The plakin Short Stop and the RhoA GTPase are required for E-cadherin-dependent apical surface remodeling during tracheal tube fusion

Drosophila tracheal development provides a powerful model for the study of lumen formation, a process that is integral to the development of tubular networks such as those found in vertebrate circulatory, respiratory and excretory organs. The Drosophila tracheal system arises from nests of cells that invaginate from the epidermis and undergo branching morphogenesis postmitotically within each embryonic hemisegment (Manning and Krasnow, 1993). The lumenal or apical surface originates as the surface of these tracheal pits and expands as tracheal cells progressively invaginate and form branches. Recent genetic analysis of tubulogenesis in Drosophila (Beitel and Krasnow, 2000) and C. elegans (Buechner et al., 1999) has identified mutations that affect lumenal morphology. However, the mechanism of apical surface regulation remains poorly understood.

The tracheal lumen is initially closed at branch tips. Concurrent with branching morphogenesis, specialized cells at branch tips, known as fusion cells, join branches into a continuous tubular network. This process of anastomosis requires each fusion cell to recognize its partner in the adjacent hemisegment and to form a lumen that connects the two branches (Samakovlis et al., 1996b; Tanaka-Matakatsu et al., 1996). Shotgun, the Drosophila homolog of the cell adhesion molecule E-cadherin is integral to the initial fusion cell contact (Uemura et al., 1996). Mutations in shotgun affect tracheal branch extension and lumen formation at anastomosis sites, as do mutations in armadillo, the Drosophila homolog of its effector β-catenin (Uemura et al., 1996; Beitel and Krasnow, 2000). E-cadherin and β-catenin control cell polarity (McNeill et al., 1990) and tube extension in culture (Yap et al., 1995; Matsumura et al., 1997; Pollack et al., 1997), suggesting an evolutionarily conserved role for cadherin-mediated cell adhesion in apical surface regulation.

Tracheal cells face the lumen with their apical surfaces, suggesting that apically localized molecules play a role in lumen formation. Mutations in discs lost (dlt), which encodes an apically localized PDZ domain family protein (Bhat et al., 1999; Tanentzapf et al., 2000), disrupt epithelial cell polarity, as do mutations in crumbs, which encodes a Dlt-associated (Bhat et al., 1999), EGF repeat family transmembrane protein (Tepass et al., 1990). crumbs mutant embryos are defective in forming mature zonula adherens (ZAs) (Tepass, 1996), structures at apical/lateral contacts between cells that contain E-cadherin.

Several studies suggest that these apical surface determinants and E-cadherin regulate the cytoskeleton and therefore control lumen formation and morphology. For example, Crumbs may attach β_H-spectrin, an F-actin cross-linker, to the apical membrane in Drosophila (Wodarz et al., 1995) and mutations in a C. elegans β_H-spectrin moderately enlarge the lumen of the excretory canal (Buechner et al., 1999). E-cadherin interacts with F-actin via multiple mechanisms (Gumbiner, 2000). These include binding to p120 catenin (p120ctn), a negative regulator of the RhoA GTPase (Anastasiadis et al., 2000; Noren et al., 2000). RhoA controls the formation of F-actin-containing focal adhesions and stress fibers in cultured cells (Ridley and Hall, 1992; Jou and Nelson, 1998). These interactions with RhoA also potentially regulate apical membrane protein targeting (Jou and Nelson, 1998), a process important for lumen development in culture (Lipschutz et al., 2000).

Little is known about the cytoskeletal structures required for lumen formation and how apical surface determinants are localized. We identify here an F-actin-rich track that is associated with E-cadherin-dependent contacts between fusion cells that appears to guide deposition of apical surface determinants and lumen formation. During anastomosis, Short Stop (Shot), an evolutionarily conserved plakin (Gregory and Brown, 1998; Strumpf and Volk, 1998; Leung et al., 1999; Lee et al., 2000a), accumulates at these contacts and transiently along the track. Mutations in shot and constitutively active alleles of the Drosophila RhoA (RhoA) GTPase specifically disrupt this contact and the associated track. Remarkably, the interactions of Shot with F-actin and its binding to microtubules are functionally redundant in organizing the track, suggesting that Shot acts with other pathways to organize F-actin and microtubules, rather than as an F-actin/microtubule cross-linker. We propose that in fusion cells, RhoA antagonizes Shot to regulate E-cadherin-associated cytoskeletal structures required for apical surface determinant localization and lumen formation.

Embryos to be stained with monoclonal antibody (mAb) 2A12 were fixed in B-5 (Kolodziej et al., 1995). Otherwise, embryos were fixed as described elsewhere (Uemura et al., 1996). The antibodies used in this study were rabbit anti-GFP (Boehringer Mannheim), rabbit anti-lacZ (Jackson Immunologicals), guinea pig anti-Shot (Strumpf and Volk, 1998), rabbit anti-Dlt (Bhat et al., 1999), rat monoclonal anti-E-cadherin (Uemura et al., 1996) and rat monoclonal anti-tubulin (Harlan Sera-Lab, UK), together with FITC-, Cy3- and Cy5-conjugated secondary antibodies (Jackson Immunologicals). Staged embryos (Campos-Ortega and Hartenstein, 1985) were filleted in 70% glycerol, stained and viewed under a Zeiss LSM 410 confocal microscope using a 100× PlanApo lens. For F-actin visualization, embryos were filleted live in Ringer’s solution and fixed in Ringer’s/4% formaldehyde for 20 minutes, before staining with fluorescently labeled phalloidin. To compensate for slight variations in specimen height, 1 μm confocal sections were obtained, channels from each section merged in Adobe Photoshop, and information from adjacent sections composed to yield a 1 μm sagittal section through the center of the tracheal lumen.

pUAST-C-Shot L-GFP contains a DNA fragment that encodes a GFP fusion with the Shot long isoforms C-terminal microtubule binding domain (C-Shot L) (Lee and Kolodziej, 2002) in the GAL4 expression pUAST vector (Brand and Perrimon, 1993). pUAST-Shot L(C)-GFP expresses a GFP fusion with a full-length type C Shot long isoform (Lee and Kolodziej, 2002). Type C isoforms do not bind F-actin. pUAST-Shot L(C)-ΔGAS2-GFP also lacks DNA sequences encoding most of the GAS2 motif (amino acids 4859 to 4905; Accession Number, AAF24343).

Flies bearing these transgenes were obtained by standard methods (Ashburner, 1989); other transgenic flies have been previously described (Lee and Kolodziej, 2002). The btl-GAL4 and esg-GAL4 enhancer trap lines (chromosome II) obtained from Mark Krasnow (Stanford University) were used to drive pUAST transgenes in tracheal and fusion cells, respectively. In most cases, pUAST transgenes were expressed in shot³ mutant embryos by crossing shot³ pUAST recombinants to btl-GAL4 shot³ or esg-GAL4 shot³ stocks. To visualize tracheal microtubule organization in wild-type and shot³ mutant embryos, shot³/+; pUAST-C-Shot L-GFP/+ flies were crossed to btl-GAL4 shot³/CyO²⁷⁶ flies. The shot³ mutant embryos lack epidermal and CNS Shot protein (Lee et al., 2000a). Second chromosome pUAST-actin-GFP and pUAST-GAP43-GFP lines were obtained from Akira Chiba (University of Illinois, Urbana-Champaign); esg-lacZ lines were obtained from Shigeo Hayashi (National Institute of Genetics, Mishima, Japan). Homozygous third chromosome pUAST-RhoA^V14, pUAST-Rac1^N17 and pUAST-Cdc42^N17 lines were obtained from Liqun Luo (Stanford University) and pUAST-RhoA^N19 was obtained from Jeffrey Settleman (Massachusetts General Hospital). These flies were crossed to btl-GAL4 pUAST-actin-GFP or btl-GAL4/+; pUAST-C-Shot L-GFP/+ flies to visualize F-actin and microtubules in tracheal cells expressing dominant negative or constitutively active Rho family GTPases.

The wild-type tracheal system arises from segmentally repeated branched networks joined through anastomosis across segment boundaries (Manning and Krasnow, 1993). Anastomosis is complete in the dorsal trunk (Fig. 1A) by stage 14, in the lateral trunk (Fig. 1A) by stage 15, and across the dorsal midline (Fig. 1C) by stage 16 (Samakovlis et al., 1996b). In shot null mutant embryos, as well as in weaker shot¹ and shot² alleles (Lee et al., 2000a), these anastomoses are disrupted (Fig. 1 and data not shown). In stage 14 or later shot³ null mutant embryos, the dorsal trunk lumen is discontinuous at 75% of anastomosis sites (n=273) (Fig. 1B,D). Lateral trunk connections (Fig. 1B) and all anastomoses at the dorsal midline (Fig. 1D) are also affected. However, the overall branching pattern within hemisegments appears normal, suggesting that shot does not affect branch formation (Fig. 1B,D). A few dorsal trunk connections form apparently normally (6% of dorsal trunk anastomoses) (Fig. 1D), or are constricted relative to wild type (19% of dorsal trunk anastomoses), suggesting that the role of shot overlaps that of other molecules.

These defects could reflect defects in the differentiation of fusion cells, the specialized cells that occupy anastomosis sites (Samakovlis et al., 1996b; Tanaka-Matakatsu et al., 1996). At stage 14, vesicular structures containing lumenal antigens are detected in cells at anastomosis sites in wild-type and shot mutant embryos (Fig. 1E,F), suggesting that shot does not qualitatively affect the ability of these cells to synthesize and package lumenal antigens. A pair of fusion cells are present at anastomosis sites in wild-type and shot mutant embryos, and express all fusion cell specific markers examined (Fig. 1G,H, and data not shown). Thus, shot mutants appear normal with respect to fusion cell viability, and specific marker and lumenal antigen expression.

shot is allelic to kakapo (Gregory and Brown, 1998; Strumpf and Volk, 1998) and encodes a family of plakin-related proteins (Lee et al., 2000a). Long isoforms of Shot provide essential cross-links between F-actin and microtubules during axon extension (Lee and Kolodziej, 2002) and stabilize microtubule arrays in neuronal support and muscle attachment cells (Prokop et al., 1998). Our observation of lumen formation defects in shot mutant embryos led us to investigate the organization of the tracheal cell cytoskeleton after anastomosis in wild-type and shot mutant embryos. We focused on cells in the dorsal trunk, the largest tracheal branch and therefore the easiest in which to visualize cytoskeletal structures.

We visualized F-actin both by labeling dissected embryos with fluorescent phalloidin and by detecting actin-GFP (Verkhusha et al., 1999) specifically expressed in tracheal cells (Fig. 2). In wild-type tracheal cells at anastomosis sites, F-actin accumulates apically so that it appears as a continuous tube (Fig. 2A,G). In shot mutant tracheal cells at abnormal anastomosis sites, F-actin accumulates apically, but surrounds two blind-ended lumens (Fig. 2D,I).

We examined microtubules in tracheal cells using anti-tubulin. In wild-type tracheal cells, microtubules appear to accumulate apically (Fig. 2B,C). In shot mutant embryos, microtubules are no longer apically concentrated, though they remain largely at the cell periphery (Fig. 2E). Similar results are obtained with C-Shot L-GFP, a fusion between the microtubule binding domain of Shot and GFP that decorates all microtubules in transfected cells (Lee and Kolodziej, 2002). In wild type tracheal cells, C-Shot L-GFP accumulates apically (Fig. 2H), but in shot mutant tracheal cells, the protein is less clearly organized (Fig. 2J).

We investigated the identity of the cells that fail to form a lumen at anastomosis sites by labeling fusion cells with the esg-lacZ enhancer trap (Tanaka-Matakatsu et al., 1996) and labeling all tracheal cells with GAP43-GFP, a membrane-associated GFP derivative (Ritzenthaler et al., 2000). After anastomosis is complete in wild-type embryos (stage 14), the fusion cells form a compact doughnut around the lumen (Fig. 2K) (Samakovlis et al., 1996b) and flanking tracheal cells are drawn close to each other at the anastomosis site (Fig. 2K). The apically concentrated actin surrounds the tube and appears continuous through the fusion cells (Fig. 2K). In stage 14 shot mutant embryos, the fusion cells remain extended (Fig. 2L), creating a wide gap between flanking tracheal cells. The fusion cells are the only tracheal cells in the gap. The posterior cell often appears intercalated into the anterior branch, so that it stretches between the two branches (Fig. 2L). All tracheal cells accumulate F-actin apically, including the fusion cells, but the apical surface in each fusion cell does not develop into a bridging tube. Their apical surfaces remain facing the blind-ended lumens in their respective branches (Fig. 2L). Fusion cells therefore fail to remodel their F-actin cytoskeletons in shot mutant embryos.

Antibodies against the long Shot isoforms (Strumpf and Volk, 1998) reveal that these proteins concentrate cortically in all tracheal cells, and also assemble in fusion cells into a transient track (Fig. 3A) that marks the future axis of the lumen bridging the two branches (Fig. 3D). Shot is not detected cortically in tracheal cells or in a track in fusion cells in shot³ null mutant embryos (Fig. 3E,H). However, anti-Shot also non-specifically labels the tracheal lumen (Fig. 3E), precluding visualization of apically localized endogenous Shot with this reagent.

In order to investigate Shot localization during anastomosis in more detail, we expressed a Shot L(A)-GFP fusion in all tracheal cells, and used anti-GFP to follow its localization with respect to E-cadherin and F-actin (Fig. 4), and also with respect to Discs Lost (Dlt), an apical surface determinant (Bhat et al., 1999) (Fig. 5). The Shot L(A) isoform contains an N-terminal F-actin binding domain and a C-terminal microtubule binding domain (Lee and Kolodziej, 2002), and Shot L(A)-GFP fully rescues anastomosis defects when expressed in tracheal cells in shot null mutant embryos.

In tracheal cells, E-cadherin localizes to an adherens junction network that encircles the lumen (Fig. 4F-J) (Uemura et al., 1996). During anastomosis, E-cadherin also localizes to an early site of contact between the fusion cells (Figs 4F, 6F) (Tanaka-Matakatsu et al., 1996). The spot of E-cadherin enlarges in one of the fusion cells, usually the anterior member of the pair (Figs 4G, 6G). Subsequently, E-cadherin forms a track extending in both cells (Figs 4H, 6H) (Tanaka-Matakatsu et al., 1996). This track is then remodeled into a ring (Fig. 4I,J) (Tanaka-Matakatsu et al., 1996). Later in anastomosis, the fusion cells become more compact and doughnut shaped (Figs 3, 6N,O) (Samakovlis et al., 1996b). Compaction of the fusion cells draws the adherens junctions that demarcate fusion cell contacts with flanking tracheal cells closer to the central ring of E-cadherin. The central ring marks the anastomosis site and encircles the bridging lumenal connection (Figs 4I,J, 6J) (Tanaka-Matakatsu et al., 1996).

Shot L(A)-GFP and F-actin both accumulate in fusion cells at E-cadherin contacts (Fig. 4). As anastomosis proceeds, the F-actin forms a track that extends beyond the E-cadherin contact, and spans the two fusion cells (Figs 4S, 6Q,R). The track is a site for new membrane deposition (compare Fig. 6L with 6M). Shot L(A)-GFP also accumulates along this track (Fig. 5L), though it is more typically concentrated near E-cadherin contacts. These results are consistent with those obtained using anti-Shot (Fig. 3).

We investigated whether Shot L(A) can associate with E-cadherin contacts via its C-terminal domain (C-Shot L) or whether these associations primarily occur via its F-actin binding domain. C-Shot L-GFP expressed in tracheal cells accumulates at sites of E-cadherin localization (Fig. 4U-X), suggesting that it also mediates interactions with E-cadherin-associated cytoskeletal structures. As C-Shot L-GFP colocalizes with microtubules in cultured cells (Lee and Kolodziej, 2002), microtubules may also be concentrated near these sites. However, we were unable to visualize microtubules consistently during anastomosis using anti-tubulin staining and a variety of fixation methods, though we were able to detect microtubules in other tissues at this developmental stage (data not shown).

The Shot- and F-actin-containing track in fusion cells (Fig. 5L,Q) appears before apical surface determinants Dlt are detectable along the track (Fig. 5G). As the lumen forms within the fusion cells, Dlt accumulates along the track and then the new apical surface (Fig. 5H-J). Dlt levels are initially lower in the fusion cells relative to the flanking tracheal cells (Fig. 5I), suggesting that new apical surface is made in the fusion cells and existing apical surfaces do not simply stretch to connect with each other. Thus, a E-cadherin-associated track of F-actin and Shot appears to mark the site of new apical surface formation in fusion cells.

In shot mutant embryos, fusion cells frequently fail to form new E-cadherin contacts (Fig. 7E) and an associated F-actin track (Fig. 7G), but sometimes form inappropriate or misoriented E-cadherin contacts with tracheal cells other than their fusion partner (Fig. 7I,M). These aberrant contacts are not detectably associated with F-actin (Fig. 7K,N). These findings suggest that Shot is required to form or maintain the F-actin-containing structures associated with E-cadherin contacts, and to direct the formation of these contacts to the appropriate part of the fusion cells.

Rearranging the cytoskeleton during axon extension requires the presence of both the N-terminal F-actin binding domain and the C-terminal microtubule-binding GAS2 motif in the same Shot protein (Lee and Kolodziej, 2002). We therefore investigated whether these domains are required for lumen formation and cytoskeletal organization in fusion cells. Tracheal expression of Shot L(A)-GFP rescues anastomosis defects in the dorsal trunk (Fig. 8A), lateral trunk (Fig. 8A), and dorsal midline (data not shown). Fusion cell expression of Shot L(A)-GFP cells using an esg-GAL4 enhancer trap also rescues these defects. However, these rescued embryos also express Shot L(A)-GFP at low levels in other tracheal cells (data not shown).

Surprisingly, tracheal expression of long Shot isoforms that lacked either the N-terminal F-actin binding domain (Fig. 8B) or the C-terminal microtubule-binding site (Fig. 8C) also restored lumen formation in fusion cells. Deleting both the F-actin binding site and the microtubule binding site abolished rescue activity (Fig. 8D), indicating that at least one cytoskeletal interaction domain must be present for fusion cells to form a connecting lumen. Expression of the Shot C-terminal domain-GFP fusion in tracheal cells does not rescue the anastomosis defects, suggesting that other domains in the Shot long isoforms are also required for activity (data not shown).

Shot L(A)-GFP accumulates apically in tracheal cells (Fig. 8E). Mutant derivatives that lacked either the F-actin (Fig. 8F) or the microtubule binding site (Fig. 8G) are largely apically concentrated, but are partially delocalized. However, Shot molecules that lack both the F-actin and microtubule-binding domains no longer accumulate apically (Fig. 8H). Thus, Shot concentrates apically via interactions with the cytoskeleton.

To investigate further the role of Shot in organizing the cytoskeleton, we examined F-actin and microtubule distributions in tracheal cells in rescued shot mutant embryos. Tracheal expression of Shot proteins that lack either the F-actin or the microtubule binding sites restores both normal F-actin distribution in fusion cells (Fig. 9A,B) and the apical accumulation of C-Shot L-GFP, a microtubule binding protein, in all tracheal cells (Fig. 9C,D). Normal Dlt localization in fusion cells was also restored in rescued embryos (data not shown). F-actin remodeling in fusion cells and apical accumulation of microtubules in tracheal cells can therefore occur provided that Shot can directly interact with either F-actin or microtubules. A Shot L(A) derivative that lacks both the F-actin and the microtubule binding domains does not rescue F-actin remodeling in fusion cells (Fig. 9E), or apical accumulation of microtubules in all tracheal cells (Fig. 9F). Thus, the F-actin- and microtubule-binding domains of Shot are required for these processes, but appear functionally redundant. The interactions of Shot with the cytoskeleton may therefore facilitate additional organizing interactions between F-actin and microtubules.

Our findings that lumen formation defects are associated with the loss of E-cadherin-associated cytoskeletal structures led us to investigate the roles of other cytoskeletal regulators in lumen formation. E-cadherin signaling antagonizes the RhoA GTPase (Anastasiadis et al., 2000; Noren et al., 2000), which controls F-actin polymerization in many developmental contexts (Lu and Settleman, 1999). Constitutively active mutations in Drosophila RhoA reduce dendrite branching (Lee et al., 2000b), a phenotype also observed in shot mutant embryos (Prokop et al., 1998; Gao et al., 1999).

We therefore examined the effect of expressing constitutively active RhoA^V14 in tracheal cells. Tracheal defects in these mutant embryos strongly resemble those in shot mutant embryos. Lumen formation is blocked at all anastomosis sites, but occurs normally in cells interior to branches (Fig. 10A). The effects of RhoA^V14 on new apical cytoskeletal structures in fusion cells are also similar to those observed in shot mutant embryos, but more penetrant. In embryos expressing RhoA^V14 in tracheal cells, F-actin (Fig. 10D) and Dlt (Fig. 10G) accumulate apically in tracheal cells, but surround blind-ended tubes terminating at anastomosis sites, rather than continuous lumenal connections. E-cadherin does not form a three ringed structure at anastomosis sites (Fig. 10J), and Shot L(A)-GFP is distributed apically around blind-ended tubes (Fig. 10L). Tracheal expression of RhoA^V14 also disrupts apical accumulation of C-Shot L-GFP throughout the trachea (Fig. 10N), another similarity with shot mutant embryos (Fig. 2J). Fusion cells in embryos that express RhoA^V14 in tracheal cells fail to form initial E-cadherin-containing contacts (Fig. 10P) and associated F-actin tracks (Fig. 10Q). RhoA^V14 expression also modestly attenuates branch migration, but does not detectably affect the pattern of primary and secondary branching (Fig. 10A).

To investigate further the role of RhoA, we examined lumen formation in embryos expressing dominant negative RhoA (Lee et al., 2000b) in all tracheal cells. In these RhoA^N19 embryos, cells in the dorsal trunk and elsewhere do not form a distinct lumenal cavity (Fig. 10B), though primary and secondary branching appears normal. Lumenal antigens recognized by mAb 2A12 (Fig. 10B) and anti-Shot (data not shown) accumulate around cells towards the center of the dorsal trunk, but not in a defined lumen. F-actin (Fig. 10E) is also more broadly and patchily distributed. Dlt is more extensively distributed, and is no longer restricted to a well-defined apical region (Fig. 10H). Surprisingly, E-cadherin localization appears normal (Fig. 10K), suggesting that these cells retain some polarity. Shot L(A)-GFP is distributed cortically, without any apical accumulation (Fig. 10M), and C-Shot L-GFP also fails to accumulate apically in tracheal cells expressing RhoA^N19 (Fig. 10O). Cells in the dorsal trunk remain continuous (Fig. 10K,M,O) though they do not form a distinct lumen. These data therefore suggest that the formation of apical cytoskeletal structures and lumen require RhoA. RhoA^N19 expression also modestly attenuates branch migration outside the dorsal trunk, suggesting an additional role in tracheal cell migration (data not shown).

RhoA, Rac and Cdc42 make up a family of related GTPases with specific roles in diverse F-actin-based morphogenetic processes (Hall, 1998). To determine whether the role of RhoA is specific, we also expressed wild-type RhoA, dominant negative Drosophila Rac1 (Luo et al., 1994) or Cdc42 (Luo et al., 1994) in tracheal cells and examined tracheal development. None of these alleles detectably affects lumen formation or apical structures (Fig. 10C,F,I and data not shown), though Cdc42^N17 and Rac1^N17 embryos exhibit modest defects in branch migration (data not shown). Thus, the role of RhoA role in controlling the formation of apical cytoskeletal structures is likely to be specific among these Rho family GTPases.

Lumen formation between tracheal branches requires fusion cells to make new E-cadherin-containing contacts with each other and to remodel their apical surfaces (Uemura et al., 1996). The results presented here provide further insights into how the cytoskeleton and associated proteins support contact formation and subsequent apical surface remodeling. The F-actin- and microtubule-binding domains of Shot are required to maintain and remodel E-cadherin contacts and to assemble a track of F-actin and Shot in fusion cells. This track initiates at the E-cadherin contact and extends outwards from it to connect with the existing apical assemblies of F-actin and Shot. We propose that the track guides new apical surface formation. Apical surface determinants and membrane appear to accumulate along the track, possibly by spreading from existing apical concentrations. This track may also enable the fusion cells to contract and to draw the existing lumenal surfaces closer, as fusion cells appear notably less compact in shot mutant embryos.

Loss-of-function shot and gain-of-function RhoA alleles have similar phenotypes in fusion cells, and RhoA disrupts Shot localization. We therefore propose that RhoA negatively regulates track assembly and E-cadherin contact remodeling by Shot. Apically organized F-actin and adherens junctions in other tracheal cells appear to develop normally in shot mutant and RhoA^V14 embryos, suggesting specific requirements for shot and RhoA during new apical surface formation in fusion cells. We propose that Shot and RhoA regulate E-cadherin-dependent cell adhesion in selected developmental contexts.

We and others have shown that shot is required in neurons for growth cone motility (Van Vactor et al., 1993; Kolodziej et al., 1995; Prokop et al., 1998; Gao et al., 1999; Lee et al., 2000a). We show here that shot is required to remodel E-cadherin-containing contacts between tracheal fusion cells. Surprisingly, Shot proteins perform these distinct morphogenetic roles using different combinations of the same cytoskeletal interaction domains. In fusion cells, the binding sites for F-actin and microtubules appear functionally redundant. The F-actin binding domain is essential when the GAS2 microtubule binding site is absent, and the GAS2 microtubule binding site is essential when the F-actin binding site is absent. By contrast, during axon extension, the Shot behaves as an F-actin/microtubule cross-linker because the cytoskeletal interaction domains are both individually essential and required in the same molecule (Lee and Kolodziej, 2002).

These observations suggest that direct interactions between Shot and cytoskeletal proteins organize the cytoskeleton in fusion cells. The F-actin and microtubule domains may directly enable the accumulation of their cytoskeletal partners at the E-cadherin contact. In support of this hypothesis, the structurally similar F-actin binding domain of plectin alters F-actin organization (Andra et al., 1998) and the GAS2 motif stabilizes associated microtubules against depolymerization (Sun et al., 2001) in cultured cells. Since Shot’s interactions either with F-actin or with microtubules suffice to organize both cytoskeletal elements, binding to either F-actin or microtubules may then enhance other organizing interactions between F-actin and microtubules.

These other interactions may involve molecules required for E-cadherin signaling. E-cadherins are physically linked to F-actin via the β-catenin/α-catenin complex (Gumbiner, 2000) and to dynein, a microtubule-based motor, via β-catenin (Ligon et al., 2001). They can further regulate actin dynamics via association with p120, a RhoA antagonist (Anastasiadis et al., 2000; Noren et al., 2000); E-cadherins also stabilize microtubule minus ends in cultured cells (Chausovsky et al., 2000). E-cadherin signaling may therefore affect other proteins mediating interactions between F-actin and microtubules. Candidates include other F-actin/microtubule cross-linkers (Fuchs and Yang, 1999; Goode et al., 2000), regulators of Rho family GTPases that bind to microtubules (Ren et al., 1998; Glaven et al., 1999; Waterman-Storer et al., 1999) and F-actin-based motors that form complexes with microtubule-based motors (Huang et al., 1999). Further analysis will be necessary to identify these other molecules in fusion cells; these other cytoskeletal regulators may permit residual anastomoses in shot mutant embryos.

Our analysis also indicates that RhoA is required for lumen formation, most probably by regulating the apical cytoskeleton or by affecting the transport of lumenal antigens. Similarities between RhoA^V14 and shot mutant phenotypes suggest that RhoA could work either to antagonize Shot activity, or through parallel pathways acting on F-actin and microtubules. RhoA has many effectors that control F-actin distribution (Hall, 1998). RhoA^V14 has been reported to stabilize subsets of microtubules in fibroblasts in culture via an F-actin-independent pathway (Cook et al., 1998). Shot localizes apically via its interactions with the cytoskeleton, and either these interactions or the cytoskeletal structures themselves may be RhoA-regulated.

In cells throughout the trachea, reduced RhoA activity disrupts lumen formation and partially disrupts Dlt localization. Tracheal expression of RhoA^N19 does not appreciably affect E-cadherin localization. In cultured epithelial cells, E-cadherin localization is also resistant to RhoA^N19 (Jou and Nelson, 1998). These findings are consistent with RhoA functioning downstream of or parallel to E-cadherin. E-cadherin-associated p120ctn negatively regulates RhoA (Anastasiadis et al., 2000; Noren et al., 2000), but whether a similar pathway operates in Drosophila is unknown.

In fusion cells, RhoA can also function upstream of E-cadherin, as constitutively active RhoA^V14 affects E-cadherin localization selectively in these cells. E-cadherin distribution is more dynamic in fusion cells than in other tracheal cells (Tanaka-Matakatsu et al., 1996; Uemura et al., 1996), and may therefore be more sensitive to RhoA^V14. RhoA^V14 also affects new E-cadherin contacts in culture (Jou and Nelson, 1998). Further experiments will reveal whether Shot, RhoA and E-cadherin function in a common, evolutionarily conserved pathway to regulate apical surface remodeling in fusion cells.

We thank David Greenstein, Kathy Gould, Chris Wright and anonymous reviewers for helpful comments on the manuscript. We thank Talila Volk for anti-Shot, Hugo Bellen for anti-Dlt; Tadashi Uemura for anti- E-Cadherin; Mark Krasnow, Akira Chiba, Liqun Luo, Shigeo Hayashi and Jeff Settleman for strains; and Andrew Cook, Ahmad Rabi and Young-Ah Shin for able technical assistance. Confocal microscopy was performed in the Vanderbilt Cell Imaging Facility. This research was supported by funds from the Howard Hughes Medical Institute, a Vanderbilt Discovery grant and NIH 1RO1 GM/HL61202-01A1 to P. A. K.