Tubulogenesis

Tubes are the fundamental structural unit of many organs, including the vascular system, exocrine glands, digestive tract, lung and kidney. As metazoans evolved into creatures larger than small clusters of cells, tubes became essential for transporting and distributing metabolites. Given the complexity of the processes that create tubes of specific sizes and shapes, it is not surprising that their development and maintenance can go awry and cause devastating diseases.

Biological tubes come in a remarkable diversity of sizes and range of complexity. They can be small, such as the unicellular tubes in small capillaries or in the C. elegans kidney cell. They can be large, but with relatively simple organization, as demonstrated by the massive multicellular tube of the intestine of a blue whale. They can also form organs of extreme complexity, such as the human lung, which includes some 10⁵ conducting and 10⁷ respiratory airways. Tubes can be composed of single or multiple cell layers with the additional incorporation of other tissues, such as connective tissue and muscle (e.g. the vascular system and esophagus).

Many tubular organs are ultimately contiguous with the external environment and develop from polarized epithelial cells of ectodermal, endodermal or mesodermal origin. However, irrespective of their cellular origin they exhibit one of two types of organization. The large majority are tubes with ‘seams’ that in transverse section contain a lumen enclosed by one or more cells connected by auto- or intercellular junctions (Lubarsky and Krasnow, 2003). Less common are ‘seamless’ tubes with no cellular junctions in the circumference of the lumen. Examples of such seamless tubes include the sinusoidal blood vessels in the mammalian kidney (Bär et al., 1984), the C. elegans kidney cell (Buechner, 2002) and the Drosophila tracheal system (Maruyama and Andrew, 2012). Importantly, an organ can contain multiple tube types and junctional organizations (Buechner, 2002; Herwig et al., 2011; Lubarsky and Krasnow, 2003). Furthermore, there are examples of tubes converting between the seamed and seamless types during development, which involves complex cell surface rearrangements to introduce or eliminate cell-cell junctions and interfaces (Dong et al., 2009; Ribeiro et al., 2004).

A common topological feature of both seamed and seamless tubes is the presence of a single lumen. Whereas a single lumen arises naturally when a tube forms by invagination (see ‘budding’ below), this outcome is not obvious when a tube forms by the coalescence of vesicles or compartments. Consistent with active processes establishing and maintaining a single lumen in a tube, mutations have been identified that disrupt lumen coherency in the zebrafish gut and Drosophila tracheal system, resulting in tubes with multiple or discontinuous lumens (Bagnat et al., 2007; Levi et al., 2006).

Although seamed tubes have basically the same organization by the end of development, there are at least five cellular mechanisms for creating what is essentially a sheet of cells rolled into a tube.

A common mechanism is termed ‘budding’, in which one or more cells invaginate from an existing sheet or tube of cells to create a short new tube that is subsequently extended by cell migration and/or cell division (Chung and Andrew, 2008). A classic example of budding is angiogenic sprouting, in which a new blood capillary sprouts or buds off an existing vessel (Iruela-Arispe and Davis, 2009).

Alternatively, in ‘wrapping’ an entire row of cells can invaginate and pinch off from a sheet of cells to immediately create a longer tube (Lubarsky and Krasnow, 2003; Sawyer et al., 2010). A highly relevant example of wrapping is neural tube formation in mammals, where incomplete closure of the developing tube results in spina bifida. An important distinction between budding and wrapping is that the lumen of a bud is contiguous with the space from which the bud invaginated, whereas wrapping typically isolates a lumen from an existing space.

Lumenal space is also captured to form a tube by cells that are not in a contiguous sheet or tube through the strategy of ‘lumen entrapment’. The best defined in vivo example of this is Drosophila heart formation in which cells migrating in from opposite sides of the embryo converge and touch at their opposing edges, but repel at their opposing centers, thus enclosing a lumen (Medioni et al., 2008; Santiago-Martínez et al., 2008). Lumen entrapment is also used by human endothelial cells in culture to form capillary-like tubes. Single cells exocytose a large vacuole and then ‘capture’ the former vacuolar space by resealing the opening with autocellular junctions (Davis et al., 2002).

A common theme of budding, wrapping and lumen entrapment is that an existing space is extended or captured. However, a lumen can also be formed by opening a space inside a cord or group of cells. ‘Cavitation’ creates space by eliminating cells that reside inside a cluster and replacing them with fluid. For example, in mammary gland development, lumenal space is created by apoptosis or autophagy of cells inside the gland (Mailleux et al., 2007). Alternatively, in ‘cord hollowing’ adjoining cells in a cord establish apposing apical surfaces and secrete fluid and matrix into the lumen. Examples of cord hollowing include zebrafish intestinal and mammalian dorsal aorta development (Herwig et al., 2011; Martin-Belmonte and Rodriguez-Fraticelli, 2009), as well as Madin-Darby canine kidney (MDCK) cell tubulogenesis in a three-dimensional matrix. Notably, there are some similarities between lumen capture by vertebrate endothelial cells and cord hollowing by MDCK cells in that apical surface is created de novo. However, there is an important topological distinction in that during cord hollowing the new lumenal space is never contiguous with the space exterior to the developing organ, whereas lumen capture converts exterior space into lumenal space.

Formation of seamless tubes proceeds analogously to cord hollowing, except that the creation of the lumenal surface occurs entirely intracellularly and thus constitutes a distinct mechanism termed ‘cell hollowing’. Well-characterized examples of cell hollowing are the development of the C. elegans excretory system (Buechner, 2002) and the Drosophila tracheal terminal and fusion branches (Levi et al., 2006).

Tubulogenesis and tube size regulation require the coordinated orchestration of multiple molecular pathways including cell surface receptors, cell matrix, adhesion molecules, the cytoskeleton and vesicular transport. Small GTPases play a central role in regulating many of these processes. Here we highlight some of the emerging mechanisms that regulate tubulogenesis.

Both epithelial and vascular lumens rely on the molecular crosstalk between cell-cell and cell-matrix adhesion pathways (Iruela-Arispe and Davis, 2009). These connections provide the information required for the acquisition of cell polarity, a crucial prerequisite for lumen formation. Polarization results in the segregation of cell-cell contacts to the lateral aspects of the cell, while associations with the extracellular matrix (ECM) become exclusive to the basal side. The distribution of these cell surface molecules is concurrent with polarization of the centrosomes and, through an as yet unclear mechanism, regulates the orientation of cell division (Taylor et al., 2010; Zovein et al., 2010). This is crucial because cell divisions in the plane of the lumen wall result in stratification, whereas division orthogonal to this plane will lead to enlargement of the tube.

Although the complete molecular framework responsible for triggering cell polarity during lumen formation is yet to be uncovered, a few key molecules have been identified. Not surprisingly, as key mediators of cell-matrix interactions, integrins are central regulators of polarity and lumen formation. Genetic or pharmacological blockade of integrins results in arrest of vascular tubulogenesis and, although the specific mechanism by which integrins control lumen formation is unclear, both the orientation of cell division and the distribution of cell-cell adhesion molecules are impaired (Drake et al., 1992; Zovein et al., 2010). For example, in the endothelium, inactivation of β1 integrin yields stratification of the endocardium due to inappropriate orientation of cell division, as well as persistent lumenal expression of cadherins and other cell surface proteins (Zovein et al., 2010). These abnormally localized cell-cell adhesion molecules engage in interactions with neighboring cells, blocking the emergence of a lumen (Zovein et al., 2010).

Cell polarity can also play a decisive role in controlling the dimensions of a tube once a lumen has formed. Changes in the activity of conserved apical-basal polarity complexes that include conserved proteins such as Crumbs and Scribble (Kerman et al., 2008; Laprise et al., 2010), or the activity of planar cell polarity (PCP) complexes (Chung et al., 2009), can change tube size by more than 50%.

Cytoskeletal proteins and their associated GTPases constitute a crucial set of lumen regulators. Both the actin cytoskeleton and microtubules participate in lumen formation by regulating vesicular trafficking and by enabling changes in cell shape that promote lumen enlargement or reduction. Dominant-negative mutants for Rac and Cdc42, which are small GTPases that modulate actin assembly, impair endothelial lumen formation in vitro (Bayless et al., 2000; Koh et al., 2008), while in mouse models endothelial deletion of Rac1, as well as its downstream effector Wave2 (also known as Scar2, Imd2 or Wasf4), results in defects in vascular development (Tan et al., 2008; Yamazaki et al., 2003). More recently, genetic inactivation of Rasip and Arhgap29, which are both regulators of Cdc42 and Rac, was shown to lead to lumen collapse in the vertebrate dorsal aorta (Xu et al., 2011).

Mechanical properties of the cytoskeleton also play a key role in tubulogenesis, as revealed by mutations in the BTB protein Ribbon, which increase apical cytoskeletal stiffness several fold and markedly reduce lumen elongation (Cheshire et al., 2008).

Finally, matrix metalloproteinases have also been shown to contribute to lumen formation. In particular, the polarized expression of specific membrane-tethered matrix metalloproteinases in vascular sprouts facilitates activation of Cdc42 and progressive invasion through the degradation of ECM molecules (Sacharidou et al., 2010).

Topologically, a lumen develops when a cell or cells delineate membrane surfaces that enclose a space. However, this does not necessarily mean that a usable tube is created because intercellular adhesion and/or intercellular matrix between apposing cells can prevent a lumen from enlarging to usable dimensions. At least two distinct mechanisms can provide the force needed to separate cell surfaces and generate a lumen between closely apposed cell surfaces. In zebrafish brain ventricle development, the Na⁺/K⁺-ATPase containing the Snakehead alpha subunit (Atp1a1) is required to ‘inflate’ the ventricle, presumably through ion transport generating hydrostatic pressure to separate cell surfaces (Lowery et al., 2009). During development of the mouse aorta, by contrast, electrostatic repulsion of heavily sialylated proteins is required to separate the lumenal faces of the aortic cells (Strilić et al., 2010).

Once an open tube is formed, organ function is dependent on lumen size, as this dictates its flow properties. Several obvious mechanisms for tube size control have been identified, including regulation of cell number, cell shape, cell orientation, lumenal matrix, membrane trafficking and pressure/flow. Importantly, although there is overlap between the molecular mechanisms that form tubes and those that regulate tube size, there are also mutations that exclusively target tube size without affecting the initial process of tube formation (see below). The extent to which this separation reflects the differential deployment of similar mechanisms at distinct development times, as opposed to fundamental differences in the formation versus maintenance of an organ, remains to be determined.

Perhaps the simplest strategy for controlling lumen diameter is cell number: add more cells around a tube and it should increase in diameter (or length); reduce cell number and the tube should get narrower (or shorter). The question of control then becomes a problem of coupling cell division and death to a system that senses lumenal size. In the mammalian vascular system, increased levels of soluble growth factors, such as vascular endothelial growth factor (VEGF), can drive increased vessel diameter or length (Lee et al., 2005). However, tube size can be controlled independently of cell number. Classical experiments with salamanders demonstrated that tube diameter (and length) could be maintained despite an approximately fivefold change in cell number (Fankhauser, 1945). Similarly, approximately fourfold changes in cell number in the Drosophila tracheal system did not dramatically alter the size of tracheal tubes (Beitel and Krasnow, 2000).

Changes in the organization of cells within a tube can also affect its size. Intercalation can increase the length but decrease the diameter of a tube. Developmental examples of this include the Drosophila Malpighian tubules (Jung et al., 2005) and hindgut (Lengyel and Iwaki, 2002). Changing the shape and orientation of cells also alters tube size. For example, in the Drosophila tracheal system, mutations in Src42 do not change the rectangular shape of tracheal cells but instead cause cells to expand in the circumferential rather than axial direction of the tube, resulting in shorter but wider tubes (Förster and Luschnig, 2012; Nelson et al., 2012). Mutations in lumenal ECM proteins have been shown to increase the length of tracheal cells and tubes (Araújo et al., 2005; Swanson and Beitel, 2006). In the mammalian vascular system, rapid changes in diameter do not involve changing endothelial cell number but are instead controlled by contraction and relaxation of smooth muscle cells that surround the vessels.

Given that most lumens are bounded by the apical surfaces of the enclosing epithelial or endothelial cells, it is unsurprising that regulation of vesicular trafficking can drastically affect lumen size. In the C. elegans renal tubule, vesicular trafficking defects cause significant dilation of the tubules (Buechner et al., 1999; Mattingly and Buechner, 2011), while apical delivery of the polarity protein Crumbs, and of other apical membrane factors, controls lumen diameter in the Drosophila salivary gland and trachea (Kerman et al., 2008; Myat and Andrew, 2002).

The ability of tubes to contend with and regulate their size is illustrative of the interplay of physical forces and tubes in biological systems. As a vital function of biological tubes is to transport liquids or gasses, pressure and flow are two physical parameters that cells can sense to regulate lumen size. These physical forces, although often dispensable during the initiation of lumen formation, are essential for keeping the lumen open and also seem to regulate lumen diameter (Jones et al., 2006). Endothelial and epithelial cells can be exquisitely sensitive to shear stress, and loss of the ability to sense flow in the kidney can result in the grossly expanded lumens of mammalian polycystic kidney disease (Wilson and Goilav, 2007). By contrast, blockade of flow in a vascular segment leads to its regression, whereas an increase in pressure results in remodeling of the vessel. Interestingly, recent work has indicated that lymph vessel size is regulated not by flow but by mechanical stress created by the swelling of tissues that are accumulating fluid (Planas-Paz et al., 2012).

At first glance, tubes in biological systems appear to be relatively simple. Sheets of cells enclosing a lumen: how complicated could these be to make? In fact, they are surprisingly complex. Research to date has revealed a plethora of cellular and molecular mechanisms that create and maintain these tubes. It seems almost as though no two tubes are the same even within a given organism or tissue. The complexity of tubulogenesis makes it difficult to draw broad conclusions about the process. Generalizations are further complicated by the fact that even in the most studied systems there is still a very limited understanding of the molecular mechanisms of tubulogenesis. Thus, although there currently appears to be little commonality between the specific mechanisms of lumen formation in different contexts, it is likely that we are at the stage of the proverbial blind men trying to describe an elephant. Every system presents a different entry point into the problem and we are unable to see how the threads weave together.

Nonetheless, a few common elements of tubulogenesis are evident. First, polarity is a key prerequisite for lumen formation. Establishment of molecularly distinguishable basal, lateral and apical sides is essential for the changes that underlie lumen morphogenesis. Second, the molecular mechanisms involve the same broad group of players, including the cytoskeleton, cell adhesion, junctional complexes and cell-matrix interactions. Yet their use and regulation are surprisingly distinct. The diversity of solutions employed in making seemingly simple structures clearly underscores the evolutionary importance of developing tubes in biological systems.

We apologize to authors whose work we could not directly cite owing to space limitations.