Transcytosis in the development and morphogenesis of epithelial tissues

Abstract

Transcytosis is a form of specialized transport through which an extracellular cargo is endocytosed, shuttled across the cytoplasm in membrane‐bound vesicles, and secreted at a different plasma membrane surface. This important process allows membrane‐impermeable macromolecules to pass through a cell and become accessible to adjacent cells and tissue compartments. Transcytosis also promotes redistribution of plasma membrane proteins and lipids to different regions of the cell surface. Here we review transcytosis and highlight in vivo studies showing how developing epithelial cells use it to change shape, to migrate, and to relocalize signaling molecules.

This review describes how epithelial cells use a specialized form of vesicular transport across their cytoplasm to adapt shape, motility, and signaling during cellular and organismal development.

Introduction to transcytosis

Epithelia are the polarized cell layers that line the exterior surfaces of organisms and the internal surfaces of tubular organs. Importantly, epithelial tissues act as selective physiologic barriers, preventing the passage of harmful or infectious material into tissues while permitting the passage of molecules congruent with tissue function (Powell, 1981). Epithelia act as barriers due to the presence of occluding junctions, specialized transmembrane protein complexes that line the junction between adjacent epithelial cells (Jonusaite et al, 2016; Zihni et al, 2016). Occluding junctions block the passage of molecules between neighboring epithelial cells and prevent plasma membrane components from diffusing laterally across the junction from one membrane face to another. Blocking lateral diffusion allows the cell to maintain different cohorts of membrane‐associated proteins and lipids at its apical (external) surface and its lateral and basal (internal) surfaces, a major component of cell polarity. Thus, for a molecule to cross the epithelium, it must either squeeze through the occluding junction or cross through the epithelial cells themselves.

Although ions and small molecules utilize channel or diffusion‐based methods to pass through cells, transcytosis serves as the primary fashion by which macromolecules can be transported across epithelial barriers (Garcia‐Castillo et al, 2017). Some of the most well‐studied examples of transcytosis involve physiologically relevant cargos such as immunoglobulins. For example, the neonatal intestine absorbs and transcytoses maternal IgG (apical to basal) to impart passive immunity (Simister & Rees, 1985; Pyzik et al, 2019). IgA is also transported across various epithelia into mucosal secretions (basal to apical) (Brandtzaeg, 1981; Rojas & Apodaca, 2002). Endothelial cells also establish tight junctions and downregulate bulk transcytosis to establish the blood–brain barrier (BBB), and then rely on receptor‐mediated transcytosis to transport select macromolecules, such as iron and insulin, across the BBB (Ben‐Zvi et al, 2014; Chow & Gu, 2017; Pulgar, 2018; Ayloo & Gu, 2019; O'Brown et al, 2019). Finally, many bacterial, viral, and even fungal pathogens hijack transcytosis pathways to cross epithelial barriers and gain access to inner regions of the body (Garcia‐Castillo et al, 2017; Ruch & Engel, 2017; Cain et al, 2019).

It is generally less appreciated that, in addition to transporting materials across cells, transcytosis also allows redistribution of membranes and proteins between the apical and basolateral surfaces within an epithelial cell. Therefore, transcytosis can have significant effects on cell shape, development, and morphogenesis. In this review, we highlight known and proposed roles of transcytosis during epithelial tissue development.

In a simplified picture of transcytosis (Fig 1), cargo is first internalized via receptor‐dependent, clathrin‐mediated endocytosis (CME) or a receptor‐independent (cargo‐nonspecific) method of endocytosis and incorporated into an early endosome (Garcia‐Castillo et al, 2017; Naslavsky & Caplan, 2018). Endosomes are dynamic, multifunctional organelles in which endocytosed cargo can be sequestered or sorted to other subcellular destinations (Fig 1). From the early endosome, some cargo is recycled back to the plasma membrane, ensuring both receptor and membrane lipid turnover so that endocytosis can continue (Naslavsky & Caplan, 2018). As endosomes mature, their lumens acidify and acquire internal vesicles, at which point they are termed late endosomes or multivesicular bodies (MVBs) (Hanson & Cashikar, 2012). In the final phase of endosome maturation, the late endosome fuses with a lysosome, resulting in the breakdown of its contents into usable macromolecules (Trivedi et al, 2020). Thus, to be successfully transcytosed, an endocytosed cargo must avoid lysosomal degradation and instead be sorted out of endosomes for secretion at the opposing membrane face. How transcytosed cargo is sorted away from cargoes with other intracellular destinations is enigmatic and appears to involve cell‐ and cargo‐specific compartments and mechanisms, highlighting a need for in vivo models that allow study in tissue context (Tuma & Hubbard, 2003; Garcia‐Castillo et al, 2017). Nevertheless, genetic perturbations of known regulators have allowed for the study of specific cargoes in living cells by blocking transcytosis at endocytosis, early, or late endosomal transport steps (see Box 1).

Figure 1. Simplified schematic of transcytosis.

The diagram shows an epithelial cell in cross section, with basal (blue) and apical (red) plasma membranes. Tight junctions between the lateral faces of adjacent cells are indicated with red ellipses. The path of transcytosed cargo (purple) is indicated by black arrows beginning with 1) endocytosis, 2) incorporation into a sorting endosome, 3) sorting to the secretory pathway, and 4) exocytosis. Depending on the cargo and cell type, transcytosis may occur in either an apical‐to‐basal (as shown) or basal‐to‐apical direction, and sorting/recycling may take place in various different endosomal compartments (Fung et al, 2018; Naslavsky & Caplan, 2018). Segmented arrows indicate pathways taken by de novo synthesized directly secreted cargo (DS), apical membrane recycling (R), and cargo bound for transport through the late endosome (or multivesicular body) (MVB) and lysosomal degradation.

Box 1. Genetic tools for the study of transcytosis.

Developmental studies of transcytosis frequently make use of genetic tools to interfere with key GTPases that mediate early or late steps of transcytosis. Dynamin promotes endocytic scission to allow vesicle internalization, and Rab5 promotes early endosome formation (Antonny et al, 2016; Yuan & Song, 2020). Rab4, Rab8, and Rab11 promote exocytosis and recycling of endosome contents back to different regions of the plasma membrane (Seachrist et al, 2000; McCaffrey et al, 2001; Linder et al, 2007; Welz et al, 2014). Temperature‐sensitive (ts) alleles of dynamin or dominant‐negative (dn) forms of Rab5 block many (though not all) forms of endocytosis, while Rab4(dn), Rab8(dn), or Rab11(dn) each block distinct pathways for recycling and exocytosis. Combined with cell‐ and tissue‐specific expression techniques, these genetic tools can be used to assess the need for endocytosis and exocytosis in a variety of developmental processes in vivo. These genetic tools are excellent for investigating the requirement for endocytosis and/or exocytosis in transport of specific cargoes, but ultimately direct visualization of membrane or cargo moving from one side of the cell to another provides the strongest evidence for transcytosis.

Transcytosis as a regulator of epithelial cell shape and tubulogenesis

The shape of epithelial cells is tailored to the shape and function of the organ they comprise. The sizes and shapes of epithelial cells are diverse, defined in part by the areas of their apical, basal, and lateral cell membranes (Fig 2). Maintenance of epithelial cell shape is critical, as dysregulation of basal/apical membrane ratio can cause defects in both tissue function and development (Wodarz et al, 1995; Wilson, 2011; Zang et al, 2015). Epithelial cells in a planar sheet can transition from a cuboidal or columnar shape in which apical and basal area are equivalent (Fig 2A) to a wedge shape (Fig 2B) by constricting or expanding one membrane face. Cells that change shape en masse can form folds or pits in planar epithelia, leading to the development of tubular organs and more complex embryonic structures (Pearl et al, 2017). Most models describe constriction or relaxation of the actin cytoskeleton as the primary contributor to such shape changes (Herrera‐Perez & Kasza, 2018). However, in addition, changes in the relative amounts of apical versus basal trafficking, endocytosis, or transcytosis can also contribute to such shape changes, because vesicle trafficking not only moves cargo, but also redistributes membranes.

Figure 2. The basal versus apical membrane ratio is a major factor in cell shape.

Schematic representations of epithelial cells with equal (A) or unequal (B) basal (blue) or apical (red) membrane area. (A) Cuboidal and columnar epithelial cells differ in height but have similar apical and basal membrane areas. (B) Wedge‐shaped epithelial cells with unequal apical and basal membranes. Cells such as these often line curved tissues. (C) Unicellular tubes are polarized epithelial cells that contain an internal apical lumen. The morphology of these tubes depends on basal/apical membrane area in a manner similar to other epithelial cells (Sundaram & Cohen, 2017). 1) The early Drosophila trachea terminal cell is unicellular tube with apical and basal membranes that elongate at a similar rate. 2) The C. elegans excretory duct cell contains an internal apical membrane that elongates to a greater length than the external basal membrane. 3) The fusion cell of the Drosophila tracheal system is a unicellular tube with a smaller apical membrane than its external, basal surface.

The role of polarized exocytosis and endocytosis in cell and tissue shaping has been extensively reviewed (Jewett & Prekeris, 2018; Xie et al, 2018). Here we highlight several studies that suggest transcytosis can reduce the area of one membrane face and redistribute internalized membrane to other growing faces of the cell.

Transcytosis redistributes endocytosed apical membrane for lateral membrane growth during Drosophila cellularization

During cellularization, the unified membrane surrounding the syncytial Drosophila embryo invaginates between nuclei to separate the syncytium into cellular units. Over the course of an hour, these furrows rapidly elongate relative to the apical surface of each growing cellular unit, eventually enclosing each nucleus to form a simple columnar epithelium (Fig 3A) (Lecuit, 2004). In one of the first studies to tie transcytosis to developmental cell shape change, Pelissier et al reported that when endocytic scission was blocked by temperature‐sensitive (ts) dynamin inactivation, coated pits (presumably incomplete endocytic events) built up at the apical surface of the newly enveloping embryo cells and the subsequent growth of their lateral membrane was inhibited (Pelissier et al, 2003). Expressing dominant‐negative (dn) versions of the GTPases Rab5 (to block early endosome function) or Rab11 (to block apical recycling and trafficking) also inhibited growth of the lateral membrane. Subsequent electron microscopy showed that the apical membrane is initially rich with microvilli that flatten as cellularization progresses. Both endocytosis and mechanical unfolding of these microvilli have been proposed to drive this change in apical membrane topology (Fabrowski et al, 2013; Figard et al, 2013; Figard et al, 2016; He et al, 2016). Other evidence suggests that Rab8 facilitates directed exocytosis to lateral membrane faces, allowing endosomal membrane to be added to these growing membrane faces (Mavor et al, 2016). Together, the data suggest a model where apical microvilli serve as a reservoir of plasma membrane, and that some of this membrane is internalized via dynamin‐dependent endocytosis and then transcytosed in a Rab8‐dependent manner to the growing lateral membrane during cellularization. In this model, the transcytosed membrane supports lateral membrane growth alongside apical membrane flow and de novo membrane export from the ER/Golgi (Fig 3A).

Figure 3. Transcytosis as a regulator of epithelial cell shape.

In each of the models diagrammed, the apical and basal membrane surfaces are designated as red and blue, respectively. The hypothesized path that transcytosed membrane takes through each cell is designated by solid black, numbered arrows. The proposed contribution of directly secreted de novo synthesized membrane (DS) from the ER/Golgi (shaded orange) is indicated by segmented arrows. (A) Pelissier et al, (2003) suggested that endocytosed apical membrane contributes to lateral membrane expansion and the generation of a columnar epithelium during Drosophila cellularization. In this model, (1) clathrin (green) and dynamin internalize apical membrane (red), which is then trafficked through (2) Rab5‐ and Rab8/Rab11‐dependent endosomes (E) to contribute to lateral membrane expansion (Mavor et al, 2016) (3). The contribution of apically endocytosed membrane seems to be incomplete, suggesting that export of directly secreted membrane (DS) from the Golgi also supports lateral membrane growth as well. The apical membrane also fuels lateral membrane ingression through the flattening of its microvilli (red arrows) (Figard et al, 2013; Figard et al, 2016). (B) Soulavie et al (2018) proposed a transcytosis model for apical lumen elongation in the C. elegans excretory duct cell. As the duct cell elongates, the apical lumen grows to a length greater than the cell itself. In this model, internalization of basal membrane (1) is completed by AFF‐1 (teal)‐mediated membrane scission followed by incorporation into RAB‐11‐dependent secretory endosomes (2) and trafficking to the apical surface (3). Potentially, directly secreted membrane (DS) from the Golgi apparatus (or unidentified intracellular sources) may also be trafficked apically via these endosomes. (C) Mathew et al, (2020) provided evidence that synchronous apical and basal membrane growth during early Drosophila terminal cell elongation is achieved by transcytosis. In this model, newly synthesized membrane (DS) and proteins are first secreted at the apical lumen of the terminal cell. Apical plasma membrane (red) along with basally bound transmembrane proteins is endocytosed in a clathrin‐dependent (green) and dynamin‐dependent fashion (1). Internalized membrane is then trafficked through early (2) and late, multivesicular endosomes (3) before being transcytosed to the basal membrane (blue) (4). Concurrent apical recycling (R) from the late endosome balances apical and basal membrane allocation. (D) Basal‐to‐apical transcytosis is necessary for the development of bile canaliculi. In HepG2 cells, apically destined membrane proteins (purple) and membrane are initially secreted to the basolateral surface (gray‐blue) (DS). This cargo is then endocytosed (1) and incorporated into basal endosomes (E) (2). Vesicles and tubular projections containing MAL2 (orange) traffic from the subapical compartment (SAC) to fuse with basal cargo‐containing endosomes (3). After fusion, the MAL2⁺ cargo‐containing tubular/vesicular bodies aggregate and return to the SAC (4) for subsequent delivery of cargo to the apical, canalicular surface (5) (de Marco et al, 2002; de Marco et al, 2006; Madrid et al, 2010).

Apical transcytosis shapes the C. elegans excretory duct tube

Tubular epithelia are ubiquitous in biology and can be made of many cells (multicellular tubes) or, in some cases, a single cell (unicellular tubes) (Fig 2C) (Sundaram & Cohen, 2017). Like other epithelial cells, unicellular tubes are polarized, having both a defined apical surface (the tubular lumen entirely contained within the cell) and basal surface (the plasma membrane comprising the cell's external face). The relative sizes of these domains can vary, contributing to different tube shapes (Fig 2C). For example, some unicellular tubes contain an intracellular lumen that is shorter than or similar in length to the cell itself (e.g., Drosophila trachea cells Fig 2C), while others contain winding lumens greater than the length of the cell (e.g., C. elegans excretory duct Fig 2C). How lumens grow within a cell is not fully understood, and both endocytic and exocytic mechanisms have been proposed for the delivery of membrane to these growing apical surfaces (Sundaram & Cohen, 2017).

The duct cell of the C. elegans excretory system is an asymmetrically shaped unicellular tube (Fig 3B) (Sundaram & Cohen, 2017). The duct first wraps toward itself to form a tube with an autocellular junction. The cell then erases this junction by fusing with itself, to form a seamless, unicellular tube spanning the cell’s cytoplasm (Stone et al, 2009). Over a period of hours, the duct apical membrane elongates, outpacing basal membrane growth and achieving a length greater than the cell itself (Fig 3B). These differing apical and basal growth rates make the duct an attractive model for studying membrane allocation in tubulogenesis (Stone et al, 2009; Soulavie & Sundaram, 2016; Sundaram & Cohen, 2017; Soulavie et al, 2018).

The Sundaram laboratory recently proposed a transcytosis model for duct apical membrane growth (Soulavie et al, 2018). They reported that the exoplasmic cell fusion protein AFF‐1 is necessary for duct autofusion and lumen elongation. They also proposed that AFF‐1 serves as a dynamin‐independent mediator of endocytic scission. Duct‐specific AFF‐1 knockdown after autofusion caused failure in lumen elongation accompanied by the accumulation of large infoldings of basal plasma membrane continuous with the extracellular environment, suggesting they were paused endocytic events. Soulavie et al also detected tagged AFF‐1 protein at sites of basal endocytosis in wild‐type cells, suggesting that AFF‐1 is poised to regulate scission of endocytic pits and membrane internalization in the duct. The duct lumen also failed to elongate in rab‐11 mutant worms, indicating that apically directed membrane trafficking is also necessary for lumen growth. Together, these results suggest that for efficient lumen elongation, basal membrane must be internalized in an AFF‐1‐dependent manner and transcytosed to the apical surface via RAB‐11⁺ recycling endosomes (Soulavie et al, 2018).

Basal transcytosis synchronizes apical and basal membrane growth during early Drosophila tracheal terminal cell development

Terminal cells of the Drosophila trachea are a remarkable example of how a single cell can form extended and complex structures. In Drosophila embryos, terminal cells elongate to a highly branched structure and generate a single tubular lumen that extends through the cell body and into each branch (Best, 2019). Building these long, narrow, lumen containing cell projections requires coordinated delivery of membrane and proteins to both apical and basal surfaces. Fittingly, these cells have served as a long‐standing model for both cell extension and tubulogenesis.

During tube morphogenesis, the terminal cell’s basal and apical (luminal) membranes are highly coordinated and elongate at the same rate, but how the terminal cell allocates newly synthesized plasma membrane and components to each of these surfaces is not completely understood (Gervais & Casanova, 2010). A recent report by Mathew et al (2020) presents a model in which de novo membrane is not transported bi‐directionally to the apical and basal surfaces of the terminal cell, but instead specifically exported to the apical membrane, endocytosed, and then redistributed to both surfaces (Fig 3C). They demonstrate that the apical membrane is subject to continuous, clathrin‐ and dynamin‐mediated endocytosis during early terminal cell elongation. Blocking endocytosis via temperature‐sensitive dynamin inactivation caused reversible failure in terminal cell elongation accompanied by an increase in the proportion of apical plasma membrane with little to no growth of the basal membrane, suggesting that during lumen and cell elongation, newly synthesized plasma membrane components are primarily delivered apically. This agrees with other reports demonstrating that dynamin mutant terminal cells accumulate cystic lumens and fail to extend branched basolateral structures later in development (Schottenfeld‐Roames et al, 2014). Consistent with de novo secretion being apically directed, basal transmembrane proteins such as FGFR and integrin were transiently detected at the apical surface during growth. Electron microscopy and 3D reconstruction of cells showed a variety of intracellular vesicles, presumptive endosomes, and larger MVBs near the growing tip of the terminal cell and its lumen. Matthew et al go on to show that these MVBs are lost when endocytosis is blocked. Furthermore, they report that blocking the function of MVBs by Snf7::GFP overexpression halted the growth of both the apical and basal membranes. Together, these results suggest a model in which new membrane and integral membrane proteins are exported first to the apical membrane, endocytosed, sorted through endosomes and MVBs, and redistributed to the apical and basal plasma membranes, respectively (Fig 3C). Thus, basal membrane and proteins are delivered via transcytosis from the apical surface.

Apical transcytosis promotes canalicular lumen formation in hepatocytes

Most studies of transcytosis in mammals have relied on cultured cells rather than developing organisms, but evidence points to an important role for transcytosis in developing hepatocytes (Bartles et al, 1987; Schulze et al, 2019). Hepatocytes are specialized epithelial cells that generate a complex network of small tubes, the canaliculi, that transport bile. In their classic study, Bartles et al used pulse‐chase metabolic radiolabeling and subcellular fractionation experiments to track the movements of various apical or basolateral integral membrane proteins in hepatocytes (Bartles et al, 1987) (Fig 3D). They found that multiple apical proteins trafficked initially to the basolateral surface before being transcytosed to the apical surface (Bartles et al, 1987). Subsequent studies have suggested that caveolin‐dependent bulk transcytosis is the predominant mode for delivery of membrane and cargo to growing apical/canalicular membranes (Schulze et al, 2019).

The tetraspan myelin and lymphocyte protein 2 (MAL2) is required for apical transcytosis in HepG2 hepatoma cells (de Marco et al, 2002). MAL2 co‐localizes with Rab11 in both subapical and more dispersed membrane compartments, and it promotes formation of large tubulovesicular bodies that eventually fuse with basolateral endosomes containing transcytosing cargo (de Marco et al, 2006). After fusion, the MAL2⁺ vesicle containing endocytosed cargo is then transported to subapical endosomes, suggesting that MAL2 is a middleman directing basal‐to‐apical transcytosis (Fig 3D). MAL2 binds to the formin INF2, which has both actin polymerizing and depolymerizing activities, and links MAL2⁺ tubulovesicles to actin filaments that appear to propel their movement (Madrid et al, 2010). Depletion of INF2 alters MAL2 distribution, reduces contacts between MAL2⁺ vesicles and actin, and disrupts both transcytosis of apical cargo and canalicular lumen formation in cell‐based assays. The Rho family GTPase Cdc42 binds to INF2 and is also required for canalicular lumen formation, suggesting it may regulate INF2 activity in actin polymerization/depolymerization. These data strongly suggest that a Cdc42‐INF2‐MAL2 pathway promotes basal‐to‐apical transcytosis during canalicular lumen formation, though the model has not yet been confirmed in vivo. It remains unclear how apically destined protein cargos are recognized and specifically mobilized by the Cdc42‐INF2‐MAL2 pathway.

There are hints that a similar Cdc42‐INF2‐MAL2 pathway might promote transcytosis (or at least apical exocytosis) in other cell types. Depletion of Cdc42, INF2, or MAL2 disrupts lumen formation in Madin–Darby canine kidney (MDCK) cells, a popular model for studies of polarized trafficking (Bryant et al, 2010; Madrid et al, 2010). Both Cdc42 and the INF2 ortholog EXC‐6 are required for proper lumen growth and shaping in the C. elegans excretory canal cell, a unicellular tube, though no MAL‐related protein has yet been implicated (Shaye & Greenwald, 2015, 2016).

Transcytosis of extracellular matrix factors and lipoproteins

The lumens of developing tubes are shaped by apical extracellular matrices (aECMs) that contain protein, carbohydrate, and lipid components (Luschnig & Uv, 2014). Many aECM components are synthesized and secreted apically by the cells that directly line the lumen, but some can be provided by other tissues and delivered to the tube lumen via transcytosis. Circulating lipoproteins in the bloodstream also rely on transcytosis to cross the endothelium and access other tissues in the body.

Transcytosis delivers the apical extracellular matrix protein Serpentine from the basal extracellular environment to the developing tracheal lumen in Drosophila

Serpentine (Serp), a chitin deacetylase, is critical for proper aECM assembly and shaping of multicellular tubes in the Drosophila trachea (Luschnig et al, 2006; Wang et al, 2006). Serp is derived in part from neighboring fat body cells, transported via the hemolymph, and then transcytosed across the tracheal epithelium to the growing lumen (Dong et al, 2014). Serp mutants can be rescued by expressing full‐length Serp from tracheal or fat body‐specific drivers. GFP‐tagged Serp expressed from a fat body‐specific promoter is detectable in the tracheal lumen as well, even when expressed after tracheal tight junctions are established. Cell‐specific RNAi knockdown of Serp indicates that trachea and fat body‐derived Serp function redundantly, suggesting that a combination of Serp natively expressed and transcytosed from the hemolymph ensures proper protein dosage during lumen development (Dong et al, 2014).

Transcytosis transports lipoproteins across the vascular endothelium and contributes to atherosclerosis

In mammals, low‐density lipoprotein (LDL) and high‐density lipoprotein (HDL) particles are the major vehicles for delivery of cholesterol and fatty acids to different parts of the body. LDL and HDL travel within the bloodstream, but then must escape the vascular system in order to reach other tissues. Several different receptor‐mediated or receptor‐independent pathways for transcytosis have been implicated in this process (Zhang et al, 2018b).

The mechanisms of LDL transport across arterial cells are of particular interest because abnormal accumulation of LDL near the basal surfaces of these cells is an initiating event in atherosclerosis. Two recent studies suggest that LDL is transcytosed basally across arteries in a manner dependent on the scavenger receptor SR‐B1, a known LDL transporter. Armstrong et al used total internal reflection fluorescence (TIRF) microscopy studies to visualize LDL transcytosis in both cultured murine aortas and human endothelial cells. They observed that transcytosis decreased after SR‐B1 depletion and increased after SR‐B1 overexpression (Armstrong et al, 2015). Subsequently, Huang et al (2019) generated mice with endothelial‐specific deletion of SR‐B1 and found that they had reduced LDL accumulation in the arterial walls. These authors also identified the Rac guanine nucleotide exchange factor dedicator of cytokinesis 4 (DOCK4) as a key partner of SR‐B1 in the process and showed that higher endothelial SR‐B1 and DOCK4 expression correlated with atherosclerotic plaques in both mouse and human samples. Together, these studies suggest that SR‐B1‐dependent LDL transcytosis directly contributes to the formation of atherosclerotic plaques and that blocking such transcytosis in arteries could have therapeutic value.

Transcytosis as a regulator of cell motility

In addition to affecting the shape of stationary cells, transcytosis can also support motility of both epithelial and non‐epithelial cells. In this case, rather than moving membrane and cargo from apical to basal surfaces (or vice versa), transcytosis moves membrane from the rear to the leading edge of the migrating cell.

Transcytosis recycles junctional complexes during collective cell migration

Transcytosis helps maintain cell junctions during collective cell migration. Astrocytes are glial cells that, similar to epithelia, contain cadherin‐based cell junctions. Peglion et al reported that photo‐labeled N‐cadherin is internalized at the rear of migrating astrocytes and concurrently recycled to the leading edge of the cells in vitro (Peglion et al, 2014). They also found that blocking dynamin or Rab5 activity reduces accumulation of N‐cadherin at the leading edge. The authors suggest a "treadmilling" model in which the endocytosis and recycling of junctional complex proteins allows disassembly of adherens junctions at the rear of the cell and concurrent establishment of new junctional complexes at the leading edge (Fig 4). This transcytotic recycling along with movement of the actin cytoskeleton establishes a sustained retrograde flow of junctional complexes along the lateral cell face, allowing the cells to move together while maintaining intact adherens junctions (Fig 4) (Peglion et al, 2014). A similar internalization and redistribution model has been proposed for integrins, proteins which mediate epithelial cells' attachment to ECM components of the basement membrane and which are redistributed during migration and polarity‐establishing events (Moreno‐Layseca et al, 2019).

Figure 4. Transcytosis facilitates junction treadmilling in collective cell migration.

As epithelial cells migrate en masse, tissue integrity requires that cell‐cell junctions be maintained. The cells shown here are migrating in the same direction. Junctional complex proteins (e.g., cadherin represented in red) are internalized and moved by vesicular transport in an anterograde direction. This allows the assembly of new junctional complexes in the direction the cells are moving and recycles those disassembled from the opposite side of the cell. This supports an actin‐dependent “retrograde flow” of cell junctions in the opposite direction of cell motion, ensuring the cells are not disconnected as they migrate (Peglion et al, 2014).

Transcytosis as a potential facilitator of non‐epithelial ameboid cell movement

Transcytosis‐driven plasma membrane flow also has been proposed as a means for single cells to generate sufficient friction to propel themselves (Bretscher, 1996, 2014; Barry & Bretscher, 2010). A recent study provided direct evidence for this model. O'Neill et al (2018) reported that optogenetic activation of RhoA in one side of cultured macrophages causes them to move in the opposite direction. In this study, RhoA activation was sufficient to drive motion of the cells across surfaces as well as through attachment free liquid environments. By time‐lapse imaging of membrane tethered beads and fluorescent markers of the inner and outer plasma membrane leaflets, the authors showed that RhoA activation at the rear of the cell causes a retrograde flow of membrane, likely stimulating this movement. The RhoA stimulated membrane flow is supported by simultaneous endocytosis at the rear of the cell and exocytosis at the front of the cell, facilitating continuous motion (O'Neill et al, 2018). Pharmacologic inhibition of either secretion or endocytosis significantly inhibited RhoA stimulated motion, suggesting that both are necessary for sustained membrane flow and motility. Interestingly, blocking direct (Golgi‐mediated) secretion to the plasma membrane with brefeldin‐A only partially inhibited motility, leaving open the possibility that other trafficking pathways contribute to the generation of membrane flow. Consistent with this, they observed endocytosed vesicles containing labeled transferrin and caveolin moving toward the migrating cell front, strongly suggesting that endocytosed membrane undergoes anterograde transcytosis to support migration and membrane flow.

Other studies in Dictyostelium and Toxoplasma models have proposed similar forms of flow‐powered cell movement (Tanaka et al, 2017; Gras et al, 2019). However, only in Toxoplasma have the data suggested that membrane lipids internalized at the cell rear are subsequently transcytosed to the leading edge of the cell, streamlining the endocytic/exocytic cycle (Gras et al, 2019). Whether or not transcytotic membrane recycling contributes to flow‐based motility in other systems remains to be investigated.

Transcytosis and intercellular signaling

For cell signaling, a ligand and its cognate receptor must meet physically. Sequestration of ligands and receptors in different compartments, such as on opposite sides of an epithelium, will therefore prevent signaling. Several studies from the Drosophila model suggest that transcytosis allows signaling ligands to cross epithelial barriers and reach their receptors (Fig 5A). Such regulation may help ensure that signaling occurs at appropriate levels or with appropriate timing to direct developmental events. Ligand transcytosis might also allow specific modifications or interactions to occur within endosomes or facilitate packaging into extracellular vesicles (EVs) for long distance signaling (Fig 5B).

Figure 5. Transcytosis as a regulator of signaling.

(A) Because many signaling molecules cannot pass through occluding junctions, epithelial tissues can act as barriers separating ligands from their cognate receptors. Transcytosis is one way a ligand can cross such epithelial barriers. In this simplified example, a signaling ligand (red) is secreted apically into a separate compartment from its basally localized receptor. Through endocytosis (1), endosomal transport (E) (2), and trafficking to the basal membrane (3) for secretion (4), the ligand can meet and activate its receptor (5). (B) Highly lipid‐modified ligands are unlikely to diffuse efficiently in the aqueous extracellular environment. Ligand‐producing cells can endocytose newly secreted ligand (red) (1) and incorporate it into early endosomes (EE) (2). In early endosomes, ligand can be sorted to late endosomes (LE) and multivesicular bodies (MVB) which are able to incorporate cargo into intraluminal vesicles. MVBs can fuse with the plasma membrane (4) and release ligand‐associated intraluminal vesicles as extracellular vesicles or exosomes (EV) (5) which may conduct long range signaling more efficiently than naked, lipidated ligand. This mechanism of secretion, endocytosis, and transcytosis for release on exosomes has been proposed for the highly lipid‐modified ligands wingless and hedgehog in the Drosophila wing imaginal disk (Gradilla et al, 2018; Parchure et al, 2018).

Wingless transcytosis across the Drosophila wing imaginal disk epithelium

The Drosophila Wnt ligand Wingless (Wg) is critical for the formation and patterning of the developing wing. In the wing imaginal disk, a precursor epithelium from which the wing develops, a subset of cells express and secrete Wg, generating a signaling gradient across the tissue (Cadigan, 2002). Several studies have suggested that, in producing cells, Wg is secreted apically, reabsorbed by endocytosis, and then transcytosed to the basal surface of the epithelium for secondary release and signaling (Fig 5A) (Gallet et al, 2008; Yamazaki et al, 2016; Routledge & Scholpp, 2019; Witte et al, 2020).

Time course imaging of tagged, inducible Wg protein supports an apical‐to‐basal transcytosis model (Yamazaki et al, 2016). Shortly after initial expression, tagged Wg is detectable in Golgi compartments, then incorporated into apical endosomes, and eventually detected apically, intracellularly, and basally, consistent with its localization under steady‐state conditions. For the apical‐to‐basal transcytosis model to hold, Wg must be endocytosed at the apical membrane. Consistent with this, blocking endocytosis with dynamin(ts) caused an apical buildup of Wg as well as basal depletion. This apical buildup was reversed once tissues were returned to a permissive temperature. A similar apical buildup and basal depletion was observed when endocytosis was blocked by Rab5(dn) expression. To test whether apical endocytosis and transcytosis was important for Wg signaling, Yamazaki et al overexpressed the Wg inhibitor Notum in hemolymph, the Drosophila blood analogue, which has access to the basal compartment of the disk epithelium, but not the apical. Deactivation of basal Wg caused downregulation of known Wg targets in the epithelium, indicating that abundant active Wg at the apical surface was insufficient to carry out signaling (Yamazaki et al, 2016). The same study reported that the Wg receptor Frizzled2 was predominantly localized in the basal compartment, suggesting that transcytosis may be needed for Wg to access this receptor.

The endosome‐associated E3 ubiquitin ligase Godzilla and its target, the snare protein Synaptobrevin, are necessary for Wg signaling and proposed to mediate apical‐to‐basal transcytosis (Yamazaki et al, 2013; Yamazaki et al, 2016). Knockdown of Godzilla or Synaptobrevin causes similar phenotypes, with intracellular buildup of Wg near the apical cell surface, reduced Wg signaling, and loss of wing margin (Yamazaki et al, 2016). Wg and Godzilla are detectable in the same Rab5⁺ early endosomes, suggesting that Godzilla interacts with Wg in this compartment, and may direct Wg to more mature endosomes for eventual transcytosis and secretion (Yamazaki et al, 2016). A more recent study reported that loss of the kinesin Klp98A causes buildup of Wg in apical endosomes, with a concomitant reduction in basal endosomes (Witte et al, 2020). Together, these studies describe Wg as first apically secreted, endocytosed, and transcytosed to the basal surface in a Godzilla‐, Synaptobrevin‐, and Klp98A‐dependent mechanism.

It is important to note that the functional importance of Wg transcytosis is still being debated. Multiple studies suggest that Wg signaling in the wing imaginal disk occurs not only at basal surfaces, but also at apical surfaces or in endosomal compartments (Hemalatha et al, 2016; Chaudhary & Boutros, 2019; Linnemannstöns et al, 2020). Furthermore, Witte et al, 2020 found that Wg signaling was not significantly perturbed after klp98A loss, despite reduced transcytosis to the basal extracellular space (Witte et al, 2020). It is possible that Wg signals from multiple locations via different Frizzled family receptors, but further study is needed to understand the contributions of each.

Hedgehog transcytosis across the Drosophila wing imaginal disk epithelium

Although produced by different cells in the wing imaginal disk, the ligand Hedgehog (Hh) also is secreted basally in a mechanism that parallels Wg (Fig 5B). Apically secreted Hh undergoes endocytic reuptake and is transcytosed via endosomes to the basal plasma membrane, where it forms an extracellular morphogen gradient and co‐localizes with its receptor Patched (Ptc) (Callejo et al, 2011; D'Angelo et al, 2015). Similar to Wg, blocking endocytosis with dynamin(ts) or Rab5(dn) causes an apical buildup accompanied by basal depletion of Hh in producing cells. Blocking recycling with Rab4(dn) or Rab8(dn) also causes an accumulation of Hh‐containing apical endosomes, suggesting that following apical endocytosis, Hh moves through apical endosomes prior to recycling or transcytosis to the basolateral plasma membrane (Callejo et al, 2011; D'Angelo et al, 2015).

The Hh transporter Dispatched (Disp) has several proposed roles in Hh transcytosis. Callejo et. al reported that loss of Disp or its binding partner Dally‐like protein (Dlp) results in an increase in the number of Hh‐containing vesicles and endosomes intracellularly and concluded that Disp and Dlp promote Hh transcytosis and secretion at the basal plasma membrane (Callejo et al, 2011). In contrast, D'Angelo et al, 2015 observed that loss of Disp results in apical accumulation of extracellular Hh and concluded that Disp primarily mediates Hh endocytosis (D'Angelo et al, 2015). In this model, endocytosis of Hh occurs between synthesis and secretion for signaling. Thus, Disp is necessary for the eventual export of Hh from producing cells, but its specific mechanistic role remains to be resolved.

Wingless and Hedgehog transcytosis and extracellular vesicles

The similar apical‐to‐basal transcytosis patterns described for Hh and Wg in the wing imaginal disk are intriguing and may reflect some common regulation or purpose. Hh and Wnt have other potentially relevant similarities. Both proteins are lipid‐modified and thus may have difficulty diffusing freely in the aqueous extracellular environment (Parchure et al, 2018). Both have been detected on extracellular vesicles, or exosomes, which have been proposed to enhance their solubility and travel through the extracellular space and facilitate long distance signaling (Korkut et al, 2009; Gross et al, 2012; Beckett et al, 2013; Gradilla et al, 2014; Matusek et al, 2014; Parchure et al, 2015; Gradilla et al, 2018; Parchure et al, 2018). The endocytosis and transcytosis models described in previous sections would grant Wg and Hh access to MVBs where they could be incorporated into intraluminal vesicles (ILVs) and eventually secreted as exosomes (Fig 5A and B). Parchure et al have reported that Hh in particular is detectable in MVBs of the imaginal disk, and show that regional knockdown of ESCRT proteins (necessary for MVB function and exosome biogenesis) reduces both Hh spread and signaling in vivo (Parchure et al, 2015). Nonetheless, the necessity of exosomes for long range Wg or Hh signaling remains to be fully described (Gradilla et al, 2018; Hessvik & Llorente, 2018; Parchure et al, 2018).

Transcytosis contributes to Delta and Notch localization in Drosophila epithelia

The Notch (N) receptor and its ligand Delta (Dl) are single pass transmembrane signaling molecules that enable juxtacrine signaling between adjacent cells (Siebel & Lendahl, 2017). Endocytosis plays a key role in Dl‐N signaling, as Epsin‐dependent endocytosis of Dl exerts a pulling force on bound N to facilitate its cleavage and activation (Langridge & Struhl, 2017). Transcytosis of either Dl or N has also been proposed to modulate Dl‐N interactions and signaling strength.

In the Drosophila neuroepithelium, Notch signaling is the key determinant between epithelial and sensory organ precursor (SOP) cell fates, as well as the fate of SOP cell progeny (Hartenstein & Posakony, 1989, 1990). In this tissue, Dl is observed at basolateral surfaces and in intracellular puncta, whereas N is observed at the apical cell surface (Benhra et al, 2010). Therefore, an appreciable amount of Dl is physically separated from N and would need to be endocytosed and transported apically for signaling. Benhra et al showed that compromising Dl endocytosis via dynamin(ts) or mutants for Epsin or neuralized (neur) led to increased basal accumulation of DI (Benhra et al, 2010). Pulse‐chase antibody labeling showed that basal Dl moves into Rab5⁺ early endosomes, and eventually Rab11 and Sec15⁺ secretory/recycling endosomes, consistent with basal‐to‐apical transcytosis, although actual movement to the apical membrane was not definitively shown. Studies of Dl trafficking in MDCK cells did reveal Neuralized‐dependent basal‐to‐apical relocalization. The functional importance of such transcytosis is difficult to assess given the other key roles for endocytosis in Dl‐N signaling, but the data suggest that basally localized Dl may serve as a reservoir of ligand that can be relocalized apically via a Neur‐stimulated transcytotic pathway.

A similar study in the Drosophila wing imaginal disk reported that Notch is endocytosed from the apical and lateral cell membranes and relocalized to the subapical region by recycling or transcytosis, likely concentrating the receptor in a membrane region conducive to signaling with adjacent cells (Sasaki et al, 2007).

Transcytosis in development: open questions and opportunities

Why use transcytosis rather than direct secretion?

The above examples show that cellular strategies for transcytosis vary widely. In some epithelial cells, like hepatocytes or Drosophila trachea terminal cells, direct secretion appears predominantly basal or predominantly apical, such that a large number of cargos must be transcytosed secondarily to reach their proper locations. In other epithelial cells, including the popular MDCK culture model, direct secretion proceeds both apically and basally, with only specific cargos being transcytosed. The underlying rationale for these different strategies remains unclear.

There are a variety of reasons why cells might use seemingly circuitous transcytosis mechanisms to localize membranes and proteins, rather than direct secretion following de novo synthesis. First, some proteins may require endosomal transport to become fully effective, undergoing conformational change in the lower pH endosomal environment, or being co‐transported with a necessary chaperone or modifying enzyme. Proteins that are destined for exosome‐mediated secretion may need transcytosis in order to access the MVB compartments from which such exosomes derive. Second, transcytosis allows cells to store pre‐made materials in an out‐of‐the way location and then quickly mobilize them when needed. This may be particularly important when cells must undergo rapid shape changes during morphogenesis or migration, or when cells must respond quickly to environmental signals and stressors. Third, redeployment of stored materials may allow for more precise regulation of how much material is delivered to a cell surface and when. For dose sensitive signaling molecules like Wg, Hh, and Dl, either too much or too little can have dramatic effects on developmental outcomes, so it is important for the cell to carefully control their presentation levels. Finally, redistribution of existing membrane represents thrifty use of the cell’s resources, allowing complex shape change with less synthetic energy expenditure.

Molecular pathways for transcytosis

The vesicle compartments and molecular players involved in transcytosis also appear diverse and cell type‐ or cargo‐specific (Tuma & Hubbard, 2003; Fung et al, 2018; Ayloo & Gu, 2019; Villaseñor et al, 2019). Cells can internalize cargos selectively using dedicated receptors for clathrin‐mediated endocytosis (CME) or non‐selectively using a variety of clathrin‐independent endocytosis (CIE) mechanisms such as caveolin‐dependent endocytosis or macropinocytosis (Thottacherry et al, 2019). Sorting of endocytosed cargo then can occur within early, or plasma membrane‐associated endosomes ("fast recycling"), as well as in more centralized or late endosomes and MVBs ("slow recycling"). Slow recycling can return material to the originating plasma membrane, to the opposite plasma membrane (transcytosis), or to the Golgi ("retrograde recycling"). In polarized MDCK cells, a "common recycling endosome" (CRE) compartment appears to sort cargos for slow recycling or transcytosis in both the apical and basal directions; however, the nature of this compartment, and how it determines where each cargo should go, remains poorly understood (Apodaca et al, 1994; Tzaban et al, 2009). Identifying the relevant sorting endosomes in different cell types and dissecting how specific cargos move into and out of these compartments will be critical for understanding transcytosis.

Sorting endosomes contain microdomains with differing protein and lipid compositions, and these may be key to their sorting activity (Sonnichsen et al, 2000; Norris & Grant, 2020). For example, in large endosomes within C. elegans coelomocytes, imaging showed one region enriched for ESCRT proteins that direct contents toward ILV incorporation and the lysosomal degradation pathway, while another region of the same endosome is enriched for Retromer and other proteins that direct contents back to the Golgi for retrograde recycling (Norris et al, 2017). Cargos may be concentrated into such microdomains that then exchange material with other endosome compartments or bud off from the sorting endosome en route to different locations.

In some transmembrane cargos, specific cytosolic sequences have been found to promote recycling versus degradation by binding to specific SNX adaptors that link that cargo to Retromer or other vesicle coat complexes (Weeratunga et al, 2020). Cargo sequences and adaptors that direct basolateral secretion or recycling also have been identified (Stoops & Caplan, 2014), but signals that direct apical secretion remain more poorly understood. Many apically secreted proteins lack transmembrane and cytosolic domains, and thus cannot bind to cytosolic adaptors; instead, glycans in their luminal domains may help direct apical trafficking, with glycan‐binding lectins serving as adaptors (Delacour et al, 2009; Zhang et al, 2018a; Levic et al, 2020). Recent work identified the Factors for Endosome Recycling and Rab Interaction (FERARI) as a likely tethering complex connecting sorting endosomes to Rab11⁺ endosomes for apical recycling (Solinger et al, 2020). These promising reports suggest that continued study will reveal further adaptor protein complexes that direct endosomal cargo into specific recycling and transcytosis pathways.

Studying transcytosis in vivo

Harnessing in vivo models will be critical to understand the mechanisms of transcytosis in tissue context. The invertebrate models highlighted above should continue to prove advantageous for future studies. Significantly, the developing tissues of flies and nematodes are relatively accessible for imaging, usually at much higher resolution than is possible in larger models. These organisms also are genetically tractable, with established, community‐driven repositories of mutants and reporter constructs that allow spatially and temporally specific manipulations of trafficking factors. These models also have a successful history as platforms for unbiased forward genetic screens (Grant & Hirsh, 1999; Fares & Greenwald, 2001; Balklava et al, 2007). Finally, flies and nematodes develop quickly relative to other multicellular model systems, facilitating the development of rapid, high‐throughput approaches to studying tissue formation.

There is a great need to develop new models for studying transcytosis in vertebrates as well. Zebrafish should be a promising model because of its tractability for live imaging, as displayed in recent studies of BBB establishment (Ben‐Zvi et al, 2014; Chow & Gu, 2017; Pulgar, 2018; Ayloo & Gu, 2019; O'Brown et al, 2019) and apical trafficking in the gut (Levic et al, 2020). So far, few in vivo studies of transcytosis in vertebrates have focused on developmental roles outside of the BBB, but hepatocytes and other cell lines hint at the potential for considerable developmental transcytosis in tissues such as the liver and gut (Tuma & Hubbard, 2003). Improved methods for imaging either in vivo or in ex vivo organ culture are needed to confirm such transcytosis in vivo and to explore its functional significance in the development and shaping of these organs.

Summary: a developmental perspective on transcytosis

In addition to mediating transport of materials across cells, transcytosis allows redistribution of membranes and proteins within a cell. The works highlighted above exemplify the potential of transcytosis to facilitate cellular and organismal development. Cells can leverage transcytosis to control or package signaling molecules and mobilize membrane to facilitate radical changes in cell shape and motility. Many additional examples likely remain to be discovered, especially in vertebrates, but recognizing them will require improved imaging methods for the study of trafficking processes in vivo or in ex vivo cultures or organoids. The molecular players and endosome compartments that mediate transcytosis are incompletely understood, and this represents a major area for future study. With new microscopy techniques and a growing understanding of endosomal sorting, it is likely that new roles for transcytosis will continue to emerge.

Conflict of interest

The authors declare that they have no conflict of interest.

The EMBO Journal (2021) 40: e106163.