Abstract
The excretory system of the nematode Caenorhabditis elegans is a superb model of tubular organogenesis involving a minimum of cells. The system consists of just three unicellular tubes (canal, duct, and pore), a secretory gland, and two associated neurons. Just as in more complex organs, cells of the excretory system must first adopt specific identities and then coordinate diverse processes to form tubes of appropriate topology, shape, connectivity, and physiological function. The unicellular topology of excretory tubes, their varied and sometimes complex shapes, and the dynamic reprogramming of cell identity and remodeling of tube connectivity that occur during larval development are particularly fascinating features of this organ. The physiological roles of the excretory system in osmoregulation and other aspects of the animal’s life cycle are only beginning to be explored. The cellular mechanisms and molecular pathways used to build and shape excretory tubes appear similar to those used in both unicellular and multicellular tubes in more complex organs, such as the vertebrate vascular system and kidney, making this simple organ system a useful model for understanding disease processes.
Introduction to the Excretory System and Unicellular Tubes
Organs are made up of networks of tubes that transport fluids or gases and mediate exchange of nutrients and waste products with surrounding tissues. Each tube type in the network has a distinctive size and shape that facilitates its particular function. Many biological tubes are constructed from polarized epithelial cells, with the apical surface forming the lumen, while the basal surface forms the outside of the tube facing other tissues or body cavity (Figure 1). Wider tubes are formed from multiple cells whose lateral surfaces bind to each other via adherens and tight junctions to maintain the patency of the lumen (Figure 1A). The smallest tubes are unicellular, made up of single cells (Figure 1, B–D). Tubes are often dynamic and can convert from one type to another during development (Ribeiro et al. 2004; Denker et al. 2013; Lenard et al. 2013). Cells within tubes can also delaminate, change identity, and give rise to other cell types (Kalluri and Weinberg 2009; Zuryn et al. 2012). Biological tubes make up a vast proportion of cells in all eukaryotic organisms, and tube abnormalities underlie many birth defects and diseases in humans, yet the mechanisms that determine tubule identity, size, shape, and connectivity are incompletely understood (Lubarsky and Krasnow 2003; Iruela-Arispe and Beitel 2013; Sigurbjornsdottir et al. 2014).
Tube topologies and models for unicellular tube formation. (A and B) In multicellular tubes (A) and seamed unicellular tubes (B), adherens junctions (in black) separate and delineate the apical and basal domains. (C and D) In seamless unicellular tubes, junctions appear only at the end(s) of the tube, at the site of connection to another cell. Some multicellular tubes form by a cord-hollowing mechanism (A) involving polarized vesicle trafficking toward the lumen (arrows) (Datta et al. 2011). The excretory pore cell (in blue) (B) forms a seamed tube by wrapping and forming an autocellular junction (AJ). The excretory duct cell (in yellow) (C) initially forms a seamed tube and then converts to a seamless tube via autofusion to remove the AJ (Stone et al. 2009). The canal cell (in red) (D) forms a seamless tube, likely through a cell-hollowing mechanism involving polarized vesicle trafficking toward the lumen. The origin of relevant vesicles is not known. The intercellular junction with the duct may provide a polarizing cue that directs vesicle targeting to this region. For all tubes, both basal and apical sides may secrete extracellular matrices. For all diagrams, lumen is shown in lighter color.
The Caenorhabditis elegans excretory system is a very simple tubular organ that consists of just three unicellular tubes (canal, duct, and pore) connected in tandem to form a continuous lumen (Nelson et al. 1983) (Figure 2). Dynamic changes in pore identity and tube connectivity occur during development (Figure 3, Figure 4). Two fused gland cells (Figure 3E) form a syncytium attached to the canal–duct junction (Nelson et al. 1983), and two canal-associated neurons (Figure 2A, Figure 3F) have axons that run along the basal surface of the canal cell and have been proposed to regulate canal function (Hedgecock et al. 1987; Forrester and Garriga 1997). The system appears to function in both osmoregulation and secretion, and it shares many molecular features with osmoregulatory or renal systems in other organisms, and with other epithelial and endothelial tube networks. The simplicity of the excretory system, and its amenability to genetic analysis and live imaging, make it a very attractive model for studying the cell biology of tubulogenesis, tube dynamics, and other aspects of organ development and physiology.
The C. elegans excretory system. (A) Diagram of cells of the C. elegans adult excretory system. (B) DIC with fluorescence photograph of qpIs11[Pvha-1::gfp] adult expressing GFP in the canal cytoplasm. Arrowheads mark anterior and posterior ends of the canals running the length of the organism.
C. elegans excretory system parts list. Schematic representations of the cell types that contribute to the excretory system, their lineal origins (Sulston et al. 1983), and major transcription factor (TF) regulators or markers. Colors represent specific cell types throughout the figures. Anterior is to the left. Junctions (as visualized with AJM-1, DLG-1, or HMR-1/cadherin reporters) are represented with heavy black rings or lines. (A) Canal cell (ventral view). Two ring-shaped junctions connect the canal lumen to the duct and gland. (B) Duct (lateral view). Ring-shaped junctions connect the duct to the canal cell (right), and pore (left). (C and D) G1, G2, or G2p pore (lateral views). An AJ (arrow) seals the tube, and ring-shaped junctions connect the pore to the duct (top) and hypodermis (bottom). (E) Binucleate excretory gland (ventral view). A small ring-shaped secretory–excretory junction (SEJ, arrowhead) connects the gland to the canal cell. (F) CAN neurons (ventral view). *, putative phosphorylation. ^, markers suitable for visualizing the mature cells.
Overview of excretory system development. (A) Excretory system cells are born in disparate locations of the developing embryo and migrate to meet at the ventral midline during ventral enclosure (Sulston et al. 1983). Colors represent cell types as in Figure 3. Dashed line, ventral midline. (B) Sequential Notch and EGFR signaling break symmetry in the ABplpa vs. ABprpa lineages that give rise to the canal, duct, G1 pore, and gland cells (Sulston et al. 1983; Moskowitz and Rothman 1996; Abdus-Saboor et al. 2011). (C) LIN-12/Notch signaling breaks symmetry in the ABplap vs. ABprap lineages to promote the G2 pore vs. W neuroblast fate (Greenwald et al. 1983; Sulston et al. 1983). (D) Near the completion of ventral enclosure, the tube progenitor cells are in contact but have not yet formed junctions or lumen (Abdus-Saboor et al. 2011). Note asymmetric position of the canal cell with respect to the ventral midline. (E) By the 1.5-fold stage, the tubes have junctions consisting of AJM-1, DLG-1, and HMR-1/cadherin (heavy black lines) and a continuous lumen (white) that extends from the canal cell, through the duct and pore, and is open to the extraembryonic environment (Stone et al. 2009). Arrowhead, canal–duct intercellular junction. The gland cells also connect to the canal and duct at this region. Arrows, duct and pore AJs. The duct AJ autofuses at this stage (dashed lines) (Stone et al. 2009). (F) By late embryogenesis, the tubes have undergone morphogenesis to adopt their characteristic shapes. (G) By late in the first larval stage (L1), the G1 pore has delaminated and lost all its prior junctions and lumen, and the G2 cell has replaced it as the pore (Sulston et al. 1983; Parry and Sundaram 2014). Note that in F and G, the canal cell arms are drawn much shorter than actual length, and rotated relative to their actual lateral positions, to show cell shape; in particular the canal arms reach the midbody by late embryogenesis and the rectum by late L1 stage.
Unicellular tubes
Excretory system tubes are unicellular, meaning that the apical or luminal domain is located inside of the cell (Figure 1, B–D). Unicellular tubes are also found in tissues as diverse as vertebrate capillaries (Bär et al. 1984; Iruela-Arispe and Davis 2009; Lenard et al. 2015; Yu et al. 2015) and glia (Simons and Trotter 2007; Jessen et al. 2015), Drosophila tracheal tubes (Samakovlis et al. 1996; Schottenfeld et al. 2010; Maruyama and Andrew 2012), plant pollen tubes (Geitmann 2010), and the contractile vacuoles of single-cell protozoa (Plattner 2015). Understanding how unicellular tubes are formed and maintained has significant health relevance, given the prevalence of such tubes in the microvasculature and nervous system and their involvement in cardiovascular disease, stroke, and neuropathies (Feihl et al. 2008; Govani and Shovlin 2009; Vallat et al. 2013; Draheim et al. 2014; Haffner et al. 2016). Several different mechanisms for forming unicellular tubes have been proposed, although none are very well understood at the molecular level.
Wrapping:
An individual cell can form a tubule of defined luminal diameter by wrapping its plasma membrane around an extracellular core of matrix or other material and forming an autocellular junction (AJ) or “seam” to itself (Figure 1B). Seamed unicellular tubes are generally open at both ends. Examples of seamed unicellular tubes include the C. elegans excretory pore cell (Nelson et al. 1983); Drosophila tracheal stalk cells, which wrap around a core of chitin (Ribeiro et al. 2004; Tonning et al. 2005; Luschnig et al. 2006); many vertebrate capillaries (Yu et al. 2015), especially in the brain (Bär et al. 1984); and many invertebrate glial socket cells that wrap around neurons (Freeman 2015; Shaham 2015). A related topology is observed in other wrapping glia, such as mammalian Schwann cells, which do not form actual junction seams, but instead adhere to themselves via secreted myelin (Simons and Trotter 2007; Jessen et al. 2015).
Wrapping and autofusion:
A seamed tube formed by wrapping can subsequently autofuse to eliminate the autocellular junction and become a “seamless” tube that has no junctions running along its length (Figure 1C). Autofusion occurs in the C. elegans excretory duct (Stone et al. 2009) and pharyngeal valve cells (Rasmussen et al. 2008), some capillaries in the zebrafish vascular system (Lenard et al. 2015), and mammalian Madin-Darby canine kidney (MDCK) cells grown on micropillar arrays in culture (Sumida and Yamada 2013, 2015).
Hollowing:
As an alternative to cell wrapping, a single cell can create a seamless tube through “cell hollowing,” in which endocytic (plasma membrane-derived) and/or exocytic (Golgi-derived) vesicles coalesce to form an internal apical surface that connects to the plasma membrane (Figure 1D). This mechanism was first described in mammalian endothelial cells grown in culture (Folkman and Haudenschild 1980; Davis and Camarillo 1996; Sacharidou et al. 2012). The initial nucleating event for intracellular lumen formation is not clear, but in vivo, lumen formation often initiates near a junction with another epithelial tube cell and then appears to grow inward from that point, suggesting polarized vesicle trafficking toward the newly established apical domain (Figure 1D), similar to exocytosis-driven cord hollowing in multicellular tubes (Datta et al. 2011; Eaton and Martin-Belmonte 2014) (Figure 1A). Seamless tubes can be closed at one end or open at both ends, depending on where nucleation of apical surface occurs. Examples of seamless tubes thought to form by a hollowing mechanism include the C. elegans excretory canal cell (Buechner 2002; Berry et al. 2003), Drosophila tracheal terminal and fusion cells (Samakovlis et al. 1996; Gervais and Casanova 2010; Gervais et al. 2012), and many capillaries (Wolff and Bar 1972; Bär et al. 1984; Yu et al. 2015), including those that fuse via anastomosis (Herwig et al. 2011; Denker et al. 2013; Lenard et al. 2013; Caviglia and Luschnig 2014).
Overview of excretory system development
The C. elegans excretory system is an excellent model for the study of unicellular tubes, since it contains tubes that form by each of the above mechanisms: (1) a seamed pore tube formed by wrapping (Figure 1B and Figure 3, C and D); (2) a seamless duct tube formed by wrapping and autofusion (Figure 1C and Figure 3B); and (3) a seamless and branched canal tube likely formed by hollowing (Figure 1D and Figure 3A). The excretory system also has been used to study other important steps of organ development, including long-range cell migration, cell fate determination, epithelial delamination, and cell fate reprogramming.
Figure 4 shows a timeline of excretory system development. Cells of the excretory system are born about one-third of the way through embryogenesis and derive from the AB lineage (Sulston et al. 1983) (Figure 3 and Figure 4, A–C). The presumptive duct, G1 pore, and glands are initially born in lateral locations far from the canal cell and then migrate to their final positions during ventral enclosure (Figure 4A). Junction and lumen formation begin only after the canal, duct, pore, and gland cells come in physical contact, around the lima bean stage of embryogenesis (Abdus-Saboor et al. 2011) (Figure 4, D and E and Figure 5A). The tubes, especially the duct and canal, subsequently elongate and change shapes, concomitant with embryo elongation (Suzuki et al. 2001; Stone et al. 2009) (Figure 4F). During this time period, the CAN neurons also migrate to their positions near the posterior canal arms (Sulston et al. 1983; Forrester and Garriga 1997), and the system begins functioning in excretion, as judged by edema formation in mutant embryos (Stone et al. 2009). During the L1 larval stage, the G1 pore delaminates and is replaced as pore by the epidermal cell G2 (Sulston et al. 1983; Parry and Sundaram 2014) (Figure 4G). In the L2 larval stage, G2 divides and its daughter G2p becomes the permanent pore.
Excretory canal cell structure and ultrastructure. (A) TEM of 420 min (comma stage) embryo canal, duct, and pore (tinted as in previous pictures) shows canal lumen developing at the point where canal contacts the duct (TEM courtesy of Shai Shaham, The Rockefeller University). Nuc, cell nuclei; Lu, lumen. (B) TEM of transverse section of an adult anterior canal, shows multitude of canalicular vesicles, some connected to the round central lumen and/or to each other, while others are disconnected. MTs surround the canaliculi, as do ER and mitochondria. The left side of this section is almost entirely gap junction, while the right side is separated from muscle by basal extracellular matrix. The terminal web is stained lightly in this section and surrounds the central lumen (M. Buechner, D. Hall, E. Hedgecock, unpublished results). Note that the membrane of the canalicular vesicles stains more darkly than does the plasma membrane or luminal membrane. (C) DIC micrograph of a section of the posterior canal in N2 adult (M. Buechner and E. Hedgecock, unpublished results). Boundary between lumen and cytoplasm (red guidelines) is highly refractive and shows up well in DIC. Boundary between cytoplasm and hypodermis (blue guidelines) is more difficult to see. (D) High-resolution TEM of canalicular vesicles in adult exc-7(rh252) mutant, in which the cytoplasmic contents appear more dilute. Each canalicular vesicle is surrounded by proteins protruding into the lumen, that appear reminiscent of the lollipop structure of ATPase (M. Buechner, D. Hall, E. Hedgecock, unpublished results).
Epithelial organization in excretory tube cells
Excretory tubes exhibit many typical epithelial characteristics (Figure 3). The apical (luminal) surface of each tube cell is marked by CRB-1/Crumbs (Bossinger et al. 2001) and the apical PAR proteins PAR-3, PAR-6, and PKC-3 (Mancuso et al. 2012; Armenti et al. 2014). The tubes are joined by ring-shaped apical junctions containing both the Cadherin–Catenin complex and the DLG-1–AJM-1 complex (Stone et al. 2009; Abdus-Saboor et al. 2011; Parry and Sundaram 2014; Pasti and Labouesse 2014). Apical junctions and submembranous domains are enriched for F-actin; this is particularly true of the pore cell, which has a strong F-actin signal along its autocellular junction (Parry and Sundaram 2014), and the canal cell, which has a prominent terminal web containing ACT-5/actin and cortical actin-binding proteins (Buechner et al. 1999; Gobel et al. 2004; MacQueen et al. 2005; Praitis et al. 2005) (Figure 5B). During tube development, the lumen is filled with a fibrous apical extracellular matrix (aECM) visible by means of transmission electron microscopy (TEM) (Buechner et al. 1999; Mancuso et al. 2012) (Figure 5A). In the duct and pore, this luminal matrix disappears prior to hatching and is replaced by a cuticular lining (Nelson et al. 1983).
Excretory system functions and phenotypes
One important role of the C. elegans excretory system is in osmoregulation. Long tubes of the canal cell extend throughout the length of the animal and are thought to collect excess water to be transported to the duct cell and then excreted through the pore cell in the head (Figure 2). Osmoregulatory function is inferred from morphological and molecular features of the canal cell (Figure 2, Figure 5, Table 1) and from laser ablations of the canal cell, duct, pore, or CAN neurons, which cause characteristic Clear (Clr) and rod-like larval lethal (Lvl or Let) phenotypes in which the pseudocoelom swells with accumulating fluid (Nelson and Riddle 1984; Forrester and Garriga 1997). Many mutations that perturb these cells cause a similar lethal phenotype (Yochem et al. 1997; Liegeois et al. 2007; Stone et al. 2009; Mancuso et al. 2012), while mutants with milder excretory system abnormalities often appear pale and slightly bloated and/or show increased sensitivity to osmotic challenge (Buechner et al. 1999; Wang and Chamberlin 2002; Hahn-Windgassen and Van Gilst 2009).
The Excretory Canal Cell
Canal cell structure and ultrastructure
The excretory canal cell is the largest and one of the most distinctively shaped mononucleate cells in the worm (Figure 2, Figure 3A). The cell body and nucleus are located just ventral to the posterior bulb of the pharynx. The nucleus is very large, which suggests that, as for intestinal nuclei, the chromosomes are endoreduplicated (Hedgecock and White 1985). The cell sends out four hollow tubules or “canals,” two posterior and two anterior, forming an “H-shape” (Nelson et al. 1983). The canals are closed at their tips and extend the entire length of the adult animal. The canal cell is a seamless tube—it does not have adherens junctions along most of its length, but only at its connection to the excretory duct cell and gland.
The lumen of the canals is easily observed via DIC (Figure 5C) or TEM (Figure 5B). The lumen diameter is narrowest near the canal tips and widens steadily toward the excretory canal cell body. The lumen is also wider in the posterior canals than in the anterior canals. In cross-section, the basolateral surface of the canals extends between the lateral hypodermis and the basement membrane, and the canal and hypodermal membranes merge where gap junctions are present (Figure 5B). The cytoplasmic terminal web is a thick filament-rich coating adjacent to the luminal membrane (Figure 5B).
Surrounding the lumen and filling most of the cytoplasm are a series of smaller tubes or vesicles called canaliculi (Nelson et al. 1983) (Figures 5, B and D). EM tomography has shown that the canaliculi can be separate vesicles or attached to each other and to the lumen (Khan et al. 2013) (Figure 5B). Canalicular vesicles are 50–100 nm in diameter and differ from other vesicles such as endosomes in having an electron-dense coat, which at high magnification exhibits “lollipop” structures characteristic of ATPases (Figure 5D). Canaliculi appear to store water, and regulated docking of the canaliculi with the apical membrane has been proposed as a major driver of both axial and circumferential canal lumen growth (Khan et al. 2013; Kolotuev et al. 2013; Armenti et al. 2014) (see below).
Surrounding and interspersed with the canaliculi are multiple microtubules (MTs) running along the length of the canals (Figure 5B). The canals are rich in TBG-1/γ-tubulin, which marks MT organizing centers (MTOCs) (Zheng et al. 1991), and Venus::TBG-1 is visible as myriad puncta throughout the canals (Shaye and Greenwald 2015) (Figure 6D). Outside of the canalicular region, other larger diameter vesicles, as well as mitochondria, endoplasmic reticulum (ER), and Golgi, are prominently visible by TEM (Hahn-Windgassen and Van Gilst 2009; Khan et al. 2013; Lant et al. 2015) toward the basal surface of the canals (Figure 5B).
Canal cell outgrowth. (A) Canal length extends greatly during the L1 stage. Diagrams to scale of animals with left canals shown in red. At hatch, the posterior canals extend only as far as the gonad (light gray). During the next 8 hr the canals extend to the anus (black diagonal line) and have reached full length by the beginning of L2 stage, while the animal has also lengthened considerably. (B) Transverse section of the animal near the canal cell body. The cell body (red) stretches across left and right muscle quadrants beneath the pharynx and extends canals to the hypodermis. Some basal extracellular matrices (basal lamina) are shown, incompletely, in gray. Canal arms must cross the hypodermal basal lamina to extend posteriorward alongside the hypodermis and the CAN neuron (brown). (C) Growth of the tip of a canal during wild-type L1 stage, adapted from fluorescence micrograph (Kolotuev et al. 2013). The basolateral surface of the canals precedes the end of the lumen (apical surface) by several micrometers. In addition, large varicosities of the cytoplasm appear at regular intervals along the canal during growth, while the lumen diameter is less variable. (D) Model for canal tip outgrowth, adapted from Shaye and Greenwald (2015) and Khan et al. (2013) (not to scale). EXC-6/INF2 formin is spread throughout the canal and concentrated at varicosities and especially at the canal tip, where it helps to form a large actin patch to push the membrane forward. EXC-6 is also associated with MTOCs that nucleate MTs growing predominantly toward the cell body. Actin filaments at the tip may extend to connect the basolateral surface to the terminal web of the apical surface, providing guidance for growth of that surface as well. Growth of the apical surface may occur through ERM-1, RAL-1, and exocyst (not shown; see Figure 7)-mediated fusion of canaliculi to the luminal surface, which allows water to pass through aquaporin AQP-8 into the canal lumen; increased osmotic pressure drives luminal expansion.
Lineal origin and Notch dependency of the canal cell
Notch signaling promotes left/right asymmetry in C. elegans (Priess 2005), as in other organisms (Kato 2011), and is required to generate the asymmetrically positioned canal cell (Lambie and Kimble 1991; Moskowitz and Rothman 1996) (Figure 4, A and B). The ABpl and ABpr lineages generate mostly bilaterally symmetric descendants (Sulston et al. 1983), with some exceptions that depend on signaling by the Delta-like ligand LAG-2 through the Notch receptors LIN-12 and GLP-1 and the downstream CSL [CBF1/Su(H)/LAG-1] family transcription factor LAG-1 (Lambie and Kimble 1991; Moskowitz and Rothman 1996; Priess 2005). LAG-2 is expressed by interior MSap descendants on the left side of the embryo and activates Notch signaling in various ABpl (but not ABpr) descendants. Notch signaling promotes the identity of ABplpapp (Moskowitz and Rothman 1996), which gives rise to the canal cell and a mix of other epithelial and neuronal descendants. In the absence of Notch signaling, ABplpapp adopts an ABprpapp-like identity and generates mostly neuronal descendants; no canal cell is generated. The only known downstream target of this signaling is the Hes-related transcription factor REF-1 in ABplpapp descendants, though ref-1 mutants do not have defects in this cell fate decision (Neves and Priess 2005).
Within the ABplpapp lineage, multiple asymmetric cell divisions generate the canal cell (Sulston et al. 1983) (Figure 4B). The canal cell’s grandmother, ABplpappa, generates a larger anterior daughter that survives and a smaller posterior daughter that undergoes apoptosis. In ceh-36; unc-30 double mutants, this asymmetry can be reversed (Walton et al. 2015). In pgl-1 ced-3 double mutants, the posterior daughter can survive to generate an extra canal cell (Denning et al. 2012). The next asymmetric division, of the canal cell’s mother ABplpappaa, generates a small anterior daughter that becomes the RMEV neuron and a larger posterior daughter that becomes the excretory canal cell. Since a noncanonical Wnt pathway involving WRM-1/β-catenin promotes many anterior/posterior asymmetries in the C. elegans lineage (Sawa and Korswagen 2013; Zacharias et al. 2015), this pathway is a good candidate for controlling such asymmetries relevant to canal cell identity.
Transcription factors important for canal cell differentiation
No single transcription factor has been found to specify canal cell identity, but the following are important for canal differentiation (Figure 3A):
LIN-26 is a nematode-specific C2H2 zinc finger protein important for epithelial differentiation and maintenance, whose loss leads to epidermal degeneration as well as variable defects in canal morphology (Labouesse et al. 1994, 1996). The final cell division that generates the canal partitions epithelial and neuronal fates (Figure 4B), and activation of LIN-26 in the canal is likely to be an important early step in epithelial determination.
PROS-1 (formerly called CEH-26 or RDY-3) is a Prospero-related homeodomain transcription factor (Kolotuev et al. 2013). In pros-1 mutants, there is a nearly complete failure of canal cell and lumen outgrowth, although many aspects of canal cell identity appear intact. Downstream (direct or indirect) targets of PROS-1 include the aquaporin gene aqp-8, the guanine nucleotide exchange factor gene exc-5, the germinal center kinase gene gck-3, and the intermediate filament gene ifb-1; the last appears to be a direct target (Kolotuev et al. 2013). The mammalian homolog Prox1 promotes development of the lymphatic system (Wigle and Oliver 1999), suggesting possible similarities between the canal tube and lymphatic vessels (Kolotuev et al. 2013).
CEH-6 is a POU-family homeodomain transcription factor related to mammalian Brn1, which is required for development and function of the distal tubule of the mouse kidney (Burglin and Ruvkun 2001; Nakai et al. 2003). ceh-6 mutants have canal cell and outgrowth defects similar to those of pros-1 mutants, but less severe. Downstream (direct or indirect) targets of CEH-6 include multiple transporter- or channel-encoding genes, including aqp-8, clh-4, mrp-2, nac-2, pgp-3, pgp-4, pgp-12, sulp-4, sulp-5, sulp-8, twk-36, vha-12, vha-13, and Y70G10A.3; aqp-8 and nac-2 appear to be direct targets (Mah et al. 2007; Armstrong and Chamberlin 2010).
NHR-31 is a nuclear hormone receptor related to mammalian HNF4α. NHR-31 is required for expression of vha genes encoding subunits of vacuolar ATPase and appears to restrain and coordinate both apical and basal membrane growth (Hahn-Windgassen and Van Gilst 2009).
Neural guidance cues and basement membrane steer canal outgrowth
To adopt its distinctive H-shape, the canal cell undergoes an extensive program of elongation and branching that is regulated by neural guidance cues such as netrins, Wnts, and FGFs and by other factors in the extracellular matrix such as collagens and laminins (Hedgecock et al. 1987; Buechner 2002; Oosterveen et al. 2007; Polanska et al. 2011). Once born, the canal cell begins to grow out left and right over the ventral muscle quadrants from the cell body to the lateral surface. Both left and right extensions branch, and each extends the wide posterior-directed canals as well as the smaller anterior-directed canals. These branches are visible in GFP-marked cells by the time of the threefold stage (Suzuki et al. 2001) (Figure 6A). At hatch, the posterior canals reach approximately the V3/V4 interval in the midbody. The canal arms reach their target positions at the V6/T boundary by the end of the L1 stage (Fujita et al. 2003). While extension along the hypoderm then ceases, the canals continue to lengthen to maintain their positions as the animal grows to adulthood.
The canals are positioned between the hypodermis and a basement membrane (Figure 6B). It is unknown whether this basement membrane is synthesized by the canal cell or if the canals insinuate themselves between the hypodermis and its own basement membrane. Supporting the second possibility, the canal synthesizes several matrix metalloproteases (Spencer et al. 2011) that could be used for cutting through the basement membrane. In addition to the guidance cues mentioned above, multiple other genes and pathways that affect the extent of canal outgrowth appear to function in the hypodermis or in other nearby tissues over which the canal grows (Polanska et al. 2011; Marcus-Gueret et al. 2012).
Relatively little is known about molecular pathways that direct basal canal branching, but a few mutants have been observed that exhibit extra canal arms at low penetrance. These include mutants for unc-6/Netrin or unc-5/Netrin receptor (Hedgecock et al. 1987); nck-1(mig-1)/Nck kinase (Mohamed and Chin-Sang 2011); and axl-1, an axin-like molecule (Oosterveen et al. 2007).
Cytoskeletal organization in the growing canals
Extracellular guidance cues likely regulate cytoskeletal organization to direct canal outgrowth, similar to what occurs during neurite outgrowth (Dent et al. 2011). The growing tip of the canals is wider than the canal behind it, in appearance similar to a growth cone of a neuron (Suzuki et al. 2001). Canal basal membrane outgrowth slightly precedes apical membrane outgrowth, such that the growing tips of each canal arm initially lack lumen (Kolotuev et al. 2013) (Figure 6C). The growing tips are marked by a cluster of MTOCs and by a patch of F-actin that accumulates basolaterally rather than apically (Figure 6D), which may be important for basal outgrowth (Shaye and Greenwald 2015).
Mutations in several cytoskeletal regulators result in shortened canals without disrupting the formation or structure of the lumen.
Integrins PAT-3, PAT-2, and INA-1 link laminins and other extracellular matrix factors to the cytoskeleton (Gettner et al. 1995; Baum and Garriga 1997). While null mutations in pat-3 are embryonic lethal, hypomorphs and canal mosaics show length defects (Hedgecock et al. 1987).
UNC-53, homologous to vertebrate NAV2, is an actin- and MT-binding protein that affects many anterior–posterior cell and axon migrations, as well as canal outgrowth (Stringham et al. 2002). UNC-53 acts cell autonomously and in a common pathway with UNC-71 (ADAM), MIG-10 (Lamellipodin), ABI-1 (Abelson Interactor-1), WVE-1 (WAVE), ARX-2 (Arp2), and the Rho-specific guanine nucleotide exchange activity of UNC-73 (TRIO) (Schmidt et al. 2009; Marcus-Gueret et al. 2012; McShea et al. 2013; Ghosh and Sternberg 2014; Wang et al. 2014b). The contactin RIG-6 also functions cell autonomously in the canal, potentially as part of the UNC-53 pathway (Katidou et al. 2013). Finally, PLR-1, a putative E3 ubiquitin ligase, interacts with UNC-53 to regulate outgrowth (Bhat et al. 2015).
Kinesins are motor proteins that transport vesicles along MTs (Caviston and Holzbaur 2006). UNC-116/kinesin-1 heavy chain and KLC-2/kinesin light chain are expressed in the canal, and mutation of either gene results in short canals (Patel et al. 1993; Shaye and Greenwald 2015).
The canal cell likely forms a seamless tube through a hollowing mechanism
Canal lumen formation initiates around the lima bean stage of embryogenesis, near the newly formed junction with the excretory duct cell (Stone et al. 2009; Abdus-Saboor et al. 2011; Mancuso et al. 2012) (Figure 4E, Figure 5A). Apical PAR proteins (Achilleos et al. 2010; Mancuso et al. 2012; Armenti et al. 2014) and the ezrin/radixin/moesin ortholog ERM-1 (Khan et al. 2013) are among the earliest markers of the nascent canal lumen, which suggests that recruitment of these proteins may be involved in lumen nucleation. One model consistent with existing data is that PAR proteins recruit the exocyst complex (Armenti et al. 2014) to promote vesicle docking at the canal–duct interface, while ERM-1 organizes actin to coat these vesicles and promote their fusion with the apical membrane (Khan et al. 2013), which then grows inward (Figure 1D). However, most studies have focused on postembryonic stages of canal outgrowth, and it remains unknown whether the aforementioned gene products are actually required for lumen nucleation.
Models for canal elongation and maintenance
Possible models for canal elongation and maintenance are shown in Figure 6D and Figure 7. Elongation requires ERM-1 and the exocyst, which recruit canaliculi and other vesicles to the apical membrane by interacting with vesicle-residing proteins such as AQP-8 or RAL-1 (Khan et al. 2013; Armenti et al. 2014). The vacuolar ATPase (V-ATPase) and the aquaporin AQP-8 drive water into canaliculi (Khan et al. 2013; Kolotuev et al. 2013). Canaliculi may fuse into the canal apical membrane (Kolotuev et al. 2013; Armenti et al. 2014), or they may dock only transiently to dump water into the lumen and thereby increase hydrostatic pressure to push the lumen outward (Khan et al. 2013). Other Golgi-derived or endocytic vesicles follow MT and/or actin tracks to reach their proper destinations and deliver apical or basal membrane and other cargo. In particular, EXC-1/5/9 and the CCM-3–STRIPAK complex regulate CDC-42 to promote recycling endosome trafficking for apical membrane maintenance. Actin filaments of the terminal web, linked to the apical membrane by ERM-1 (Gobel et al. 2004; Khan et al. 2013) and SMA-1/β-spectrin (Buechner et al. 1999; Praitis et al. 2005), influence trafficking and also cooperate with intermediate filaments (Woo et al. 2004; Kolotuev et al. 2013) to provide structural support and restrain lumen diameter, guiding growth in the axial direction. At the growing canal tips, apical and basal membranes are linked by cytoskeletal filaments and the formin EXC-6 to coordinate apical vs. basal outgrowth (Shaye and Greenwald 2015).
Model for vesicle trafficking in the canal cell. Speculative model of factors allowing extension and maintenance of the luminal surface. Intermediate filaments, ACT-5/actin, ERM-1, and SMA-1/βH-spectrin surround the luminal surface (Gobel et al. 2004; Praitis et al. 2005; Khan et al. 2013; Kolotuev et al. 2013), where they maintain the smooth diameter of the lumen. ERM-1 interacts with AQP-8 on canaliculi (Khan et al. 2013), and RAL-1 GTPase on canaliculi and the PAR complex on the apical surface attract the exocyst to fuse canaliculi to the apical surface (Armenti et al. 2014), which could allow the activity of vacuolar ATPase to drive water into the lumen via AQP-8 (Figure 6D, Figure 8) (Khan et al. 2013). Various marked endosomes can be viewed moving anteriorward and posteriorward throughout the length of the canals (H. Al-Hashimi and M. Buechner, unpublished results). The canal cytoplasm is rich in MTs (Shaye and Greenwald 2015), which presumably form tracks for this movement. By analogy to its role in vertebrate kidney proximal tubules (Chou et al. 2016), EXC-4/CLIC may mediate vesicle movement in growing canals. In addition, two sets of protein cascades promote trafficking through recycling endosomes: EXC-9/CRIP (Tong and Buechner 2008), EXC-1/IRG (K. Grussendorf, D. Hall, M. Buechner, unpublished results), and EXC-5/FGD (Mattingly and Buechner 2011) activate CDC-42 (Olson et al. 1996; Gao et al. 2001) to stimulate growth of actin filaments and/or transport of endosomes along those filaments. Similarly, CCM-3 and the STRIPAK complex also activate CDC-42 to promote trafficking, possibly toward or from the Golgi (Lant et al. 2015). Presumably, some feature of a weakened or growing luminal cytoskeleton activates EXC-9 and/or CCM-3 to strengthen the cytoskeleton at these points. The recycling endosomes are hypothesized to allow maintenance and/or growth of the lumen, presumably by supplying lipid and/or membrane proteins either directly to the luminal membrane, or indirectly via the canalicular vesicle membrane.
During periods of rapid canal growth, such as during the L1 larval stage or during recovery from hyperosmotic shock, basally protruding cytoplasmic bulges, variously called “pearls,” “varicosities,” or “periluminal cuffs,” form at regular intervals along the length of the canal arms (Figure 6C). Pearls are rich in organelles such as ER and Golgi (Hahn-Windgassen and Van Gilst 2009), and most docking of canaliculi to the apical membrane occurs in these pearl regions (Kolotuev et al. 2013). Osmotic stress stimulates pearl formation, canalicular docking, and aquaporin-mediated water influx into the lumen, thereby providing a mechanism to speed lumen growth in response to physiological demands (Khan et al. 2013; Kolotuev et al. 2013). This proposed role of aquaporin-induced water influx is similar to that of blood flow in stimulating lumen growth in vascular capillaries (Herwig et al. 2011).
These models are based on studies of the following groups of genes, which are important for canal lumen growth, shape, or maintenance. Only mutants with the most severe canal defects are lethal, whereas many mutants are viable but have cystic lumens (Exc phenotype).
Gene products required for most or all lumen growth
There are only a few known gene products in this category, and their analysis so far has been limited to larval stages.
PROS-1: The transcription factor PROS-1 upregulates many genes important for canal lumen growth, including the intermediate filament gene ifb-1 (Kolotuev et al. 2013) (see above).
ERM-1: ERM proteins are membrane-actin linkers with broad roles in cortical membrane organization (Neisch and Fehon 2011) and intracellular tubulogenesis (JayaNandanan et al. 2014; Jiang et al. 2014). ERM-1 localizes near the apical membrane and is essential for terminal web organization and for both apical and basal canal outgrowth (Khan et al. 2013) (Figure 6D, Figure 7). ERM-1 is necessary and sufficient to recruit canaliculi containing the Aquaporin AQP-8 to the apical membrane (Figure 6D), but also has AQP-8-independent functions, which may involve recruiting other classes of vesicles (Khan et al. 2013).
RAL-1 and the exocyst: The small GTPase Ral helps direct assembly of the exocyst, a multiprotein complex that tethers exocytic vesicles to the plasma membrane for subsequent SNARE-mediated membrane fusion (Liu and Guo 2012). RAL-1 on canaliculi was proposed to interact with the exocyst on the apical membrane to promote canalicular fusion and lumen growth (Armenti et al. 2014) (Figure 7). The canal lumen appears short and misshapen in ral-1 or sec-5 zygotic mutants, and nearly absent in ral-1 maternal-zygotic mutants. Conversely, ectopic lumen pockets form under conditions of RAL-1 overexpression. RAL-1 may also play a role in production or stabilization of canaliculi, since ral-1 mutants have severely reduced numbers of canaliculi, and some canaliculi appear aberrantly docked at the basal plasma membrane.
Cytoskeletal factors and the terminal web
The cytoskeleton helps direct lumen growth to appropriate locations and provides structural support to maintain lumen shape and integrity.
ACT-5 (actin) is expressed almost exclusively in the terminal webs of the canal cell and intestine (MacQueen et al. 2005) (Figure 6D, Figure 7). In erm-1(RNAi) conditions that slow lumen formation, ACT-5 coats large-diameter vesicles as they coalesce (Khan et al. 2013). No phenotype has yet been attributed to act-5 loss, probably due to redundancy with other actin paralogs, but act-5 genetically interacts with other genes such as erm-1 and sma-1 (Gobel et al. 2004).
SMA-1 is homologous to the F-actin cross-linker βΗ-spectrin (Bennett and Healy 2008). SMA-1 is located at the apical membrane of the canal cell (Praitis et al. 2005) and is needed to maintain the terminal web at the apical membrane (Buechner et al. 1999) (Figure 6D, Figure 7). Strong sma-1 mutants often exhibit a canal lumen that appears like a very large sac, a single cyst surrounded by myriad canalicular vesicles. sma-1 genetically interacts with both erm-1 and act-5 for this phenotype (Gobel et al. 2004). sma-1 mRNA is bound at its 3′ UTR by EXC-7, a member of the ELAV/Hu family of RNA-binding proteins, which regulate RNA stability and translation (Fujita et al. 2003; Simone and Keene 2013). Null exc-7 mutants have the same canal phenotype as does mild knockdown of sma-1 (Buechner et al. 1999; Fujita et al. 2003). These data suggest a model in which EXC-7 binding stabilizes sma-1 mRNA to be carried to the canal distal tips to be translated, analogous to the role of Drosophila ELAV and mammalian Hu proteins in stabilizing neural mRNAs (Bronicki and Jasmin 2013; Simone and Keene 2013).
EXC-6 is homologous to human INF2, a formin with both actin polymerization and MT-binding activities (Shaye and Greenwald 2015). Human mutations in INF2 cause inherited forms of the kidney disease focal segmental glomerulosclerosis (Brown et al. 2010) and can also cause Schwann cell actinopathy (Mathis et al. 2014). EXC-6 associates with MTs and affects both actin localization and MT dynamics at the growing tips of the canal arms (Shaye and Greenwald 2015) (Figure 6D). exc-6 mutants have short canals with multiple lumens (Buechner et al. 1999), consistent with EXC-6 promoting basal membrane growth and restricting apical membrane growth to the leading edge.
TBB-2 is a β-tubulin expressed in the canals (Spencer et al. 2011). tbb-2 mutants exhibit a similar canal phenotype to that of exc-6, while tbb-2 gain-of-function mutations suppress the effects of exc-6 mutation (Shaye and Greenwald 2015).
Intermediate filament proteins in the cytoplasm provide mechanical strength to tissues (Carberry et al. 2009). IFB-1 is located at the periapical region of the canal and is required for canal outgrowth and integrity (Woo et al. 2004; Kolotuev et al. 2013). IFB-1 has two protein isoforms, IFB-1A and IFB-1B; both are expressed in the canal cell, but IFB-1B appears more critical for canal cell function (Woo et al. 2004). IFB-1 forms heterodimers with multiple other intermediate filament proteins, including IFA-1/2/3/4 (Karabinos et al. 2003). At least two of these, IFA-1 and IFA-4, are also expressed in the excretory canal cell (Karabinos et al. 2003).
Regulators of recycling endosome trafficking
Recycling endosomes are important for later stages of canal outgrowth and for maintenance of the terminal web. The relevant cargo in these endosomes, or from where it originates, is unknown, but a reasonable model is that apical membrane proteins and lipids are being continuously endocytosed and recycled, as has been shown in many epithelia (Eaton and Martin-Belmonte 2014). Recycling endosomes may return material directly back to the apical membrane or may traffic material to the Golgi for subsequent sorting and exocytosis (Figure 7).
CDC-42 is a Rho-family GTPase that plays widespread roles in apical cell polarity and cytoskeletal organization, in part by regulating vesicular transport to and from the Golgi network (Harris and Tepass 2010; Park et al. 2015). Both in C. elegans and mammalian cells, CDC-42 is colocalized with the recycling endosome marker RME-1 and promotes retrograde transport of certain cargoes from recycling endosomes to Golgi (Balklava et al. 2007; Bai and Grant 2015). In the canal cell, CDC-42 is critical for lumen extension and maintenance and appears to act via effects on endocytic recycling (Mattingly and Buechner 2011; Lant et al. 2015). CDC-42 activity is regulated by at least two sets of gene products, EXC-1/5/9 and CCM-3/STRIPAK.
cdc-42 dominant-negative mutants or cdc-42(RNAi) animals have shortened canals with lumen dilations or cysts that are depleted for Golgi and recycling endosome markers (Mattingly and Buechner 2011; Lant et al. 2015), consistent with a role for CDC-42 in endocytic recycling (Figure 7). Expression of constitutively active CDC-42 results in a convoluted lumen phenotype, in which a long lumen of near-normal diameter forms within a canal cell of greatly reduced basal surface area (Mattingly and Buechner 2011), which suggests that CDC-42 also influences the balance between apical-directed and basolateral-directed trafficking.
EXC-1, EXC-5, and EXC-9: EXC-1 is homologous to the mammalian immunity-related GTPase (IRG) family, a set of dynamin-related GTPases that are found at intracellular membrane compartments, some of which target parasite-containing vacuoles for autophagy and destruction (K. Grussendorf and M. Buechner, unpublished results) (Howard 2008; Petkova et al. 2012). EXC-5 is homologous to the mammalian FGD family of guanine nucleotide exchange factors (GEFs), regulators of CDC-42 activity that are mutated in faciogenital dysplasia (FGD) and Charcot-Marie-Tooth (CMT) disease type 4H (Olson et al. 1996; Gao et al. 2001; Suzuki et al. 2001; Delague et al. 2007). FGD5 also is required for capillary formation in culture (Kurogane et al. 2012). EXC-9 is homologous to mammalian cysteine-rich intestinal protein (CRIP), a cytoplasmic LIM-domain protein of unclear biochemical function (Lanningham-Foster et al. 2002; Tong and Buechner 2008); a vertebrate homolog is expressed in the pronephros (Hempel and Kuhl 2014) and a C. elegans homolog, VALV-1, appears to aid in the function of several epithelial valves (Tong and Buechner 2008).
exc-1, exc-5, or exc-9 loss-of-function mutants have similar phenotypes in which canal length is reduced and the canal lumen becomes progressively dilated or cystic starting in late embryogenesis (Buechner et al. 1999). In exc-5 mutants, cyst regions show reduced apical accumulation of F-actin, and the leading edge of the canal arms also lacks its usual basolateral F-actin patch, consistent with actin organization being a key downstream readout of EXC-5 function (Shaye and Greenwald 2015). The cystic regions are enriched for the early endosome marker EEA-1, but depleted of the recycling endosome marker RME-1, while Golgi markers remain unchanged (Mattingly and Buechner 2011). Overexpression of exc-1, exc-5, or exc-9 all result in a convoluted-lumen phenotype (Suzuki et al. 2001; Tong and Buechner 2008; Mattingly and Buechner 2011; K. Grussendorf and M. Buechner, unpublished results) similar to that of constitutive CDC-42 activation. Epistasis experiments based on the loss-of-function and overexpression phenotypes indicate that EXC-9 acts upstream of EXC-1, and both act upstream of EXC-5. Yeast-2-hybrid results additionally indicate that EXC-9 binds to wild-type EXC-1. These results suggest that EXC-1/5/9 act upon CDC-42 in a common pathway for apical membrane recycling (Mattingly and Buechner 2011; K. Grussendorf, C. Trezza, A. Salem, B. Mattingly, H. Al-Hashimi, D. Kampmeyer, L. Khan, V. Göbel, D. Hall, B. Ackley, M. Buechner, unpublished results), but do not affect trafficking to Golgi (Figure 7).
CCM-3 is the worm homolog of CCM3/PDCD10, a scaffold protein that is part of several protein complexes, including the striatin-interacting phosphatase and kinase complex (STRIPAK) (Hwang and Pallas 2014; Lant et al. 2015). A key role of CCM-3 is to recruit GCKIII family kinases to promote Golgi assembly and polarization (Fidalgo et al. 2010; Kean et al. 2011). Mammalian mutations in CCM-3 cause an autosomal-dominant form of cerebral cavernous malformation (CCM), a neurovascular disease associated with swelling and rupture of brain capillaries (Draheim et al. 2014). Mutants for ccm-3, gck-1, cash-1/striatin, farl-11/STRIP, or mrck-1/MRCK (myotonic dystrophy-related CDC-42-binding kinase) have shortened canals and a cystic lumen phenotype similar to that of exc-1, exc-5, or exc-9 mutants, although the phenotype originates later in development (Lant et al. 2015). The cyst regions in ccm-3 mutants do not accumulate early endosome markers, but are depleted for Golgi, ER, and recycling endosome markers and are enriched in canaliculi and other abnormal vesicles. ccm-3 mutation also results in reduced activation of CDC-42. CCM-3 is therefore concluded to act with GCK-1 and other STRIPAK homologs to activate CDC-42 (Lant et al. 2015). Based on the differences in early endosome and Golgi marker accumulation, CCM-3 and STRIPAK may act on CDC-42 at a later step of recycling than do EXC-1/5/9 (Figure 7).
Other potential trafficking regulators
The following genes also may influence trafficking, but the specific step(s) affected is unknown.
EXC-4 is homologous to mammalian chloride intracellular channel (CLIC) proteins (Berry et al. 2003). Mutants have severely truncated canals with cystic lumens (Buechner et al. 1999). Vertebrate CLIC proteins have been proposed to act in many capacities, including as: chloride channels (allowing counter-ion transport, during vesicle acidification in hollowing endothelial cells) (Ulmasov et al. 2009); enzymes (homologous to GST) (Littler et al. 2010); cytoskeletal adaptors (associating with ERM proteins) (Jiang et al. 2014); or a combination of these, since the proteins can convert between cytosolic and integral membrane forms (Littler et al. 2010; Jiang et al. 2014). Mouse CLIC4 is needed for endolysosome formation during Cdc42-dependent lumen formation in kidney proximal tubule cells, prior to becoming progressively enriched at the luminal surface (Chou et al. 2016), which may represent a conserved role for EXC-4/CLIC4 in promoting vesicle trafficking during luminogenesis (Figure 7). C. elegans EXC-4 is constitutively membrane-associated, acts as an ion channel at pH below 6 (Littler et al. 2010), is required during early lumen elongation (Berry et al. 2003), and is found at the luminal membrane of adult canals (Berry et al. 2003; Berry and Hobert 2006). The site of its action in excretory cell luminogenesis has not been determined; an attractive hypothesis is that EXC-4 might acidify vesicles for transport to promote trafficking-mediated lumen elongation (Figure 7). An alternative, and not mutually exclusive, model is that EXC-4 at the luminal surface provides counter-ions to V-ATPase-driven protons that promote osmotic-mediated lumen elongation (Kolotuev et al. 2013), and this function is also required for lumen maintenance (Figure 8).
Figure 8Speculative model of osmoregulatory function of the excretory canals. Excess organismal liquid may flow from the hypodermis into the canals via the many gap junctions between the two tissues and/or may cross the basement membrane from the pseudocoelom to the canals. Canalicular vesicles contain vacuolar ATPase and aquaporin AQP-8 (Khan et al. 2013; Kolotuev et al. 2013). Depending on environmental osmolarity, canaliculi may store water or transport water into the lumen. ATPase pumps protons into the lumen, which may attract a counter-ion into the lumen via EXC-4/CLIC or another associated channel (Berry et al. 2003; Berry and Hobert 2006; Ulmasov et al. 2009). The sodium-proton exchanger (NHX-9) is hypothesized to prevent acidification of the lumen, increasing osmolarity of NaCl sufficient to draw water into the lumen of the channel. Water then flows through the excretory duct and pore to the outside environment. In this model, the duct cuticle and membrane are impermeable to water. Pumps and channels in the duct might retrieve the NaCl to prevent its loss; or pumps and channels in the hypodermis could replenish Na+ and Cl− from the environment.
DAF-6 and CHE-14 are homologs of Patched and Dispatched, respectively, transmembrane proteins involved in Hedgehog signaling, endocytosis, and polarized secretion (Tukachinsky et al. 2012; D’Angelo et al. 2015). DAF-6 and CHE-14 are localized to apical membranes and function redundantly to promote canal lumen extension (Perens and Shaham 2005). Although C. elegans does not have a standard Hedgehog (Hh) signaling pathway, it does contain various “Warthog” and “Groundhog” proteins related to Hedgehog (Burglin 1996), several of which are expressed in the canal cell (Hao et al. 2006). Based on molecular identities, DAF-6 and CHE-14 might regulate endocytosis or secretion of specific apical factors, although it is not known if the Hh paralogs are relevant cargoes.
RDY-2 is a nematode-specific tetraspan protein located in apical membranes of the canal, duct, and pore cells (Liegeois et al. 2007; H. Gill and M. Sundaram, unpublished results). In addition to canal defects, rdy-2 mutants share several additional phenotypes with vha-5 mutants, including defects in amphid channel morphology and hypodermal alae formation, which indicates a possible role in apical trafficking.
Canalicular factors
Canaliculi docking with the apical membrane presumably increase lumen growth (Figure 6D), but the origin of canaliculi is unknown and further studies are needed to determine if they contribute apical membrane or only water. Besides RAL-1 (above), the following gene products are localized to canaliculi:
AQP-8 is an aquaporin, a multipass transmembrane protein that forms pores or channels permeable to water but impermeable to ions and other charged molecules (King et al. 2004). AQP-8 is located on canalicular vesicles and may draw water into those vesicles, which is later released into the lumen upon ERM-1-dependent docking with the apical membrane (Khan et al. 2013) (Figure 6D, Figure 7, Figure 8). Significant expansions in lumen diameter occur when ERM-1 or AQP-8 are overexpressed. aqp-8 null mutants have only a mildly shortened canal lumen, however, which suggests that aquaporin-mediated water flux may not be the key driver for lumen growth under stable physiological conditions.
Vacuolar ATPase (V-ATPase) is a proton pump important for acidification of exocytic and endocytic vesicles (Lee et al. 2010). It is composed of two multisubunit sectors: V0 embedded in the membrane and the ATP-utilizing V1 proton pump (Lee et al. 2010). The V0 sector also has a V1-independent function in apical secretion (Liegeois et al. 2006). C. elegans V-ATPase subunits (encoded by vha genes) are highly expressed in the canal cell and excretory duct (Liegeois et al. 2006, 2007; Hahn-Windgassen and Van Gilst 2009; Knight et al. 2012). The V0-a subunit VHA-5 was identified by means of immunoEM in canalicular vesicles (Kolotuev et al. 2013). In vha-5 mutants, pearl formation and canaliculi density are increased, and multiple separate lumens form (Liegeois et al. 2006, 2007; Hahn-Windgassen and Van Gilst 2009). One interpretation of this phenotype is that VHA-5 promotes efficient docking or fusion of canalicular vesicles with the mature apical membrane to create a central lumen. Another interpretation is that the V-ATPase plays a broader role in inhibiting both apical and basal membrane production and canal cell growth.
Ion transport and fluid flow
Studies of AQP-8 (above) showed that water influx can drive canal lumen expansion (Khan et al. 2013). Flow from the canal lumen into the lumens of the adjacent duct and pore tubes is also important for lumen shape, as mutations in genes such as let-4 and let-653, which are not expressed in the canal cell, but affect duct and pore tube patency, cause severe dilations and cysts in the canal (Buechner et al. 1999; Mancuso et al. 2012) (H. Gill, J. Cohen and M. Sundaram, unpublished results) (Figure 9B). Finally, ion channels or channel modulators also affect lumen shape, possibly by affecting fluid flow (Figure 8).
Excretory duct and pore structure and ultrastructure. (A) Early L1 larva expressing apical junction marker AJM-1::GFP (Koppen et al. 2001). The duct and pore are located just anterior and ventral to the posterior bulb of the pharynx (ph). Arrow, pore cell junctions; arrowhead, junction between duct and canal. (B) let-653(cs178) early L1 larva. The spherical dilation near the duct–canal junction is characteristic of mutants with a block in duct or pore lumen continuity (Stone et al. 2009; Mancuso et al. 2012). (C) Duct and G1 pore marked with dct-5promoter::mCherry (cell bodies) and AJM-1::GFP (apical junctions). (D) Duct marked with lin-48promoter::mRFP (cell body) and LET-653::GFP (lumen, line) (H. Gill, J. Cohen and M. Sundaram, unpublished results). Photos in C and D are courtesy of Fabien Soulavie (University of Pennsylvania). (E) TEM of duct lumen in an L4 hermaphrodite. Arrowheads point to membrane stacks (lamellae) that surround the apical membrane. Line indicates lumen, surrounded by darkly staining cuticle, which has detached from the apical membrane in this specimen. TEM image is courtesy of John Sulston (Medical Research Council). (F) TEM of duct and G1 pore in a 420-min (comma stage) embryo. Lines indicate fibrous apical ECM in the lumen. TEM image is courtesy of Shai Shaham (The Rockefeller University).
WNK-1, GCK-3, and CLH-3: Mammalian with-no-lysine (WNK) serine/threonine kinases are critical regulators of ion transport in the kidney and other tissues and are mutated in familial hypertension syndromes (McCormick and Ellison 2011; Alessi et al. 2014). One important set of Wnk substrates are GCK-VI kinases, which in turn regulate various ion channels (McCormick and Ellison 2011; Alessi et al. 2014). WNK-1 is the sole Wnk-family kinase in C. elegans, and it positively regulates the GCK-3 kinase, which negatively regulates CLH-3 chloride channel activity (Denton et al. 2005; Hisamoto et al. 2008). This pathway appears to function in multiple tissues to affect cell volume (Denton et al. 2005; Choe and Strange 2007). wnk-1 mutants have shortened canals, while gck-3 mutants exhibit short and wide canal lumens despite normal basal outgrowth, consistent with a role for ion channel activity in modulating canal lumen growth (Hisamoto et al. 2008). However, clh-3 mutants have normal canal morphology, and cell autonomy of wnk-1 or gck-3 canal phenotypes has not been rigorously established.
Future questions
Many questions remain about how the canal cell is formed and shaped. For example, how is the intracellular apical domain initially established and connected to that of the duct tube? What is the origin of canaliculi or other vesicles that contribute to the apical membrane? On what specific vesicle compartments do the EXC and CCM-3 gene products act? Studies of earlier embryonic stages and live imaging of the relevant vesicular compartments will be important for answering these questions. Most of the key players involved in canal outgrowth have homologs in vertebrates, many of which affect tubule shape and integrity in the vascular system, kidney, or glial Schwann cells, so a better understanding of how these genes function in the canal cell is likely to lead to broadly relevant insights.
The Excretory Duct and Pore Cells
Duct and pore tube structure and ultrastructure
The excretory duct and pore are smaller unicellular tubes that connect the canal cell lumen to the exterior of the animal to allow fluid excretion and secretion (Figure 3, B–D, Figure 4, Figure 9). The duct tube channel is essential for viability: in mutants lacking a duct or having a discontinuous duct lumen, the upstream canal lumen becomes grossly dilated (Figure 9B), and larvae die with a turgid, rod-like morphology (Stone et al. 2009; Mancuso et al. 2012). On the other hand, mutants can survive quite well without a pore cell, because the duct can connect directly to the external hypodermis to provide an open channel for excretion (Abdus-Saboor et al. 2011).
The mature duct is an elongated, seamless tube with an asymmetric shape and a narrow cellular process that connects it to the pore (Figure 3B, Figure 9, C and D) (Nelson et al. 1983; Stone et al. 2009). The duct lumen takes a looping path through the cell (Figure 9D). The duct apical membrane is attached to surrounding parallel stacks of lamellar membranes (Figure 9E) that greatly increase its potential surface area and have been speculated to facilitate selective reuptake of ions (Nelson et al. 1983). The vacuolar ATPase subunit VHA-5 localizes to these lamellar stacks (Kolotuev et al. 2010).
The pore has an autocellular junction, a relatively simple, conical shape, and few or no lamellae (Nelson et al. 1983; Stone et al. 2009) (Figure 3, C and D, Figure 9C). The pore tube further has the amazing property that its structure and function is preserved through several changes in pore cell identity (Figure 4G; see The G1-to-G2 Pore Swap below).
EGF-Ras-ERK signaling specifies the duct vs. G1 pore fate
The duct and G1 pore are another example of left/right asymmetry within the ABp lineage (Figure 4, A and B) (Sulston et al. 1983), but in this case the asymmetry is due to EGF-Ras-ERK signaling (Abdus-Saboor et al. 2011). Asymmetry of EGF-Ras-ERK signaling depends on the earlier, Notch-dependent asymmetry in canal cell position, because the canal cell is a major source of LIN-3/EGF (Figure 10A).
EGF-Ras-ERK signaling promotes duct vs. G1 pore fate. (A) The biased competition model postulates that asymmetry in canal position results in the left member of the duct/G1 equivalence group receiving more canal-derived LIN-3/EGF signal and therefore becoming the duct cell (Abdus-Saboor et al. 2011). An unknown lateral inhibitory mechanism (Sulston et al. 1983) prevents the cell on the right from receiving or responding to the EGF signal. (B) The EGF-Ras-ERK pathway acts via LIN-1/Ets and EOR-1 to promote duct fate (Howard and Sundaram 2002; Abdus-Saboor et al. 2011). Asterisks indicate potential ERK-dependent phosphorylation (Jacobs et al. 1998; Howell et al. 2010). (C and D) Reduced EGF-Ras-ERK signaling causes a duct-to-G1 pore fate transformation. Constitutive signaling causes a G1 pore-to-duct fate transformation; the two ducts fuse to make a binucleate duct (Yochem et al. 1997; Abdus-Saboor et al. 2011). (E) In mosaics lacking let-60/Ras activity in the ABpl lineage, the duct and G1 pore progenitors switch fates (Yochem et al. 1997). (F) Physical or genetic ablation of the canal cell prevents duct and G1 pore tubulogenesis (Abdus-Saboor et al. 2011).
The logic behind the duct vs. G1 pore fate decision is similar to that of tip vs. stalk cell fate decisions in more complex, branched tubular organs such as the vertebrate vascular system, lung, or kidney, where cells compete for a tip cell fate in response to growth factor signaling (Siekmann et al. 2013). The duct and G1 pore are left/right lineal homologs (Figure 4B) that are initially equivalent (Sulston et al. 1983). These cells are born in lateral positions in the embryo and migrate toward each other and the excretory canal cell during ventral enclosure (Figure 4A). As the duct and G1 pore progenitors migrate ventrally, both cells express LET-23/EGFR and appear to compete for access to the LIN-3-secreting canal cell (Abdus-Saboor et al. 2011) (Figure 10A). Whichever cell “wins” this competition becomes the duct and adopts a canal-proximal position, sinking dorsally into the interior of the embryo; the remaining cell becomes the G1 pore and adopts a more ventral position on the outside surface of the embryo. In wild-type embryos, the competition is biased such that the left precursor always wins and becomes the duct, probably because it starts out closer to the canal cell and thus gets earlier and/or higher levels of LET-23/EGFR stimulation. Signaling of LET-23/EGFR through LET-60/Ras and MPK-1/ERK is both necessary and sufficient for the duct fate (Figure 10B). In the absence of signaling, both cells adopt a G1 pore fate, whereas in the presence of constitutive signaling, both cells adopt a duct fate (Yochem et al. 1997; Abdus-Saboor et al. 2011) (Figure 10, C and D).
In wild-type animals, there is always one duct and one G1 pore, but if the presumptive duct is removed via laser ablation, the presumptive G1 pore instead adopts a duct fate (Sulston et al. 1983). Furthermore, in mosaic animals, loss of LET-60/Ras from the ABpl lineage, but not ABpr lineage, causes these cells to switch fates: the presumptive ABpl-derived duct adopts the G1 pore fate, and the presumptive G1 pore adopts the duct fate (Yochem et al. 1997) (Figure 10E). Together, these data show that the duct prevents G1 from also adopting a duct fate through lateral inhibition.
Lateral inhibition presumably prevents G1 from sensing or responding to LIN-3/EGF signaling (Figure 10A), but the molecular basis of the inhibition is not known. Unlike many other lateral inhibition mechanisms in C. elegans (Sundaram 2005), or in vertebrate tip vs. stalk cell decisions (Siekmann et al. 2013), Notch signaling does not appear to be involved; instead, morphological data suggest that the duct physically prevents G1 from contacting the canal cell and in this way reduces its access to the LIN-3/EGF ligand (Abdus-Saboor et al. 2011).
Ras signaling is required at or before the 1.5-fold stage to specific duct identity, but has additional roles beyond this time. Experiments with a temperature-sensitive (ts) allele of sos-1 (encoding the Ras guanine nucleotide exchange factor) (Chang et al. 2000) showed that upshifting to restrictive temperature after the 1.5-fold stage does not perturb duct identity, but causes other abnormalities in duct lumen morphology and integrity (Abdus-Saboor et al. 2011). One later requirement for EGF-Ras signaling is in proper execution of the G1-to-G2 pore swap (Parry and Sundaram 2014) (see below).
Transcription factors important for duct or G1 pore differentiation
The following transcription factors are important for duct and/or G1 differentiation (see also Figure 3, B and C, Figure 10B):
CEH-36/Otx and UNC-30/Pitx are related bicoid-type homeodomain factors expressed in the ABp(l/r)p lineages that give rise to the canal, duct, and G1 pore cells (Walton et al. 2015). CEH-36 and UNC-30 function redundantly to upregulate mls-2 to promote proper migration of the duct and pore cells to the ventral midline.
MLS-2 is an Hmx-family homeodomain transcription factor expressed in the ABp(l/r)paaa lineages that give rise to the duct and G1 pore (Abdus-Saboor et al. 2012). After the final set of divisions in this lineage, MLS-2 protein disappears from the DB1 and DB3 neurons, but remains present in the duct and G1 pore cells during their migration and early differentiation. In mls-2 mutants, the duct and pore cells show variable defects in migration to the ventral midline and in final tube shape (Abdus-Saboor et al. 2012; Walton et al. 2015). MLS-2 cooperates with Ras signaling to express duct-specific genes such as lin-48/Ovo (Abdus-Saboor et al. 2012). MLS-2 is also important for glial differentiation in C. elegans (Yoshimura et al. 2008), suggesting molecular similarities between duct/pore and glial tube formation.
LIN-26, the general epithelial C2H2 factor, is expressed in both the excretory duct cell and G1 pore, and its loss leads to variable defects in duct and pore morphology (Labouesse et al. 1994, 1996). As for the canal cell (see above), the final division that generates the duct and G1 pore partitions epithelial from neuronal cell fates (Sulston et al. 1983) (Figure 4B). Expression of lin-26 in the duct and G1 is likely to be one important early step in epithelial determination.
NHR-23 and NHR-25 are nuclear hormone receptors expressed in most external (cuticle-lined) epithelia, including the duct and pore (Gissendanner and Sluder 2000; Kostrouchova et al. 2001; Murray et al. 2012; M. V. Sundaram, unpublished data). In the epidermis, NHR-23 and NHR-25 are required for cyclical expression of collagens and other secreted ECM factors, and for proper molting (Gissendanner and Sluder 2000; Hayes et al. 2006; Kouns et al. 2011). These transcription factors are likely to play similar roles in the duct and pore.
LIN-1 and EOR-1 are two MPK-1/ERK targets that, in the presence of Ras signaling, function redundantly to promote excretory duct vs. G1 pore fate specification (Figure 10B). LIN-1 is an Ets transcription factor related to mammalian Elk1 (Beitel et al. 1995; Jacobs et al. 1998). EOR-1 is a BTB-Zinc Finger protein related to mammalian PLZF (Howard and Sundaram 2002; Howell et al. 2010). eor-1 lin-1 double mutants resemble let-60 ras mutants and have two G1 pore-like cells and no duct (Abdus-Saboor et al. 2011). In the absence of Ras signaling, LIN-1 also has an opposing function to inhibit the duct fate (Figure 10B). lin-1 single mutants resemble let-60(gf) mutants and have a binucleate duct cell and no G1 pore (Abdus-Saboor et al. 2011).
LIN-48 is a C2H2 zinc finger transcription factor related to Drosophila and mammalian Ovo (Johnson et al. 2001), which play roles in epithelial differentiation and shaping (Menoret et al. 2013; Roca et al. 2013; Aue et al. 2015). It is expressed continuously in the duct from mid–late embryogenesis through adulthood and is therefore a convenient duct fate marker. Ras signaling and MLS-2 cooperate to stimulate lin-48 expression in the duct (Abdus-Saboor et al. 2012). lin-48 mutants have subtle changes in pore position, possibly due to changes in duct shape (Wang and Chamberlin 2002). Interestingly, C. briggsae lacks lin-48 duct expression, and its pore cell is located posterior to the position of the C. elegans pore; forced expression of lin-48 in C. briggsae moves its pore cell anteriorly (Wang and Chamberlin 2002, 2004).
CES-2 and ATF-2 are bZIP factors related to the mammalian proline- and acid-rich (PAR) family of bZIP transcription factors (Metzstein et al. 1996). CES-2 and ATF-2 function are required for correct timing of lin-48 expression, and mutants in either ces-2 or atf-2 exhibit subtle changes in pore position, similar to the phenotype of lin-48 mutants (Wang and Chamberlin 2002; Wang et al. 2006).
The duct and G1 pore form tubes through a wrapping mechanism
Following ventral migration and contact with each other and with the canal cell, and concomitant with their cell fate decisions, the duct and G1 pore cells form tubes with autocellular junctions (Stone et al. 2009; Abdus-Saboor et al. 2011) (Figure 1, B and C, Figure 4E). Mechanisms of duct and G1 pore wrapping are not understood, but wrapping is closely coupled to de novo junction formation.
The excretory canal cell is required for duct cell and G1 pore cell wrapping and stacking (Abdus-Saboor et al. 2011). Removal of the canal cell by means of laser ablation or Notch mutation prevents tube formation; the duct and G1 pore still epithelialize but do not form tubes, and instead take adjacent positions and adopt an epidermal-like morphology (Figure 10F). Partial reduction of let-60/Ras activity (in hypomorphic mutants) can also lead to adjacent epithelial cells, only one of which forms a tube (Abdus-Saboor et al. 2011), which suggests that Ras signaling also influences the wrapping process.
AFF-1-dependent duct autofusion converts an autocellular tube to a seamless tube
Although the duct initially has an autocellular junction, this junction disappears at or soon after the 1.5-fold stage (Figure 4E). Autocellular junction removal requires the plasma membrane fusogen AFF-1 (Stone et al. 2009). AFF-1 is both necessary and sufficient for several cell–cell fusions that generate syncytia (Sapir et al. 2007). In the case of the duct, AFF-1 mediates autofusion to convert a seamed autocellular tube into a seamless tube (Figure 1C). The junctions connecting the duct to the adjacent pore and canal cells are not affected by the autofusion.
EGF-Ras-ERK signaling promotes aff-1 expression to promote duct autofusion (F. Soulavie and M. V. Sundaram, unpublished results). Normally, only the duct expresses aff-1 and autofuses, whereas the G1 pore retains its autocellular junction (Figure 4). In the absence of signaling, neither cell expresses aff-1, and both retain their autocellular junctions (Figure 10C). In mutants with constitutive Ras signaling, both cells express aff-1, and the two cells fuse to make a binucleate duct (Abdus-Saboor et al. 2011) (Figure 10D).
Luminal matrix maintains lumen shape and tube integrity during morphogenesis
Subsequent to tubulogenesis and autofusion, the duct elongates and adopts its characteristic asymmetric looping shape (Figure 4F). The pore also elongates, but to a much lesser degree. Molecular pathways involved in duct cell and lumen elongation have not yet been described. erm-1 does not appear to be expressed or have phenotypes in the duct (Gobel et al. 2004; Mancuso et al. 2012), so the mechanisms of lumen growth and maintenance may differ significantly between the excretory duct and canal cells.
Duct morphogenesis occurs in the context of a fibrous luminal matrix that is visible via transmission electron microscopy (Mancuso et al. 2012) (Figure 9F). At the completion of morphogenesis, this luminal matrix disappears, and the duct and G1 pore secrete a luminal cuticle (Figure 9E) that is contiguous with the epidermal cuticle that covers the outside of the animal. This cuticle is shed and resynthesized at each larval molt (Nelson et al. 1983). The luminal matrix and rigid cuticle in the duct and pore may serve analogous functions to the terminal web in the canal cell, to maintain apical domain shape and integrity. Several mutations that disrupt luminal matrix organization lead to duct lumen breaks and dilations (Figure 9B) and to physical separation of the duct and G1 pore cells around the time of hatch (Mancuso et al. 2012; H. Gill, J. Cohen and M. Sundaram, unpublished results). Proteins involved in duct luminal integrity and shape include:
LET-653 is a Zona Pellucida (ZP)-domain and mucin-related glycoprotein (Jones and Baillie 1995). ZP proteins form fibrils and contribute to many apical matrices (Plaza et al. 2010), including luminal matrices of the Drosophila tracheal system (Jazwinska et al. 2003), mammalian vascular system (Govani and Shovlin 2009), and kidney (Rampoldi et al. 2011). Mucins form gel-like protective barriers within many tubular epithelia (Johansson et al. 2013; Lillehoj et al. 2013). LET-653 is one key component of the early luminal matrix in the duct and pore (H. Gill, J. Cohen and M. Sundaram, unpublished results).
LET-4 and EGG-6 are extracellular leucine-rich repeat only (eLRRon) proteins (Mancuso et al. 2012), part of a family of secreted and transmembrane glycoproteins with diverse roles, including collagen fibril assembly, synaptic adhesion, and signaling regulation (Brose 2009; Merline et al. 2009). LET-4 and EGG-6 are located in or near the duct and pore apical membrane and are predicted to extend their LRR domains into the lumen. let-4 and egg-6 mutants have defects similar to those of let-653 mutants along with widespread defects in apical ECM organization and cuticle barrier function, which suggests that these transmembrane proteins may interact with luminal matrix factors and help organize matrix deposition or cross-linking.
LPR-1 is a lipocalin (Stone et al. 2009); these are cup-shaped secreted proteins that bind lipophilic cargoes and have a broad range of functions related to cargo transport, sequestration, and signaling (Flower 1996; Sun and Kawaguchi 2011; Paragas et al. 2012). lpr-1 mutants strongly resemble let-653, let-4, and egg-6 mutants, but lpr-1 is capable of acting cell nonautonomously when provided from outside of the excretory system (Stone et al. 2009), and therefore may act indirectly to affect membrane or matrix integrity.
Future questions
The duct and pore provide opportunities to study important aspects of tube development such as molecular control of tube wrapping and autofusion, roles of receptor tyrosine kinase signaling in tube shaping, and the role of luminal matrix in tube shaping and integrity. Furthermore, because the duct is a permanent epithelial cell, while the G1 pore undergoes reprogramming, these cells also provide a model for understanding the factors that determine fate plasticity.
The G1-to-G2 Pore Swap
Pore identity changes several times during development (Figure 3, C and D). In the middle of the L1 stage, the G1 pore cell delaminates from the excretory system (Figure 4G), reenters the cell cycle, and divides to generate two neuronal daughters, the RMHL and RMHR motor neurons (Sulston and Horvitz 1977; Sulston et al. 1983; Parry and Sundaram 2014). As G1 departs, the G2 epidermal cell undergoes wrapping tubulogenesis around the central lumen cuticle to form a replacement pore (Figure 4G). During the L2 stage, G2 divides and its posterior daughter (G2.p) becomes the permanent excretory pore (Sulston and Horvitz 1977; Sulston et al. 1983). These dynamic changes in pore identity require significant remodeling of cell–cell and cell–matrix attachments within the excretory organ.
Lineal origins and Notch-dependent fate specification of G2 and W
G2 and W provide another example of Notch-dependent left/right asymmetry in the C. elegans embryo (Figure 4C), again to make an epithelial vs. neuroblast fate decision. G2 and W are derived from left/right lineal homologs that are initially equivalent and compete for the G2 fate (Sulston et al. 1983). LIN-12/Notch promotes the G2 (left) over the W (right) fate; in lin-12(lf) mutants, both cells adopt a W neuroblast fate, whereas in lin-12(gf) mutants, both cells adopt a G2 pore cell fate (Greenwald et al. 1983).
Dynamic behavior of G2 and W parallels that of G1
G2 and W start out as embryonic epidermal cells, but change their identities to pore tube or neuroblast, respectively, in the L1 stage (Sulston et al. 1983). During ventral enclosure, G2 and W cells migrate to the ventral midline in unison with other ventral epidermal cells and with the duct and G1 cells (Figures 4, A and D). As the duct and G1 cells form tubes, G1 inserts itself between the encroaching G2 and W cells and the anterior epidermis to contact the outside environment (Figure 4E); eventually, the ventral apical junction ring of the G1 pore contacts those of G2 and W at a tricellular junction (Abdus-Saboor et al. 2011) (Figure 11A). At embryonic stages after 1.5-fold, G2 and W are easily recognizable, because they are the only ventral epidermal cells in the anterior central region of the embryo that do not fuse to join the hyp7 syncytium (Podbilewicz and White 1994). These cells retain long, rectangular apical junction domains throughout much of embryogenesis (Figure 11B), but their apical junctions shrink to small rings by the time of hatching (Figure 11C).
G2 and W epidermal cells and the G1-to-G2 pore swap. (A–C) Schematic ventral views of apical junctions in the pore region, with G2 and W flanking the G1 pore opening, based on Abdus-Saboor et al. (2011). (A) Lima bean-stage embryo, when the pore tube first forms. (B) Threefold embryo. G2 and W have rectangularly shaped apical junctions. (C) Early L1. G2 and W junctions are ring shaped. (D) At mid-L1, during the G1-to-G2 swap, G1 appears to stretch dorsally and “unzip,” while G2 concomitantly forms an autojunction and “zips up” to form a replacement pore (Parry and Sundaram 2014). An unknown Ras-dependent signal (?) from the duct may promote G1 junction loss (Parry and Sundaram 2014). (E–G) L4 excretory system. (E) In wild type, G2p is the pore. (F) In the absence of G2, as in lin-12(lf) mutants, there is no longer a pore after G1 delaminates, and the duct connects directly to the hypodermis (Abdus-Saboor et al. 2011). (G) In the absence of G1, as in let-60(gf) mutants, the duct connects directly to the hypodermis (see Figure 10D). G2p usually wraps around the base of the duct tube and occasionally inserts to form a true pore (Abdus-Saboor et al. 2011).
At mid-L1, W (also known as P0.a) delaminates to become a neuroblast and undergoes several rounds of division to generate four neuronal descendants, including the VB2 (W.aap), VA1 (W.pa), and VD1 (W.pp) ventral cord motor neurons, plus one of the AVF interneurons (W.aaa) (Sulston and Horvitz 1977). Shortly after W delaminates and divides, G1 also delaminates, and G2 becomes the replacement excretory pore tube (Sulston et al. 1983; Parry and Sundaram 2014) (Figure 4G, Figure11D).
Cellular events of the G1-to-G2 pore swap
G1 delamination and G2 wrapping tubulogenesis occur in the context of a cuticle-lined pore channel, and the cells appear to migrate over the luminal cuticle as they change positions (Parry and Sundaram 2014); therefore, remodeling of apical cell–matrix connections is likely to be important for delamination. The presence of the cuticle may also help the channel to remain open throughout the remodeling process.
Changes in the actin cytoskeleton precede G1 delamination (Parry and Sundaram 2014). In the early L1 stage, F-actin localizes apically along the G1 autocellular junction, but just prior to G1 delamination, this F-actin disperses. Since F-actin strengthens junctions (Huveneers and de Rooij 2013), actin relocalization may loosen the adherens junction to allow subsequent dynamic changes. After actin relocalization, G1 appears to stretch dorsally and “unzip,” converting its autocellular junction to an intercellular junction with G2 (Parry and Sundaram 2014) (Figure 11D). G2 concomitantly forms an autojunction and “zips up” beneath G1, such that a region of two-celled tube is formed at their intersection. Subsequently, G1 junctions gradually shrink and disappear, and G1 then divides. Relatively little is known yet about the molecular control of these various G1 and G2 cell behaviors, but Ras signaling in the duct is required for G1 junction loss (Figure 11D; see below). Gene expression changes that must accompany the epithelial-to-neuronal transition remain to be characterized.
G1 and G2 behave independently
Although G1 delamination and G2 intercalation are closely coordinated during wild-type development, G1 and G2 can execute these behaviors independently. For example, in lin-12(lf) mutants, there is no G2 cell (Greenwald et al. 1983), but G1 still delaminates, and the duct then connects directly to the hypodermis to maintain a functional system (Abdus-Saboor et al. 2011) (Figure 11F). In let-60(gf) mutants, there is no G1 cell to delaminate, but G2 still wraps around the base of the duct to form a pore tube (Abdus-Saboor et al. 2011) (Figure 11G). In let-60(lf) mutants, where two G1 cells but no duct are present, both G1 cells appear to initiate delamination behaviors, which suggests that the duct is not needed to instruct these behaviors (Abdus-Saboor et al. 2011; Parry and Sundaram 2014). G1 and G2 may be executing intrinsically programmed behaviors, or they may be responding to cues from the canal cell or from somewhere outside of the excretory system.
A cell-nonautonomous role for LET-60/Ras signaling in G1 pore delamination
Ras signaling in the duct cell is required for successful execution of the final steps of G1 delamination (Parry and Sundaram 2014) (Figure 11D). SOS-1 is the guanine nucleotide exchange factor that activates LET-60/Ras during most Ras signaling events (Chang et al. 2000). In temperature-sensitive sos-1 mutants that are upshifted to restrictive temperature as early L1 larvae, the pore swap initiates normally, but the G1 cell junctions do not shrink and disappear, and G1 remains stuck within the excretory system (Parry and Sundaram 2014). This delamination defect can be rescued by expressing constitutive LET-60(gf) in the duct, but not in the G1 pore, showing that signaling acts cell nonautonomously. Potential models are that Ras signaling in the duct triggers production of a secreted factor that then acts on G1 or its matrix (Figure 11D), or that Ras signaling in the duct acts more directly to modulate duct–G1 adhesion. These results suggest that Ras signaling might also have nonautonomous effects on epithelial-to-mesenchymal transitions during tumor metastasis.
Why swap?
Why does the worm bother with these complex and potentially dangerous cellular gymnastics? The functions of the G1 and G2 pores do not differ in any known way. The function of the RMH neurons is still unknown—they presumably act in the motor circuit based on their pattern of synaptic connections (White et al. 1986), but let-60(gf) mutants, which lack G1 (Abdus-Saboor et al. 2011) (and thus presumably lack the RMH neurons), are viable and have not been reported to have motility defects. Presuming that the RMH neurons do have some more subtle yet important function, why generate them in this way? The rationale is not clear, but many neuroblasts in C. elegans and other organisms are generated from epithelial progenitors (Sulston et al. 1983; Colman 1999; Technau et al. 2006; Jarriault et al. 2008; Lassiter et al. 2014), as are many other cell types in mammals (Kalluri and Weinberg 2009), suggesting that epithelial cells have a general propensity for such reprogramming. Along with other similar examples in the worm (Zuryn et al. 2012), the excretory pore provides a simple and tractable model for studying epithelial reprogramming and delamination.
The Excretory Gland
Excretory gland structure and ultrastructure
The excretory gland is an A-shaped binucleate cellular syncytium that connects to the canal cell immediately adjacent to the canal–duct junction; this connection point has been termed the secretory–excretory junction (SEJ) (Nelson et al. 1983) (Figure 3E, Figure 12, A and B). As seen under TEM, the gland cytoplasm contains many electron-dense putative secretory granules in close proximity to Golgi complexes and the SEJ; these granules disappear in dauer larvae. The gland presumably secretes material into the duct channel for passage to the outside environment and/or secretes material into the canal lumen for transport to other parts of the body, but the nature of any such secretions is unknown. The anterior gland processes appear to make synapses with the nerve ring, suggesting possible neuronal control of secretion (Nelson et al. 1983).
The excretory gland and CAN neurons. (A) The A-shaped excretory gland (gl) visualized with B0403.4promoter::GFP (Hunt-Newbury et al. 2007). Image is courtesy of Don Moerman (University of British Columbia). (B) The gland empties into the canal sinus at the secretory junction. Note luminal matrix (gray). Black arrowheads, junctions. TEM image of late threefold embryo is courtesy of Richard Fetter and Cornelia Bargmann (The Rockefeller University). (C) CAN neuron (lateral view) visualized with ceh-23promoter::GFP (Forrester and Garriga 1997).
Lineal origin and development of the excretory gland
The gland is derived from two left/right lineal homologs that migrate to the ventral midline during ventral enclosure and then fuse at two separate points (Nelson et al. 1983; Sulston et al. 1983) (Figure 3E, Figure 4). The first fusion event occurs between processes linking the cells near their junction to the canal and duct. The second fusion event occurs between the tips of two narrow processes that the gland cells extend toward the nerve ring (these form the tip of the A-shape). It is not yet known if gland fusion requires the fusogens AFF-1 or EFF-1.
The LIM homeodomain transcription factor LIM-6 is expressed in the excretory gland cell and serves as a useful marker for its morphology, but the impact of lim-6 mutation on the gland is unknown (Hobert et al. 1999) (Figure 3E).
Possible functions of the excretory gland
One hypothesized function for the gland is the production of hormones. The gland stains with the neurosecretory marker paraldehyde-fuschin (PAF) (Nelson et al. 1983) and with the sterol-binding agents dehydroergosterol (DHE) and filipin (Merris et al. 2003), consistent with steroid hormone production. In the absence of hermaphrodites, male plep-1 mutants attempt to copulate with the excretory pore of other males (Noble et al. 2015), consistent with some attractive secretion emanating from the pore, and possibly originating in the gland. Singh and Sulston (1978) and Nelson and Riddle (1984) investigated the effects of gland ablation in larvae and found no disruption in osmoregulation, molting, dauer formation, or dauer exit, but other aspects of the animal’s physiology or behavior were not examined. Several other functions have been proposed for the gland in other nematodes (see below), including secretion of the epicuticle (Page et al. 1992a), but this function remains untested in C. elegans.
The Canal-Associated Neurons
CAN structure
The canal-associated neurons CANL and CANR are named for their close physical association with the excretory canal cell arms (White et al. 1986) (Figure 2, Figure 3F). The two CANs are born in the head region (Figure 4A) and then migrate posteriorly during late embryogenesis to their final positions in the midbody, immediately adjacent to the two posterior canal arms (Hedgecock et al. 1987). CAN neurons extend one axon into the head, and a second posteriorward into the tail, between the excretory canal and basement membrane (White et al. 1986) (Figure 6B, Figure 12C).
At birth, the CAN cell bodies resemble those of other neurons. After migration, the cell bodies swell to a much larger size (Figure 12C) and may form gap junctions to the excretory canals (White et al. 1986). This observation, plus the fact that the CANs do not make obvious synapses with other neurons, suggests that the function of these cells may be neurosecretory (White et al. 1986).
Transcription factors important for CAN differentiation
Transcription factors expressed in the CANs (Figure 3F) include:
CEH-10 is a paired-type homeodomain factor (Svendsen and McGhee 1995) required for CAN differentiation (Forrester et al. 1998; Manser and Wood 1990). In ceh-10 null mutants, the CANs fail to express CEH-23::GFP and do not migrate to their correct positions; it is not clear if the mutant CANs adopt an alternative identity or simply fail to differentiate.
CEH-23 is a divergent homeodomain factor (Wang et al. 1993). CEH-23::GFP is often used as a CAN marker (Figure 12C), but ceh-23 deletion mutants are viable and do not have apparent defects in CAN development or function (Altun-Gultekin et al. 2001).
UNC-39 is a Six5-related homeodomain factor required for CAN migration and axon pathfinding (Manser and Wood 1990; Yanowitz et al. 2004).
Regulators of CAN migration and axon outgrowth
Many studies of the CAN neurons have focused on their long-range migration during embryogenesis. Mispositioning of both CAN cells, or more specifically, failure of the posterior CAN axons to reach their normal posterior regions, leads to a characteristic withered tail (Wit) phenotype in which the posterior half of the animal is much thinner than the anterior half; this phenotype has facilitated genetic screens for migration mutants (Manser and Wood 1990; Forrester et al. 1998). Many CAN migration mutants also have defects in axonal extension and/or pathfinding, but the two defects are separable (Lundquist et al. 2001; Yanowitz et al. 2004).
Among the genes important for CAN migration are many that are also important for excretory canal outgrowth, such as epi-1/laminin alpha, ina-1/integrin alpha, unc-34/Enabled, mig-10/Lamellipodin, unc-73/Trio, and vab-8/Kinesin-like (Manser and Wood 1990; Forrester et al. 1998). Multiple Rac genes are also important and likely act downstream of UNC-73 (Lundquist et al. 2001). Proper CAN migration also requires the ROR family receptor tyrosine kinase CAM-1 and multiple Wnt-related ligands (Forrester et al. 1999).
Possible functions of the CAN neurons
Laser ablation or genetic removal of a single CAN cell has no obvious effect on the animal, but removal of both CAN cells causes a rod-like lethal phenotype (Forrester and Garriga 1997; Forrester et al. 1998). Based on the similarity of this lethal phenotype to that observed after ablation of the excretory canal, duct, or pore cells, the CANs were proposed to regulate excretory system function, but this has not been demonstrated.
Functions of the Nematode Excretory System
Adenophorean vs. Secernentean systems
The functions of the nematode “excretory–secretory” (ES) system have been a source of debate for well over a century, since not all nematode species contain the canal cell, but it appears that all species contain an excretory gland cell (Bastian 1866; Schneider 1866; Chitwood and Chitwood 1974). A simpler “Adenophorean” system is found particularly in marine species such as Enoplus and Monhystera (van de Velde and Coomans 1987; Turpeenniemi and Hyvarinen 1996), which live in an environment where environmental osmolarity is generally stable. In these animals, the ES system consists of a single large ventral gland (or “renette” cell) leading directly to a cuticle-lined duct and/or pore located ventrally anywhere from behind the terminal bulb forward to just behind the buccal cavity, depending on the species (Chitwood and Chitwood 1974; van de Velde and Coomans 1987) (Figure 13).
Adenophorean ES system. Diagram of ES system from Monhystera disjuncta, based on drawings by Bird and Bird (1991) and van de Velde and Coomans (1987).
The C. elegans excretory system is an example of the more complicated “Secernentean” ES system found in most terrestrial species that experience large changes in osmolarity (Chitwood and Chitwood 1974). Secernentean species all contain a branched canal cell, but variations between species include the number and position of canals (Chitwood and Chitwood 1974; de Grisse 1977; Bird and Bird 1991), the number of excretory cell nuclei (Wharton and Sommerville 1984), fusion of the duct and pore cells (Narang 1972), and even the presence of muscles associated with the system in the parasite Brugia malayi (Moreno et al. 2010).
Secretory function of the ES system
The ubiquitous presence of gland cells suggests that secretion is a critical function of the ES system in all nematodes. Excretion of mucilaginous material has been observed from the excretory pore of many species (Maggenti 1962; Premachandran et al. 1988; Bird and Bird 1991). Proposed functions of this material include:
To help digest prey (Lee 1970; van de Velde and Coomans 1987).
To effect changes in the animal’s life cycle (Bird and Bird 1991), such as molting (Singh and Sulston 1978) or entrance or exit from the dauer or dispersive phase of nematode growth (Nelson and Riddle 1984).
To coat the cuticle with physically or osmotically protective material (Blaxter et al. 1992; Page et al. 1992a,b).
To help parasitic nematodes infect a host organism and evade its immune response (Blaxter et al. 1992; Moreno et al. 2010). Consistent with such a role, ES system-associated muscles in Brugia malayi appear to be the relevant target of the nematocide ivermectin (Moreno et al. 2010).
Efforts in the parasitic nematode research community have characterized secreted ES products aiding in host infection (Hewitson et al. 2009), but whether these originate from the ES gland or from other bodily sources has been addressed in only a few cases (Page et al. 1992a; Moreno et al. 2010). A better understanding of excretory gland development and function in C. elegans could have important implications for the role of the gland in other nematodes and suggest new ways to prevent and treat parasitic nematode infections.
Osmoregulatory function of the excretory system
Nematodes maintain a hyperosmotic coelomic fluid to inflate the body cavity and keep the hypodermis rigid for effective muscle attachment and movement (Harris and Crofton 1957; Nelson and Riddle 1984; Wright 2004). A body cavity osmolarity higher than that of the environment attracts water that leaks through the tissues exposed to the environment, especially cuticle (Perry 1977; Wharton et al. 1988; Bird and Zuckerman 1989) and intestinal cells (Stephenson 1942; Wright and Newall 1976), so the animal must eliminate excess water to maintain homeostasis.
The excretory canals are ideally located to collect this excess water, as they stretch along the entire body adjacent to the pseudocoelom, separated solely by a basement membrane. Water may enter the canal cell by diffusion or via gap junctions with the hypodermis (Nelson et al. 1983; Altun et al. 2009), and then be transported via canaliculi to the lumen for excretion (Figure 8). Since water follows osmotic gradients, the canaliculi and lumen must concentrate ions to attract water. The canal cell expresses multiple types of membrane channels and transporter proteins for regulated movement of water and ions (Spencer et al. 2011) (Table 1). Ion-selective electrical measurements at the excretory pore show effluxes of Mg++, K+, H+, Na+, and Ca++ (Teramoto et al. 2010; Adlimoghaddam et al. 2014; Wang et al. 2014a), as would be expected from concentration and release of these ions by the excretory system.
A speculative model for transport of water through the canaliculi and lumen is presented in Figure 8, in analogy to the well-studied vacuolar ATPase/acid-driven excretion of water by the tubular water vacuole of the freshwater protozoan Paramecium (Plattner 2015). The V-ATPase proton pump concentrates cations, while anion channels provide anions to neutralize the acidic charge while increasing osmolarity. AQP-8/aquaporin channels allow excess water to cross membranes easily from the cytoplasm to the areas of high osmolarity. Upon docking with the apical membrane, canaliculi release the water and ions into the lumen.
The morphological structure of the canals shares features with the vertebrate lymphatic system, in that both contain narrow tubes reaching throughout the organism to collect fluid, and both require the PROS-1 transcription factor (Kolotuev et al. 2013). In addition, the osmoregulatory function of the canals shares features with the vertebrate renal system. Although the primary pump driving ion homeostasis in the kidney nephrons is the sodium/potassium exchanger (Ross et al. 1974; Geering 1997), V-ATPase plays critical roles in intercalated cells especially of the tubular epithelia of the collecting duct (Brown et al. 2012). The protons pumped into the lumen are used to drive water to be excreted from the plasma into the collecting duct lumen via aquaporins through the neighboring principal cells. Regulation of water permeability is modified via regulated trafficking of vesicles carrying aquaporin to the plasma membrane (Kwon et al. 2009; Nedvetsky et al. 2009). Loss of either collecting duct aquaporin or V-ATPase function causes renal tubular acidosis and osteopetrosis (Borthwick and Karet 2002). Other C. elegans excretory system proteins with direct vertebrate renal homologs include proteins discussed above: NHR-31 (homolog to human HNF-4α) (Hahn-Windgassen and Van Gilst 2009; Menezes et al. 2012); CEH-6 (mouse BRN1) (Burglin and Ruvkun 2001); EXC-6 (formin INF2) (Shaye and Greenwald 2015), and channels listed in Table 1.
In addition to the excretory system, the hypodermis and other tissues in C. elegans play key roles in fluid homeostasis, a topic that has been reviewed separately (Choe and Strange 2007). The physiology of the excretory system, and how it functions coordinately with other tissues, are relatively unexplored areas that warrant further study.
The C. elegans Excretory System as a Model for Understanding Human Diseases
The morphology of the C. elegans excretory system is very different from that of the kidney or other vertebrate multicellular tubular organs. Excretory tubes do not have cilia, and their shaping does not involve cell division or cell death, so the system cannot be used to model those important processes. Nevertheless, as highlighted above, many aspects of signaling, cytoskeletal and matrix organization, and vesicle trafficking are conserved between different tube types, and between C. elegans and vertebrates. Furthermore, unicellular excretory tubes are topologically similar to endothelial capillaries and to wrapping glia such as Schwann cells, and are particularly good models for studying the mechanisms that form, shape, and stabilize these types of narrow tubes.
Examples of specific disorders that might be better understood through study of the excretory system include:
Ion channel-associated diseases, as listed in Table 1.
Achlorhydria: Loss of hydrochloric acid secretion of the stomach parietal cells is caused by knockdown of the ERM-1 homolog ezrin (Tamura et al. 2005; Yoshida et al. 2016). In this condition, tubulovesicular structures of these cells fail to fuse with the central canaliculus to secrete stomach acid, reminiscent of the effects of erm-1 mutation on canalicular vesicle fusion.
Cerebral cavernous malformation (CCM): is a stroke-predisposing syndrome associated with abnormal brain capillary bed structure (Draheim et al. 2014). Studies in both C. elegans and Drosophila suggested that unicellular tubes (such as 75% of brain capillaries) (Bär et al. 1984) are particularly sensitive to CCM3 loss (Song et al. 2013; Lant et al. 2015). CCM-3 orthologs act with STRIPAK to affect Golgi organization and/or vesicle trafficking in multiple cell types (Fidalgo et al. 2010; Kean et al. 2011; Song et al. 2013; Lant et al. 2015). The precise roles of CCM in mediating CDC42 activity and apical membrane recycling, however, are still unclear.
Hereditary hemorrhagic telangiecstasia (HHT): is another stroke disorder with more widespread defects in capillaries throughout the body (Govani and Shovlin 2009). Mutations in the luminal ZP protein endoglin are a frequent cause of HHT (McAllister et al. 1994). A key role of endoglin is to bind and influence signaling by TGFβ family receptors (Guerrero-Esteo et al. 2002). The mechanism by which the ZP protein LET-653 and other luminal factors affect excretory duct tube integrity is not yet known, but studies in both C. elegans and Drosophila suggest that ZP proteins within luminal matrices are especially critical for maintaining patency of small unicellular tubes (Jazwinska et al. 2003; H. Gill, J. Cohen and M. Sundaram, unpublished results).
Focal segmental glomerulosclerosis (FSG): is a kidney disease associated with loss and scarring of glomeruli, where capillary beds interface with kidney podocytes (D’Agati et al. 2011). Mutations in the EXC-6 homolog INF2 are one cause of FSG (Brown et al. 2010). Studies of EXC-6 confirmed that the disease-causing mutations are dominant gain of function and showed that both actin and MT-binding activities are important for function (Shaye and Greenwald 2015).
Charcot-Marie-Tooth (CMT) disease: These peripheral neuropathies often involve aberrant structure of glial Schwann cells that ensheath peripheral neurons (Vallat et al. 2013). Mutations in the EXC-5 ortholog FGD4 cause a particular CMT syndrome in which Schwann cell growth is initially normal, but fails during the period of rapid nerve growth during puberty (Delague et al. 2007). The requirement for EXC-5 in recycling endosome trafficking and apical domain maintenance (Mattingly and Buechner 2011) suggests that FGD4 may also play a similar role. Mutations in the EXC-6 ortholog INF2 also cause CMT in addition to glomerulosclerosis (Mathis et al. 2014).
Though few in number, the cells of the C. elegans excretory system use a wealth of conserved mechanisms and molecules to create a complex and vital organ. Continued studies to determine how the nematode integrates all of the elements discussed above to specify, form, maintain, and use these cells will inform our knowledge of basic cell biological processes as well as our understanding of many disease mechanisms.
ACKNOWLEDGMENTS
We thank David Hall and the Center for C. elegans Anatomy (Yeshiva University, Albert Einstein College of Medicine), Edward Hedgecock (Johns Hopkins University), Don Moerman (University of British Columbia), Erik Lundquist (University of Kansas), John Sulston (Medical Research Council), Shai Shaham (The Rockefeller University), and Richard Fetter and Cornelia Bargmann (The Rockefeller University) for providing archival images; anonymous reviewers; and John Murray, Amin Ghabrial, and members of our labs for comments on the manuscript. M.S. was supported by National Institutes of Health grant GM58540 and National Science Foundation (NSF) grant 1257879. M.B. is supported by the NSF.
Footnotes
Communicating editor: M. Labouesse
- Received December 1, 2015.
- Accepted March 7, 2016.
- Copyright © 2016 by the Genetics Society of America