Insane in the apical membrane: Trafficking events mediating apicobasal epithelial polarity during tube morphogenesis

Abstract

The creation of cellular tubes is one of the most vital developmental processes, resulting in the formation of most organ types. Cells have co-opted a number of different mechanisms for tube morphogenesis that vary among tissues and organisms; however, generation and maintenance of cell polarity is fundamental for successful lumenogenesis. Polarized membrane transport has emerged as a key driver not only for establishing individual epithelial cell polarity, but also for coordination of epithelial polarization during apical lumen formation and tissue morphogenesis. In recent years, much work has been dedicated to identifying membrane trafficking regulators required for lumenogenesis. In this review we will summarize the findings from the past couple of decades in defining the molecular machinery governing lumenogenesis both in 3D tissue culture models and during organ development in vivo.

Introduction to Epithelial Cell Polarity

Generation and maintenance of cell polarity is one of the most fundamental properties of multicellular organisms and failure of any aspect of cell polarization can lead to severe consequences at both the cellular and organismal level. Cells can both permanently differentiate into polarized cell types, such as an epithelium or neuron, or produce transiently polarized structures to aid in function. Despite these differences, the core polarity machinery appears to be highly conserved and shares many common themes. There are several types of cell polarity commonly found in epithelial cells. Planar cell polarity involves the coordinated alignment of molecules across the XY plane of a sheet of cells¹ and will be described in detail elsewhere in this issue. Radial cell polarity occurs when molecules are arranged around the cell perimeter as is seen for apical constriction during Drosophila mesoderm invagination². Columnar epithelial cells also exhibit apical basal cell polarity occurring in the Z axis, which will be the focus of this review. Since epithelial cells line surfaces that come in contact with the external environment, their opposing plasma membranes are specialized into two distinct domains: the apical surface must respond to gaseous or aqueous mediums from the external environment, whereas the basolateral domain faces internally and is surrounded by neighboring cells or an extracellular matrix.

The apical and basolateral domains are separated by a group of scaffolding proteins that form a structure called the tight junction (TJ)³. TJs act as a diffusion barrier to prevent mixing of apical and basolateral membrane components, function as an intercellular seal, and can also form paracellular pores. TJs are composed of a multi-protein complex, which can be characterized by three main protein types: transmembrane proteins, peripheral scaffolding proteins, and cytoskeletal linker proteins⁴. The transmembrane proteins include claudins, Tight Junction-Associated Marvel domain-containing proteins (TAMPs, such as occludin), and Junctions Adhesion Molecules (JAMs). Claudins appear to be the core component of TJs and form multimers arranged in anti-parallel double rows to create paracellular pores⁵. The central peripheral scaffolding protein, zonula occludens (ZO-1, ZO-2, ZO-3) can bind a combination of transmembrane proteins, cytoskeletal linkers such as Cingulin, and actin, thus providing a structural and signaling platform^3,4.

Just basal to the TJs is another junctional complex called the adherens junction⁶. Cadherins are integral membrane proteins that serve as the functional unit of adherens junctions and form homophilic interactions in the intercellular space. Cadherins are linked to the actin cytoskeleton through catenins, which allows for transmission of forces across sheets of cells. Even more basal in the cell are other junctional complexes such as desmosomes and gap junctions, although these appear to be less important in apical-basal polarity and will not be described further.

Because polarized cells are physically separated into distinct domains, there is a need for selective delivery of molecules to either the apical or basolateral domain. There are two delivery pathways utilized by cells undergoing polarization (Fig. 1): recycling from the plasma membrane followed by sorting and delivery to a polarized surface, and trafficking of newly synthesized proteins from the Trans-Golgi Network (TGN) to a polarized surface⁶. In both cases, this is accomplished by polarized membrane trafficking via domain specific organelles regulated by Rab GTPases⁷. Rab proteins are a family of small monomeric GTPases that serve as master regulators of membrane transport. They cycle between an active GTP-bound, membrane-associated form and an inactive GDP-bound, cytosolic form, regulated by proteins called Guanine-nucleotide Exchange Factors (GEFs) and GTPase Activating Proteins (GAPs), respectively⁸. Typically, when Rabs are GTP-bound, they recruit proteins called effectors which together target a vesicle to a specific location within the cell^8,9. Although there are close to 70 different Rab proteins in mammalian cells, only a dozen or so have been implicated in regulating apicobasal polarity¹⁰. Interestingly, while there is some overlap between Rabs that regulate apical basal polarity in epithelial monolayers versus 3D epithelial structures, a subset of Rab proteins appear to be specific to generating polarity in one system or the other¹⁰. In addition, it is becoming increasingly clear that membrane trafficking events may occur through “Rab cascades”, whereby one Rab protein will recruit an effector, which is a GEF activating a second Rab protein, which in turn recruits another effector, and so on^11,12. Thus, cell polarization through Rab-driven membrane trafficking is a highly coordinated event, controlled by the spatiotemporal regulation of multiple proteins.

Figure 1. Cartoon of polarized epithelial cell.

Polarized epithelial cells in vertebrates show unique features. At the apical side, distinguished by a primary cilium, the tight junctions and more basal adherens junctions connect multiple cells through a sheet. The nucleus is located basally within the cell, with the Golgi just apical to the nucleus. In addition, polarized microtubule networks run along the apical-basal axis with the plus ends oriented apically. These microtubule networks serve as trafficking routes for polarized vesicle transport. To establish apical basal polarity, cells can utilize two different trafficking pathways: 1) a direct biosynthetic pathway (dashed arrows) where newly synthesized proteins from the trans-Golgi Network are delivered to a polarized cellular domain; or 2) an indirect recycling pathway (solid arrows) whereby cargo is internalized from cell membranes then travels through an intermediary recycling endosome compartment before maturing into an apical recycling endosome for delivery to the apical surface.

Polarized epithelial cells show unique characteristics in terms of membrane composition, cytoskeletal structure, and organelle arrangement which all lend themselves to the specialized role epithelial cells perform. First, there are three highly conserved polarity complexes that are key to setting up proper cellular asymmetry. The first complex is composed of Crumbs/Pals1/PatJ and is the most apically-localized¹³. The second is the Par complex containing Par3/Par6/aPKC/Cdc42 and localizes to TJs¹³. The third complex localizes basolaterally and includes Scribble/Dlg/Lgl¹³. These three polarity complexes are essential for cell polarization and as will be described later, disruption of any of these proteins complexes results in severe cellular consequences.

The second characteristic of polarized epithelia, is the arrangement of organelles within the cell (Fig. 1). In columnar cells, the nucleus sits towards the base of the cell near the basolateral membrane, with the Golgi localized just apical to the nucleus. In addition, the microtubule network is distinctively arranged such that microtubules can nucleate from the basolateral side and run towards the apical cell surface providing a road for vesicles to traffic along⁶. There has been a lot of work identifying the mechanisms driving cell polarity in 2D monolayers, and several excellent reviews have recently been published on this topic^6,14. Thus, for the remainder, we will focus on establishment of cell polarity in 3D, also called lumenogenesis.

The ability of an organism to form a lumen, or cavity of hollow space, is one of the most vital developmental processes, resulting in the formation of most organ types. Cells have co-opted a number of different mechanisms for tube formation that vary between different tissues and organisms. However, despite this diversity, the fundamental concept of cell polarity is the cornerstone to successful lumenogenesis. Polarized membrane transport has emerged as a cellular pathway that is central not only for establishing individual epithelial cell polarity, but also for coordination of epithelial polarization during apical lumen formation and tissue morphogenesis. Thus, in recent years much effort has been concentrated in identifying the membrane trafficking regulators required for lumenogenesis. Consequently, this has been a very rapidly changing and sometimes controversial field of study that has already identified numerous mechanisms that mediate apical lumen initiation, formation, and expansion. In this review we will summarize the latest progress in defining the molecular machinery governing lumenogenesis in vitro and in vivo. While we will briefly discuss common types of lumen formation utilized by all cell types, the remainder of the review will focus on lumenogenesis in epithelial cells and will not consider models from other cell types such as neural tube development or blood vessel formation.

Types of Lumen Formation

Studies from multiple laboratories over the last couple of decades have identified many distinct methods of apical lumen formation. This incredible diversity of mechanisms mediating cell polarization is not unexpected, since epithelial cells are one of the oldest ancestral cell types identified in the first multicellular organisms. All types of lumenogensis can be generally grouped into two major categories: the formation of lumens using already established cellular sheets (budding, wrapping, entrapment) and de novo lumen formation where cells create a hollow space where there was no preexisting cavity (hollowing, cavitation) (Fig. 2). Lumen formation is often associated with the differentiation of cells from a mesenchymal or endodermal lineage into epithelial as they undergo polarization. It is important to note that these mechanisms of lumenogenesis are not mutually exclusive and it is common that an organ type will use several mechanisms to fine tune different aspects of the system. In addition, it is becoming increasingly clear that mechanisms can compensate for one another to ensure the presence of a fail-safe process during epithelial tissue morphogenesis¹⁵.

Figure 2. Schematic of common types of lumenogenesis.

Cavitation (top row) results from apoptosis of inner cells (marked with a red X) to clear out a luminal space. Cord hollowing (middle row) occurs through targeted delivery of apical endosomes to the AMIS (Apical Membrane Initiation Site) and subsequent cell divisions to expand the luminal space. Budding, (bottom row) begins when a subset of cells in a sheet bud off and exit the sheet to create a tube.

Budding

Budding is a very common lumenogenesis mechanism that occurs during branching of pre-existing epithelial tubes or sheets. Typically, a localized subset of polarized epithelial cells begins to sprout off and exit the tube or sheet to create a new branch (Fig. 2). Examples of this type of tube formation include the mammalian lung and kidney, as well as the Drosophila salivary glands, trachea, hindgut, and dorsal appendages^16,17.

Wrapping

A second type of tube formation is called wrapping and is usually used to transform a flat epithelial sheet into an epithelial tube. In this case, a group of cells spanning the entire length of an epithelial sheet undergo apical constriction driving an invagination that deepens until a new tube of cells is molded and the cellular sheet is restored. The most prominent example of wrapping is neural tube formation in mammals^16,17.

Entrapment

Perhaps the least characterized mechanism of lumen formation is entrapment. This method utilizes a combination of cell migration and repulsive forces to essentially capture a lumen between cells and has been described for development of the Drosophila heart and mouse aorta^16,18.

Hollowing

Unlike budding and wrapping, the remaining lumenogenesis mechanisms described below do not use pre-existing epithelial sheets/tubes but rather coincide with differentiation into epithelial cells, thus often referred to as de novo lumen formation. Perhaps the most well studied type of de novo lumenogenesis is via a hollowing mechanism (Fig. 2), which can function at the level of a single cell to create a seamless tube or at a multicellular level creating a tube that spans multiple cellular junctions. These processes are termed “cell hollowing” and “cord hollowing”, respectively, and they both utilize trafficking machinery to create a nascent apical lumen. Examples of cell hollowing include the C. elegans excretory canal and Drosophila trachea terminal cells, while cord hollowing is used by the vertebrate vasculature, zebrafish notochord, and MDCK tubules grown in 3D culture^16,17.

Cavitation

Lumens may also form through cavitation. In this de novo lumenogenesis mechanism, cells begin as a solid cord but hollow out the inside through either apoptosis and/or autophagy thereby creating a luminal space (Fig. 2) such as in the mammalian mammary ducts or MCF10A cells grown in 3D cultures^16,19.

Lumen Formation In Vitro

The majority of research into the molecular machinery that governs de novo lumen formation originated from work using 3D tissue culture. Madin-Darby canine kidney (MDCK) cells are an epithelial line that establishes apical basal polarity in a dish, and when a single cell is seeded in a 3D extracellular matrix it will instinctively form a lumen. These epithelial spheres of cells surrounding a central lumen are often termed cysts. While much work has been done on MDCK cell monolayers, the field is increasingly finding that there is only moderate overlap in polarity cues between 2D and 3D cultures. Consequently, utilizing MDCK cysts to study apical basal polarity in the context of tubulogenesis provides a more accurate picture. While these 3D MDCK tissue culture models pose some limitations as will be described later, they have greatly expanded our knowledge of the molecular mechanisms leading to de novo lumen formation. This section will describe recent work in this area. While the focus of this review is polarized membrane transport, it is almost impossible to examine membrane trafficking events without also considering the cytoskeleton. The two processes are intertwined on many levels from providing tracks along which vesicles travel throughout the cell to deforming membrane to allow segregation and maturation of endocytic compartments. Thus, while this review will focus on the mechanisms of membrane trafficking during lumenogenesis, we will include details about the role of the cytoskeleton in these events to complete the story.

Cell division-linked polarization during apical lumen formation

During lumen formation, cell division events are often linked to the first signs of apicobasal polarity establishment. Until recently, it was unknown how cell division provided a polarizing cue, but studies of the midbody have provided some insights into this process. Currently, the midbody-dependent cord hollowing model^20,21 is perhaps the best accepted model in the field. It begins with cell division-mediated formation of a midbody that marks the site for the future lumen. Thus, the mitotic midbody may function as a symmetry-breaking polarity cue that allows targeted endocytic vesicle delivery of apical membrane and cargoes effectively establishing the apical membrane initiation site (AMIS). Upon further trafficking and membrane rearrangements, the AMIS matures into a pre-apical patch (PAP) which is distinguished from the AMIS by spatial separation of tight junction and apical plasma membrane markers^22,23. While at the PAP stage the two cells now have distinct apical membranes, the luminal space is still not resolvable by confocal microscopy²³. Then once a small lumen is established through a combination of polarized vesicle exocytosis, hydrostatic pressure, and contractile forces, subsequent oriented cell divisions allow for the expansion of the apical luminal space (Fig. 3). Interestingly, this mechanism of cell-division linked lumenogenesis appears to be conserved in other specialized epithelial cell types, such as hepatocytes²⁴ and human intestinal cells²⁵.

Figure 3. Midbody-dependent cord hollowing model of lumen formation.

A non-polarized cell divides and forms a midbody (blue), which functions as a spatial cue for creation of an apical domain (Step 1). Apical endosomes (red) are then transported to the midbody where they deliver apical cargo, establishing the apical membrane initiation site, or AMIS (Step 2). The exocytosis of membrane and apical proteins results in formation of a nascent lumen (Step 3). Subsequent cell divisions expand the luminal space and increase the size of the cyst (Step 4). These cell divisions are oriented such that the mitotic spindle is perpendicular to the apical lumen (top left cell) and cleavage furrow ingression begins at the basolateral membrane and proceeds apically towards the midbody (top right cell).

The role of the extracellular matrix and polarity complexes during lumenogenesis

In addition to the midbody, the extracellular matrix also emerged as providing functional cues for cell polarity. There are two types of extracellular matrices MDCK cells are grow in to form lumens. A collagen I matrix, often with the addition of extracellular matrix components, was used early on, but more recent experiments use Matrigel, which is a mixture of laminin-1 and collagen. It was demonstrated that a collagen I matrix controls the location of basal plasma membrane by signaling through integrins and the Rac1 small monomeric GTPase²⁶. When grown in collagen I, MDCK cells expressing dominant-negative Rac1 mutants show a polarity inversion phenotype, such that the apical membrane faces the extracellular matrix and a lumen does not form inside the sphere of cells²⁶. Surprisingly, this phenotype is rescued if cells are embedded in a Matrigel matrix, presumably due to the addition of laminin, suggesting that Rac1 controls extracellular laminin secretion and assembly in the basement membrane. Extracellular laminin can then signal in an integrin-dependent manner to mediate the formation of the basal plasma membrane²⁶. Rac1 also activates PI3-kinase to regulate lipid composition at distinct membranes²⁷. As will be described further, it appears that the integration of multiple pathways is necessary for polarity establishment during lumen formation.

The polarity complex proteins were also early players studied in 3D cultures. A number of these proteins including Par3²⁸, PALS1²⁹, PATJ³⁰, Lgl³¹, and aPKC²⁷ are all necessary for single apical lumen formation, although the mechanisms remain to be fully understood. The first characterization of molecular machinery mediating the role of polarity complexes during lumenogenesis came from Martin-Belmonte and colleagues³². They found that the phosphatase PTEN is required for PtdIns(4,5)P₂ enrichment specifically at the apical membrane. PtdIns(4,5)P₂ then recruits Annexin2 which can interact with active Cdc42 to bind and localize the Par6/aPKC complex to the apical membrane. This apical localization of Par6/aPKC is independent of Par3, since Par3 mediates the formation of the canonical Par3/6 complex at the TJs^32,33. Further studies revealed that the role of Par3 appears to be more complex, possibly due to its dynamic interaction with aPKC. Indeed, recent reports have shown that Par3 is required for establishment of apical membrane identity during the earliest steps of lumen formation and only later becomes restricted to the TJs^22,34. Perhaps this discord in results arises from differences in the timing of imaging Par3 localization during cyst formation, or the methods used to visualize Par3. Moreover, early Par3 recruitment may be transient but significant to establish an initial apical signal, and necessary again later for TJs formation.

Polarized endocytic transport and lumen formation

Despite this early work on polarity proteins, how initial establishment of an apical domain (also referred to as AMIS) eventually leads to apical lumen formation and expansion remained unclear. Work from Ferrari and colleagues elegantly contributed to this answer by showing that a combination of forces act in parallel to drive opening of a lumen²³. The first force is driven by ROCK-mediated contractility, whereby Myosin II activity is down-regulated to allow cellular changes to occur. The second force, hydrostatic pressure, is dependent on establishment of TJs to create a diffusion barrier and subsequent targeted insertion of ion and water channels into the apical membrane. Liquid then flows out of the cell and into the luminal space to drive opening. Interestingly, as the lumen inflates, there is a marked decrease in cell volume, suggesting that the system is isochoric, where volume is conserved between the cells and the luminal space.

The identification of apically-localized ion and water channels as key drivers of lumen formation and expansion raised the question of how these transporters and channels are targeted specifically to the apical membrane during lumenogenesis. This requirement of targeted transport in establishment of polarized membrane domains led to a number of studies that began to elucidate the apical trafficking machinery. The central players in this process are the Rab family of small monomeric GTPases, which interact with specific effectors to coordinate the timing and targeted delivery of proteins to apical plasma membrane. A Rab11a-dependent pathway was first implicated in apical membrane formation through delivery of Crumbs3 via the microtubule network during the first symmetry-breaking cell division³³. Crumbs3 in turn recruits aPKC to reinforce apical membrane identity³³. Work from another group suggested that Podocalyxin (an apically localized glycoprotein also known as Gp135) recycling and delivery to the apical surface is mediated by transport through Rab11a-containing endosomes²². Additionally, it was shown that Rab11 initiates a Rab cascade by recruiting Rabin8, a known GEF for Rab8, thereby activating Rab8 and creating a specialized Rab11/Rab8-containing apical endosomes²². These apical endosomes then dock and fuse at the apical membrane through tethering with the Exocyst complex²². This work also placed Annexin2 and Cdc42 proteins on Rab8a/Rab11a vesicles mediating the interaction with aPKC²² (Fig. 4).

Figure 4. Molecular trafficking pathways of apical lumen formation.

During late telophase, Rab11/FIP5 endosomes (red) and their binding partners (proteins listed in red) are transported to the AMIS (blue) where they are tethered by several complexes, including ZO-1/Cingulin, the Exocyst, and Rab35, mediating vesicle fusion with the AMIS. In a second pathway, Rab27/Slp4a endosomes (green) and their binding partners (listed in green) are also targeted to the AMIS through tethering with Slp2a and Syntaxin-3. It remains unclear how these two pathways are connected, and whether Rab8 has a unique or overlapping function. Dotted lines denote vesicle tethering targets. Proteins localized at the AMIS are in blue.

In addition to Rabin8, a number of other Rab11 effectors have since been implicated in lumenogenesis (Fig. 4). The Rab11 effector FIP5 sequentially interacts with sorting nexin 18 (SNX18), Kinesin-2, and Cingulin to mediate apical vesicle formation, transport, and targeting to the AMIS^35,36. This process is also regulated by Myosin-5b, presumably through binding to Rab11 and possibly Rab8^37–39. A feed-forward mechanism for reinforcement of apical identity and TJ formation is mediated by Cingulin²¹. Cingulin directly binds to the central spindle microtubules, by interacting with glutamylated tubulin C-terminal tails and also is recruited to the AMIS through interaction with ZO-1. Cingulin then activates Arp2/3-dependent branched actin polymerization, which in turn recruits more ZO-1 (and consequently Cingulin) as well as mediates Rab11-FIP5 vesicle tethering²¹. FIP2, another FIP family member, is also necessary for single lumen formation. Phosphorylation of FIP2 and interaction with the clathrin-adaptor protein Eps15 are important in establishing cellular junctions during lumen formation^40,41, although the precise mechanism remains unclear. It is thought that the different FIP family members (there are 5 in total) give Rab11 vesicles different identities and function in different cellular processes. Consistent with this, apical Rab11-FIP2-based transport is independent of both Rab11a and Myosin-5b, suggesting it functions in a distinct pathway from Rab11-FIP5 apical recycling endosomes⁴⁰.

With all the molecular players involved in lumenogenesis, the temporal regulation of apical vesicle transport must be tightly controlled. For example, during cell division, apical cargo transport needs to be delayed until late telophase to allow formation of the midbody-associated AMIS. In part, this delay in apical transport is regulated by FIP5 phosphorylation³⁶. During metaphase and anaphase FIP5 is phosphorylated, which retains Rab11/FIP5 endosomes at the centrosomes by inhibiting FIP5 binding to SNX18. In late telophase, FIP5 is de-phosphorylated allowing it to interact with SNX18 and promote budding and transport of apical endosomes to the newly formed AMIS³⁶. Thus, it appears that at least in part, post-translational modifications coincident with stages of the cell cycle may regulate the timing of apical membrane delivery.

In the last few years several other endocytic transport pathways regulating lumen formation have been identified. Microarray-based expression analyses revealed several transcripts that were upregulated specifically in 3D MDCK cultures when compared to cells grown in 2D monolayers⁴². In focusing on those genes that were also downregulated in epithelial cancers, Gálvez-Santisteban and colleagues identified the synapotagmin-like protein Slp2-a, which is a known regulator of Rab27-trafficking events. They placed Slp2-a on Rab27a/b-containing transport vesicles, targeting them for apical delivery through binding membrane enriched in PtdIns(4,5)P₂. This study also identified the closely related family member Slp4-a as a Rab27a/b, Rab8a/b, and Rab3b effector to specify interaction with the apical SNARE syntaxin-3⁴² (Fig. 4). It is interesting to note, that while Rab8 is involved in both the apical Rab11 pathway and this Rab27-centric pathway, it is unclear how the two pathways are related. Nonetheless, the combination of three Rab proteins on the same vesicle may serve as an example of a Rab code to further impart specificity for the target membrane.

To achieve precise delivery of vesicles to the apical cortex, there are a number of tethering factors that reside at the AMIS such as Cingulin, the Exocyst complex, and Syntaxin-3^21,22,42. Recent work has shown that, in addition to cargo trafficking, Rab proteins may also serve as apical tethers as may be the case for Rab35^10,43 (Fig. 4). While two bodies of work looking at Rab35 during lumenogenesis agree that functional disruption results an unusual phenotype where apical basal polarity is reversed causing inverted cysts^10,43, they disagree as to the mechanism by which this occurs. Klinkert and colleagues argue that Rab35 sits at the AMIS and directly binds the cytoplasmic tail of Podocalyxin on vesicles that also contain aPKC, Cdc42, and Crumbs3 (presumably Rab11-apical recycling endosomes)⁴³. In contrast, Mrozowska and Fukuda argue that polarity inversion is instead caused by improper β1-integrin recycling resulting from the inability of Rab35 to bind its effector ACAP2 which in turn acts as a GAP for Arf6¹⁰. Thus, further work in this area will be necessary to nail down the precise mechanism of Rab35. Finally, to add to an ever expanding repertoire of Rabs that regulate apical membrane biogenesis, recent studies suggest that Rab14 may act upstream of trafficking events to regulate membrane lipid composition at the apical surface by modulating Cdc42 activation and/or midbody positioning during cell division⁴⁴. Thus, while Rabs are canonically thought to serve as master regulators of membrane trafficking events, it appears that during lumen formation they not only label vesicles containing apical cargo but may also moonlight in other non-traditional roles.

Several other important players in apical lumen formation have recently come to light. The chloride intracellular channel (CLIC) family of proteins have recently been unearthed as potential novel regulators of membrane trafficking by defining distinct retromer-dependent vesicular compartments⁴⁵. CLIC4 is a cytoplasmic protein and has been implicated in lumenogenesis both in 3D MDCK cultures and the developing mouse kidney⁴⁵. Its function appears to be highly conserved as mutations in the C. elegans homolog of CLIC4 results in a cystic excretory canal (equivalent to the mammalian urinary system)⁴⁶. CLIC4 regulates retromer-mediated apical delivery of Rab11a and PTEN and indirectly controls branched actin polymerization necessary for proper endocytic sorting and recycling⁴⁵. Another player, EFA6 performs a dual role in lumenogenesis. During the early 2–4 cell stages, EFA6 likely acts as a GEF to activate Arf6, which is known to function in apical trafficking⁴⁷ and tubulogenesis⁴⁸. Then, once a nascent lumen has formed, EFA6 interacts with alpha-actinin 1 which modulates actomyosin contractility to help lumen coalescence and enlargement⁴⁹. Finally, while Podocalyxin is a commonly used apical membrane marker, it may play a more active role in lumenogenesis. As part of the CD34 family of transmembrane sialomucin proteins, its composition makes it both negatively charged and slippery due to glycosylation. Thus, Podocalyxin may help opposing apical membranes open by electrostatic repulsion and anti-adhesion^50,51. In summary, the trafficking events mediating apical membrane establishment and lumen formation require an intricate and coordinated network of trafficking machinery, polarity complexes, and membrane proteins.

Manifestation of multilumenal phenotypes

The complexity of lumen formation and requirement for coordinated timing and localization of many molecules provides ample opportunity for various luminal phenotypes. One phenotype mentioned previously that occasionally arises is a polarity inversion, such that the apical proteins become localized to the basolateral side and vice versa, effectively turning the entire extracellular matrix into a lumen. This phenotype is seen when Rac1 or Rab35 are functionally disrupted^10,26,43. There are several potential mechanisms through which polarity inversion may occur. First, there may be defects during endosomal recycling, such that apical proteins are either never internalized from basolateral membranes or they get internalized but stuck in endosomal compartments and are never delivered apically. Alternatively, cues from the extracellular matrix may play a role in defining membrane identity and disrupting these signaling pathways may cause the cell to misinterpret apical and basolateral identity.

The most common phenotype observed in 3D tissue culture is a multilumenal phenotype which can arise from various mishaps throughout the steps of lumenogenesis. The first opportunity for problems to occur is during initial establishment of the AMIS. Functional disruption of components of the apical trafficking or tethering machinery results in the inability of a cell to properly establish and localize an apical domain initially and after each cell division^{10,21,22,42,52,53}. A second occasion where mistakes can occur is during subsequent cell divisions that expand and maintain a lumen. The cells undergo oriented divisions such that the cleavage furrow forms perpendicular to the luminal space (Fig. 3). Cleavage furrow ingression is asymmetric, whereby the furrow begins at the side furthest from the lumen and proceeds towards the lumen²¹ (Fig. 3). Thus, it is conceivable that if the cells divide in the wrong plane, either due to misalignment of the microtubule spindle or cleavage furrow ingression direction, erroneous lumens may arise from each cell division. There are a number of proteins known to regulate spindle orientation during lumen expansion, including Cdc42 and its GEFs and GAPs^54–57, polarity complex proteins^55,58, phosphatidylinositol phosphatases⁵⁹, microtubule binding proteins⁶⁰, and other small GTPases⁶¹. Consequently, while apical membrane biogenesis is crucial for lumen formation, proper orientation of later cytokinesis events is also essential for expansion and maintenance of a single lumen.

Summary and future directions

The number of molecular players involved in lumenogenesis is astonishing and raises the question of the biological relevance of all these different pathways. One explanation is that the need for multiple targeting regulators ensures a multifaceted approach to single lumen formation, a process that is absolutely imperative to proper development. Perhaps each Rab pathway is carrying a different piece of the puzzle and a cell requires all pieces to complete the puzzle/establish an apical domain. While colocalization experiments have attempted to answer this, the ultimate experiment would be to immunoprecipitate intact vesicles labeled by different Rab proteins during lumen formation and perform proteomics on their contents. Not surprisingly, this is technically difficult and time consuming, so the development of novel techniques to determine vesicle cargo will greatly advance our understanding. An alternative, but not necessarily mutually exclusive model is that various apical targeting and lumen formation mechanisms are employed by different epithelial tissues to accommodate variances in the regulation and spatiotemporal properties of these tissues. Finally, since the formation of epithelial ducts and sheets are at the core of tissue function, multiple apical targeting machineries may provide redundant mechanisms to ensure proper organ morphogenesis. Unfortunately, the last two concepts cannot be easily tested using in vitro 3D tissue culture systems but will require coordinated use of both in vitro and in vivo experimental models.

Lumen Formation In Vivo

Three-dimensional tissue culture systems are instrumental for defining the pathways that mediate lumenogenesis, as well as identifying the molecular machinery that regulates the timing and location of lumen formation. However, developing organs typically do not originate from a single cell suspended in an extracellular matrix. For example, the mouse pancreas and kidneys and zebrafish intestine begin as a clump of epithelial cells that initiate isolated mini-lumens between cells. As development progresses, these mini-lumens fuse and cellular remodeling helps create a single continuous lumen. In addition, channels and pores may open allowing ion and fluid flow to generate pressure in the system and further drive lumen expansion. Studying this lumen coalescence and maturation poses an outstanding question that 3D tissue culture models are not presently equipped to answer. Additionally, many of the multilumenal phenotypes seen from in vitro cultures are not recapitulated when the same proteins are disrupted in vivo, especially in vertebrate models. This suggests that organisms may have compensatory pathways at work, which is not surprising considering how crucial formation of epithelial tubes and sheets are in many aspects of development. Finally, MDCK 3D cultures recapitulate the cord hollowing method of lumen formation, yet organisms often use a combination of methods for lumenogenesis. Thus, moving into model organisms to complement 3D culture methods is essential for a complete understanding. In this section, we will summarize the work on lumenogenesis during development from several model organisms, highlighting at least one example of each mechanism of tube formation described previously.

C. elegans excretory system

Perhaps the simplest example of tubulogenesis in vivo comes from work in the C. elegans excretory system, which is somewhat equivalent to the mammalian urinary system. In C. elegans, this system is made up of three cells⁶², and each use a different mechanism to form a unicellular tube⁶³. The pore tube forms by wrapping such that the single pore cell will form an autocellular junction with a luminal space. The duct tube forms similarly in that it utilizes the wrapping mechanism, but it then proceeds to dissolve the cell seam through an autofusion event. In contrast, the canal cell is an example of cell hollowing. It is composed entirely of a single cell that sends out hollow tubules to create an H-shape extending the length of the worm⁶². Despite the differences in tube formation, all three cells exhibit characteristic epithelial traits including apical Crumbs⁶⁴ and PAR complex proteins^65,66 at the luminal side. In addition, the tubes are connected through cellular junctions composed of both the Cadherin and Dlg complexes⁶³. In the canal cell, lumen formation begins when specialized vesicles at the apical surface tether and fuse, aided by the PAR proteins and Exocyst complex⁶⁶. The ezrin/radixin/moesin ortholog ERM-1 regulates actin organization at the apical surface and promotes the insertion of the aquaporin AQP-8 into the membrane to expand the lumen through fluid pressure⁶⁷. Additionally, loss of the Cdc42 GEF FDG1 results in cystic excretory canals due to disruption of apical cytoskeleton organization⁶⁸. Once the tubes have formed, a number of trafficking proteins have been implicated in tube extension and maintenance, but much detail is lacking about the content and molecular makeup of these vesicles⁶³. For a more detailed analysis of development of the C. elegans excretory system, an excellent review was recently written by Sundaram and Buechner⁶³.

Drosophila tracheal system

The trafficking events mediating Drosophila tracheal system formation may be the best characterized in the field. This organ system is the critical provider of oxygen to every cell in the fly, and as such is highly branched and spans the entire organism. There are four types of tubes that make up the system, and like C. elegans utilize multiple mechanisms to form lumens. While the primary branches of the trachea system are made up of multicellular tubes that form through budding, the precise mechanism is unknown¹⁷. Thus, we will focus mainly on lumenogenesis in the terminal cells and fusion cells. The terminal cells are located at the ends of the tracheal branches and come in contact with individual tissues. It remains controversial whether they use an exocytic vesicle centered-cell hollowing method¹⁷ or an alternative process where apical membrane extends intracellularly and unidirectionally from the site of adhesion to the neighboring cell⁶⁹. Regardless, several trafficking pathways have been implicated in tracheal terminal cell tube formation. The Exocyst is required for vesicle tethering and fusion during lumen formation, and it is directed to the proper cellular location by the PAR polarity complex^70,71. Rab35 and its GAP Whacked (mammalian homolog TBC1D10) localize to the apical membrane and regulate where and to what degree the terminal cell membrane grows⁷². Whether Rab35 in this sense has overlapping functions with 3D tissue culture models remains to be seen. In addition, disruption of both the Rab11 recycling endosome-derived trafficking pathway and the Rab10 Golgi-derived pathway are necessary for severe defects in terminal cell outgrowth, suggesting that unlike tissue culture models, the two vesicle trafficking pathways are redundant⁷¹.

Fusion cells are located at the end of migrating trachea branches. Through filapodial extensions, they make a connection with another fusion cell and then form a continuous lumen⁷³. Interestingly, when the fusion cells make the initial contact with another cell, they become bipolar with two apical domains at opposing sides of the cell and a cytoskeletal network running between⁷⁴. These apical domains are labeled with many of the same proteins found in tissue culture, including aPKC, Crumbs, PATJ, subapical Rab11 vesicles, and Exocyst components^75–78. How these two apical membranes come to fuse into a single lumen was controversial until recently. Both a direct plasma membrane invagination/fusion model had been proposed, along with an indirect mechanism aided by an intermediary trafficking compartment⁷⁴. Recent work by Caviglia and colleagues found evidence for the latter, suggesting that the intermediary trafficking compartment is a specialized secretory-lysosome organelle⁷⁹. These secretory lysosome compartments are downstream of the activity of a fusion cell-specific GTPase Arf-like 3 (Arl3) and contain Rab39, Rab7, Syntaxin 1A, and the C2-domain protein Munc13–4. Synaptobrevin (also known as VAMP3) was found on the membrane of the cells, and presumably is the SNARE complement to Syntaxin 1A aiding in secretory-lysosome fusion. Intriguingly, this membrane fusion event is calcium-dependent⁷⁹. After a continuous lumen has formed, apical secretion results in a chitin-based luminal matrix that helps modulate tube size and maintain structural integrity of the luminal space^80,81. In addition, other housekeeping Rab trafficking pathways are required to modulate lumen size and maintenance. It is interesting to note that unlike MDCK cell culture, in the Drosophila trachea, lumen expansion occurs in the absence of cell division events¹⁷.

Drosophila heart

Drosophila heart formation is one in vivo model of entrapment and the genes regulating heart development are highly conserved throughout mammals⁸². In terms of trafficking events, there is not a lot known in this system; however, these cells use a unique mechanism to form the heart. Prior to lumen formation, there are two rows of cardiac cells located on opposite sides of the embryo running along the head to tail axis⁸³. These cells undergo a mesenchymal to epithelial transition and migrate towards each other meeting at the midline⁸². Then, the two cells opposite each other make contact first at the dorsal leading edge and then at the ventral edge, forming a lumen in between⁸⁴. When each cardioblast in the row performs this dual contact with its partner in synchrony, a tube is formed along the head to tail axis of the embryo, often termed the dorsal vessel. Accordingly, E-Cadherin is expressed at the dorsal and ventral cell contact surfaces and removed from the apical luminal surface, mediated by Slit and Robo^85–87. Concurrently the cytoskeleton must be rearranged to accommodate the newly formed lumen. In this system, Cdc42 appears to be a master regulator of both processes, as it mediates both formin-based actin polymerization at the apical surface, Arp2/3-based actin polymerization for E-Cadherin endocytosis, and luminal localization of Slit⁸⁴. In the future, it will be intriguing to see what role vesicular trafficking events play in models of entrapment.

Mouse mammary gland

The mouse mammary gland is an example of lumen formation by cavitation, whereby cells begin in a solid rod then the internal core cells undergo apoptosis to hollow out the interior of the rod. While MCF-10A cells cultured in an extracellular matrix form cysts similar to MDCK cells, the mechanism is quite different. In culture, the cells in the center of the sphere undergo apoptosis from lack of extracellular matrix attachment. Apoptosis in this sense is often called anoikis¹⁹. A similar mechanism seems to occur in vivo during mouse mammary development¹⁹. Conceivably, with apoptosis as the mechanism of lumen formation, the role of membrane trafficking may be less significant. However, several papers have implicated some of the usual players in this process. The polarity protein Par3 is necessary for normal morphogenesis of branched ducts in vivo, and Par3 knockdown mammary glands showed increased levels of apoptosis and proliferation⁸⁸. Par3 interaction with aPKC helps restrict Par3 localization to TJs and promotes apical aPKC localization⁸⁸, again suggesting a complicated relationship between Par complex proteins. A second piece of work implicates Huntington (HTT) as a molecular scaffold that links Par3-aPKC localization with Rab11a vesicles through kinesin-dependent microtubule transport⁸⁹. While loss of Huntingtin (HTT) specifically in mammary luminal cells results in disrupted ductal morphogenesis, the experiments showing it acts through the Rab11a pathway were performed in 3D MDCK cell culture, and thus will need to be validated in vivo. Regardless, the relationship between establishment of apical-basal polarity and apoptosis in formation of lumens in mammary cells is unclear, and this work provides a potential link.

Zebrafish intestine

Zebrafish, Danio rerio, are becoming an increasingly useful model organism for studying human disease. Despite some differences in lumen formation between zebrafish and mammalian gut development, the molecular character of enterocytes (gut cells) and injury response pathways are conserved⁹⁰. The zebrafish intestine starts out as a solid cord of endodermal cells at the midline that acquire epidermal fate, begin to polarize, and form many small lumens de novo. The cells are highly proliferative at this point, without significant apoptosis⁹¹. Just after 3 days post fertilization, the small lumens fuse to form a single lumen running from mouth to cloaca⁹¹. Lumen formation occurs in the anterior to posterior direction⁹¹, such that the posterior intestine is the last to fuse into a single lumen.

Like many other models, early work in zebrafish showed that the apical polarity protein aPKC is necessary for single intestinal lumen formation⁹². Because zebrafish must fuse many small lumens that are several cells apart into a single continuous lumen, it provides an interesting model to study lumen coalescence. Work by Bagnat and colleagues showed a unique role for the TJ Claudin protein in this process⁹³. When Claudin-15 function is disrupted, gut cells fail to fuse into a single lumen, but surprisingly, polarity and TJ formation remain unaffected. Based on these observations, it is proposed that Claudin-15 forms paracellular pores through which ions are pumped by the Na⁺/K⁺ ATPase into the lumen to create an electrochemical gradient between the cells and the luminal space. This ion gradient is followed by fluid flow, generating a hydrostatic force to drive lumen expansion. However, this hydrostatic force is not enough to push lumens into coalescence, as aPKC and Rab11a mutants still show multilumenal phenotypes^92,94. Thus, it is reasonable to consider that as lumen coalescence occurs, it might be followed by cellular rearrangements to re-establish junctional complexes and membrane polarity. Just prior to fusion, bridge contacts between cells of adjacent lumens show basolateral markers and Cadherin contacts⁹⁴. During fusion, there are two different mechanisms observed for bridge rearrangement⁹⁴. The most frequent is apical membrane expansion, where Podocalyxin is observed creeping into the bridge membrane from one side and Cadherin contacts are broken, possibly through endocytosis but this has not been directly shown. Rab11a is necessary for recycling bridge proteins, including Cadherin. Interestingly these bridge membranes are mosaics of apical and basolateral proteins, suggesting that apical membrane identity is not fully established. Alternatively, although less frequently, Podocalyxin can be seen creeping into the bridge contact from both sides, indicating bipolar apical character of bridge cells. Bipolarity is also seen in Drosophila tracheal fusion cells⁷⁴. It will be interesting to see if this combination of hydrostatic forces and cellular rearrangements also drives lumen coalescence in other systems.

Studying zebrafish gut development has also uncovered a new player in lumen formation, Plasmolipin⁹⁵. Plasmolipin is a MARVEL domain-containing transmembrane protein, expressed specifically in the posterior portion of the zebrafish intestine where cells are highly endocytic. Plasmolipin regulates endocytosis by interacting with the clathrin adaptor EpsinR to regulate apical SNARE recycling, and is required for restricting Crumbs localization to TJs in MDCK cysts⁹⁵. Thus, this work nicely utilizes a combination of in vivo lumen formation with 3D tissue culture to elucidate both the molecular mechanism of Plasmolipin and its contribution to gut development.

Mouse intestine

Fish and mammalian intestinal systems form lumens slightly differently. In mice, the endoderm that will form the gut initially begins as a flat layer of cells. The anterior and posterior ends fold over creating a small pocket. The anterior pocket then moves towards the posterior end while the posterior pocket moves towards the anterior end until they close the tube, which is aided by the embryo turning at E9.0⁹⁶. Coincident with tube formation, the cells take on epithelial character⁹⁶. Once mice have formed a lumen, the cells then undergo further morphogenesis, expanding the apical surface to create villi. Walton and colleagues recently published an excellent review on this process⁹⁷, so we will not cover it here.

Because the intestinal tract is a highly specialized type of lumen, there are apical membrane protrusions called microvilli that line the luminal surface to enhance surface area and aid in nutrient absorption. While the trafficking pathways that regulate microvilli formation are still being revealed, Rab8 and Rab11 play an important role^98–100. The interplay between cytoskeletal networks, cellular adhesion, and apical trafficking during intestinal microvilli formation is beyond the scope of this article, but is nicely reviewed by Crawley and colleagues¹⁰¹.

Vertebrate kidneys

Vertebrate kidney development is an excellent model for lumen formation because it is comprised of a number of different tubes that form through different mechanisms. For example, the s-shaped body develops through de novo lumenogenesis, whereas the ureteric bud as the name implies forms through a budding event. These two adjacent tubes will eventually fuse to form a continuous lumen and the s-shaped body turns into the epithelial part of the nephron, while the ureteric bud becomes the collecting duct. Several excellent reviews have been written on apical-basal polarity during kidney morphogenesis^102,103.

Mouse pancreas

Bearing resemblance to a head of cauliflower, the mammalian pancreas is a highly branched organ made up of tubes of various diameters. The branching morphogenesis events leading to pancreas development in mice has become a recent topic of investigation and yielded some unique cellular dynamics during lumenogenesis. Prior to lumen formation, a single layer of epithelial cells stratifies into a multilayered epithelium, such that only the top layer of cells shows apical character (E8.75–10.5). Concurrently, these cells bud out creating a central lumen, which will become the main tube of the “pancreatic tree”. The branches begin to form through de-stratification of the epithelium, which involves polarity remodeling such that inner cells not in contact with the lumen acquire apical markers and the luminal top layer of cells also acquire basolateral polarity¹⁰⁴. Interestingly, it appears that a single cell will first acquire apical polarity through vesicle exocytosis, and this cell will then induce its neighbors to polarize through Cdc42-mediated junctional coupling¹⁰⁵. As a result, many mini-lumens form de novo in the newly polarized multilayered epithelium and through cell rearrangements eventually fuse to create a highly branched tubular monolayer of cells by E15.5^104,105. In contrast to Cdc42, which is required for mini-lumen formation in the pancreas, Afadin plays a role in lumen fusion to create a continuous tubular network¹⁰⁶. Afadin appears to be especially important in the cells at the branch tips of the pancreas where de novo lumenogenesis occurs as the tip epithelium grows outward¹⁰⁶. Loss of Afadin disrupts apical exocytosis of Rab8-positive vesicles¹⁰⁶, although the precise mechanism is unclear. Luminal fusion in the central region of the pancreas is dependent on RhoA-driven actomyosin forces, although in the absence of RhoA, Afadin can compensate, again through an unknown mechanism¹⁰⁶. Taken together, these data paint a complicated picture of pancreas development, where cells utilize several types of lumen formation along with cell shape changes, rearrangements, membrane trafficking, and luminal fusion to form an intricate network of highly branched tubules.

Summary and future directions

Despite the differences in lumen formation between cultured cells and model organisms, several underlying themes prevail. First, the polarity complex proteins are master regulators of apical basal polarity throughout all systems, and as such need to be localized to the proper cellular domain. Second, in terms of membrane trafficking, the Rab11/Rab8 recycling endosome pathway appears to be critical for delivery of apical cargo across all types of tissues and tube formation. Additionally, the function of proteins such as the Exocyst complex and SNAREs, seems to be highly conserved. Furthermore, once an initial apical domain and mini-lumen is established, a number of different forces, including hydrostatic pressure, electrostatic repulsion, and cytoskeletal mechanics emerge to expand and maintain the luminal space. Finally, it is becoming clear that most epithelial tissues use several distinct lumen formation methods during organogenesis. While it is still vague which signaling pathways are associated with each type of lumen formation, a common theme emerges: large epithelial tubes predominantly form by wrapping or budding, whereas networks of smaller tubes typically found at branch tips use de novo lumenogenesis. Future work in this field will hopefully reconcile discrepancies between in vitro and in vivo models of lumenogenesis and shed light on the need for an overwhelming number of proteins and mechanisms involved in apical lumen formation.

Synopsis Statement.

The creation of cellular tubes is one of the most vital developmental processes, resulting in the formation of most organ types. The key to successful tubulogenesis is generation and maintenance of apical and basolateral cell polarity. In recent years, much work has been dedicated to identifying membrane trafficking regulators required for polarity establishment during tubulogenesis. In this review we will summarize these findings in both 3D tissue culture models and during organ development in vivo.