A modifier screen identifies regulators of cytoskeletal architecture as mediators of Shroom-dependent changes in tissue morphology

Tissue architecture is typically defined during specific stages of embryonic development and errors in these processes can result in human disease. One example is formation of the vertebrate neural tube. The neural tube is formed via the concerted effort of many cellular pathways that functionally convert a plate of neural ectoderm into a closed tube. Errors in this process can result in birth defects such as spina bifida, exencephaly, or craniorachischisis (Nikolopoulou et al., 2017). One cellular pathway that controls this process is regulated by the Shroom3 cytoskeletal adaptor protein (Haigo et al., 2003; Hildebrand and Soriano, 1999). Shroom3 controls neural tube morphogenesis via the formation of apically positioned contractile networks of actomyosin and these networks facilitate neural tube closure by inducing apical constriction and the anisotropic contraction of actin filaments (Haigo et al., 2003; Hildebrand, 2005; McGreevy et al., 2015). This is accomplished via the modular nature of Shroom3. Shroom3 localizes to the apical compartment of epithelial adherens junctions via a direct interaction with F-actin (Hildebrand and Soriano, 1999). This interaction is mediated by the Shroom Domain (SD) 1, a unique actin-binding motif present in most Shroom proteins characterized to date (Dietz et al., 2006; Hildebrand and Soriano, 1999; Yoder and Hildebrand, 2007). Shroom3 function is also dependent on Rho-kinase (Rock), such that Shroom3 directly binds to Rock and regulates both its localization and catalytic activity (Das et al., 2014; Nishimura and Takeichi, 2008; Zalewski et al., 2016). The interaction between Shroom and Rock has been elucidated at the molecular level and is mediated by the conserved SD2 region of Shroom and a conserved coiled-coil region of Rock (Mohan et al., 2013, 2012; Zalewski et al., 2016). The interaction between Shroom and Rock results in the localized activation of non-muscle myosin II (myosin II) contractility, which provides the mechanical force needed to facilitate neural tube morphogenesis. The regulation of myosin II activity by Rock and other cellular pathways has been well described (Heissler and Sellers, 2016). Rock modulates myosin II activity in two ways. First, Rock can directly phosphorylate the associated regulatory light chain (RLC), which modulates the actin-associated ATPase activity and the conformation of myosin II (Amano et al., 1996). Secondly, Rock negatively regulates the phosphatase that dephosphorylates the RLC, thus preventing the inactivation of myosin II (Feng et al., 1999).

Shroom proteins are required for numerous biological processes and are associated with several human diseases. In mammals, there are three definitive Shroom proteins, Shroom2, Shroom3, and Shroom4, each of which contains an N-terminal PDZ domain, the centrally located SD1, and the C-terminally located SD2 (Dietz et al., 2006; Hildebrand and Soriano, 1999; Yoder and Hildebrand, 2007). All three proteins can directly interact with F-actin and regulate cell morphology via Rock (Dietz et al., 2006; Farber et al., 2011; Hildebrand, 2005; Yoder and Hildebrand, 2007). In humans, SHROOM2 has been linked to neural tube morphogenesis, colorectal cancer, and medulloblastoma (Chen et al., 2018; Dunlop et al., 2012; Shou et al., 2015), while in vitro studies indicate it is important for cell migration, vasculogenesis, metastasis, and melanosome biogenesis (Fairbank et al., 2006; Farber et al., 2011; Yuan et al., 2019). SHROOM3 mutations have been implicated in chronic kidney disease, heart morphogenesis, and neural tube closure in humans (Deshwar et al., 2020; Durbin et al., 2020; Köttgen et al., 2009; Lemay et al., 2015; Matsuura et al., 2020; Tariq et al., 2011). Using model organisms or cell culture, Shroom3 has been shown to control neural tube closure, axon growth, intestine architecture, eye morphogenesis, thyroid budding, and kidney development (Grosse et al., 2011; Hildebrand and Soriano, 1999; Khalili et al., 2016; Loebel et al., 2016; Plageman et al., 2010; Taylor et al., 2008; Yeo et al., 2015). Finally, SHROOM4 mutations have been associated with X-linked mental defects (Armanet et al., 2015; Danyel et al., 2019; Zapata et al., 2017).

We have shown that the Shroom gene is conserved in Drosophila and encodes multiple protein isoforms that have different subcellular distributions and activities in vivo (Bolinger et al., 2010). The most highly conserved region of Drosophila Shroom is the SD2, the region that binds to Drosophila Rho-kinase (Rok) (Mohan et al., 2012). Drosophila Shroom also contains a divergent SD1 motif and this appears to mediate localization to adherens junctions in polarized epithelia (Bolinger et al., 2010). Consistent with the known activities of mammalian Shroom3, expression of Drosophila Shroom in epithelial cells induces apical constriction in a Rok and myosin II dependent manner (Bolinger et al., 2010). While Shroom3 is essential for mouse and human development, Shroom is not absolutely essential for Drosophila viability, as Shroom null flies can be recovered, albeit with significantly reduced frequency (Simoes Sde et al., 2014). In Drosophila embryos, Shroom is planarly distributed and works in a complicated network with RhoA, Rok, and myosin II to control convergent extension movements (Simoes Sde et al., 2014). These elegant studies showing the role of Shroom in regulating directional contractility are supported by observations that Shroom proteins can be polarly distributed in mammalian tissues and cells (Farber et al., 2011; McGreevy et al., 2015; Muccioli et al., 2016).

To better understand the mechanisms that control Shroom-regulated changes in cell and tissue morphology, we have established tools to perform genetic screens for modifiers of Shroom activity in Drosophila. Shroom gain-of-function phenotypes in the eye and wing can be suppressed or enhanced by known components of the Shroom pathway. Using a candidate approach, we have identified several cytoskeletal regulators, including Short stop and Enabled, as participants in Shroom-mediated changes in cell morphology. Shroom regulates the distribution of Ena and this is likely mediated by conserved proline-rich sequences in Shroom and the EVH1 domain of Ena. We further show that while Ena is required for the Shroom gain-of-function phenotypes, apical recruitment of Ena is not sufficient to cause changes in cell morphology. Additionally, by using an isoform of Shroom that does not bind Ena, but still engages Rok, we show that apical constriction can be modulated by different cellular pathways depending on the context.

We hypothesized that imaginal discs from Drosophila melanogaster are a relevant model for assessing Shroom function during cell and tissue morphology since they are comprised of polarized epithelia and endogenously express the ShroomA protein (Fig. 2), the isoform most similar to mammalian Shroom3 (Bolinger et al., 2010). To perform these studies, we generated lines expressing ShroomA in the wing or eye imaginal discs under control of A9-gal (A9) or lozenge-gal4 (lz), respectively. Over expression of ShroomA in the eye or wing imaginal discs results in rough eye and crumpled wing phenotypes in the adult flies. Wings from A9>ShroomA/+ flies are significantly smaller and show dorsal curling at the margin (Fig. 1A). There do not appear to be defects in the differentiation of the various cell types of the wing, as the bristles and veins exhibit normal morphology and positioning. Wing phenotypes were quantified by measuring the total area of the wing blade that is derived from the wing pouch of the imaginal disc (Fig. 1B).

Adult male lz>ShroomA/+ flies exhibit a rough eye phenotype and the eye is typically smaller (Fig. 1C). To quantify the severity of the eye phenotype, images were cropped to include the center portion of the eye and converted to a binary image representing the reflections of each ommatidia (Fig. 1D). The binary image was processed to a Voronoi tessellation using the reflections of the ommatidia as the generators. All steps were performed using ImageJ. This and similar approaches have been used to determine the distribution and geometry of tissue structures and cells both in vitro and in vivo (Barriga et al., 2013; Cheng et al., 2014; Matsushima et al., 2015). The Voronoi cells generated by this conversion are representative of the distribution of ommatidia in the original sample, as they capture features such as fused and misaligned ommatidia (Fig. 1D, merge). The area of the Voronoi cells was measured and these values used as a quantitative measure of the rough eye phenotype (Fig. 1E).

Previous studies suggest the Gal4-UAS expression system is more effective at higher temperatures (Brand and Perrimon, 1993; Carvajal-Gonzalez et al., 2016). Additionally, because the transgenes used for these studies are located on the X-chromosome, we hypothesized there could be phenotypic variation in males versus females due to differences in expression. We therefore assessed the Shroom-induced phenotypes at different temperatures in males and females. Wings from A9>ShroomA/+ flies exhibit a continuum of phenotypes depending on the sex and temperature, with males raised at higher temperatures exhibiting the most severe phenotype (Fig. 1A,B). We observed a similar gradation of the eye phenotype of lz>ShroomA/+ male flies raised at different temperatures (Fig. 1C). These observations indicate that the eye and wing phenotypes can be utilized in modifier screens since they can be enhanced or suppressed under different conditions.

We have shown that Shroom proteins induce apical constriction when expressed in polarized epithelia in both cell culture and Drosophila embryos (Bolinger et al., 2010; Hildebrand, 2005). To determine if the observed phenotypes result from aberrant tissue morphology, we isolated imaginal discs from third instar larva from crosses of homozygous A9>ShroomA or lz>ShroomA females to homozygous shotgun-GFP (dE-cadherin) males and stained the resulting F1 heterozygous larva to detect Shroom and GFP (Fig. 2). In the wing pouch of control shotgun-GFP/+ imaginal discs, we observe endogenous Shroom protein co-localized with GFP in cell–cell and tri-cellular junctions, with highest expression in rows of cells that border the anterior portion of the wing margin and the anterior-posterior boundary (Fig. 2A–A″′). In A9>ShroomA/+; shotgun-GFP/+ discs, ectopic ShroomA is primarily expressed in the dorsal portion of the wing pouch, the future hinge region, and more sporadically in the ventral compartment (Fig. 2B–B‴). In the ventral portion of the wing imaginal disc, cells expressing ShroomA have smaller apical area, consistent with Shroom's ability to initiate apical constriction (Fig. 2B‴, arrows, G). Collectively, these data indicate that ShroomA expression alters epithelial cell shape in the imaginal disc, resulting in the observed wing phenotype.

In the control shotgun-GFP/+ eye imaginal disc, endogenous Shroom is expressed in the morphogenetic furrow and the pre-cluster rosettes of the developing ommatidia (Fig. 2C–C″, bracket and arrowheads, respectively). Shroom expression becomes restricted to R3/4 (Fig. 2C″, asterisk) and is eventually lost from the ommatidia. The expression pattern and subcellular distribution of Shroom is similar to that previously described for myosin II in these cells (Escudero et al., 2007; Robertson et al., 2012), consistent with the role of Shroom in modulating myosin II contractility. In lz>ShroomA/+; shotgun-GFP/+ eye discs, ectopic ShroomA is detected in cells posterior to the morphogenetic furrow and is excluded from the ommatidia (Fig. 2D). As seen in the wing imaginal discs, ShroomA-expressing cells appear apically constricted relative to neighboring cells (Fig. 2D′, arrows). Overall, formation of the ommatidia does not appear drastically altered, although some ommatidia are irregular in shape (Fig. 2D′,D″).

To determine if the observed rough eye phenotype is consistent with defects in cellular organization or differentiation, we stained pupal retinas to detect ShroomA and Elav (Fig. 2E,F). Endogenous Shroom protein is expressed in the pigment cells and these cells are organized in a hexagonal array consistent with ommatidia organization (Fig. 2E′). Within these cells, Shroom is localized to cell–cell and tricellular junctions. Elav is expressed in highly organized clusters of photoreceptor cells (Fig. 2E″). In lz>ShroomA retinas, ectopic ShroomA is expressed in the pigment cells, with lower expression observed in the cone cells (Fig. 2F′). In these retinas, the regular hexagonal array of pigment cells is disrupted and the ommatidia exhibit irregular shapes and fusions (Fig. 2F and F, dotted outline). The individual pigments cells also appear apically constricted relative to the controls (Fig. 2F′, arrows). This disorganization is further supported by the erratic positioning of the photoreceptors as evidenced by Elav staining (Fig. 2F″). These data indicate that ectopic ShroomA causes errors in formation of the pupal retina that are consistent with the adult rough eye phenotype but does not appear to alter differentiation during eye development.

We hypothesized that A9>ShroomA and lz>ShroomA flies could be utilized to assess the role of other proteins in Shroom-regulated processes. As a proof of principle, we tested if reducing the gene dosage of known components of the Shroom pathway, including Zipper (Zip, myosin II), Rho-Kinase (Rok), Spaghetti Squash (Sqh, RLC), and Flapwing (Flw, PP1β9C encoding a PP1β type protein also referred to as PP1c) could modify the ShroomA-induced phenotypes described above (Fig. 3A). To accomplish this, we performed a heterozygous F1 modifier screen using mutant alleles of the above genes (see Fig. S1). To assess the role of myosin II in the rough eye phenotype, we crossed homozygous lz>ShroomA females to Zipper/CyO heterozygous males at 28°C, collected the resultant heterozygous F1 male progeny, and analyzed the rough eye phenotype of lz>ShroomA; CyO/+ (control) versus lz>ShroomA; Zipper/+ (experimental) flies from the same cross. To eliminate potential complications from background effects, we utilized multiple mutant alleles of Zipper. As predicted by our model, reducing the dosage of myosin II protein suppresses the rough eye phenotype caused by ShroomA over expression (Fig. 3B,D).

The location of Rok, Sqh, and Flw on the X-chromosome precluded analysis using lz>ShroomA because of the dominant marker B associated with the FM7 balancer. We instead used A9>ShroomA to assess the ability of these genes to modify the Shroom phenotype. For X-linked candidates, we crossed females heterozygous for the candidate allele (i.e. rok¹/FM7) to A9>ShroomA males and collected trans-heterozygous F1 females of the following genotypic classes: A9>ShroomA/candidate allele (experimental, i.e. A9>ShroomA/rok¹) and A9>ShroomA/balancer (controls, i.e. A9>ShroomA/FM7). Consistent with the above results, reducing the dosage of either zipper, rok, or sqh suppresses the phenotype caused by ShroomA overexpression in the wing imaginal disc (Fig. 3C,E). Conversely, reducing the dosage of PP1c (Flw) enhances the Shroom phenotype in the wing. One concern was that the dominant CyO locus present in most chromosome 2 balancers might interfere with our analysis. However, our data indicate this is not the case as A9>ShroomA/FM7, A9>ShroomA; CyO/+, and A9>ShroomA/+ females all exhibit equivalent wing phenotypes (Figs 1 and 3). It is interesting to note that while all three alleles of Zipper acted as suppressors, they did not produce identical results, with the Zip¹ allele being the weakest. Both Zip1 and Zip2 are caused by truncation mutations and reported to be loss-of-function alleles (Franke et al., 2010). This suggests that genetic background may influence phenotype modification and that testing multiple alleles will likely be an important step in the screening process. In total, these data indicate that the A9>ShroomA and lz>ShroomA phenotypes can be modified by second site enhancers and suppressors.

To identify other proteins involved in Shroom-mediate changes in cell shape, we performed a heterozygous F1 modifier screen of candidate genes known to encode regulators of cytoskeletal architecture and cell morphology. Crosses were performed as described above (see also Fig. S1). Wings were isolated from F1 heterozygous control and experimental females and measured to define wing areas. To determine if a given candidate enhanced or suppressed the wing phenotype, we calculated the ratio of the areas of experimental (i.e. A9>ShroomA; Zipper/+) to control (i.e. A9>ShroomA; CyO/+) wings. A ratio of 1 indicates there is no difference between the experimental and control groups, a value less than 1 indicates the wing area of the experimental group is smaller than control (enhanced wing phenotype), and a value greater than 1 indicates the wing area of the experimental group is larger than controls (suppressed wing phenotype). We identified 12 genes (in addition to those tested above) as potential modifiers of Shroom function from a pool of 67 unique candidates, with most acting similarly in both the wing and eye disc (Fig. 4; Table S1). While this frequency seems high, we anticipate this is the result of testing candidate genes and this frequency would be much lower on a genome-wide scale. The majority of identified candidates act as suppressors and are predicted to function as regulators of cytoskeletal organization. It is interesting to note that mediators of processes such as cell adhesion or apical-basal polarity did not modify the Shroom-induced phenotypes using this approach. We identified multiple enhancers of Shroom, two tyrosine kinases, EGFR and Src42A, and Rap1, although EGFR was found to modify only the rough eye phenotype. The parental Rap1¹ allele used in these studies causes a rough eye phenotype and could not be accurately evaluated in combination with lz>ShroomA. EGFR signaling has been shown to act in several aspects of eye development, including both cell fate specification and cell morphology (Brown et al., 2006; Freeman, 1996; Kumar, 2012) while Src42A has been implicated in epithelial morphogenesis via the regulation of adherens junction dynamics (Shindo et al., 2008; Takahashi et al., 2005). Previous work suggests that Rap1 works in cooperation with several factors to define epithelial morphology in Drosophila and is required for Shroom3 activity in Xenopus neural tube closure (Bonello et al., 2018; Haigo et al., 2003; O'Keefe et al., 2012).

Enabled (Ena) and short stop (shot) were identified as suppressors in our initial screen and both regulate cytoskeletal architecture in epithelia. Previous work in Drosophila identified Ena as a determinant of actin dynamics that is required for epithelial morphology (Gates et al., 2007). Additionally, it has also been suggested that Ena/VASP are important for Shroom3 induced apical constriction in MDCK cells (Plageman et al., 2010). Shot is a member of the spectraplakin family of actin-microtubule crosslinking proteins and is required for microtubule organization in axons and follicular epithelia in Drosophila (Applewhite et al., 2010; Bottenberg et al., 2009; Nashchekin et al., 2016; Takacs et al., 2017). Microtubules have also been implicated in Shroom-regulated cellular architecture (Lee et al., 2007). To further investigate the role of Ena and Shot, we assessed the ability of different alleles to modify the phenotypes in the eyes and wings (Figs 5 and 6). In the eye, we see significant restoration of normal ommatidia organization and distribution (Figs 5A and B, 6A and B), while in the wing we observe a significant increase in area and decrease in the crumpled morphology (Figs 5C and D, 6C and D). As was observed for Zipper, the Shot and Ena alleles exhibit different degrees of suppression of the Shroom phenotypes, relative to each other, although all are predicted to be amorphic or loss-of-function alleles. In the case of Ena, both alleles used in this study are nonsense mutations, potentially resulting in truncated proteins, if translated. In the case of Shot, the molecular natures of shot^SF20 and shot³ are unknown but both are predicted to be strong alleles, while shot^V104 encodes a C-terminally truncated protein (Bottenberg et al., 2009; Strumpf and Volk, 1998). These data again point to the importance of verification using multiple alleles.

In an effort to understand the cellular basis for the observed genetic interaction, we stained wing imaginal discs to detect Shroom and Shot. In control discs, Shroom and Shot are both expressed in the anterior wing margin (Fig. 5E). At the subcellular level, while both are apically enriched, they exhibit complementary distribution, with Shot more medial and Shroom more peripheral (arrowheads, Fig. 5E′–E‴). This is consistent with previous descriptions of Shot localization in epithelial cells (Nashchekin et al., 2016). This distribution is largely maintained in cells that express ectopic ShroomA, although there appears to be elevated levels of Shot in cells expressing excess ShroomA (arrows, Fig. 5F–F‴). This indicates that Shroom and Shot are in the same subcellular domain and could cooperate in regulating apical architecture.

To verify that our analysis is not complicated by CyO, we balanced the Ena⁴⁶ and Zip² alleles over In(2LR)Gla, wg[Gla-1] PPO1[Bc] (referred to as GlaBC) and repeated the analysis of the wing phenotype in the F1 progeny (Fig. S2). We observe similar phenotypic suppression in both instances. We utilized this same cross to ensure that suppression is not the result of decreased ShroomA expression. We separated F1 larva based on the presence or absence of the dominant PPO1[Bc] marker and stained the wing imaginal discs to detect ShroomA. ShroomA expression in the dorsal wing pouch is comparable in A9>ShroomA/+;Ena⁴⁶/+ and A9>ShroomA/+;GlaBC/+ discs. We also observe that the dorsal compartment of the wing pouch from the A9>ShroomA/+;Ena⁴⁶/+ discs is restored to a more regular morphology (Fig. 6E, dotted lines).

To further investigate how Ena and ShroomA cooperate to regulate cell and tissue morphology, we stained W1118, lz>ShroomA/+, and A9>ShroomA/+ imaginal discs to detect Ena and Shroom. In W1118 imaginal discs, endogenous Shroom and Ena proteins exhibit extensive overlap in adherens junctions of the lines and arcs that are formed following passage of the morphogenetic furrow (Fig. 7A, arrowheads). Shroom and Ena colocalization is maintained at the periphery of the pre-clusters of cells that will form the ommatidia (Fig. 7A, arrows). As Shroom expression becomes restricted to R3/4, Ena protein is seen in the cell junctions of ommatidia (Fig. 7A, asterisks). In lz>ShroomA/+ eye discs, we observe increased Ena staining in cell–cell junctions of those cells that express ectopic ShroomA but typically express lower levels of Ena and are not incorporated into the ommatidia pre-clusters (Fig. 7B, arrow).

Endogenous Shroom and Ena are widely expressed in the wing disc and both are more highly expressed in the anterior region of the wing margin (Fig. 7C). At the subcellular level, both are localized to apically positioned tri-cellular and cell–cell junctions (Fig. 7C′–C′′′, arrowheads). In A9>ShroomA/+ imaginal discs, there is a dramatic increase in Ena localization to cell–cell junctions throughout the wing disc. As seen in the eye disc, this increase occurs only in those cells that express ectopic ShroomA (Fig. 7D). To further investigate the role of Ena function in Shroom-induced phenotypes, we stained W1118 or A9>ShroomA/+ imaginal discs to detect Ena and F-actin (Fig. 7E,F). Consistent with the known activities of both Ena and Shroom, cells that exhibit increased levels of junctional Ena also display increased levels of junctional F-actin. Analysis of the fluorescent intensity of ShroomA versus Ena and Ena versus F-actin suggests there is a direct correlation between the localization of ShroomA, Ena, and F-actin (Fig. 7G,H). Taken together, these data suggest that ShroomA regulates the distribution of Ena and that ShroomA and/or Ena may function to regulate the amount of junctional F-actin.

The above results suggest that Ena activity is necessary for Shroom-induced apical constriction. Additionally, previous studies indicate that the EVH1 domain of Drosophila Ena binds to proline-rich ligands of the sequence LPPPP, which is slightly different from the canonical ‘FPPPP’ sequence (Chen et al., 2014). The primary amino acid sequence of ShroomA contains two such motifs, SPELPPPP and DEPLPPPPP, within a stretch of approximately 25 amino acids (Fig. 8A). Based on these observations we wanted to determine if Ena recruitment is dependent on these sequences and if increased recruitment of Ena is sufficient to cause apical constriction. To test this, we expressed variants of Shroom that contain different domains of the protein using the A9-gal4 driver (Fig. 8B). These include full-length ShroomA, ShroomAΔSD2, a version of ShroomA that retains the actin and putative EVH1 binding sites but lacks the Rok binding motif, and ShroomB, a naturally occurring isoform that retains the SD2 but lacks the actin and putative EVH1 binding sites (Bolinger et al., 2010; Mohan et al., 2012). As expected, A9>ShroomA induces a strong wing phenotype and this coincides with increased Ena distribution in the dorsal compartment where ShroomA is expressed (Fig. 8D). Similarly, the ShroomAΔSD2 deletion variant is highly expressed in the dorsal compartment and there is a commensurate increase in Ena localization (Fig. 8E). However, ShroomAΔSD2 is largely inert as the A9>ShroomAΔSD2 wings exhibit normal morphology. These data suggest that the SD2 is essential for apical constriction while Ena recruitment is needed for ShroomA-induced apical constriction. This is further supported by the observation that ShroomB causes a severe wing phenotype but does not induce significant changes in Ena distribution (Fig. 8F).

To determine if endogenous Ena distribution is regulated by Shroom, we reduced the level of Shroom protein using RNAi. In control wing discs, Shroom and Ena are co-expressed and co-localize in the anterior margin (Fig. 8G, arrow). As expected, the Shroom RNAi effectively reduces the amount of Shroom protein in the wing imaginal disc (Fig. 8H′). Coincident with the loss of Shroom protein, there is a decrease in the amount of Ena protein in the anterior portion of the wing margin (Fig. 8H″, arrow). However, the localization of Ena protein is not perturbed in other regions of the wing disc. Based on the presence of potential EVH1 binding sites in ShroomA, we tested the ability of the EVH1 domain to co-localize with ShroomAΔSD2 in Cos7 cells (Fig. 8I–K). In these cells, ShroomAΔSD2 is localized to the cell cortex and cytoplasmic puncta as previously described (Bolinger et al., 2010) while the EVH1 is largely cytoplasmic. When co-expressed, we observe extensive re-distribution of the EVH1 to ShroomA cytoskeletal compartments (Fig. 8K). Together, these data suggest that ShroomA can regulate the localization of Ena in vivo and that this is likely mediated by the EVH1 domain of Ena and the LPPPP sequences of ShroomA.

This study describes a genetic approach to identify cellular pathways that participate in tissue morphogenesis. This method takes advantage of the observation that ectopic Shroom protein can utilize the endogenous contractile machinery within epithelial cells to induce apical constriction and disrupt normal tissue morphology. While the work described here focuses on candidate genes that encode known regulators of epithelial and tissue architecture, we predict these tools can be used to perform unbiased, genome-wide screens to identify novel participants in Shroom-mediated cellular processes. We have utilized two different tissues, eye and wing imaginal discs, for these studies and are confident that these screens can identify factors that are used in a wide range of tissues and cells to control cell dynamics. This is based on the observations that ShroomA induces similar cellular phenotypes in both types of imaginal discs and the phenotypes can be modified in both tissues. A powerful aspect of this screen is that these processes are functionally conserved in vertebrate cells and tissues. Additionally, the simplified nature of the Drosophila genome makes these screens possible. Due to genetic and functional redundancy, we predict that the analysis performed here would be more complicated using vertebrate or cell culture model systems. Drosophila have single genes for Shroom, Rok, myosin II, and Ena while mammals possess gene families for these factors. In support of this, we have previously shown that both Rock1 and Rock2 must be inhibited to prevent Shroom3-mediated apical constriction in cell culture (Mohan et al., 2013). This screening approach should allow for the identification of novel genetic interactions in Drosophila that can be further verified in mammalian model systems to define their potential role in human disease.

Most of the modifiers we identified participate in defining actin or microtubule architecture. Of these, several regulate actin dynamics at the level of polymerization or stability, including Ena, Diaphanous, Chickadee, and Slingshot. Interestingly, three of these proteins can be linked, directly or indirectly, to neural tube formation in mice (Grego-Bessa et al., 2015; Lanier et al., 1999). It should be noted that several classes of actin regulators did not appear to modify the Shroom phenotypes, including nucleators, binding proteins, or adaptors, suggesting that specific types of actin organization are required for Shroom-induced perturbation of cell architecture. This is further supported by the observation that Tropomyosin was also identified in the screen. Tropomyosin regulates the structure of actin filaments and the binding of other proteins, including myosin II and cofilin, that in turn modulate cell architecture or behavior (Gunning et al., 2015). It is particularly intriguing to note that Tropomyosin mutations can suppress phenotypes caused by the loss of Flapwing, presumably caused by increased myosin II activity (Vereshchagina et al., 2004). In addition to the actin cytoskeleton, these studies also support a role for microtubules in Shroom-induced phenotypes. This is consistent with the role of microtubules in apical constriction in Drosophila (Corrigall et al., 2007; Fernandes et al., 2014; Ko et al., 2019). Recent evidence indicates that apical-medial microtubules play an important role in ventral furrow invagination and this is mediated by Patronin, a protein known to interact with Shot (Ko et al., 2019; Nashchekin et al., 2016). These studies show that microtubules stabilize the connection of contractile networks to cell junctions to facilitate tissue morphogenesis. These studies are consistent with our results in relation to Shroom function and Shot distribution in the wing epithelium. It will be interesting to determine if the identified proteins act upstream or downstream of Shroom. While our data suggest Ena acts downstream of Shroom, proteins such as Tropomyosin could function upstream by regulating the amount of Shroom that can bind to F-actin or downstream by modulating the amount of myosin II that can be recruited or activated by the Shroom-Rok complex. It was surprising that determinants of cell adhesion or polarity, such as cadherins or Par complex proteins, were not identified in this screen. It is possible that these proteins are present in sufficient quantity and reducing the dosage is unable to modify the Shroom overexpression phenotype and thus other genetic approaches will be needed to assess the role of these pathways.

Our data show that endogenous Shroom protein is expressed in epithelial cells during wing and eye development, suggesting it functions in these tissues under normal circumstances. Shroom null flies that survive to adults do not exhibit significant defects in the eyes or wings, although null embryos do exhibit defects in convergent extension (Simoes Sde et al., 2014) and perhaps this could contribute to the observed reduction in viability. In embryos, Shroom is important for the polarized distribution of contractile myosin II needed for convergent extension. It is possible that Shroom activity in disc epithelial cells is redundant to other pathways that regulate Rok and myosin II and Shroom normally functions to make these pathways more robust or function with higher fidelity. Uncovering these subtle interactions will require additional genetic approaches. The localization of Shroom in the eye and wing disc appears to be highly regulated and is reminiscent of that exhibited by myosin II and phosphorylated Sqh, particularly in the eye imaginal disc (Escudero et al., 2007; Robertson et al., 2012). We observe a dramatic increase in Shroom protein in cells that are exiting the morphogenetic furrow and forming the pre-clusters that will give rise to the ommatidia. As the ommatidia form, Shroom expression becomes restricted to the R3/4 cells and eventually is lost from these cells. This distribution is essentially the inverse to that of E-cadherin, which is highest in the radial junctions and lower in the circumferential junctions (Fig. 2). This could reflect differences in adhesive interactions between the ommatidia pre-clusters and the inter-ommatidia cells, which facilitates rotation of the ommatidia. This hypothesis is supported by previous studies demonstrating that differential adhesion generates specific cellular organization and compartmentalization in the developing eye (Hayashi and Carthew, 2004; Warner and Longmore, 2009). Interestingly, the PCP protein Flamingo is also expressed in R3 and R4 and we have previously identified interactions between the Shroom3 and PCP pathways in the neural tube (Ho et al., 2010; McGreevy et al., 2015). As eye development continues, we observe Shroom expression in the pigment cells of the pupal retina. In both the imaginal disc and the retina, Shroom distribution is restricted to specific cell junctions, suggesting there are differential adhesive or contractile forces associated with these membranes.

In the wing imaginal disc, we observe expression of Shroom protein in rows of cells that border the anterior half of the wing margin. Consistent with the genetic interactions, we observe a similar expression pattern for both Ena and Shot in these cells. It is currently unclear if the co-expression of Shroom, Ena, and Shot is controlled pre- or post-transcriptionally. It is possible that the expression of Shroom, Ena, and Shot is coordinately regulated in a gene network. Alternatively, the stability or apical localization of these proteins may be interdependent or closely orchestrated. This expression pattern in the anterior wing margin is similar to members of the Irre cell Recognition Module (IRM), including cell surface receptors Roughest, Hibris, and Kirre, which help position the sensory organs (Linneweber et al., 2015). This is particularly interesting in light of the fact that the vertebrate orthologs of these genes, Neph and Nephrin-1, and Shroom3 are all involved in formation of podocytes in the glomerulus of the mammalian kidney (Kestilä et al., 1998; Khalili et al., 2016; Matsuura et al., 2020; Sellin et al., 2003; Yeo et al., 2015). It will be exciting to apply genetic analysis to investigate if these pathways cooperate to regulate tissue morphology.

Ena and Shroom show extensive co-expression and colocalization in both the wing and eye imaginal disc, although Ena is more widely expressed than Shroom (Figs 2 and 7). In both the wing and eye imaginal disc, Ena is expressed in most cells and is localized primarily in the tricellular junctions with lower expression in the adherens junctions. However, as seen in the wing margin and the morphogenetic furrow, cells that express Shroom protein also exhibit high levels of Ena in the cell junctions. Importantly, reducing the amount of Shroom protein perturbs the localization of Ena in the anterior wing margin. The relationship between Ena, Shroom, Rok, and myosin II in defining cell shape is likely to be complicated. This stems from the observations that these factors could be placed both upstream and downstream of Shroom. For example, we have previously shown that Shroom distribution to the apical adherens junctions is mediated, at least in part, by direct binding to F-actin. However, it has also been established that RhoA and Rok regulate F-actin architecture to influence Shroom distribution, which then facilitates the polarized distribution of Rok and myosin II (Simoes Sde et al., 2014). Ena has been shown to have multiple roles in Drosophila development, including axon guidance, collective cell migration, and epithelial morphogenesis (Gates et al., 2007; Gertler et al., 1995; Jodoin and Martin, 2016; Myat et al., 2019). The role Ena plays in Shroom-mediated apical constriction is unclear. Our data suggest that Ena functions downstream of Shroom and is recruited to adherens junctions via an LPPPP-EVH1 interaction. Ena is primarily defined as a modulator of F-actin dynamics that facilitates the formation of long filaments by competing with barbed-end capping and promoting the addition of actin monomers to the barbed end (Bear and Gertler, 2009). This activity may be important for providing the substrate for activated myosin II to drive cell contraction. This is consistent with studies in vertebrate cells showing that Diaphanous 1, is also required for contractility in adherens junctions (Acharya et al., 2017) and that we also identified Dia as a potential modifier of Shroom activity.

Elegant studies from several groups have identified many other signaling pathways that control the distribution of contractile myosin II networks during Drosophila development, including the Fog, PCP, HH, Dpp, EGF, Toll, and integrin signaling pathways (Corrigall et al., 2007; Dawes-Hoang et al., 2005; Fernandes et al., 2014; Kolesnikov and Beckendorf, 2007; Paré et al., 2019, 2014; Robertson et al., 2012; Winter et al., 2001). How all these signaling pathways are orchestrated and converge on myosin II at the cellular and tissue level is a fascinating question. It has been shown that the above processes use a variety of methods to regulate the small GTPase RhoA, which activates Rok, including several GTP exchange factors or GTPase Activating Proteins (Kolesnikov and Beckendorf, 2007; Mason et al., 2016; Nikolaidou and Barrett, 2004). It should be noted that other GTPases such as Rap1 or CDC42 also regulate apical constriction (Sawyer et al., 2009; Spahn et al., 2012). Our work has shown that Shroom3 may activate Rock independent of RhoA, suggesting that there as mechanisms to bypass small GTPases in the activation of myosin II (Mohan et al., 2013; Zalewski et al., 2016). It will be informative to utilize this screening approach to further test how these pathways might work with ShroomA to control cell morphology.

Fly stocks were maintained on standard media at 25°C, unless indicated otherwise. Shroom gain-of-function was achieved using the Gal4-UAS system (Brand and Perrimon, 1993). The Gal4 expression drivers, A9-gal4 and lozenge (lz)-gal4, and the UAS-Shroom lines have been described previously (Bolinger et al., 2010; Crew et al., 1997; Sun and Artavanis-Tsakonas, 1997). A9>ShroomA and lz>ShroomA were generated by selecting recombinants between UAS-ShroomA and either A9-gal4 and lz-gal4, all of which map to the X-chromosome (Fig. S1), and were maintained as homozygous stocks.

The heterozygous F1 modifier screen was performed as follows. For genes located on the second or third chromosomes, males heterozygous for the candidate alleles were crossed to homozygous A9>ShroomA or lz>ShroomA females at either 22, 25, or 28°C. The resulting heterozygous F1 progeny were collected and analyzed by measuring the wing blade area or ommatidia organization. For phenotypic analysis, A9>ShroomA; balancer/+ or lz>ShroomA; balancer/+ serve as the controls while A9>ShroomA; mutant/+ and lz>ShroomA; mutant/+ are the experimental samples (see Fig. S1 for sample crosses). In all cases, control and experimental categories were also divided based on sex. For candidates on the X-chromosome, females heterozygous for the candidate allele were crossed to A9>ShroomA males and F1 progeny collected. A9>ShroomA/balancer are the controls and A9>ShroomA/mutation are the experimental group. Modification of the wing phenotype was established by calculating the ratio of the areas of the experimental wings to the control wings.

The following alleles were identified as modifiers in these studies:

Other lines used include In(2LR)Gla, wg[Gla-1] PPO1[Bc] (GlaBC), y1 w*; TI{TI} shgGFP (Bloomington Stock Center), Shroom RNAi (VDRC v100672, P{KK106863}), and Armadillo-Gal4 (a gift from M. Rebeiz, University of Pittsburgh). A complete list of candidate genes and alleles that were used in the preliminary screen can be found in Table S1.

For wing analysis, adult flies were collected in 70% EtOH and washed through a graded series of EtOH:glycerol to a final solution of 30% glycerol (in PBS). Wings were removed, mounted on microscope slides in 30% glycerol, and imaged with a Leica DMR compound microscope (5X/0.15 air objective) and Leica DFC300F digital camera with Leica Acquire software. Wings were measured and processed using ImageJ and Photoshop. For adult eyes, flies were collected, frozen at −20°C, and imaged using a Leica S8APO microscope with a ring light and polarizing filter. Images were captured using a MC170HD digital camera and Leica Acquire Software and processed using ImageJ and Photoshop. For electron microscopy, adult flies were dehydrated through a grades series of EtOH: hexamethyldisilazane (HMDS) into a final solution of 100% HMDS, sputter coated, and imaged with a Jeol JSM6390LV SEM. To analyze imaginal discs, third instar larva were isolated in PBS/0.1% tween, cut in half, turned inside out, and fixed in 4% paraformaldehyde in PBS/0.1% triton for 20 min. Fixed larva were stained with primary antibodies at 4°C overnight in PBS/0.1% triton, washed in PBS/0.1% triton, and stained with fluorescent secondary antibodies for 2 h at room temperature. Individual imaginal discs were removed from the larva, mounted in Vectashield, and imaged using an Olympus Fluoview confocal microscope with either 40X/1.30 or 100X/1.40 oil immersion objectives. Retinas were dissected 48 h after pupation as described (Hsiao et al., 2012). Antibodies and fluorescent reagents used include Rat anti-Shroom (R11, detects ShroomA and ShroomB, R22, detects ShroomA, 1:250, Bolinger et al., 2010), mouse anti-Myc mAb 9E10, (1:100, a gift from Dr Ora Weisz), mouse anti-Enabled mAb 5G2 (1:200, Developmental Studies Hybridoma Bank), mouse anti-Elav 9F8A9 (1:400, Developmental Studies Hybridoma Bank), mouse anti-Shot mAbRod1 (1:400, Developmental Studies Hybridoma Bank), TRITC-phalloidin (Sigma-Aldrich), and Alexa 488 and 568 goat anti-mouse and goat anti-rat IgG (Invitrogen).

Cos7 cells were maintained at 37°C, 5% CO₂ in DMEM, supplemented with 10% FBS, pen/strep, and L-glutamine. Cells were transiently transfected with expression vectors using Lipofectamine 2000, plated onto coverslip, and stained to detect Shroom or the myc-tag 24 h post transfection. Sequences encoding the Ena EVH1 domain were amplified by PCR and cloned into pCS3MT. The ShroomA expression vector has been described previously (Bolinger et al., 2010). Cells were imaged using an Olympus Fluoview confocal microscope using a 60X/1.42 oil immersion objective.

We thank Drs B. Stronach, M. Rebeiz, and M. Peifer for providing Drosophila stocks and all Neural Defects students and instructors. The antibodies obtained from the Developmental Studies Hybridoma Bank were developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.