Smith and Schoenwolf, 1987; Smith and Schoenwolf, 1988). Apical constriction combined with basal nuclear positioning results in the formation of the wedge-shaped cells observed at the “hinge-points”, where the neural tube folds during neurulation. The neural tubes of teleosts as well as the caudal portion of amniote neural tubes form from an unpolarized cord of cells through a process often referred to as secondary neurulation (Schoenwolf, 1983; Schoenwolf, 1984; Schoenwolf and Delongo, 1980).
The Drosophila salivary gland and trachea begin as polarized epithelia that form during cellularization of the earlier blastoderm stage embryo. During cellularization, the nuclei migrate to the periphery of the embryo and are separated into individual cells by the inward growth of surface membrane (Foe and Alberts, 1983; Knoblich, 2000; Mazumdar and Mazumdar, 2002). During the cellularization process, the ~6000 cells at the surface, which will eventually give rise to nearly the entire embryo, become polarized with their apical surfaces free and facing out and their basal surfaces in and facing the yolk (Knoblich, 2000; Mavrakis et al., 2009). Within this polarized monolayer of cells, the salivary gland and trachea are specified through the combinatorial action of early patterning genes (Kerman et al., 2006).
The Drosophila salivary gland forms from the ventral ectodermal cells in the most posterior region of the developing head. The cells on each side of the ventral midline will form the salivary duct tubes and the more laterally-positioned cells on each side will give rise to the simple, unbranched secretory tubes. Once specified, salivary gland cells cease dividing, greatly simplifying morphogenetic studies in this system. Similarly, programmed cell death plays no role in salivary gland formation. Morphogenesis of the secretory portion of the salivary gland has been extensively analyzed by staining and analysis of whole embryos, detailed analysis of histological sections, scanning and transmission electron microscopy, confocal imaging and, more recently, live two-photon imaging (Sonnenblick, 1950; Panzer et al., 1992; Myat and Andrew, 2000a; Myat and Andrew, 2000b; Cheshire et al., 2008). Like the neural tube, the first morphological change is the elongation of the precursors along their apical-basal axes to form the columnar cells of the salivary gland placode. The cells internalize through a budding type mechanism beginning with a small group of cells in a dorsal posterior position in the placode, followed by dorsal anterior, ventral anterior and, finally, ventral posterior cells (Myat and Andrew, 2000b). Internalization occurs by apical constriction wherein the nuclei move to a basal position and the apical surface constricts. As has been proposed for primary neurulation, apical constriction and basal nuclear localization can be uncoupled during salivary gland budding; mutations in fork head (fkh), which encodes a FoxA transcription factor required for apical constriction, do not affect basal nuclear positioning (Myat and Andrew, 2000a). The salivary gland primordia completely fail to internalize in the fkh mutants, however, providing a direct link between apical constriction and invagination/budding.
The Drosophila trachea forms from 10 groups of approximately 40 cells each on both sides of the embryo in regions corresponding to the second thoracic segment through the eighth abdominal segment. Tracheal cells also elongate along their apico-basal axes to form tracheal placodes. Similar to the secretory tubes of the salivary gland, tracheal cells invaginate through an apical constriction mechanism, beginning with 2-3 cells along the anterior-posterior midline of each placode, with ventral cells invaginating slightly ahead of the more dorsal cells. Tracheal cells undergo their final round of division during invagination and recent live imaging studies of tracheal invagination suggest that both cell rearrangement and oriented mitotic divisions also contribute to tracheal cell internalization (Brodu and Casanova, 2006; Nishimura et al., 2007). Tracheal cell invagination absolutely requires the Trachealess (Trh) bHLH-PAS transcription factor. In trh mutants, tracheal cells fail to undergo apical constriction, fail to enrich F-actin at the apical surface and completely fail to internalize (Isaac and Andrew, 1996; Llimargas and Casanova, 1999). Trh functions in part through transcriptional activation of the rhomboid gene, which encodes a transmembrane protease essential for processing of the EGF ligand (Lee et al., 2001), and a downstream EGF effector, crossveinless-c (cv-c), which encodes a Rho-GAP linked to cytoskeletal changes during invagination (Boube et al., 2000; Zelzer and Shilo, 2000; Brodu and Casanova, 2006). rhomboid is the spatially limited and thus regulating component for EGF signal activation in Drosophila since the EGF ligand, receptor and downstream effectors are expressed relatively ubiquitously. Although rhomboid mutants and embryos mutant for other EGF pathway components and downstream effectors show some delay in invagination, most cells eventually internalize, demonstrating that other factors downstream of Trh are also key to the tracheal invagination (Llimargas and Casanova, 1999; Brodu and Casanova, 2006). Once internalized, subsets of tracheal cells will migrate away from each other in a stereotypical fashion to form the primary branches. The formation of primary branches involves the same type of budding mechanism that is used to internalize the salivary secretory glands and the trachea as well as to elaborate the tubes of the mammalian lungs, kidney and pancreas (Kim and MacDonald, 2002; Ghabrial et al., 2003; Kumar and Melton, 2003; Metzger et al., 2008; Nigam and Shah, 2009).
Wrapping and budding are driven in part by apical constriction, a cell shape change that promotes the inward bending of a polarized epithelium (Sawyer et al., 2009). Apical constriction requires the localized activation of Rho kinases at the apical surface, a process driven by transcriptional activation of signaling cascades as well as the apical localization of key actin binding proteins, such as Shroom3 in vertebrates and folded gastrulation and DRhoGEF2 in Drosophila (Barrett et al., 1997; Hacker and Perrimon, 1998; Haigo et al., 2003; ; Dawes-Hoang et al., 2005; Hildebrand, 2005; Kolsch et al., 2007; Nishimura and Takeichi, 2008; Sawyer et al., 2009). Apically localized Rho kinase activity (Rok in Drosophila, and ROCK1 and ROCK2 in mammals) leads to the localized activation and accumulation of non-muscle myosin. In the amniote neural tube as well as in other Drosophila tissues that undergo apical constriction, the Abelson tyrosine kinases function to recruit and organize filamentous actin at the apical surface (Koleske et al., 1998; Fox and Peifer, 2007). Myosin contraction results in apical constriction through the tethering of the actinomyosin cytoskeleton to the circumferential apical adherens junctions. Apical constriction has long been proposed to occur through a “purse-string” mechanism wherein apical constriction is driven by a continuous purse-string like contraction of the actinomyosin belt underlying the adherens junctions (Rodriguez-Diaz et al., 2008). Recent live imaging studies on Drosophila mesoderm formation, however, suggest that apical constriction may instead occur through a ratcheting mechanism wherein repeated asynchronous myosin contractions at the medial apical cortex pull the adherens junctions inward (Martin et al., 2009). These constrictions, which are maintained between the pulses of myosin contraction by the actinomyosin belt underlying the adherens junctions, serve to incrementally reduce apical area, thus driving the cell shape changes required for remodeling epithelia to form tubes (Martin et al., 2009).
Though tubes can undergo significant morphogenesis while remaining highly polarized, many tubes polarize after morphogenesis or go through cycles of polarization, depolarization and repolarization. This creates the challenge of building an epithelial tube from unpolarized precursors. Three general mechanisms have been described for the formation of tubes from unpolarized rudiments: cord hollowing, cell hollowing and cavitation (Lubarsky and Krasnow, 2003). Cord hollowing occurs with MDCK cells grown in 3-D matrices in conditions favoring rapid polarization (Martin-Belmonte et al., 2008), during formation of the zebrafish gut (Bagnat et al., 2007) and during secondary neurulation in the chick and mouse embryo (Schoenwolf and Delongo, 1980; Schoenwolf, 1984). Cord hollowing involves the formation of several small lumena separating the apical domains of subsets of cells within the primordia as the cells acquire polarity (Figure 3C) (Lubarsky and Krasnow, 2003). The small apical lumena subsequently coalesce into a single large common lumen, a process that for the zebrafish gut is likely mediated by paracellular ion transport driving accumulation of lumenal fluid (Bagnat et al., 2007).
The related process of cell hollowing has been described for several models of tube development, including cultured HUVEC cells grown in 3-D matrices (Bayless et al., 2000; Bayless and Davis, 2002), the C. elegans single excretory (kidney) cell (Buechner, 2002), zebrafish blood vessels (Kamei et al., 2006; Blum et al., 2008), and the Drosophila tracheoles (Guillemin et al., 1996). Lumens initiate as small cytoplasmic vesicles that fuse with each other to produce large intracellular lumena (Figure 3D). These lumena ultimately fuse with the apical plasma membrane to connect with the lumena of adjacent cells within the tube. The distinction between cord hollowing and cell hollowing is in whether the luminal space first forms extracellularly by direct fusion of vesicles with the plasma membrane (cord hollowing) or intracellularly by fusion of vesicles with each other (cell hollowing). Since both processes involve the formation and fusion of vesicles eventually destined for the apical plasma membrane, the underlying molecular mechanisms are likely to be quite similar. Tubes that form by budding also require the formation and fusion of vesicles with the apical membrane for growth, suggesting very similar mechanisms for luminal expansion during tube development (Seshaiah et al., 2001; Myat and Andrew, 2002; Behr et al., 2007; Jiang and Crews, 2007; Tsarouhas et al., 2007; Kakihara et al., 2008; Kerman et al., 2008; Shaye et al., 2008).
The third mechanism for forming tubes from unpolarized rudiments, cavitation, begins with the polarization of cells on the periphery of the condensation, followed by the apoptotic death of cells in the center to create a lumenal space (Figure 3E). Examples of tube formation by cavitation are seen with mammary (MCF-10A) cells grown in 3D matrices (Debnath et al., 2002), with MDCK cells grown in matrices under conditions that delay polarization (Martín-Belmonte et al., 2008) and is the proposed mechanism for tube formation in the caudal half of the amniote neural tube (Schoenwolf, 1984). Importantly, apoptosis has been observed in vivo during lumen formation in both the mammalian mammary and salivary glands (Humphreys et al., 1996; Jaskoll and Melnick, 1999). Based on studies in both the kidney (MDCK) and mammary (MCF-10A) cell models of tube formation, the role of programmed cell death is not in establishing polarity or even in creating an apical lumen but rather in the clearing and maintenance of an open lumenal space. In the MDCK cell model, lumen formation can occur either through cord hollowing or by cavitation; the choice of mechanism depends on the efficiency of cell polarization (Martín-Belmonte et al., 2008; Martín-Belmonte and Rodriguez-Fraticelli, 2009). When MDCK cells are grown in the appropriate type of ECM, they polarize quickly and form an open lumen with little or no cell death. When cells are grown on a less differentiated matrix, the cells must first synthesize an ECM that favors polarization and then polarize. As a consequence of the delay in polarization, far more cells are present and thus trapped in the lumen when the cyst polarizes; these cells subsequently die by apoptosis. Thus, in this system, programmed cell death provides a mechanism for clearing cells that fail to polarize and contribute to the epithelium quickly enough (Martín-Belmonte et al., 2008).
Similarly, antibody blocking and siRNA knockdown of E-cadherin in cultured mammalian salivary gland explants, a treatment that should delay or block polarization, results in the formation of aberrantly dilated lumens filled with apoptotic cells (Walker et al., 2008). As mentioned earlier, MCF-10A cells form cysts in culture by first polarizing cells in the periphery and subsequently clearing cells in the interior by apoptosis (Debnath et al., 2002). In this system, apoptosis also functions to maintain an open lumen; stimulating cell division after the cyst has formed does not affect cyst architecture since the central cells are simply removed by apoptosis. Stimulating cell division under the same conditions, but blocking apoptosis by the expression of apoptotic inhibitors, results in the filling of the lumenal space (Debnath et al., 2002). In the MCF-10A model of mammary gland morphogenesis, apoptotic cell death appears to be induced in cells only after the peripheral ductal cells have polarized and established contact with a basement membrane (Reginato et al., 2005). In this and related cell culture models of tubulogenesis, the lumenal cells may die by a form of programmed cell death known as anoikis, which occurs when epithelial cells lose contact with the ECM (Meredith et al., 1993; Frisch and Francis, 1994; Ruoslahti and Reed, 1994). Interestingly, even in the mammary glands, where programmed cell death is thought to be a key event in tube formation (Debnath et al., 2002), blocking apoptotic cell death in vivo does not prevent polarization of the tube epithelia and only delays lumen clearing; the lumenal cells are eventually removed through caspase-independent cell death (Mailleux et al., 2007). Thus, cavitation could be viewed as a special case of cord hollowing, which requires the extra step of clearing the lumen, a process typically occurring through programmed cell death.
With all mechanisms of tube formation, cell polarization is the critical step, even if it occurs several days or hours before lumen formation, as is observed with the formation of tubes from an already polarized epithelial primordia. In forming tubes from an unpolarized rudiment, cells use a variety of strategies to create a polarized epithelium of the correct size and spacing (Eaton and Simons, 1995; Nelson, 2003; Mostov et al., 2005; Nelson and Bissell, 2006; Chung and Andrew, 2008; Martin-Belmonte et al., 2008; Martin-Belmonte and Rodriguez-Fraticelli, 2009). In MDCK cells, polarization is linked to differentiating signals provided by the ECM (Yu et al., 2005). In zebrafish neural keel and human intestinal epithelial cells, polarization is linked to cell division (Tawk et al., 2007) (Jaffe et al., 2008). Whereas, in the morphogenesis of bile ducts and lung aveoli, cell division is completely absent and apparently unnecessary (Tanimizu et al., 2009; Yu et al., 2007).
Despite the investments of tube precursors to form the initial polarized epithelia, many tissues will lose that polarity, at least partially, as tubes grow and are remodeled during different developmental stages. The mammary gland iteratively polarizes to simple organization and depolarizes to a stratified organization throughout the lifecycle (Hinck and Silberstein, 2005; Watson and Khaled, 2008; Ewald et al., 2008). Only a few examples of tissues are known that once polarized maintain this polarity through tube elongation and migration; well-studied examples include the Drosophila salivary gland and the vertebrate kidney (Figure 6A; (Kerman et al., 2006; Drummond, 2003). Although the Drosophila trachea, which also forms through a budding mechanism, will largely maintain its polarity, the process of fusion, wherein the connections between the tracheal segments form, involves de novo formation of an ectopic apical domain on the basal sides of fusing cells (Figure 6B; Lee and Kolodziej, 2002; Jung et al., 2005; Tanaka et al., 2004; Jiang and Crews, 2007; Kakihara et al., 2008). Similarly, forming the fine lumena of the tracheoles requires de novo formation of apical domains in terminal cells through a cell hollowing type mechanism (Guillemin et al., 1996). As discussed in the subsequent sections, growth of many tubes involves either partial loss of polarity or a lag in polarization following tube elongation.
Recent studies of de novo tube formation in two very different settings illustrate the range of potential strategies for creating polarized epithelial precursors. The first example of an unexpected mechanism for achieving epithelial polarity links cell division to the acquisition of apical basal polarity (Figure 3F). The zebrafish neural tube precursors first form a plate of condensed unpolarized cells that converge to the dorsal midline. Upon their arrival at the dorsal midline, the neuronal precursors undergo a mediolaterally-oriented cell division (Kimmel et al., 1994). The daughter cell that arises from what was originally the more superficial side of the cell, will cross the midline, intercalate and stretch to ultimately contact the future basal surface on the side opposite its origin. This cell division creates mirror image daughter cells with regards to apicobasal polarity and cell morphology (Tawk et al., 2007). Apicobasal polarity is established prior to completion of cell division, demonstrated by the localization of the apical marker protein Pard3-GFP to the site of the future cleavage furrow. Pard3-GFP is maintained at this position, the future apical domain. In planar cell polarity (PCP) mutants, where convergence of the neural plate is delayed, the cell divisions take place on both sides of the dorsal midline leading to the formation of duplicate neural tubes with mirror-image apicobasal polarity. Importantly, blocking or delaying cell division rescues the neural tube defects of PCP mutants (Ciruna et al., 2006; Tawk et al., 2007). These studies suggest that the polarity established during cell division can serve to localize apical patterning information. This mechanism may be quite general, since Caco-2 cells grown in 3-D matrices also establish a discrete apical domain at the cleavage plane between dividing daughter cells, with the apical domains remaining positioned near the center of the developing cysts (Jaffe et al., 2008). Linking polarity to cell division occurs in a number of cell types, including in single cell organisms such as Dictyostelium, where PTEN (the enzyme normally enriched at the apical surface of polarized epithelia that converts PIP3 to PIP2 (Martin-Belmonte et al., 2007); localizes to the cleavage furrow and PI3-Kinase (the enzyme that converts PIP2 to PIP3) localizes to the distal ends of dividing cells (Janetopoulos et al., 2005).
Another interesting example of cells acquiring the correct polarity to form lumena is seen in organotypic cultures that simulate bile duct formation (Tanimizu et al., 2009). The bile ducts form tubes from a monolayer of polarized epithelial cells that surround the portal vein. Recent work with an in vitro model of bile duct formation suggests a novel mechanism for creating a centralized lumen surrounded by a correctly polarized epithelium. In the organotypic culture model, cells are first grown to a monolayer on a collagen-matrigel matrix. Once an epithelial monolayer has formed, a second layer of the same matrix is applied to the top. In response to this second layer of matrix, a subset of cells within the epithelial monolayer loses polarity and extrudes from the epithelia. These cells subsequently migrate over the epithelial cells that remain in the monolayer to form a second cell layer (Tanimizu et al., 2009). This second layer of cells then repolarizes in reverse orientation to the original epithelium and a lumenal space is created between the two cell layers. The repolarization of a subset of these epithelial cells suggests that polarity cues are provided by the collagen-matrigel matrix. The loss and subsequent recovery of polarity provides a mechanism for maintaining contact between the basal surfaces of epithelial cells and the ECM while at the same time creating a localized apical domain facing away from the ECM. The formation of tubular structures in this system occurs in the absence of cell death or cell division, consistent with in vivo bile duct formation, where neither extensive death nor division are observed during the tube forming stages (Tanimizu et al., 2009). As live imaging techniques are applied to more examples of in vivo tissue morphogenesis and organotypic culture, we are likely to discover even greater diversity in the mechanisms through which appropriate tube architecture is achieved.
The growth or elongation of tubes can occur through several distinct cellular mechanisms including changes in cell shape, cell arrangement, cell division, and cell recruitment (Figure 4). There is also a diverse range of morphologies and configurations at the end of elongating epithelial tubes. The tip of the tube can be composed of single, few or many cells. The elongating tube can have partial or complete basement coverage. Importantly there can be limited, extensive or no forward oriented protrusions (Figure 5). Frequently, tube elongation is accompanied by tube migration, wherein the tube moves to a new position in the embryo, or branching morphogenesis, wherein the tube elaborates branch-like structures of varying complexity. The clearest examples of tube growth by cell shape change concurrent with cell migration are seen with the Drosophila salivary gland and trachea (Myat and Andrew, 2000b; Bradley and Andrew, 2001; Myat and Andrew, 2002; Kerman et al., 2008). The role of cell shape change is quite easy to study in these systems because development occurs relatively rapidly and can be imaged live in an intact embryo (Figure 6A,B; (Cheshire et al., 2008; Jazwinska et al., 2003; Kakihara et al., 2008; Kato et al., 2004; Ribeiro et al., 2002). Moreover, these cells completely stop dividing either once they are specified (salivary gland) or once they complete internalization (trachea), highlighting the role of cell shape change in tube elongation. Examples of cell rearrangement fueling tube growth include the vertebrate neural tube, most primary branches in the Drosophila trachea, as well as the Drosophila salivary duct, hindgut and kidneys (Bradley et al., 2001; Caussinus et al., 2008; Jung et al., 2005; Lengyel and Iwaki, 2002). Growth by cell division is the predominant mechanism for tube elongation in vertebrate systems and a role for oriented mitoses in controlling tube shape has been demonstrated for the mouse kidney (Karner et al., 2009; Saburi et al., 2008), gut (Matsuyama et al., 2009), and chick neural tube (Sausedo et al., 1997). Tube growth by cell recruitment has been clearly demonstrated in the Drosophila kidney, known as the Malpighian tubules, a system which seems to utilize all of the known mechanisms for its elongation (Jung et al., 2005). Tube growth through cell recruitment may also play a critical role in the development of vertebrate epithelial organs, but the details of these processes await further analysis.
Tube elongation in the Drosophila salivary gland occurs immediately following invagination. As soon as the cells internalize, they elongate driving the elongation of the entire salivary gland tube. The characterization of the molecular changes downstream of two key transcription factor genes required for full salivary gland elongation, huckebein (hkb) and ribbon (rib), reveals that molecular and cellular events at the apical membrane surface are essential for tube elongation through cell shape change (Bradley and Andrew, 2001; Myat and Andrew, 2002; Cheshire et al., 2008; Kerman et al., 2008). Loss of either transcription factor decreases the levels of Crumbs, an apically localized transmembrane protein that specifies and expands the apical membrane domain in multiple different contexts in Drosophila and in vertebrates (Knust, 1994; Gosens et al., 2008). Hkb is also required for salivary gland expression of klarsicht (klar), which encodes a “pioneer” protein that mediates delivery of vesicles to the apical membrane domain. Klar is thought to directly interact with dynein, a minus-end directed microtubule motor protein (Welte et al., 1998; Mosley-Bishop et al., 1999), consistent with the minus-ends of microtubules being enriched in the apical domain of the salivary gland (Myat and Andrew, 2002). Finally, Rib, which is also required for tube elongation in the trachea, functions to decrease phosphorylation of Moesin, an Ezrin-Radixin-Moesin family protein whose active phosphorylated form cross-links the apical plasma membrane to the underlying actin cytoskeleton. The effects of wild-type Rib function on Moesin would be to diminish those cross-links, thus allowing for easier deformation of the apical membrane during tube elongation. The characterization of Hkb, Rib and their downstream transcriptional targets reveal that both the synthesis and apical delivery of membrane as well as the deformability of that membrane is key to tube elongation (Myat and Andew, 2002; Cheshire et al., 2008; Kerman et al, 2008). The regulation of cell shape change through regulation of apical membrane mechanics is thus a major factor for tube elongation in systems where cell divisions no longer occur. Not surprisingly, and as will be discussed later, limiting tube elongation in the Drosophila salivary gland and trachea is also linked to events at the apical surface.
Although all of the Drosophila tracheal branches elongate in part by cell shape change, most primary tracheal branches also elongate by cell rearrangement through a process that has been dubbed “stalk cell intercalation or SCI” (Caussinus et al., 2008). In this process, two cells that start in a side-by-side arrangement slide past one another to become arranged end-to-end, effectively doubling tube length (Jazwinska et al., 2003; Ribiero et al., 2004). In the side-by-side configuration, the apical surfaces of two cells, connected by intercellular junctions, form the lumenal surface. In the end-to-end configuration, the apical surface of only a single cell, which is attached to itself through intracellular junctions along the length of the tube and to its proximal and distal neighbors by intercellular junctions at its ends, surrounds the lumen. Thus, the SCI process requires the orderly replacement of the intercellular junctions located along the length of the tube with autocellular junctions. SCI begins with one cell reaching around the lumen to contact itself on the proximal side of the tube and the other cell reaching around the lumen to contact itself on the distal side of the tube. The autocellular junctions initiated at these contact sites then zipper up as the cells slide past one another, stopping when the only remaining intercellular contacts are at the ends of the cells. SCI begins with cells at the most proximal end of the tracheal branch and is fueled (apparently entirely) by the directed migration of the “tip cells” at the distal end of the elongating branch (Caussinus et al., 2008). Two proteins that localize to the apical lumen are required to prevent the cells undergoing SCI from completely separating: Piopio (Pio), a secreted Zona Pelucida (ZP) domain protein and Dumpy (Dpy), a very large transmembrane protein containing hundreds of EGF repeats interspersed with 21-residue “DPY” domains on its lumenal face (Jazwinska et al., 2003). Pio and Dpy are proposed to form part of a lumenal scaffold that runs along the length of the tube preventing the zippering up of the autocellular junctions to continue to the point where the two cells separate.
Cell rearrangement is also key to elongation of the Drosophila hindgut, a tube that elongates in the absence of cell division and of cell death. The hindgut large intestine, which starts as a tube of approximately 50 cells in circumference, elongates to form a tube of only 12 cells in circumference (Iwaki et al., 2001). This cellular rearrangement is driven, in large part, by a gradient of JAK/STAT (Janus Kinase/Signal Transducer and Activator of Transcription) signaling, with the source of the JAK/STAT ligand – Unpaired (Upd) – being expressed in the the small intestine, a structure found just anterior to the large intestine (Johansen et al., 2003). Loss-of-function mutations in the ligand (Upd), receptor (Domeless/Dome), kinase (Hopscotch/ Hop) or the responding transcription factor (Stat92E) result in tubes that elongate only 40-50% of wild type, resulting in tubes with ~20-30 cells surrounding the central lumen. Loss-of-function of the ligand can only be rescued by its localized expression in anterior cells; expression of the ligand or an activated form of the receptor throughout the hindgut results in a phenotype very similar to the loss-of-function mutants. Thus, it is the gradient of signal that is required for elongation; how this gradient is transduced into the directional rearrangement of cells remains to be discovered but will clearly involve changes in cell adhesion and the dismantling and reforming of junctional components, much like occurs during germ band elongation at earlier stages of Drosophila embryogenesis (Bertet et al., 2004; Blankenship et al., 2006).
Cell rearrangement is also important in the elongation of the vertebrate neural tube, where cells undergo a type of rearrangement very similar to the one we have described for the Drosophila hindgut. In these systems, the process is known as convergent extension, since the cells rearrange to converge in one dimension and extend in the other, simultaneously narrowing and elongating the tube (Keller et al., 2000). In the neural tube, this process is driven by medially-directed protrusive activity that exerts traction forces, allowing cells to pull towards one another (Elul and Keller, 2000; Wallingford and Harland, 2001). The Planar Cell Polarity (PCP) or non-canonical Wnt signaling pathway, controls cell polarity in an axis perpendicular to the apical-basal axis (Adler, 2002; Klein and Mlodzik, 2005; Bastock and Strutt, 2007; Seifert and Mlodzik, 2007; Wang and Nathans, 2007). Over the past several years, the PCP pathway has been shown to function directly in convergent extension during neural tube elongation in all of these systems (Jessen et al., 2002; Wallingford and Harland, 2002; Ciruna et al., 2006; Wang et al., 2006a; Tawk et al., 2007; Torban et al., 2007; Ybot-Gonzalez et al., 2007; Etheridge et al., 2008). PCP regulated convergent extension is not only required for tube lengthening, it is also required for neural tube closure since PCP mutants often fail to completely seal off the neural tube due to the increased distance between the tube edges (Wallingford and Harland, 2002; Wang et al., 2006; Ybot-Gonzalez et al., 2007). In the past year, PCP signaling has also been implicated in tube elongation in both the mouse kidney and gut (Saburi et al., 2008; Karner et al., 2009; Matsuyama et al., 2009). In these systems, the PCP defects were reportedly manifest through effects on oriented cell division (Saburi et al., 2008; Karner et al., 2009); a more recent study on the role of Wnt9b and the PCP pathway in the mouse kidney, however, suggests that although oriented cell divisions are important in maintaining tube length control in the adult, the earliest PCP defects are in the convergent extension that occurs in the embryonic kidney to establish initial tube diameter (Karner et al., 2009). Importantly, regulators of PCP signaling localize to a subapical domain in all epithelia that have been examined, again highlighting the importance of the apical domain in regulating tube dimensions.
As mentioned ealier, the Drosophila Malpighian tubules elongate through a combination of regulated cell division, cell elongation and cell rearrangement (Ainsworth et al., 2000; Jung et al., 2005). The Malpighian tubules also provide the clearest example wherein cell recruitment occurs during tube elongation (Denholm et al., 2003). The Malpighian tubules are four elongated tubes that begin as epithelial buds of six to ten cells each that form at the junction between the hindgut and midgut epithelium. The mature tubes are only two cells in diameter and contain a total of 584 cells (+/− 1 cell): 484 principal cells, which secrete potassium ions into the lumen through the combined action of a vacuolar-H+-ATPase and a K+/H+ exchanger, and 110-111 stellate cells, which contain channels that allow for the flow of chloride ions and water into the lumen (O’Donnell et al., 1996; O’Donnell et al., 1998; Linton and O’Donnell, 1999). Based on cell marking studies, the stellate cells derive from a population of caudal mesoderm precursors that lie directly over the epithelial precursors that form the principle cells of the Malpighian tubules (Denholm et al., 2003). The stellate precursor cells first attach to the tube and then insert themselves into the polarized tube epithelium. Once intercalated, these cells are fully polarized, with the correct localization of polarity markers, apical microvilli, and polarized actin polymerization, although they never express the Crb protein, consistent with the lack of Crb expression in all tissues of mesodermal origin. The mammalian kidney also derives from both epithelial cells (the ureteric bud) and surrounding mesenchyme; reciprocal signaling between these two cell populations is key to kidney differentiation (Constantini, 2006; Dressler, 2006; Nigam and Shah, 2009). The cellular details, however, regarding how both cell types become integrated into the tubular structures of the mature kidney remain to be elucidated, although recent advances in renal organ culture and methods for live imaging of renal cells should facilitate such studies (Watanabe and Costantini, 2004; Chi et al., 2009a). As discussed in the next section, insight on kidney tube growth and elongation in vertebrates is emerging from studies of the more primitive zebrafish kidney (Vasilyev et al., 2009) and from organ culture of mouse ureteric buds (Watanabe and Costantini, 2004; Chi et al., 2009a; Chi et al., 2009b).
There is massive growth of the embryo during mammalian development. Many externally developing organisms have a significant phase of early embryonic development in which cell division is limited or absent and the major role of morphogenesis is to reconfigure and rearrange its constituent cells, rather than create new ones. By contrast, the formation of each of the major mammalian organ systems is accompanied by massive cell proliferation and the organ itself expands significantly in size as it is being built. For example, the mammary epithelial network is built during embryonic development as a rudiment a few millimeters in length (Figure 6C). During pubertal branching morphogenesis these millimeter sized ducts elongate to multiple centimeters. Clearly, cell division must play a major role in the outgrowth of many mammalian tubes. What is very interesting, and less clear at present, is the extent to which cell division is the major mechanism for the morphogenesis of these mammalian tissues; cell division could simply provide a source of new cells with tubular elongation largely accomplished by cell shape change, cell rearrangement, or cell migration.
During tube elongation by cell division, some tubes remain fully polarized, whereas other tubes either polarize sequentially or remain largely unpolarized until the organ has nearly achieved its full dimensions. Tube elongation by cell division has been documented by live imaging in several systems and by staining with either BrdU or specific phospho-histone antibodies in others. An example of an organ that remains fully polarized during its outgrowth and remodeling is the zebrafish kidney (Vasilyev et al., 2009). The zebrafish kidney is an interesting example of tube elongation in that although cell division contributes to tubule outgrowth, it is cell migration fueled by active fluid transport that primarily drives tube elongation. The zebrafish kidney comprises two pronephric ducts, maintains adherens junctions and apical brush borders during the entire process of tube elongation, indicating that these cells are highly polarized. The pronephric cells also have small distally-directed protrusions and the basal surfaces stain positive for antibodies recognizing phosphorylated Focal Adhesion Kinase (FAK), basal indicators of active cell migration. Importantly, a basal lamina is present throughout the process of pronephros elongation. Physical obstruction of fluid flow in the developing pronephros significantly inhibits tube elongation as well as the convolution of the proximal tubes (Vasilyev et al., 2009). This form of tube elongation is morphologically quite similar to elongation of the Drosophila salivary gland and tracheal dorsal trunk, wherein the tubes elongate while maintaining a fully polarized epithelium (Cheshire et al., 2008; Kerman et al., 2008). In the case of the salivary gland and trachea, however, there is no visible basement membrane during elongation and traction forces are more likely to be fueling tube movement. Nonetheless, secretion of an apical fluid filled matrix is important for diametrical tube expansion of the Drosophila trachea (Tsahouras et al., 2007). Similarly, expansion of the mammalian lung during neonatal development also requires Cl− mediated secretion of water into the lumena (Olver et al., 2004), indicating a more general role for active fluid transport in tube expansion.
In contrast to the zebrafish kidney and Drosophila salivary glands and trachea, the Müllerian ducts of birds and mammals are excellent examples of tube length increases driven by cell division (Guioli et al., 2007; Orvis and Behringer, 2007). In females, the Müllerian ducts give rise to the oviduct, uterus and (upper) vagina. In males, the Müllerian ducts degenerate due to the presence of Müllerian Inhibitory Substance (aka Anti-Müllerian Hormone) produced and secreted by the Sertoli cells of the testes. Müllerian ducts initially form by a mechanism very similar to that of the early Drosophila salivary gland and trachea, wherein the specified primordia form placodes that subsequently invaginate from the coelomic epithelium to form paired tube-like structures, one on each side of the coelomic cavity, lateral to the Wolffian ducts. Invagination of the tubes continues until their distal tips directly contact the Wolffian ducts. At this stage, the Müllerian ducts become dependent on the Wolffian ducts for further elongation (Carroll et al., 2005; Kobayashi et al., 2005). This dependency, which was originally interpreted by some to indicate that the Wolffian ducts actually contribute cells to the Müllerian ducts, instead reflects a requirement for Wnt9b expression in the Wolffian duct for Müllerian duct elongation (Carroll et al., 2005). Labeling of Müllerian duct cells with BrdU and/or phospho-Histone H3 antibodies reveals that cell division occurs at similar levels throughout the Müllerian duct tubes, suggesting that cell division along the entire length of the tube is contributing to its elongation (Guioli et al., 2006; Orvis and Behringer, 2007). Surgical removal of the Müllerian ducts rostral to the distal tips reveals, however, that the tip cells are capable of populating the caudal portions of the Mullerian ducts (Orvis and Behringer, 2007). Histological analysis of the Müllerian ducts at different stages and at different rostral-caudal positions suggests that the Müllerian ducts form epithelial tubes, with wedge-shaped cells surrounding a central lumen, immediately behind the actively dividing distal tip. The lack of robust expression of polarity markers in the Müllerian ducts, including E-Cadherin, Vimentin and cytokeratins, relative to the expression levels of the same markers in the Wolffian ducts suggests that as the Müllerian duct tube is forming, it is not “as polarized” as the neighboring Wolffian duct (Orvis and Behringer, 2007). The reduced expression of these markers could also either reflect their relatively earlier developmental stage or indicate that other related proteins contribute more to their polarization.
Recent advances in organ culture and in cell marking techniques have provided additional insights into the cellular basis of tube elongation and elaboration in several tissues, including the mammalian mammary and salivary glands, and the ureteric buds of the developing kidney (Watanabe and Constantini, 2004; Patel et al., 2006; Ewald et al., 2008; Chi et al., 2009). The mammary epithelium consists of two cell layers, a luminal epithelial layer, which forms the ducts and secretory alveoli, and a basal myoepithelial layer, which provides the forces for secretion. The epithelium is embedded in a fat pad composed of adipocytes, blood vessels, fibroblasts and immune cells. Mammary development occurs during three separate stages: the embryonic stage when the duct placodes first form and elaborate a rudimentary gland, the pubertal stage in which the gland undergoes extensive duct elongation and secondary branch formation, and the adult/pregnancy stage, when the gland forms tertiary branches and the luminal epithelium undergoes rapid proliferation and differentiation of the secretory alveoli. It is the pubertal stage mammary gland that has been used extensively for studying epithelial branching morphogenesis (Figure 6C, Watson and Khaled, 2008). During puberty, terminal end buds (TEBs) transiently form at the ends of each duct. These structures have multiple luminal epithelial cell layers, are highly proliferative and contain stem cells (Hinck and Silberstein, 2005; Watson and Khaled, 2008). TEBs are responsible for building the bilayered ducts, but are not themselves bilayered (Figure 6D-F). Organotypic culture techniques have been developed that enable many aspects of mammary branching morphogenesis to be modeled in vitro, within 3D ECM gels (Figure 7D) (Simian et al., 2001; Wiseman et al., 2003; Fata et al., 2007; Ewald et al., 2008).
Long-term confocal imaging of primary organotypic 3-D cultures of mouse mammary glands treated with growth factors reveals that the mammary ducts elongate by a form of collective epithelial migration, wherein the polarized epithelial cells reorganize into a multilayered epithelium that migrates collectively without leading edge projections (Ewald et al., 2008). This multilayered epithelium is a specialized organizational state that is necessary for mammary tube elongation. Mammary ducts are bilayered tubes, but they are not built by a bilayered elongation intermediates. The functional unit of elongation and branching is a multilayered tube without basal protrusions and with extensive, though possibly incomplete, basement membrane coverage. The actively dividing terminal end buds leave in their wake a highly polarized bilayered epithelium comprising an inner duct epithelium surrounded by an outer layer of myoepithelial cells (Figure 6F). As will be discussed later, initial gaps in coverage of the luminal epithelial cells by the myoepithelial cells may determine sites of epithelial outgrowth. Interestingly, branching morphogenesis in these cultures is completely dependent on growth factor mediated proliferative signals. If proliferation is blocked by inhibition of MAP kinase (Fata et al., 2007) or DNA synthetase (Ewald et al., 2008), the epithelial fragments are unable to organize into a multilayered epithelium and instead arrest as highly polarized cysts. By contrast, inhibition of Rac or of myosin light chain kinase blocks tube initiation and elongation, but does not prevent the transition into a multilayered organization. So growth factor dependent proliferation is required but is not sufficient for tube elongation; there is a requisite actively motile phase that depends on small GTPase activity (Ewald et al., 2008).
Based on the above examples, it is clear that tubes utilize many varied mechanisms for their elongation: cell shape change, cell rearrangement, cell division and cell recruitment, often accompanied by active migration. Importantly, the use of any one mechanism clearly does not exclude the utilization of additional mechanisms even within a single organ. Indeed, to achieve the rapid growth necessary during early development and/or in response to hormonal cues of puberty and pregnancy, it may be necessary to use multiple strategies in a coordinate fashion. It is thus likely that studies to date have highlighted only the major mechanisms by which each of the tube types achieve their final dimensions. It is likely if those major pathways were blocked, we would discover that cells also either can or do use other means to build appropriately sized organs. Indeed, studies in the Drosophila trachea reveal that if the trachea develops with only half the number of progenitor cells, a fully patterned trachea of correct dimensions is formed (Beitel and Krasnow, 2000). Thus, cells can adjust their individual shapes and sizes in the interest of achieving an entire organ of correct dimensions.
Much of our understanding of tube size control comes from the relatively simple model system of the Drosophila embryonic trachea. A seminal genetic screen identified several mutations that affect final tracheal tube dimensions (Beitel and Krasnow, 2000). The mutations fall into two different categories: genes encoding the proteins required to assemble and modify a transient chitinous matrix in the lumena of the nearly fully formed trachea as well as genes encoding components of the insect pleated septate junction (pSJ), a junctional structure located just basal to the adherens junctions, which has a similar function and utilizes homologous protein components with the more apically-positioned tight junctions of vertebrate epithelial cells. pSJs provide barrier function, preventing the diffusion of water and solutes between epithelial cells, and serve a fencing function, separating the apical and basolateral domains of epithelial cells. In general, mutations in genes encoding proteins required for the synthesis, secretion or maturation of chitin result in defects in tube diameter, with regions of tube dilation and constriction (Araujo et al., 2005; Devine et al., 2005; Moussian et al., 2006; Tonning et al., 2005), whereas mutations in the septate junction protein components most often lead to tubes that are longer than those of wild type (Behr et al., 2003; Llimargas et al., 2004; Paul et al., 2007; Wu et al., 2007). In wild-type cells, chitin forms a transient cable network that may provide a scaffold that regulates lumen expansion. Based on findings that the apical localization of two chitin deacetylases is disrupted in the septate junction mutations affecting tube length, it has been proposed that secretion and assembly of a chitinous matrix coordinates uniform radial tube expansion and length increases (Luschnig et al., 2006; Wang et al., 2006b). Modification of this matrix by the chitin deacetylases is subsequently required to prevent further tube length increases. Roles for a secreted apical matrix in the regulation of tube uniformity and size control have also been revealed with the Drosophila salivary gland (Seshaiah et al., 2001; Abrams et al., 2006; Jayaram et al., 2008), suggesting shared mechanisms for maintaining uniformly sized tubes of correct dimensions. Whether a similar mechanism exists for tube size control in vertebrate systems is unknown.
Tube architecture can vary enormously: some tubes are quite simple, with little to no branching, whereas others can be quite elaborate, forming tubes of great complexity. Clearly, tube architecture is determined by organ function. Simple secretory organs, such as the Drosophila salivary gland, do not require the elaborate structure that is necessary for example in the vertebrate lung, vasculature or kidney, where high surface area is critical for the exchange of gases, nutrients, metabolites or waste products. Similarly, tube architecture is important for the efficient packaging of internal organs, so that they can fit within a protected space and be surrounded by the tissues they require for optimal function. Tube architecture can be hardwired, based on the uniformity in structure of different organs observed among members of the same or closely related species, for example in the mammalian lung (Metzger et al., 2008). However, branching in other organs, primarily those less limited by tissue packing constraints, appears to be somewhat more stochastic. For example, the organ structure and epithelial branching pattern of the mammary gland is highly variable in both humans and mice.
The question arises: how are the various levels of complexity in tube architecture achieved? Are the tubes themselves intrinsically different or can just a few adjustments be made to allow, for example, unbranched tubes to branch or to reduce the number of branches an organ normally forms? Is the information for branching cell autonomous or is it mediated through external signals? These are early days, but with new tools and techniques becoming available, and as multiple systems are compared, we are beginning to fill in the gaps in our understanding.
The simplest and best-characterized branching organ is the Drosophila trachea, which undergoes very stereotypical branching patterns during embryogenesis to form six primary branches through a budding type mechanism (Figure 6B,B’) (Manning and Krasnow, 1993; Ghabrial et al., 2003; Uv and Samakovlis, 2005; Kerman et al., 2006). Secondary branching of the trachea subsequently occurs at the ends of the primary branches to form the long subcellular tracheoles (Guillemin et al., 1996). The patterning of secondary branching, which occurs toward the end of embryogenesis and in the larval stages, is not genetically predetermined and is instead controlled by local demands for oxygen (Jarecki et al., 1999). Nonetheless, in both primary and secondary branching, the driving force for tracheal tube outgrowth is fibroblast growth factor (FGF) signaling (Klambt et al., 1992; Reichman-Fried et al., 1994; Sutherland et al., 1996; Jarecki et al., 1999; Centanin et al., 2008). The developing trachea expresses one of the two Drosophila FGF receptor genes, known as breathless (btl), and simply migrates out in response to local sources of its FGF ligand, which is encoded by the branchless (bnl) gene (Klambt et al., 1992; Reichman-Fried et al., 1994; Sutherland et al., 1996). In embryos missing function of either btl or bnl, tracheal cells invaginate but completely fail to migrate, forming only internalized sacs of polarized tracheal precursors. In embryos expressing an activated form of the Btl receptor throughout the trachea or expressing the Bnl ligand in all ectodermal cells, ectopic branching is observed (Lee et al., 1996; Sutherland et al., 1996). In the wild-type trachea, the one to two cells at the ends of each branch that receive the highest levels of FGF signal, form the tip cells, which migrate out in direct response to Bnl, with a variable number of tracheal precursors following, in a FGF-independent manner, to form what will eventually be the branch stalks (Ghabrial and Krasnow, 2006). In the early embryo, sites of bnl expression are determined by patterning genes, whereas in later stage larvae and adults, bnl expression is activated in direct response to transcription factors that are activated in response to hypoxia (Jarecki et al., 1999). A recent study has shown that not only does hypoxia induce expression of bnl in the oxygen-deprived tissues, but it also leads to upregulation of btl expression in nearby tracheal cells (Centanin et al., 2008).
Although FGF signaling is key to primary branching in the Drosophila embryonic trachea, it is not the only signaling pathway required for branch outgrowth. Other signaling pathways, including TGFβ/BMP, Wnt, and Robo/Slit, are required for specification of branch identity and/or regulation of subsequent outgrowth of specific branches (Vincent et al., 1997; Chen et al., 1998; Chihara and Hayashi, 2000; Llimargas, 2000; Englund et al., 2002; Gallio et al., 2004; Lundstrom et al., 2004). For example, the TGFβ/BMP pathway, activated by localized expression of the Dpp ligand, is required for specification of branches that migrate dorsally and ventrally (Vincent et al., 1997; Chen et al., 1998). Dpp signaling appears to function in parallel with the Bnl/Btl FGF pathway to activate expression of transcription factors required for branch-specific elongation of dorsal and ventral branches (Myat et al., 2005). Wnt signaling is required for the specification of cells in the major artery of the trachea that runs the length of the embryo, the dorsal trunk (DT). Loss-of-function mutations in Wnt pathway components result in a loss of DT identity and the subsequent mismigration of DT precursors with the cells of the trachea that oxygenate the viscera (Chihara and Hayashi, 2000; Llimargas, 2000). Correspondingly, ectopic activation of Wnt signaling results in a failure of the visceral branch precursors to migrate away from the DT (Chihara and Hayashi, 2000; Llimargas, 2000). How these other pathways coordinate with FGF signaling to control tracheal branch specific outgrowth remains to be discovered.
FGF signaling has also been implicated in the branching of many vertebrate organs, including the lung, kidney, mammary, salivary and lacrimal glands (for review, see Lu and Werb, 2008). In almost all of these systems, knock-out or reduction of Fgfr2 (the vertebrate receptor most similar to the Drosophila Btl protein) or specific FGF ligands (FGF10 and/or FGF7) either reduces or completely eliminates branching, whereas supplying exogenous FGF to organotypic cultures of these organs promotes branching (de Boer et al., 1996; Sekine et al., 1999; Ohuchi et al., 2000; Steinberg et al., 2005; Lu et al., 2008; Mailleux et al., 2008). As with the Drosophila trachea, other signaling pathways also impinge on branching morphogenesis in each of these organs, including for example GDNF/Ret and Wnt signaling in the kidney (Costantini, 2006; Dressler, 2006), and TGFβ/BMP, EGF and Shh signaling in the lung, salivary and mammary glands (Patel et al., 2006; Sternlicht et al., 2006; Warburton et al., 2005). In the Drosophila trachea, it is clear that the effects of FGF signaling on branching morphogenesis are entirely through branch migration since these cells cease dividing prior to branch outgrowth. In many mammalian systems, where branch outgrowth is accompanied by extensive cell proliferation, it is unclear whether FGFs have their effects on branching through migration or proliferation, although studies in kidney organ cultures reveal that the two processes can be uncoupled and thus potentially regulated by independent mechanisms (Fisher et al., 2001; Watanabe and Costantini, 2004).
Growth factor signaling can regulate tube formation in Drosophila through effects on proliferation as well. The Drosophila air sac is a specialized branch of the trachea that forms in the adult. Both cell division and active migration are required to build the air sac and FGF signaling acts specifically in branch migration, whereas EGF signaling mediates the mitogenic effects (Cabernard and Affolter, 2005). A major challenge to understanding the molecular regulation of branching in mammalian organs is the multiplicity of signaling pathways involved. Whereas Drosophila tracheal development is predominantly dependent on the actions of a single FGF receptor, mammalian systems frequently require multiple RTKs. For example, mammary development requires ErbB1 (EGFR), ErbB2, FGFR1 and FGFR2 signaling (Howlin et al., 2006; Lu and Werb, 2008). Understanding the epistasis relationships between multiple RTK pathways is complex, especially when coupled with the technical difficulties inherent in producing interpretable compound mutants in these pathways.
In a set of very elegant cell marking studies carried out in the mammalian kidney and Drosophila trachea, it was revealed that the elongating tubes in these systems comprise two distinct populations: Tip cells (also known as leader cells or terminal cells) and trunk cells (also known as follower or stalk cells) (Cabernard and Affolter, 2005; Shakya et al., 2005). Tip cells in Drosophila absolutely require receptor tyrosine kinase (RTK) signaling for outgrowth, whereas the follower cells do not. Moreover, it is the cells that receive the highest levels of RTK signaling that become the tip cells and they do so by moving into the leader position, passing by cells in which RTK signaling levels are relatively lower (Ghabrial and Krasnow, 2006).
The concept of a restricted requirement for RTK signaling in leader cells has been demonstrated in the mammalian kidney as well (Shakya et al 2005). The investigators used chimeric animals to explore how the Ret receptor tyrosine kinase (RTK) signaling pathway affects ureteric bud outgrowth. Ret signaling is essential for ureteric bud development; mutations in the receptor Ret, co-receptor GFRα1, both of which are expressed in the ureteric bud epithelium, or in the ligand GDNF, which is expressed in the surrounding metanephric mesenchyme, result in either a complete or partial failure in ureteric bud formation (Costantini and Shakya, 2006). For their analysis, Shakya et al (2005) generated two kinds of chimeric animals: those in which the GFP-labeled cells were wild type and those in which the GFP-labeled cells were mutant for the ret gene. The unlabelled host cells in both types of chimeras were wild type. The group discovered that, whereas wild-type GFP labeled cells populated all portions of the Wolffian duct and branching ureteric bud, the ret-/- GFP cells failed to populate the distal tip cells of the emerging ureteric bud, often referred to as the ampulla. Indeed, ret-/- cells were not found in ureteric ampullae even when additional bud sites were induced. In subsequent branching generations, GFP-labeled ret-/- cells were again found in stalks but not in buds, consistent with the later restricted expression of Ret in only the distal tips of the branching ureteric buds.
In a follow-up study by this group, time-lapse imaging of chimeric organ cultures in which both mutant and wild-type cells were marked revealed that the ret+ ureteric bud precursors actively move to the site where the ampulla will emerge (Chi et al., 2009). Moreover, by adjusting the relative levels of Ret signaling by either generating chimeras consisting of ret null cells mixed with ret hypomorphic cells or by generating chimeras in which the marked cells were null for an inhibitor of RTK signaling (Sprouty1-/-), the group demonstrated that some minimum difference in Ret signaling activity is required for the preferential localization of cells in the tip and that it is the cells with the highest levels of signaling that move to the distal bud (Chi et al., 2009). This finding nicely parallels studies in Drosophila tracheal cells, demonstrating a requirement for FGF signaling in the leader but not follower cells and for relative levels of FGF signaling being important in sorting the leaders (tip cells) from followers (stalk cells) (Cabernard and Affolter, 2005; Ghabrial et al., 2006). As discussed earlier, studies from Caussinus et al (2008) reveal that it is the leader cells in the embryonic trachea that generate the traction forces required for the cell rearrangements of embryonic tracheal tube outgrowth.
Recent work in the mammary gland has revealed a similar logic operating in the requirement for FGFR2 (Lu et al., 2008). The investigators generated labeled, mosaic deletions of FGFR2 using both transgenic and adenoviral approaches and demonstrated that FGFR2 is required in the terminal end bud, but not in the trailing duct behind it. Deletion of FGFR2 in a high fraction (>90%) of cells in the epithelium produced a developmental delay during which the terminal end bud reorganized to reconstitute itself from an FGFR2 positive population that escaped recombination. This FGFR2 population then drove the elongation of a trailing duct with a significant fraction of FGFR2 negative cells (Lu et al., 2008).
Tube elaboration in the Drosophila embryo is highly stereotyped and often genetically determined. In wild-type embryos, the secretory portions of the salivary gland are always large single cell layer unbranched tubes, whereas the duct is a simple Y-shaped structure connecting the secretory tubes to the alimentary canal (Bradley et al., 2001; Kerman et al., 2006). The kidney system always comprises four elongated, unbranched tubules, two extending anteriorly towards the head and two extending posteriorly (Jung et al., 2005). Each tracheal segment shows a nearly identical pattern of branch outgrowth during primary branching (Ghabrial et al., 2003; Kerman et al., 2006). Thus, tube architecture, at least at early stages in these simple systems, is highly programmed. Nonetheless, very minor perturbations in that program can alter branching patterns. One clear example is observed with mutations in a transcription factor, encoded by the hairy gene, whose major job in the Drosophila salivary gland is to quickly shut off expression of two early-expressed salivary gland genes: hkb, the transcription factor that transiently activates expression of klar and stabilizes Crb, and klar itself (Myat and Andrew, 2002). In hairy mutants, expression of hkb and of klar is higher and persists longer than in wild-type salivary glands. As a consequence, invagination and subsequent elongation of the primordia occurs at multiple positions instead of only the single dorsal-posterior site in WT glands. Due to internalization occurring at these ectopic sites, the salivary gland either branches, if the invagination sites are separated, or becomes bulbous, if the invagination sites are close. Thus, one small change in the early program of gene expression and the ultimate morphology of the tube – simple linear versus branched or bulbous – is altered. This study suggests that tubes have the capacity to adopt a wide range of morphologies and that even simple and subtle change in the timing and positioning of specific cell behaviors can have profound effects on overall tube architecture.
Recent studies of the branching program of the mouse lung reveal that the branching patterns of complex mammalian tubular organs animals can be highly stereotypical. Metzger and colleagues (2008) examined hundreds of fixed mammalian lungs to generate a 3-D reconstruction of mouse lung morphogenesis at different developmental stages (Figure 6G). They demonstrated that from animal to animal, the pattern of lung branching is highly reproducible and involves only three modes of branching used reiteratively in only three different subroutines to pattern the entire lung. The modes of branching were described as (1) domain branching, (2) planar bifurcation, and (3) orthogonal bifurcation. Domain branching is the most complicated and involves daughter branches forming sequentially along the proximal to distal axis of the parent branch in a distinct order, eventually spanning the entire circumference of the parent branch. First, branches begin to form at regular intervals along a single circumferential position. Once a few branches have formed, a second row of branches arises approximately 90° offset from the original row and with each branch displaced slightly distal to the corresponding branches in the first row. Subsequently, a third and then a final row of branches form, again with a 90° circumferential and slightly distal displacement of branches relative to the previously formed row. This pattern of branching creates a “bottle-brush” type structure that both maximizes the distance between branches and takes advantage of all of the usable branching space. In planar bifurcation, the end of the branch simply splits into two branches in the same plane as that of the previous branching event. Orthogonal branching occurs when the end of the branch also splits into two branches, but in this case, the division is in a plane 90° rotated with respect to the plane of the previous branching event. Interestingly, although theoretically, thousands of possibilities exist regarding the order in which the branching modes might be deployed, only three sequences of deployment could describe the branching events of the entire lineage (Metzger et al., 2008).
In this same study, Metzger et al., (2008) identified three mutations that disrupt the branching pattern. First, they examined inversus viscerum (iv), a mutation in the dynein heavy chain gene (Dnahc11). The iv mutation sometimes reverses global left-right patterning completely and also frequently reversed the left-right branching pattern of the lung. Second, mutations in Sprouty2 (one of a family of genes that down-regulates FGF and other receptor tyrosine signaling pathways) resulted in extra proximally-positioned branches forming in the ventral domains of specific lobes. Third, mutations in a gene the authors dubbed shifty resulted in a distal shift in the position of daughters sprouting from some parent branches. Thus, the branching patterns of tubular organs in higher organisms can also be stereotypical and, through the utilization of only a few genetically programmed and reiterative behaviors, can generate elaborately complex and beautiful structures (Metzger et al., 2008). It will be important to determine the degree of genetic programming of spatial patterning information that is typical in mammalian organs. The lung has been examined significantly more thoroughly than the other organs, but its stereotypic branching program may partly reflect the tight spatial constraints in the chest cavity. Mammary development, at least superficially, appears much more heterogeneous with mammary branching patterns from estrus-matched sisters often appearing quite different in detail.
Recent live imaging of branching morphogenesis in organ cultures has provided mechanistic insight into how branches form in different tissues. Such studies have been done with the mammalian kidney, pancreas, salivary gland and mammary gland using tissue-specific GFP constructs (Watanabe and Costantini, 2004; Larsen et al., 2006; Puri and Hebrok, 2007; Ewald et al., 2008; Chi et al., 2009a). Studies in the mammalian kidney support the idea that complex tube architecture can be achieved through a set of simple reproducible branching modes (Figure 7B; Watanabe and Constantini, 2004). Starting with very early kidneys in which the primordia have just emerged from the Wolffian duct, Watanabe and Constantini (2004) were able to follow branching dynamics for about five generations of branching. In this work, the authors described three modes of branching, the most common being bifurcation at the ends of branches, followed by trifurcation at the ends of branches, and, finally, lateral (or side) branching, which may be related to the domain branching observed in the developing lungs. As with the lungs, some patterns emerged in the branching process. For example, the first branching event was most often a bifurcation whereas the second branching event was most often a trifurcation. Lateral branching occurred only during the second and third generation of branching and only in tube segments that had achieved a minimal length. Importantly, the live imaging revealed that the branching events that were inferred from still images could occur by more than one mechanism. For example, a trifurcation in which one of the resulting branches grows more slowly than the other two could appear morphologically identical to a lateral branching event coupled with a later bifurcation. Live imaging of the kidney combined with fixed imaging of cell membrane markers has revealed that, like the zebrafish kidney, the mammalian kidney also maintains its epithelial polarity throughout branching morphogenesis, although when the ureteric bud first emerges from the Wolffian duct, the bud epithelium transiently forms a pseudostratifed epithelium (Chi et al., 2009a; 2009b).
Ex vivo organotypic culture of mammalian salivary glands have also provided insight into the cellular mechanisms underlying the reiterative branching that occurs in glandular organs. Ex vivo culture of mammalian salivary glands, primarily the submandibular gland (SMG), has been done since the 1950s and has proven an excellent system for revealing the roles of signaling pathways, epithelial-mesenchymal interactions and the extracellular matrix (ECM) in branching morphogenesis (for review, see Patel et al., 2006). As with multiple other branching tubular organs, FGF signaling plays a major role in SMG development. SMG buds form but degenerate in FgfR2 null embryos, and FGFs 10, 8 and 7, as well as heparin sulfates that modify FGF signaling, have all been implicated at different stages of branching and differentiation (Figure 7A,C). As with all known branching tubular organs, SMG growth and morphogenesis requires function of several additional signaling pathways, including EGF, BMP and Shh, as well as a number of transcription factor genes.
Signaling from the ECM plays a critical role in sculpting the salivary glands. For example, not only are specific laminins required for branching but they seem to work in part by regulating the levels of FGF signaling (Patel et al., 2006). Similarly, several secreted matrix metalloproteases result in decreased branching when their function is disrupted; it is unclear, however, whether the effects on branching are through their effects on signaling pathways or through more direct effects on ECM turnover. Recent studies suggest that changes in the ECM provide the driving force for branching of the SMG, which occurs by cleft formation. As globular buds emerge, clefts or small invaginations form and, as these clefts deepen, the bud is separated into two parts. Levels of some ECM proteins, such as collagen III, are higher in clefts, suggesting that increased stiffness in the cleft ECM may prevent or inhibit outgrowth, allowing for expansion in only the flanking regions. Recent studies have implicated the ECM protein fibronectin as a major player in the process of clefting. In an attempt to identify molecular differences in clefting versus nonclefting regions of the SMG, laser micro-dissection was used to isolate SMG clefts versus buds and to profile differences in mRNA levels (Sakai et al., 2003). Unexpectedly, this study revealed high levels of fibronectin mRNA in the cleft explants. Similar differences were observed in situ with antibodies to the fibronectin protein. These authors went on to show that siRNA knockdown or antibody blocking of fibronectin as well as antibody blocking of its integrin receptor, α5β1, led to significant inhibition of branching in SMG ex vivo cultures. Correspondingly, the addition of fibronectin increased branching and at optimal concentrations effectively doubled both the number of clefts and buds. Importantly, the authors showed that the addition of pre-aggregated fibronectin lead to a decrease in E-cadherin levels at sites adjacent to the fibronectin complexes. These discoveries led the authors to propose a model wherein increased levels of cellular engagement with the ECM through fibronectin-integrin complexes results in the displacement of cell-cell interactions through E-cadherin, a process necessary to form the deep clefts. These authors suggest that the increased levels of collagen III in clefts might be a consequence of increased fibronectin, since fibronectin is known to regulate collagen polymerization. Although both kidney and lung budding are thought to occur through a mechanism of bud outgrowth or extension, Sakai et al., (2003) showed that fibronectin also accumulates at sites of branching in these tissues and that depleting fibronectin with either antibodies or siRNA also decreased branching in lung and kidney, suggesting a very general role for cell-ECM engagement at tube branch points.
Live imaging of SMG morphogenesis further supports a role for cell-ECM interactions in cleft formation and branching (Larsen et al., 2006). GFP cell labeling revealed highly dynamic cell movements during branching morphogenesis that probably drive bud outgrowth. The active cell migration occurred during the early stages of branching morphogenesis, but was largely absent in mature glands. Interestingly, the patterns of cell movement were completely unrelated to the sites of cleft formation. Instead, labeling of fibronectin revealed its early accumulation at the base of cleft sites, with new fibronectin being laid down as the clefts deepened. This study further supports a mechanism wherein active assembly of fibronectin and its association with integrins provides a wedging force that drives the separation of the bud primordia into two separate populations, allowing for bud formation (Larsen et al., 2006).
Recent live imaging of mammary ductal morphogenesis in 3D culture has provided additional insight into the budding process in this organ (Figure 6C-F, 7E,E’; Ewald et al., 2008): (1) ductal elongation occurs through a process of collective cell migration without actin-rich cellular protrusions at the leading edge of the duct, (2) the elongating front includes an actively-dividing and dynamically rearranging population of incompletely polarized luminal epithelial cells, (3) ductal branches most often form at sites with little myoepthelial cell coverage and outgrowth ceases once the luminal epithelial cells attain full apico-basal polarity and simple epithelial organization. Mammary branching morphogenesis appears to result from a dynamic interaction between two distinct epithelial populations, the luminal and myoepithelial cells. Based on colocalization and dynamic cell behaviors, it appears likely that the luminal epithelial cells are promoting ductal elongation and the myoepithelial cells act to limit and pattern ductal elongation. Myoepithelial cells could play key roles in regulating duct outgrowth either through contractile forces or through signaling interactions. A central aspect of the morphogenesis of the mammary gland appears to be transitions between different states of epithelial organization, from bilayered to multilayered to elongate and elaborate the tube, then back to bilayered to repolarize and redifferentiate the tube (Figure 7E). Importantly, the multilayered organization observed during normal mammary gland development is strikingly similar to that observed in invasive tumors, suggesting conserved mechanisms for normal and neoplastic growth (Ewald et al., 2008). This organ culture system coupled with live imaging holds much promise for revealing the molecules and mechanisms driving branching morphogenesis and for learning what distinguishes normal outgrowth from cancerous invasion.
While epithelial morphogenesis is critical for normal development, if it occurs in the wrong time or place it can also drive catastrophic dysfunction. Mortality in human cancer is largely attributable to the local invasion of tumors into adjacent normal tissues and the dissemination and metastasis of tumor cells to distant sites (Nguyen et al., 2009). Both of these processes can be viewed as epithelial morphogenesis gone awry. It is important to note that many of the core cellular properties and behaviors that we study in the normal development of epithelial tubes are modulated in epithelial cancers. Carcinomas frequently display a loss of tubular architecture, loss of apical-basal polarity, excess proliferation, and loss of differentiation. It has been known for a long time that decreased apico-basal polarity and lower differentiation correlate with a more advanced tumor and poorer patient prognosis (Ewing, 1932; Nguyen et al., 2009; Nguyen and Massague, 2007). There is likely something very important about the normal 3D organization of epithelial tubes, as this architecture is lost early and across diverse human cancers.
As developmental biologists, it is our working hypothesis that cancer recapitulates normal cellular and molecular mechanisms occurring in the wrong time and place or to the wrong extent, but it is critically important to determine the extent to which that is true for individual human malignancies. Cancer isn’t one disease, it is many, even among cancers of a particular tissue of origin. The basic cellular mechanisms of cell proliferation, cell death, cell shape change, etc. are used to build both organs and tumors. However, each normal organ has a developmental program that represents a specific configuration of these cellular mechanisms in a specific temporal sequence and spatial organization. The building blocks are likely conserved, but the spatial and temporal configuration of these cellular mechanisms is potentially quite different among various organs and especially among model systems. The interesting question for branching morphogenesis, as applied to epithelial cancers, is whether the normal developmental program for a specific organ is recapitulated during cancer progression or whether tumors grow and invade using a novel configuration of these core mechanisms. It is possible that some tumors recapitulate normal developmental programs in great detail whereas others start with the same molecular genetic toolkit but access, through mutation and selection, combinations of cellular processes that never occur together in the normal animal. This is why we seek to understand the developmental program for both normal and neoplastic growth. Does A=B? We don’t fully understand A or B yet and so we can’t answer the question. A better understanding of which tumors most closely recapitulate normal developmental mechanisms would enable higher impact collaborations between basic scientists and clinicians in areas where existing knowledge from model systems has yet to be translated to clinical utility.
The task ahead is to resolve branching morphogenesis to the level of single cells and individual molecular activities. Critical aspects of this work will be to understand more fully the coupling of differentiation and morphogenesis and to understand the degree to which the morphogenetic potential of a tissue is determined by its differentiation state. It will also be important to develop a more mechanistic understanding of the coupling of polarity and proliferation during morphogenesis. We have examples at both extremes currently, with Drosophila salivary gland completing all proliferation prior to tube morphogenesis and elongating and branching as a fully polarized simple epithelium (Kerman et al., 2006; Cheshire et al., 2008; Kerman et al., 2008; Figure 6A-B’). Alternately, mammary ductal elongation is accomplished by a transient, specialized structure with high levels of proliferation and a complex, stratified epithelial organization (Ewald et al., 2008; Lu and Werb, 2008; Figure 6D-F).
Several major open questions relate to pattern formation during tube morphogenesis. How are branch points determined? How does an epithelium decide to start, stop, and restart morphogenesis? It will be important to answer these questions in individual model systems and also to validate the results across model systems so that we can understand the generality of cellular mechanisms: across size scales within an organ, among different epithelia during morphogenesis and between epithelial morphogenesis and neuronal or fibroblastic morphogenesis. Of primary interest is to determine the extent of conservation of the molecular mechanisms regulating these processes. This will require some patience as we are only just beginning to understand the cellular basis of tube morphogenesis in some systems. It will also require a skeptical eye, as there are a limited number of major molecular pathways; at some trivial level the same sorts of genes will be involved across different systems (e.g. FGF, Wnt, Notch). More interesting than whether FGF signaling is involved is to determine whether the genetic regulatory logic is similar and whether the types of events being regulated by a given set of genes are mechanistically similar. It is an exciting time in the study of epithelial morphogenesis with new tools and techniques enabling powerful new experimental approaches. Many questions that were easily articulated at the founding of the Society for Developmental Biology 70 years ago are just now becoming answerable. We look forward to the results.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.