|Home | About | Journals | Submit | Contact Us | Français|
How do animal cells assemble into tissues and organs? A diverse array of tissue structures and shapes can be formed by organizing groups of cells into different polarized arrangements and by coordinating their polarity in space and time. Conserved design principles underlying this diversity are emerging from studies of model organisms and tissues. We discuss how conserved polarity complexes, signalling networks, transcription factors, membrane-trafficking pathways, mechanisms for forming lumens in tubes and other hollow structures, and transitions between different types of polarity, such as between epithelial and mesenchymal cells, are used in similar and iterative manners to build all tissues.
The defining feature of metazoa is that their cells are organized into multicellular tissues and organs. Although almost every eukaryotic cell is spatially asymmetric or polarized, polarity must be coordinated in space and time for individual cells to form a tissue1. Cell polarity involves the asymmetric organization of most of the physical aspects of the cell, including the cell surface, intracellular organelles and the cytoskeleton2,3. Analysis of the polarization of unicellular eukaryotes, such as yeast, has yielded enormous insights into the mechanisms that underlie the polarity of individual cells3. Formation of a tissue, however, requires an ensemble cast; the emergent properties of the tissue result from the combined roles of the individual cells that are involved. Accordingly, several biological processes, including cell division, cell death, shape changes, cell migration and differentiation, must be coordinated with the polarity requirements of a tissue to form an organ4.
Evolutionarily, epithelia are the most archetypal polarized tissues in metazoa, with ~60% of mammalian cell types being of epithelial or epithelial-derived origin5. Accordingly, the best studied polarized tissue is the simple epithelium of the mammalian intestine and kidney, the cells of which are columnar in shape (that is, they are taller than they are wide). The apical surfaces of these cells provide the luminal interface and are specialized to regulate the exchange of materials, such as nutrients from the intestine. The lateral surfaces of these cells contact adjacent cells and have specialized junctions and cell–cell adhesion structures3,6 (FIG. 1a). The basal surfaces of these cells contact the underlying basement membrane, extracellular matrix (ECM) and, ultimately, underlying blood vessels. The basal and lateral surfaces are fairly similar in composition and organization and are often referred to together as the basolateral surface. The apical and basolateral surfaces, however, have very different compositions. In vertebrates, tight junctions (TJs) are found at the apical-most portion of the lateral surfaces, where the TJs form barriers both between the apical and basolateral surfaces and between adjacent cells, limiting paracellular permeability7 (FIG. 1a).
Many epithelial organs make use of interconnected tubular networks, although the basic design principles (as defined by Rafelski and Marshall8) are the same: a series of tubes terminates in a spherical ending or cap, which is referred to as an acinus, end bud, alveolus or cyst in different tissues. Tubular networks can either arise independently and then become interconnected, or can be branching trees that form via new sprouts from existing tubes. Many conserved morphogenetic processes give rise to these structures, including mechanisms of lumen formation and expansion, tubulogenesis, branching morphogenesis, mesenchymal–epithelial transitions (MET) and epithelial–mesenchymal transitions (EMT).
Cellular specialization through polarization occurs in almost all cell types. Neural synapses have specialized sites for neurotransmitter release and uptake9 (FIG. 1b). The apical membranes of photoreceptor epithelium undergo light-sensing activity, whereas the basal surfaces connect to underlying neurons (FIG. 1c). Migrating cells, such as neutrophils or Dictyostelium discoideum amoebae, exhibit asymmetric front–back polarity as they move towards attractive cues10 (BOX 1). With a core requirement for cellular asymmetry in biological function, understanding how cells polarize and coordinate this process to form a tissue is a central question.
Conserved, core protein complexes are involved in the generation and maintenance of all types of polarization. Three major polarity complexes, the PAR (CDC42–PAR3–PAR6–aPKC), Crumbs (Crb–PALS–PATJ) and Scribble (Scrib–Dlg–Lgl) complexes function in such diverse contexts as asymmetric cell division, epithelial and neuronal polarization, chemotactic migration and cell proliferation2,17 (see figure). Although these complexes perform seemingly different functions in different cell types and contexts, some overarching themes can be drawn.
Polarity complexes distribute asymmetrically in cells, promoting the expansion of the membrane domain they associate with. In epithelial cells (panel a), the PAR and Crumbs complexes promote apical polarity, whereas the Scribble complex operates at the basolateral surface. The PAR complex can also be divided into two subcomplexes: apical CDC42–PAR6–aPKC and tight junction (TJ)-localized PAR3–aPKC, which recruits the lipid phosphatase PTEN (BOX 2). Polarity complexes can be mutually antagonistic (panel b)2,107, a design principle that allows the establishment of axes of asymmetry8. Inappropriate movement of the Scribble complex into the apical domain is antagonized by the phosphorylation of Lgl by aPKC: phosphorylation of Lgl dissociates the protein from the cell cortex108. Similar reciprocal exclusion mechanisms between apical and basolateral complexes maintain this asymmetry17, allowing apical and basolateral regions to become discrete, non-overlapping domains — a possible example of zero-order ultrasensitivity109. Several signalling receptors that disrupt or promote the formation of polarized adhesion appear to do so by modulating this balance68,110,111. Asymmetric polarity complex distribution also polarizes other cell types2 (for example, migratory cells; panel c), although the Crumbs complex is apparently specific to epithelia and epithelial-derived cell types, such as neurons.
Phosphatidylinositol-phosphates (PtdInsPs) are phospholipids that are singly or multiply phosphorylated on the 3, 4 and/or 5 positions on an inositol head group113 PtdIns(3,4,5)P3 can be generated from PtdIns(4,5)P2 by a family of PtdIns3-kinases (PI3K), and PtdIns(4,5)P2 can be generated from PtdIns(3,4,5)P3 by PTEN (a 3-phosphatase). The balance between these two lipids is crucial to polarity homeostasis, and recent evidence reveals that PtdInsP and associated proteins have core roles (see figure, panelsa–c) in membrane identity and polarity generation.
Asymmetric PtdIns(4,5)P2:PtdIns(3,4,5)P3 distribution occurs in various cell types, including migrating neutrophils114, polarized kidney epithelia46 and Drosophila melanogaster photoreceptors115. In polarized MDCK cells (panel b), PtdIns(3,4,5)P3 localizes exclusively at the basolateral membrane, whereas PtdIns(4,5)P2 is enriched apically46,116, like in D. melanogaster epithelia117,118. The PAR complex (PAR3–aPKC) modulates asymmetric PtdIns distribution by recruiting the phosphatase PTEN to tight junctions (TJs)118–120, potentially restricting PtdIns(3,4,5)P3 from moving across the TJ into the apical membrane. Asymmetry of PtdIns(4,5)P2:PtdIns(3,4,5)P3 can also occur without TJs, such as in migrating neutrophils (panel a), where the PtdInsP and associated proteins (G12, G13, EBP50, phospho-ERM (pERM)) control ‘backness’ and ‘frontness’114. PtdInsP distribution can also be asymmetric in a single membrane domain, such as in the apical membrane of D. melanogaster ommatidia (panel c), where PTEN controls levels of PtdIns(3,4,5)P3115.
Asymmetric PtdIns(4,5)P2: PtdIns(3,4,5)P3 distribution is fundamental to the maintenance of cell polarization. Addition of exogenous PtdIns(3,4,5)P3 to basal membranes results in their expansion into the surrounding matrix; apical addition induces rapid loss of apical identity, transcytosis of basolateral membrane to the apical surface and projections from the apical surface121. Addition of PtdIns(4,5)P2 to the basal surface results in rapid redistribution of apical proteins towards the basal membrane46. The opportunistic pathogen Pseudomonas aeruginosa takes advantage of at the this process by inducing ectopic PtdIns(3,4,5)P3 at the apical membrane, converting it into basolateral-like membrane and facilitating entry into the epithelium from the luminal surface122. Actin-regulatory proteins and membrane-trafficking pathways that are influenced by PtdInsP, such as the vesicle-regulating exocyst complex123, may therefore require PtdIns(4,5)P2:PtdIns(3,4,5)P3 asymmetry to ensure correct vesicle targeting, cytoskeletal organization and maintenance of polarity. ZA, zonula adherens.
Despite the integral involvement of polarity complexes in morphogenesis, the mechanisms through which they promote asymmetry are still largely unclear. Recent evidence reveals that these complexes are central organizing platforms that modulate the microtubule cytoskeleton, membrane traffic and phosphatidylinositol-phosphate regulation2 (BOX 2), in part by controlling Rho GTPase activation through guanine nucleotide-exchange factors and GTPase-activating proteins 2,39,60. Disruption of polarity complexes also has marked effects on cellular proliferation, revealing that these complexes have key roles in tumour suppression67,112.
Although many biological processes contribute to the formation of an organ, we will focus on how cell polarity is controlled and contributes to morphogenesis in the context of whole tissues. We discuss the molecular control of tissue polarization in in vivo organs and in in vitro organotypic models, including the establishment, transcriptional control and molecular regulation of tissue polarization, control of polarity orientation, and regulation of polarity by ECM and Rho GTPase signalling. We emphasize the role of epithelial lumen and tube formation and expansion, as epithelial tissues have provided many fundamental insights into how polarity is coordinated at the cellular, tissue and organ level.
The organization of cells into tissues involves the concerted integration of polarizing cues from various interdependent biological processes. First, cells must sense their environment, including where they are in relation to their neighbours. This can be mediated by direct interaction of cells with the ECM through various receptors, such as integrin, dystroglycan and proteoglycan molecules11,12. Cells can sense and modify the chemical composition, assembly, stiffness and other mechanical properties of the ECM13,14. Cells can also communicate with other cells through an array of adhesion molecules, such as cadherins15, and through the sensing of diffusible factors, such as morphogens, chemoattractants and chemorepellants16. These combined cues provide instructions that enable cells to orientate their polarity and begin to assemble into groups. Second, cells in a forming tissue must coordinate the asymmetrical distribution of polarity complexes17 to establish and enforce the generation of an axis of asymmetric organization (BOX 1). Concurrently with this second step, the cytoskeleton and membrane-trafficking systems organize asymmetrically2. These basic steps allow individual cells to become asymmetrically polarized (see the reviews by Bornens and by Nelson and Mellman in this issue). Examples of polarized epithelial and migratory cell polarity and of the formation of polarity complexes are presented in BOX 1.
An important design principle is that polarization must be coordinated between all cells in a tissue. For example, although the organization of polarity complexes, the cytoskeleton, membrane-trafficking events and adhesive junctions must be asymmetric in a single cell, the orientation and organization of this asymmetry must be coordinated between neighbouring cells. In addition to apico–basal polarity, some behaviours, such as cell division and migration, can also be polarized in an orthogonal axis; that is, in the plane of the tissue. Although this planar cell polarity (PCP) is extremely important in tissue formation — for example, it regulates the expansion of a tissue along a particular axis — we will only briefly touch on it here as it has been reviewed excellently elsewhere18,19.
Although the signalling mechanisms to induce and regulate morphogenetic movement during tissue formation are well characterized, how these processes are regulated at the cellular level has only recently started to become clear. The identification of conserved, core polarity-regulating complexes that operate in various contexts has given many insights into these processes (BOX 1). Recent progress in understanding the orientation of tissue polarity, the molecular regulation of epithelial lumen generation and the moulding of biological tubes is discussed below. In addition, key roles for phosphatidylinositol-phosphates (PtdInsPs; BOX 2) and Rho-family GTPases in many of the aforementioned processes have recently become clear, and their contributions to cell and tissue polarity are discussed below.
One design principle underlying morphogenesis is that cells can switch between different types of polarity. In EMT, epithelial cells switch their polarity to that of mesenchymal cells (FIG. 2a). EMT occurs in many developmental processes, such as during gastrulation and during formation of the neural crest and primitive streak, whereby well-polarized epithelial sheets convert to motile mesenchymal cells and give rise to another tissue type20. Loosely defined, EMT involves the disruption of polarized adhesion, such as epithelial (E)-cadherin-based junctions, disruption of apico–basal polarity, reorganization of the cytoskeleton, altered basement-membrane composition and organization, and adoption of motile behaviours and invasion into surrounding tissue. As many aspects of EMT are reminiscent of tumour formation, inappropriate recapitulation of EMT has received much attention as a mechanism for metastasis of epithelial cells (this has been excellently reviewed elsewhere20,21).
Conversely, MET occurs through condensation of mesenchymal cells into tightly adhesive groups, generation of apico–basal polarity, adoption of epithelial characteristics and transition to a polarized epithelial tissue21 (FIG. 2a). MET contributes, for example, to certain sections of the developing kidney22.
The exact definition of EMT is not clear-cut, perhaps because different forms of morphogenesis may use some, but not all, aspects of complete EMT. It is also unclear whether EMT involves only changes in motility, adhesion and polarity, or whether EMT involves changes in cell fate and differentiation20,23. There are examples of partial EMT (pEMT) whereby epithelial cells become motile, but do not completely lose adhesion and polarity24–28 and do not seem to lose epithelial differentiation and fate. It is important, therefore, to emphasize that cells undergoing different forms of EMT do not lose all polarity; rather, they may simply substitute epithelial polarity and characteristics for mesenchymal polarity. Many migratory cells, such as neutrophils or D. discoideum, display strong front–back asymmetry and exhibit polarized migration towards attractive cues10. Some of the core polarity complexes that modulate apico–basal polarity (the PAR complex; see BOX 1) are also involved in the organization of front–back asymmetry2. Morphogenesis of polarized tissues, which involves the contribution of both stable and motile phases during formation, can therefore be seen as movement along a continuum of MET–EMT stages. Such transitions between polarity states are fundamental for shaping many metazoan tissues, and alteration to polarity provides a mechanism to change cell behaviour without necessarily changing cell fate.
Asymmetric morphogen gradients can provide instructive cues for a tissue to undergo morphogenesis, for example, by inducing transcription factors (TFs) to drive these processes16. The Snail and ZEB families of TFs, for example, are potent inducers of, and are often required for, EMT events both in vivo and in vitro29. How these transcriptional modulators induce cellular outputs to regulate tissue polarity has only recently started to become clear.
The traditional view is that the loss of E-cadherin and of other junction proteins induces altered cellular polarity and EMT21. Recent evidence, however, suggests that some EMT-associated TFs also control morphogenesis by directly repressing transcription of molecules that are involved in polarity complexes30,31, membrane-trafficking systems32, the cytoskeleton32 and the basement membrane33 (FIG. 2b). For example, Snail and ZEB1 repress key components of both the scribble and Crumbs polarity complexes30,31, and re-expression of scribble and Crumb complex proteins partially rescues epithelial polarization. Although loss of apico–basal polarity is an inherent part of the definition of EMT, loss of E-cadherin is not always induced by or correlated with expression of some of these TFs32. In these cases, alterations to polarity complexes or to the ECM may suffice to allow for altered cellular morphogenesis. Indeed, during metastasis of colon carcinoma, invasive fronts of cells only occur at sites of lost basement-membrane integrity33, showing that ECM signalling is a key regulator of EMT and thus of tissue polarity.
Stimulation of epithelial cells with various cytokines, such as hepatocyte growth factor (HGF) or transforming growth factor-β (TGFβ), can induce expression of Snail- or ZEB-family members21,29. Snail expression can induce co-expression of ZEB factors34, which further enables additional TGFβ expression, initiating a positive-feedback loop21. Interestingly, this EMT-inducing module is also under the control of a negative-feedback loop that involves endogenous microRNAs (miRNAs)35,36 (FIG. 2b). The miR-200 family of miRNAs promotes epithelial differentiation via downregulation of ZEB1. Loss of the miR-200 family is found in tumour samples that have lost epithelial polarity, and forced re-expression of the miR-200 cluster restores epithelial polarization and differentiation. This downregulation, however, is also reciprocal; miR-200 members are targets for direct transcriptional repression by ZEB1. Whether cells become polarized into an epithelial tissue (MET) or become motile (EMT) therefore depends on a balance between mutually antagonistic miRNA and TF modules, which controls epithelial polarization, adhesion and the ECM. Unravelling the factors that promote the expression of the miR-200 family will be key to understanding the control of epithelial polarization, and may involve the local signalling microenvironment, such as the ECM.
Cells in tissues are often surrounded by ECM. Recent studies have revealed another design principle: the ECM, more than providing a structural scaffold, can define positional information and differentiation cues for tissues, ultimately influencing tissue polarity13. Transduction of these cues via ECM receptors, such as integrins, dystroglycan and proteoglycans, can ultimately lead to changes in cell polarity and shape through various mechanisms. Modulation of the cytoskeleton and signalling through Rho-family GTPases appears to be key to regulating ECM-directed polarity specification12,13,37,38.
A fundamental design principle for forming a polarized tissue is that cells must interpret an initializing cue to polarize. When isolated cells, such as neutrophils or D. discoideum, are stimulated with a chemoattractant, they typically polarize and migrate towards the source of the molecule10. This orientation of polarity is set up by an asymmetric gradient of the initiating cue, such as a bacterium or chemokine. The PAR complex (BOX 1) is involved in such front–back polarity in migratory cell types, in part by controlling asymmetric orientation of the microtubule cytoskeleton39. What governs the orientation of apical and basolateral polarity in epithelial tissues, on the other hand, is less clear, partially because of the systems that are used to study this process. Polarized epithelial cells that are cultured on an artificial support, such as a dish or filter, receive strong, asymmetric initiating cues for polarization — the rigid substratum provides a cue, while the ‘free’ medium provides another — and there is consequently an axis from which the cells can collectively orientate their polarity1,40.
How is the orientation of polarity defined when cells are completely surrounded by ECM, such as in vivo? Studies using epithelial cells grown in 3D culture have provided some answers (BOX 3). Normally, MDCK (Madin–Darby canine kidney) cyst structures form as a simple epithelium surrounding a central luminal space40. Inhibition or loss of β1-integrin or the downstream GTPase RAC1 results in inversion of this orientation such that the apical surface becomes orientated towards the matrix12,37. A similar inversion of glandular epithelial polarity also occurs in a subset of invasive ductal breast carcinomas41, demonstrating that inversion of polarity can also occur in vivo during tumorigenesis. By contrast, inversion of polarity orientation is promoted by, and dependent on, the inappropriate activation of the RhoA GTPase and its downstream effectors, RoCK1 and myosin II38. Despite this inversion, cell–cell junctions form and cells maintain some level of polarity; that is, proteins and lipids are asymmetrically distributed. This suggests that the establishment and correct orientation of polarity are not obligate partners and can be molecularly uncoupled. It further suggests that signalling at the cell–ECM interface is a primary determinant of the axis along which epithelial cells orientate their polarity.
Organotypic three-dimensional (3D) culture of epithelial cells in extracellular matrix (ECM; matrigel, a basement membrane-like tumour product, or COLI gels) provides a way to model epithelial morphogenesis in vitro. Many types of epithelial cells form organoid structures in 3D culture24,40,124–126, recapitulating varying levels of tissue polarity and architecture. For MDCK cells, at least, this is analogous to a mesenchymal–epithelial transition (MET). Motile cells that are embedded in matrix either migrate together and adhere, or proliferate from an individual precursor to form stable adhesive complexes. Cells then undergo morphogenesis and polarization to eventually form simple epithelial cysts that surround a hollow, fluid-filled lumen. Stimulation with growth factors, particularly hepatocyte growth factor (HGF), induces branching and tubulogenesis of structures, modelling some aspects of embryonic kidney development127,128. Notably, the tubules produced in this system closely resemble the morphogenesis of the mouse embryonic nephric duct (C. Mendelsohn, personal communication). 3D culture has proved to be indispensable for examining aspects of cell polarization, lumen formation, oncogene function and the influence of the ECM on tissue organization, which remain technically challenging to manipulate in vivo. Notably, mutants in many polarity regulators or oncogenes show vastly different phenotypes in 3D culture than in traditional 2D culture1,24,37,46, even in instances where no phenotype is observable in 2D culture.
On the basis of 3D culture, researchers have established a simple set of rules that epithelial cells follow during polarization1,40. Each cell strives to have three types of surface: a basal surface, which contacts the ECM, a lateral surface, which contacts other cells, and an apical surface, which faces the lumen. A cell that does not contact the ECM undergoes apoptosis, whereas a cell that lacks an apical surface will generate a lumen at a region of contact with other cells, or even within itself. Notably, such parameters can be incorporated into an in silico model of epithelial morphogenesis and can yield remarkably life-like ‘cysts’ during simulations129. Computational and systems biology approaches are likely to play an increasingly important role in the development and analysis of in vitro tissue systems, such as in regenerative medicine or stem-cell-derived differentiation of epithelial tissue types, in which induction and maintenance of a polarized epithelium will be crucial.
Maintenance of polarization and migration of neutrophils towards a chemoattractant requires the continued sensing of the polarizing cue, such as by G-protein-coupled receptor-mediated activation of PtdInsP signalling cascades10. Analogously, maintenance of epithelial polarity in tissues may require the continued sensing of polarizing cues, such as detecting the ECM through integrins and detecting cell neighbours through cadherins. Notably, both of these receptors can control the generation of PtdIns(3,4,5)P312,42–44, which has a fundamental role in generating and maintaining basolateral membrane identity (BOX 2), at least in MDCK cells. Cadherin and integrin receptors can signal through small GTPases, such as to the Rho and Arf families, which can function both upstream and downstream of PtdInsP. Receptor–GTPase–PtdInsP signalling modules (whether the receptor is integrins, cadherins or a G-protein-coupled receptor) may therefore be key to both the generation and maintenance of tissue polarity in diverse contexts.
The three prototypical Rho GTPases, RhoA, RAC1 and CDC42, play integral roles in cytoskeletal arrangement, membrane-trafficking pathways and ECM interactions, and all of these roles are crucial for cell polarization45. CDC42 function is crucial for epithelial lumen formation46–48 (see below and BOX 1), whereas RAC1 is associated with both integrin and cadherin signalling12,44 and controls the orientation of polarity in epithelial and migrating cell types12,49. RhoA is associated with both apical and basal membranes in epithelial cells50 and at the rear of migrating cells51 and appears to regulate cell shape in some systems45.
Various connections between the PAR complex and Rho GTPases have recently emerged, emphasizing their key roles in cell polarity. Unique functions of Rho GTPases occur at discrete subcellular locales, regulated by guanine nucleotide-exchange factor (GEF) and GTPase-activating protein (GAP) proteins. For instance, binding of CDC42 to PAR6 is required for proper function of this apical determinant52. In C. elegans embryos undergoing radial polarity generation, CDC-42 targets the atypical protein kinase C (aPKC)–PAR-3–PAR-6 polarity complex to non-junctional membranes53. This requires a GAP (PAC-1) to exclude CDC-42 from junctions and a (as yet unidentified) GEF protein to activate CDC-42 at apical membranes. Deletion of this GAP causes ectopic mistargeting of polarity complexes and polarity defects. PAR-3 binds the Rac GEF TIAM-1 to regulate cell–cell junctions. TIAM-1, through aPKC, apparently controls microtubule organization, providing a mechanism for the PAR complex to regulate asymmetric cytoskeletal organization39. In certain Drosophila melanogaster and chick embryonic tissues, GEF-mediated activation of RhoA at the apical surface, coupled to GAP-mediated inactivation of RhoA at the basal surface, allows apical cell constriction and remodelling of the epithelium while maintaining apico–basal polarity54–59.
The PAR6–aPKC subcomplex also regulates RhoA signalling through direct interaction with and modulation of p190A RhoGAP activity60. RhoA–ROCK signalling can conversely disrupt aPKC–PAR6–PAR3 interaction and function by direct phosphorylation of PAR3 (REF. 61). This emphasizes the emerging notion that PAR–Rho GTPase complexes act as discrete multimeric signalling ‘hubs’ in different regions of the cell, controlling various aspects of asymmetric cellular organization and polarity. Identifying unique functions for Rho GEFs and GAPs will be crucial to understand Rho GTPases and PAR complex function at such discrete locales. As PAR3–TIAM1 complexes can control microtubule organization39, and as the PAR6–CDC42 subcomplex controls recycling from endocytic compartments62, both the cytoskeleton and membrane trafficking systems may be direct targets of such complexes. Similarly, given the key roles of Rac, CDC42 and Rho in cell polarity, it will be important to dissect overlapping or potential novel roles of the remaining Rho GTPase family members in tissue formation and polarity. For a more extensive discussion of the role of GTPase signalling in cell polarity, see the review by Iden and Collard in this issue.
During the morphogenesis of an epithelial tissue, cells often organize into biological tubes. Such tubes provide the basic plumbing that is crucial for organ and organismal function, and their formation is therefore a fundamental event in the generation of diverse tissues during metazoan development63. For example, vascular tubes allow for transport of O2 and nutrients throughout the body, the digestive system lumen allows absorption of food and mammary tubes allow the secretion of milk64. Although classic embryology has shown that there is an enormous diversity of mechanisms for tube formation, some common themes and molecular regulators have emerged.
The paramount requirement for a biological tube is that a lumen must form, and the lumen must be enclosed and unobstructed. Many tubes form by rearrangement of existing epithelial sheets. Such sheets can be deformed by evagination, invagination or similar folding to give rise to tubes that bud off64, such as the branching of the ureteric bud during embryonic kidney development. A variant of this occurs when an epithelial sheet folds or rolls up along an axis that is parallel to the plane of the sheet, such as in the neuroepithelium of the developing chick. By merging the epithelium only at specialized contact points, a tube with radial tissue symmetry and a central lumen can be formed63. As such morphogenetic movements have been reviewed elsewhere50,63,64, we concentrate here on more recent developments in the different lumen- formation mechanisms of cavitation, hollowing and membrane repulsion. TABLE 1 lists the molecular machinery that is currently implicated in these processes. In these contexts, groups of poorly polarized cells can begin to tightly adhere to one another and generate lumens de novo (see below; FIG. 3). For cavitation and hollowing, such adherence is essentially a MET event and occurs, for example, as part of the condensation of the metanephric mesenchyme during kidney development22.
Cavitation occurs when a group of cells proliferate in an adhesive, but initially only moderately polarized, manner (FIG. 3a). The selective apoptosis of cells that are not in contact with the ECM gives rise to an outer epithelial layer surrounding a now hollow lumen. This process occurs, for example, in three-dimensional (3D) models of mammary acini, and in in vivo mouse mammary end buds24,65,66. In these situations, pro-apoptotic BCl2-family factors have key roles in luminal cell apoptosis, although additional mechanisms, such as autophagy, appear to contribute to luminal clearance in such tissues65,66. The PAR, scribble and Crumbs complexes have important roles in suppressing cell proliferation in D. melanogaster tissues67, and both the PAR (aPKC–PAR6) and scribble complexes promote apoptosis of luminal cells during cavitation68, thus contributing to the formation of a polarized tubular epithelium.
During hollowing of rapidly polarizing groups of cells, intracellular vesicles (varying in size between systems) are delivered to the cell surface at a coordinated point between two closely apposed cells, creating a luminal space de novo (FIG. 3b). These vesicles are thought to contain fluid that is taken up by endocytosis, and to contain apical proteins that are destined for delivery to the lumen. Their movement to the cell surface results in the generation of space between two (or more) polarized cells, and concomitant cell-surface delivery of the apical, luminal membrane. The surrounding cells now exhibit apico–basal polarity and are orientated around a lumen, and the whole tissue is subsequently expanded in a highly polarized manner. This mechanism has been observed in 3D organotypic models of kidney and vascular development, as well as in blood vessels in vivo46,69,70 (although its observation in blood vessels is controversial71).
A molecular understanding of this process has recently become clear46 and involves concerted integration of PtdInsP, Rho-family GTPases and the PAR polarity complex. Signalling from the ECM, through integrin receptors, initially orientates apico–basal epithelial polarity in newly adhering groups of cells (see below). Enrichment of PtdIns(4,5)P2 at the apical plasma membrane by the lipid phosphatase PTEN results in the apical recruitment of the small GTPase CDC42 via the PtdInsP-binding protein annexin-2. Activated CDC42 in turn binds the PAR6–aPKC polarity complex, thereby ensuring targeting to the apical membrane (FIG. 3b; BOX 1). Scaffolding of this complex to nascent Lumina is required to efficiently generate apico–basal polarity and, consequently, a single polarized lumen.
Interestingly, when vacuolar exocytosis is inhibited, lumens eventually form by cavitation47, which emphasizes the robust drive of epithelial cells to form a hollow lumen. The CDC42–PAR6–aPKC polarity complex appears to be a master regulator of lumen formation; its disruption results in either multiple small lumens or in the accumulation of apoptotic cells in the lumen46,72,73. Interestingly, in 3D organotypic culture of mammary acini, activation of the ERBB2 oncoprotein, which promotes cell survival and lumen filling in human cancers, also results in a lumen-filling phenotype by inducing uncoupling of PAR3 from the PAR6–aPKC complex68. The identities of the downstream effectors of this complex in lumen formation are still unclear, although regulation of both glycogen synthase kinase-3β(GSK3β) and microtubules have recently become good candidates39,73. Thus, the CDC42–PAR6–aPKC complex appears to regulate lumen polarity, formation and maintenance in diverse tissue contexts.
During D. melanogaster heart-tube formation, membrane repulsion appears to regulate lumen formation. Here, two rows of myoendothelial cells line up along the midline and extend membrane process towards the mirroring cell on the other side of the midline. In epithelial cells, this would normally result in adhesion between cells along the entire lateral contact. In these cells, however, junctions occur only at the dorsal- and ventral-most regions, resulting in the formation of an enclosed luminal tube between the rows of cells (FIG. 3c). Interestingly, Slit–Robo signalling, a ligand–receptor coupling that governs repulsive signalling in other cellular contexts, occurs at the site where the future lumen will form74,75, excluding cadherin complexes from the region and promoting luminal development between cells. As cells must convert adhesive regions between cells into nascent lumens during hollowing, it will be important to determine whether repulsive forces or anti-adhesive molecules (such as gp135 (also known as podocalyxin)76) play analogous roles in other tissues. Likewise, although the fly homologue of PAR3 localizes to adhesive interfaces in developing lumens, whether any of the PAR, scribble or Crumbs polarity complexes play a role in this mechanism of lumen formation is currently unclear.
Once rudimentary luminal structures are formed, tubes must lengthen and become interconnected networks. In some systems, such as the mammalian lung and mammary ducts, this is achieved primarily through budding of new sprouts and branches63. This design principle can make repeated use of such budding mechanisms, with the branches following simple patterns77. Many rounds of such iterative branching give rise to an extensive tree. In other networks, such as the developing D. melanogaster trachea, different parts of the tubular network arise in an initially unconnected manner and sprouting branches extend towards each other, fusing and creating an anastomozing network78. Complex rearrangements of apical polarity and cell–cell adhesion must occur when two branches of a network eventually undergo fusion to maintain lumen integrity and generate an interconnected lumen79. In some cases, lumens form in single cells, although these often connect up to lumens between multiple cells (FIG. 4). In D. melanogaster airways, for example, this occurs via switching between intercellular (between two or more cells) and autocellular junctions80 (in a single cell). specialized adhesion molecules, luminal matrix proteins and modulation of endocytic recycling pathways are involved in such rearrangements80–83.
Most, if not all, morphogenetic mechanisms involve alterations in cell polarity at some level, although how polarity is regulated and remodelled during morphogenesis is poorly understood. In simple epithelia, the PAR, scribble and Crumbs polarity complexes specify apico–basal polarity (BOX 1). In pseudo-stratified or multilayered epithelia, or in Lumina in single cells, organization of the tissue, such as location of cell–cell junctions and biogenesis of the apical membrane, is more complex, and how polarity is controlled in these situations remains unclear. For example, budding of the mammary gland tubular tree has recently been shown to involve an intermediate step whereby the buds are filled with multilayered cells84; the polarization of multilayered cells is different from and much less well understood than that of simple (monolayered) epithelia. When MDCK cysts are treated with HGF to form tubules, cells undergo a pEMT, passing transiently through an intermediate stage of chains and cords of cells, which have some properties of migrating mesenchymal cells, before returning to complete epithelial polarity25,26. Such chains and cords can also be viewed as small regions of multilayered cells, somewhat analogous to the multilayering seen in mammary gland buds. The polarity requirements of different regions of a tissue, therefore, changes during morphogenesis; cells can switch, even transiently, between polarized states.
Once a lumen is formed, how is it expanded to the appropriate physiological diameter? Several diseases are caused by tubes that are either too wide85 (for example, polycystic kidney disease (PKD)) or too narrow80 (for example, vascular stenosis)86. In the cavitation model of lumen formation, proliferation of the entire tissue to the required dimension followed by regulated cell death of inner cells allows for the generation of appropriate lumen diameter24,65. In other circumstances, such as in lumens formed by hollowing, sheet wrapping or folding, various mechanisms for lumen expansion exist. In the embryonic zebrafish gut, multiple small lumens are remodelled into a single tube in a process that is dependent on the accumulation of luminal fluid, regulated by certain pumps (Na+/K+-ATPase) and modulated by the ion permeability of TJs (claudin-15)87. Similar regulation of lumen expansion is also observed in MDCK cysts and in thyroid 3D cultures where chloride ion secretion is involved in determining lumen diameter87,88. Notably, inhibition of the CFTR chloride channel inhibits lumen overexpansion in both MDCK cyst and mouse models of PKD, suggesting that chloride transport may be a key regulator of lumen size89. Moreover, coordinated regulation of the Na+/K+-ATPase and septate junctions (the invertebrate equivalent of TJs) also appears to be key for appropriate lumen expansion in the D. melanogaster trachea90,91. Interestingly, this function of the Na+/K+-ATPase appears to be independent of pump activity90, suggesting that there are some differences in the mode of actions between species or tissues. Whether multiple small lumens occur as obligatory precursors to a single lumen and whether fluid accumulation drives lumen expansion in other biological tubes remain to be determined.
At least two additional processes account for changes in lumen diameter in the tracheal tube and salivary gland of D. melanogaster (FIG. 4a). Initially, the apical surfaces of cells on opposite sides of the tracheal tube are close together, and the lumen is narrow. A rapid burst of membrane traffic to the apical surface and secretion into the lumen occurs, concomitant with lumen expansion92,93. In the trachea, this corresponds to an increase in lumen diameter but not in length (which is apparently controlled by other mechanisms), whereas in the salivary gland both the lumen diameter and length are increased92. At roughly the same time an intra-luminal ECM, comprised of fibrils of the Polymer chitin, is assembled in the trachea. The synthesis and subsequent modification of chitin fibrils is thought to provide a mould over which the precise dimensions of the tube can be modelled (reviewed in REF.94). The burst of exocytic traffic around this time controls some of the chitin-regulating enzymes, and thus likely controls some of the lumen-expansion mechanism. Subsequently, and in the case of the trachea before gas enters into the lumen, the lumen is rapidly cleared via endocytosis at the apical surface of tracheal cells93. Perturbation of exocytosis or endocytosis halts the processes of lumen expansion and clearance, demonstrating that there is an integral role for membrane trafficking in modulating lumen morphogenesis. It remains unknown whether trafficking of chitin-modifying enzymes or of additional lumen-destined cargo is the most crucial contributor to lumen expansion. However, these studies raise the possibility that rather than responding to a pre-existing apical ECM, tubular epithelial cells may transiently generate their own apical matrix, which may act to regulate lumen expansion.
As there is no homologue of chitin in vertebrates, the extent to which these principles can be directly extended to mammalian tubular systems is unclear. Instead, secretion of other ECM molecules, such as proteoglycans into lumens, which is observed, for example, in the D. melanogaster retina, may play an analogous role82,83,95 (FIG. 4b). The potential participation of an apical matrix in lumen formation and expansion thus remains an attractive concept. Similarly, membrane-trafficking pathways to the apical surface will likely play a key role in lumen formation and tissue polarity.
Although we have begun to understand how TFs can promote the loss of polarized epithelial characteristics, how do transcriptional regulators promote the expression of genes that induce polarization of tissues? Some insight has come from studying D. melanogaster tubular epithelia, such as salivary glands, in which a network of TFs, including hairy, huckebein and ribbon, control expression of genes involved in apical polarity and luminal development, among other targets. These targets include components of the Crumbs complex (BOX 1), as well as the apical transport machinery, such as the molecular motor dynein and certain luminal ECM-modifying enzymes96,97, all of which have varied roles in creating the apical luminal structure. Similarly, in the developing zebrafish gut, the TCF2 TF regulates the expression of certain TJ proteins and ion pumps87, which regulate apical lumen expansion.
During branching of the D. melanogaster airways, the transcription factor Spalt (induced by Wnt–Wingless signalling) promotes expression of RAB11 and RIP11, two members of a membrane-recycling pathway that is involved in surface delivery of D. melanogaster E (DE)- cadherin complexes. RAB11 and RIP11 promote inter-cellular adhesion and the formation of lumens between two or more cells. Interestingly, repression of Spalt by the repressor Knirps (induced by TGFβ–decapentaplegic signalling) in adjacent regions of the network results in downregulation of this cadherin-recycling pathway, promoting the formation of autocellular junctions and lumens in single cells80. Notably, the RAB11–RIP11 complex is also involved in D. melanogaster retina development98 (FIG. 4c), and recycling from these endosomes is regulated by the PAR6–CDC42 complex62, suggesting that diverse lumen-formation contexts are a core requirement for this pathway. Thus, in addition to transcriptional promotion or repression of junction proteins, polarity complexes and basement-membrane proteins, traditional ‘fate inducing’ signalling molecules, such as Wnt and TGFβ, can induce tissue formation through the regulation of membrane-trafficking pathways. Further examination of the transcriptional control of membrane traffic, of which little is known, should yield enormous insights into the regulation of cell polarity and morphogenesis during tissue formation.
Although much has been learned about how the polarity of individual cells is established and maintained, we are still in the early days of understanding how polarized cells are put together to make tissues. Communication between cells, both through cell–cell contact and via the ECM and diffusible factors, and using both chemical and mechanical signals, is at the heart of polarized tissue and organ formation.
Over 85% of fatal malignancies in adults in the USA arise from epithelial tissues99, and loss of polarity is a hallmark of increased malignancy. The mechanistic bases for this connection are rapidly being elucidated, as is described in several recent reviews20,67,100. Acute injury of major epithelial organ systems is collectively one of the most important causes of death worldwide101–103. Understanding polarization of epithelia, therefore, is important in analysing the response of a tissue to acute injury and in generating prospects for regenerative medicine. Many organs, such as the kidney, lung and liver, can recover from even severe or acute injury, provided that the patient survives the initial insult. In the case of the kidney, at least, this involves the local proliferation of epithelial cells, which replace their dead neighbours in denuded regions of the tubules through a process that appears to involve a pEMT104,105. Repeated injury, however, leads to a permanent EMT, whereby epithelial cells become fibroblastic and contribute to a fibrotic response, which ultimately destroys organ function106. Learning how to improve the response to acute injury, as well as how to avoid fibrosis and EMT after chronic injury, such as by controlling the polarity state of cells, offers enormous possibilities to enhance human health. As we begin to understand how polarization occurs and is controlled at the tissue level, we move closer to being able to translate such research potential into medical reality.
This work was supported by National Institutes of Health grants to K.E.M. and a Susan G. Komen Foundation Postdoctoral Fellowship to D.M.B. We thank R. Metzger, C. A. Hunt and A. J. Ewald for comments on the manuscript and members of our laboratory for discussions. This paper is dedicated to the memory of our colleagues S. Ross and P. Kolodzeij.
aPKC | CDC42 | claudin-15 | Crb | Dlg | ERBB2 | PAR-3 | PAR6 | PATJ | RAC1 | RhoA | ROCK1 | Scrib | Snail | ZEB1
Mostov lab homepage: http://anatomy.ucsf.edu/mostovlabpage/Homepage.html