Cadherin 99C regulates apical expansion and cell rearrangement during epithelial tube elongation

Epithelial tissues in multicellular organisms are manifested by apicobasal polarity, which is essential for their correct form and function (Laprise and Tepass, 2011; St. Johnston and Ahringer, 2010; Tepass, 2012). In polarized epithelia, the apical domain faces the outside world or the tube lumen, and is separated by junctional structures from other membrane domains. The basal and lateral domains, often referred to collectively as the ‘basolateral’ domain, are distinct, with the lateral domain contacting neighboring cells through adhesion molecules and the basal domain contacting an underlying extracellular matrix (ECM) or cells of other tissues.

Drosophila epithelia have been instrumental in identifying polarity regulators and revealing the mechanisms underlying the establishment and maintenance of polarity. The known polarity regulators localize in complexes located in specific regions along the lateral cell surface. Apical determinants, including the Crumbs (Crb/Sdt/Patj/Lin7) and atypical Protein kinase C (aPKC/Par3/Par6) complexes, are enriched in the sub-apical region (SAR), the small region of cell-cell contact apical to the adherens junctions (AJs) (Tepass, 1996,, 2012; Wodarz et al., 2000). The Scribble complex (Scrib/Lgl/Dlg), which specifies basolateral identity, localizes to the septate junction (SJ) region, immediately basal to the AJs, along with known SJ components (Bilder and Perrimon, 2000; Strand et al., 1994; Woods and Bryant, 1991). Mutually competitive, negative feedback between these apical and basolateral regulators limits each polarity domain (Tepass, 2012). Interestingly, only very few apical surface proteins have been identified (Baumgartner et al., 1996; Jaźwińska et al., 2003; Zhang and Ward, 2009) and their importance in cell polarity and organ shape has not been addressed.

Drosophila tubular epithelia include tissues that differentiate into organs with specific functions, such as secretion (e.g. the salivary gland, SG) or gas exchange (e.g. the trachea) (Maruyama and Andrew, 2012). These tubular organs form directly from already polarized ectodermal epithelia, which exhibit discrete subcellular localization of polarity complexes and other junctional proteins. The SGs form from two primordia of ∼150 surface ectodermal cells found in the ventral region of parasegment two, the most posterior region of the head. Over a period of a few hours, the SG primordia invaginate, elongate and collectively migrate to form simple unbranched tubes (Maruyama and Andrew, 2012). The large cell size and simple structure, as well as the absence of cell division and cell death, make the Drosophila SG ideal for examining how changes in cell shape, adhesion and position affect tube morphology.

Cadherin 99C (Cad99C), the Drosophila ortholog of the human Usher syndrome-linked protein protocadherin 15, localizes to the apical surface of several simple undifferentiated epithelia, including ovarian follicle and wing imaginal disc cells (D'Alterio et al., 2005; Schlichting et al., 2005; Schlichting et al., 2006). Cad99C regulates microvillar length in follicle cells, with Cad99C loss reducing microvillar length and overexpression increasing microvillar length (D'Alterio et al., 2005; Schlichting et al., 2006). We understand very little about how Cad99C controls membrane dynamics and we do not know if or how this protein functions in other epithelial tissues. Similarly, the role of the Stranded at Second (SAS) protein, which also localizes to the apical surface of many epithelial tissues (Schonbaum et al., 1992), is unknown. Here, we explore the roles of both proteins in tubular epithelia. We show through loss and overexpression studies that Cad99C affects cell rearrangement during tube elongation in the Drosophila SG and that overexpression of either Cad99C or SAS confers apical identity on other membrane domains.

Staining of Drosophila embryonic SGs with known polarity components revealed that SAS localizes most apically, occupying the apical surface, which directly contacts the lumen (Fig. 1A). Along the lateral surface, the most apical staining was observed in the SAR with Crb (Fig. 1B) and other components of the Crb complex (data not shown). E-Cadherin (E-Cad) staining – along with other AJ components – was observed immediately basal to the SAR (Fig. 1B; data not shown). SJ markers, such as Coracle (Cora) (Lamb et al., 1998), and the basolateral determinants Scrib, Dlg and Lgl (Bilder and Perrimon, 2000; Strand et al., 1994; Woods and Bryant, 1991) localized just basal to the AJ (Fig. 1C,D; data not shown). Finally, the lateral membrane-associated protein α-Spectrin (α-Spec) localized along the entire lateral domain (Fig. 1D).

Cad99C mRNA is maternally contributed and, at later embryonic stages, is expressed in several tubular organs, including the SG, trachea, foregut and hindgut (Fig. 1E). As observed with SAS, Cad99C protein localized to the apical surface in the SG and other tubular organs, distinct from the sub-apical localization of Crb (Fig. 1F) and consistent with its localization in follicle cells and imaginal discs (D'Alterio et al., 2005; Schlichting et al., 2005; Schlichting et al., 2006). Super-resolution images revealed that Cad99C and SAS are mostly non-overlapping, indicating that the two proteins occupy distinct regions on the apical surface (Fig. 1G).

Cad99C expression in developing tubular organs led us to examine Cad99C function in the SG. Most embryos from homozygous mutant females (Cad99C^M) are not recoverable, whether the females are crossed to Cad99C mutant or to wild-type males, as Cad99C is required for eggshell integrity (D'Alterio et al., 2005; Schlichting et al., 2006). Surprisingly, most (>90%) of the Cad99C^M/Z embryos we did recover (from crosses of homozygous females and males) developed relatively normally but had longer, thinner SG lumens than wild type (Fig. 2A,B); only a small subset of the recovered Cad99C^M/Z embryos showed severe morphological defects (supplementary material Fig. S1). Embryos missing only maternal (Cad99C mutant females crossed to wild-type males) or only zygotic function (heterozygous Cad99C mutants crossed to each other) did not have overt defects, suggesting that either maternal or zygotic function of Cad99C is sufficient for embryonic development (data not shown).

To investigate the cellular basis for the longer thinner SGs observed in Cad99C^M/Z embryos, we carried out transmission electron microscopy (TEM) analysis using the high-pressure freezing (HPF) method that allows better preservation of cellular membranes than traditional fixation methods (McDonald and Auer, 2006). We also carried out confocal imaging with a variety of polarity and tissue-specific markers. No phenotypic differences were detected at the ultrastructural level in comparisons of stage 15/16 Cad99C^M/Z mutant and wild-type SGs (Fig. 2C). Moreover, overall polarity of Cad99C^M/Z mutant SGs appeared unaffected and the mutant SGs migrated normally, contacting the same tissues at all stages as wild-type SGs (Fig. 2E; data not shown). Total cell numbers were not significantly different (Table 1), and the apical domain size and elongation ratio of individual cells were also unchanged (Fig. 2D,F). Indeed, the only difference was that fewer cells surrounded the lumen in cross-sections of the Cad99C^M/Z mutant SGs compared with wild type (Fig. 2E; Table 1). Thus, the longer, thinner SG lumens in Cad99C^M/Z mutants reflect differences in cell rearrangement during tube elongation.

Overexpression of full-length Cad99C protein (Cad99C-FL; D'Alterio et al., 2005) in otherwise wild-type SGs using fork head (fkh)-Gal4 (Henderson and Andrew, 2000) resulted in a dramatic expansion of luminal width with differences first visible during invagination (Fig. 3A,B) and more obvious in late embryos (Fig. 3C-F). The apical domain of individual cells in Cad99C-FL-overexpressing SGs was expanded significantly (Fig. 3F), with the apical membranes extending deep into the lateral regions on the apical side, where staining of all tested apical markers was observed, including SAS, Crb, aPKC and β_H-spectrin (Fig. 3D,E; data not shown). TEM analysis of stage 15/16 Cad99C-FL-overexpressing SGs revealed increases in apical, microvilli-like structures extending into the lumen, as well as in apical membrane protruding deep into the regions between cells (Fig. 3G). Interestingly, the apical ECM appeared less regular in Cad99C-overexpressing SGs, with large regions of very sparse electron-dense material (Fig. 3G). Importantly, confocal cross-sections of Cad99C-FL-overexpressing SGs revealed many more cells surrounding the lumen than in wild type (Fig. 3H; Table 1), indicating that cell rearrangement during tube elongation is also affected by Cad99C overexpression. Overall luminal length of Cad99C-FL-overexpressing SGs was not significantly different from wild type despite the apical expansion of individual cells, likely reflecting some length compensation due to the defects in cell arrangement observed in Cad99C-FL overexpressing SGs relative to wild type (Fig. 3F).

Intriguingly, morphological changes associated with Cad99C-FL overexpression were not confined to the apical domain. Throughout embryogenesis, Cad99C-overexpressing SGs had irregular basal boundaries, with indentations between individual cells, distinct from the smooth basal surfaces of controls (Fig. 3B). By late embryogenesis, several Cad99C-overexpressing cells had lost the typical columnar shape and were either cuboidal or round, with minimal contact between neighbors (Fig. 3D). The rounded cells always had mislocalized Cad99C signals in the basal domain. Overexpression of either Cad99C-FL or Cad99C-FL-GFP (Schlichting et al., 2006) in the trachea using breathless (btl)-Gal4 (Shiga et al., 1996) resulted in similar phenotypes, including apical membrane expansion and cell rounding that correlated with mislocalized Cad99C in the basal membrane, suggesting a conserved role for Cad99C in different tubular epithelia (supplementary material Fig. S2).

Rounded basal domains were invariably associated with Cad99C mislocalization to the basal membrane, and with weak/punctate overlapping SAS staining (Fig. 4B-E). These phenotypes are reminiscent of the overexpression phenotype of Crb, a known apical determinant, in epidermal cells (Wodarz et al., 1995). Crb overexpression in SGs also expanded the lumen and caused the basal domain to acquire apical character, based on SAS mislocalization (Fig. 4F). Interestingly, Cad99C was not detected in the ‘converted’ domain where Crb and SAS mislocalized (Fig. 4F). Similarly, Crb was barely detectable in the ‘converted’ apical domain observed with Cad99C overexpression (Fig. 3D), suggesting that the mislocalized SAS observed with Cad99C overexpression is not caused by Crb mislocalization.

To investigate the function of the extra- and intracellular domains of Cad99C, we expressed each using fkh-Gal4 (Fig. 4A; supplementary material Fig. S3A). Cad99CΔCyt (Schlichting et al., 2006) overexpression resulted in apical SG membrane expansion (Figs 3F and 4G), but overexpressed Cad99CΔEx (Schlichting et al., 2006) did not affect SG morphology (supplementary material Fig. S3C). Confocal and TEM analyses revealed that Cad99CΔCyt expression in SGs, like Cad99C-FL, increases microvilli-like structures and causes apical membranes to protrude into the lateral domain (Fig. 4G,H).

Some phenotypes of Cad99CΔCyt overexpression were distinct from those resulting from Cad99C-FL overexpression. The tubes were longer than with Cad99C-FL overexpression (Fig. 3F) and the average number of nuclei per cross-section was not significantly different in Cad99CΔCyt-overexpressing compared with wild-type SGs, suggesting normal cell rearrangement (Fig. 4I; Table 1). Strikingly, in Cad99CΔCyt-expressing SGs, Cad99C signals were highest in the basal region (Fig. 4G). As the Cad99CΔEx protein localized apically (supplementary material Fig. S3), we conclude that the cytoplasmic domain mediates apical targeting and/or retention. Cad99CΔCyt-localizing basal domains were always associated with rounded cell morphology and weak SAS mislocalization (Fig. 4G). As endogenous Cad99C protein was not observed in the basal domain when stained with a cytoplasmic region-specific antibody, and as Cad99CΔCyt overexpression in a Cad99C-null background resulted in the same phenotype as in a wild-type background, the observed phenotypes are not due to redistribution of the endogenous protein (data not shown). Altogether, these data suggest that the extracellular (and transmembrane) domain of Cad99C is sufficient to confer apical character on membranes but does not affect cell rearrangement during tube elongation.

We expressed truncated versions of Cad99C that lack different parts of the cytoplasmic domain. Cad99CΔPBD, which lacks only the putative PDZ-binding domains, correctly localized to the apical membrane, suggesting that this domain is not required for localization (data not shown). Interestingly, Cad99CΔCyt-31, which deletes most of the cytoplasmic region leaving the 31 residues immediately after the transmembrane domain intact (D'Alterio et al., 2005), also correctly localized to the apical membrane, suggesting that this small juxtamembrane cytoplasmic region is sufficient for apical accumulation (Fig. 4A,J).

Cad99C overexpression causes a huge increase in vesicular structures both in the apical domain and throughout the cytoplasm, suggesting increased apical targeting (supplementary material Fig. S4B). Indeed, Cad99C overexpression resulted in dramatic increases in staining in the apical region of Rab11 and Rab6 (supplementary material Fig. S4C-F), two Rab GTPases implicated in apical trafficking (Coutelis and Ephrussi, 2007; Del Nery et al., 2006; Kerman et al., 2008; Januschke et al., 2007; Mallard et al., 2002; Martinez et al., 1997,, 1994; Opdam et al., 2000; Pelissier et al., 2003; Saraste and Goud, 2007; Satoh et al., 2005; Strickland and Burgess, 2004). Consistent with the increased Rab11 and Rab6 staining, TEM analysis revealed an approximate fourfold increase in vesicles per cell slice in the apical region of Cad99C-overexpressing SG cells compared with wild type (supplementary material Fig. S4G-I). Triple labeling of Cad99C-overexpressing SG cells revealed that both Cad99C and Cad99CΔCyt vesicular staining largely overlap with Rab6 (supplementary material Fig. S4J-M).

To elucidate the mechanisms of the cell shape changes associated with Cad99C overexpression, we examined the distribution of junctional proteins, the actin cytoskeleton and the basal ECM. Staining with the AJ markers E-Cad and Armadillo (Arm), the Drosophila β-Catenin (Peifer et al., 1993), revealed that Cad99C-FL- and Cad99CΔCyt-expressing SGs have stretched AJs and wider apical domains in each cell compared with wild type (Fig. 5A-C; data not shown), consistent with increases in apical domain size (Fig. 3F). AJ markers were also detected in the expanded apical region that protrudes into the lateral domain between neighboring cells (Fig. 5B,C). Interestingly, Cad99C-FL overexpression resulted in reduced E-Cad/Arm at the apical boundaries and their dispersion throughout the apical region near the Cad99C signals (Fig. 5B; data not shown). This finding suggests that Cad99C-FL overexpression increases apical membranes at the expense of the AJ domain. Cad99CΔCyt overexpression resulted in milder affects on E-cad/Arm localization, either due to its reduced localization to the apical domain or to the absence of the cytoplasmic region. Unlike Cad99C overexpression, Crb-overexpressing SGs had punctate/spotlike AJ marker staining (Fig. 6B), further supporting the idea that Cad99C and Crb expand apical membranes by distinct mechanisms.

Cora and Dlg SJ staining patterns were narrower and extended further along the lateral membrane in both Cad99C-FL- and Cad99CΔCyt-overexpressing SGs, with more notable changes associated with Cad99CΔCyt overexpression (Fig. 5D,E). Similar patterns with SJ markers were observed in Crb-overexpressing SGs (Fig. 6A,B). Interestingly, the Cad99C- or Crb-mislocalized domains and the Cora/Dlg-localized domains were entirely non-overlapping (Fig. 5E; Fig. 6A), suggesting mutual exclusion of the ‘apicalized’ domain and the remaining ‘lateral’ domain (hereafter referred to as SJ domain). Consistent with this idea, the SJ domain excluded the PH-domain of PLCδ (von Stein et al., 2005), a PtdIns(4,5)P₂ sensor enriched in the apical domains of fly tubular organs (Rousso et al., 2013; data not shown). SJ domains also excluded other proteins, including lateral E-Cad and F-actin; weak E-Cad and phalloidin signals were observed uniformly along the entire lateral domain in wild-type cells but were nearly absent in the SJ domains of Cad99C-overexpressing cells and were higher in the ‘converted’ apical domain (Fig. 5C,E). The more rounded Cad99C-overexpressing cells appeared to contact neighboring cells only through the SJ domain (Fig. 5E), based on a nearly complete overlap in SJ marker and α-Spec staining (supplementary material Fig. S5). Indeed, TEM analysis revealed rounded basal surfaces and wider spaces between neighboring cells in Cad99C-FL and Cad99CΔCyt-overexpressing SG cells than in wild type, indicating a loss of cell-cell contact, nearly everywhere except where the SJ proteins presumably localize (Fig. 5F). By contrast, most Crb-overexpressing SG cells maintained tight contacts with their neighbors along the entire lateral domain (Figs 4F and 6A), perhaps through homophilic adhesion between the Crb proteins on neighboring cells (Tanentzapf et al., 2000). These data support the idea that Cad99C does not mediate homophilic adhesion (D'Alterio et al., 2005) and suggest that membrane accumulation of Cad99C specifically disfavors cell-cell contact, leading to cell rounding.

The morphological changes in the basolateral domain of SG cells also correlated with changes in ECM organization, based on localization of Nidogen (Ndg), a glycoprotein with important linker functions in basement membranes (Hynes and Zhao, 2000). In Drosophila, Ndg-positive ECM, which is deposited by a subset of mesodermal cells (Broadie et al., 2011), is detected in thin sheets attached to the surface of many internal organs (Wolfstetter and Holz, 2012), with noticeable intensity around SGs at later stages of development (Fig. 5G). Compared with the strong Ndg localization surrounding the basal side of wild-type SGs (Fig. 5G), Cad99C-FL- or Cad99CΔCyt-expressing SGs with rounded basal domains showed only trace amounts of Ndg staining (Fig. 5H; data not shown). Notably, Crb-overexpressing SGs had largely intact Ndg-positive ECM structures surrounding the tissue (Fig. 6C).

As the ‘converted’ apical Cad99C domains always had mislocalized SAS, we asked whether SAS was required for conferring apical character by Cad99C expression. SGs in embryos missing either sas alone or missing both sas and zygotic Cad99C were largely normal (data not shown). Overexpression of Cad99CΔCyt in an sas-null background (transheterozygotes of sas protein null mutant and deficiency) showed the same SG phenotype as in a wild-type background (Fig. 7A), suggesting that the Cad99C overexpression phenotypes are independent of SAS. Surprisingly, SAS overexpression also expanded apical membranes (Fig. 7B,C). TEM analysis revealed both increased microvilli-like apical structures and apical membrane expansion into the lateral domain (Fig. 7E). E-Cad staining was somewhat reduced and irregular in the AJ domain, and was increased in basolateral domains in the full-length SAS (SAS-FL)-overexpressing SG cells, except in the SJ domain (Fig. 7C; data not shown). SAS-FL overexpression with a stronger line caused more severe morphological defects with rounded cells that correlated with mislocalized basal SAS (Fig. 7D). Very strong F-actin signals were observed in the apical membranes that extended into the lateral domain (Fig. 7D) and elongated SJ domains were observed in regions devoid of SAS (Fig. 7D), reminiscent of both Cad99C- and Crb-overexpression phenotypes. The SAS-mislocalized domain did not stain with antibodies to Cad99C or Crb (data not shown), suggesting that all three proteins are capable of ‘apicalizing’ membrane independently. As with Cad99C, overexpressed SAS-FL signals colocalized with Rab6 (Fig. 7F).

SAS contains several motifs involved in protein-protein interaction, including von Willebrand factor (vWF) type C domains and fibronectin type 3 (FN3) domains in its large extracellular domain; no known functional motifs have been discovered, however, in the small cytoplasmic domain (consisting of only 37 residues). Overexpressed SASΔCyt (entire cytoplasmic region deleted) localized apically and almost completely phenocopied the SAS-FL overexpression phenotypes (supplementary material Fig. S6). Interestingly, SASΔCyt-overexpressing SG cells have Crb dispersed along the apical surface, unlike with the SAS-FL-overexpressing cells, where Crb primarily localized to the normal SAR domain (supplementary material Fig. S6B,C).

Unexpectedly, strong Ndg signals were observed not only along the basal boundary but also in the apical ECM of SAS-FL-or SAS-ΔCyt-overexpressing SGs (Fig. 7G; supplementary material Fig. S6E,F). As no changes in ndg mRNA levels and/or localization were detected (supplementary material Fig. S7), these data suggest abnormal ECM protein turnover in SAS-overexpressing SGs. Moreover, apical Ndg was never observed with overexpression of Cad99C or Crb, suggesting that this is unique to SAS. Co-overexpression of Cad99C-FL and SAS-FL resulted in substantial expansion of apical membranes at the expense of AJ regions, indicated by the dispersed E-Cad signals (Fig. 7H), suggesting additive effects. Loss of Cad99C did not affect the SAS-FL overexpression phenotype, further supporting the idea that Cad99C and SAS expand apical membranes independently of each other (Fig. 7I).

Altogether, these studies reveal that two apical surface transmembrane proteins – Cad99C and SAS – can confer apical properties to other membrane domains. These properties include a loss of cell-cell contact, ectopic accumulation of apical markers with a corresponding loss of the junctional markers that are characteristic of lateral domains, as well as changes in the distribution of ECM markers that are characteristic of basal domains. Our studies also suggest that the SJ, the site where known basolateral determinants are found, is the membrane domain most refractory to apical conversion.

Here, we describe previously undiscovered roles for apical surface proteins in regulating apical membranes and impacting overall tissue architecture in Drosophila tubular epithelia. As loss and overexpression of Cad99C affect cell rearrangement during tube elongation and as the apical ECM is also partially disrupted in the Cad99C-overexpressing SGs, we propose that adhesion between the apical cell surface and apical ECM affects cell rearrangement in Drosophila SGs (Fig. 8). We propose that the adhesion between the apical surface and apical ECM through Cad99C serves to counteract the as yet un-described forces driving the cell rearrangement events that elongate the SG tube. With loss of Cad99C, adhesion is weakened, allowing cells to rearrange more easily, resulting in longer SGs with fewer cells surrounding the tube circumference. Correspondingly, too much Cad99C increases adhesion, making it more difficult for tubes to elongate by cell rearrangement, resulting in wider glands with more cells surrounding the tube circumference. Normal cell rearrangement with Cad99CΔCyt overexpression (the same number of cells in cross section as WT) suggests a role for the cytoplasmic domain in this process. Supporting the role for Cad99C in adhesion between the apical surface and the apical ECM, Cad99C loss significantly rescues the irregular apical membrane phenotype observed in SGs mutant for the secreted AdamTS-A metalloprotease (Ismat et al., 2013). Interactions between the apical ECM and membrane-spanning apical cell surface proteins have also been suggested from studies of the roles of zona pellucida (ZP) domain-containing proteins in Drosophila tracheal development (Bökel et al., 2005; Jaźwińska et al., 2003); loss of these proteins can lead to a complete loss of connections between tracheal cells undergoing tube elongation by cell rearrangement.

Overexpression of Cad99C regulates total apical area, consistent with the function of Cad99C in follicle cells in regulating microvillar length (D'Alterio et al., 2005; Schlichting et al., 2006); the extra membrane associated with Cad99C overexpression in tubular epithelia, however, appears in two forms: microvilli-like projections and expansion of the apical surface into the lateral domain, apparently at the expense of apical lateral structures, such as AJs (Fig. 8). Interestingly, SAS also drives apical area expansion when overexpressed, causing both increases in microvilli and expansion into the lateral domain.

Several lines of evidence suggest that Cad99C and SAS, like Crb, drive apical conversion in tubular epithelia (Fig. 8). First, Cad99C mislocalization always correlates with mislocalization of SAS in the same domain. Second, the actin cytoskeleton is reorganized. Strong F-actin signals are characteristic of the apical domain, and F-actin is much more enriched in the Cad99C- or SAS-mislocalized basal domain than in wild-type cells. Third, the Cad99C- and SAS-mislocalized domains are mutually exclusive with the remaining lateral domains, where SJ proteins localize. The elongated SJ protein distribution could be due to the SJ structures being disrupted because a dye exclusion assay revealed compromised barrier function in the completely rounded Cad99C- or SAS-overexpressing SG cells (data not shown) and because mislocalization of SJ proteins into the entire basolateral domain has been reported in many SJ mutants (Laval et al., 2008; Nelson et al., 2010; Paul et al., 2003; Wu et al., 2004). Alternatively, the elongated SJs could be due to cells establishing new SJ domains in more basolateral regions in response to the new apical domains being established nearby.

The cell rounding caused by Cad99C and SAS mislocalization probably reflects the loss of cell-cell adhesion, which we suspect normally imposes the columnar/cuboidal shapes on epithelial cells. Indeed, one feature of the apical surface is favoring cell-matrix over cell-cell contact. Moreover, the loss of basal ECM (Ndg) observed with Cad99C overexpression also suggests a loss of basal character; the cells either no longer properly localize the basal proteins that bind Ndg and/or other basal ECM components or the basally localized Cad99C somehow interferes with this binding. The abnormal Ndg localization to both apical and basal surfaces in SAS-overexpressing SGs supports some level of basal ECM internalization (Godyna et al., 1995; Memmo and McKeown-Longo, 1998; Wienke et al., 2003) and recycling to the membrane, given that SG cells do not synthesize this protein (supplementary material Fig. S7). Indeed, co-overexpression of Cad99CΔCyt and SAS results in a complete loss of Ndg staining in the region of the SG (data not shown). The differential effects of Cad99C and SAS on Ndg accumulation suggest unique activities for each protein in establishing/maintaining the apical surface. Cad99C and SAS may have a common role in binding to and organizing the apical ECM, but the exact components they bind both inside and outside the cell are likely to be distinct.

Our finding that two apical membrane components – Cad99C and SAS – can confer apical character on other membrane domains suggests that multiple avenues exist for establishing and/or maintaining overall epithelial polarity. Learning how these and the previously known determinants function should provide new insight into the contributions of cell-cell and cell-matrix interactions towards specifying polarized membrane domains. Importantly, our findings on the role of Cad99C in apical membrane expansion – potentially mediated through interactions with the apical ECM – may provide additional insight into how mutations in the human Usher proteins contribute to progressive sensory loss.

Fly strains used in this study were: Oregon R; Cad99C^57A, Cad99C^120B, UAS-Cad99C-FL-GFP, UAS-Cad99CΔCyt, UAS-Cad99CΔEx (Schlichting et al., 2006); UAS-Cad99C-FL, UAS-Cad99CΔCyt-31-GFP (D'Alterio et al., 2005); fkh-Gal4 (Henderson and Andrew, 2000); btl-Gal4 (Shiga et al., 1996); ubi-RFP-Rab6 (Januschke et al., 2007); UAS PLCδ-PH-GFP (von Stein et al., 2005); and sas¹⁵ (Lewis et al., 1980; Lee et al., 2013).

To generate the UAS-Cad99CΔPBD transgenic flies, the Cad99C open reading frame (ORF) deleting the putative PDZ-binding domains [nine amino acid residues (SEVETTTEL) from the C-terminal region of Cad99C, underlined residues indicate consensus of Class I PDZ domain-binding sites] was amplified using the UAS-Cad99C-FL genomic DNA as a template. To generate the UAS-SAS-FL and UAS-SASΔCyt transgenic flies, the entire SAS ORF and the fragment deleting the cytoplasmic domain (37 amino acids) were amplified, respectively, using LD44801 EST clone as a template. The constructs were subcloned into the pUAST vector (Brand and Perrimon, 1993) using the Gateway system (Carnegie Institution).

Primary antibodies included anti-β-gal (Promega, 1:500), anti-GFP (Molecular Probes, 1:10,000), rabbit anti-Cad99C (Cad99C-RB; peptide antibody specific for the C terminus of the Cad99C protein; C. Dahmann, Max Planck Institute of Molecular Cell Biology and Genetics, Germany; 1:3000), guinea pig anti-Cad99C (Cad99C-GP; polyclonal antibody specific for the extracellular region of the Cad99C protein; D. Godt, University of Toronto, Canada; 1:3000), rat α-Cad99C (Cad99C-RT; polyclonal antibody specific for the intracellular region of the Cad99C protein; D. Godt; 1:25), anti-Crb (DSHB, 1:10), anti-SAS (D. Cavener, Penn State University, PA, USA; 1:500), anti-α-Spec (DSHB, 1:1), anti-aPKC (Santa Cruz, 1:200), anti-β_H-spectrin (G. Thomas, Pennsylvania State University, USA; 1:100), anti-DE-Cad (DSHB, 1:10), anti-Arm (DSHB, 1:100), anti-Cora (R. Fehon, University of Chicago, IL, USA; 1:2000), anti-NrxIV (H. Bellen, Baylor College of Medicine, Houston, TX, USA; 1:2000), anti-Dlg (DSHB, 1:500), anti-Rab11 (S.C. and D.J.A., unpublished) and anti-Ndg (A. Holz, Institut für Allgemeine und Spezielle Zoologie, Germany; 1:2500). Fluorescence-labeled secondary antibodies were used at a 1:500 dilution (Invitrogen). Phalloidin-546 was used at a 1:250 dilution (Invitrogen).

Embryo were fixed and stained following standard protocols, except for the α-E-Cad and α-Arm stainings, for which embryos were fixed in 4% paraformaldehyde in PBS and devitellinized with ethanol. For phalloidin staining, embryos were fixed in 1:1 formaldehyde:heptane for 40 min and hand devitellinized. All confocal images were obtained with Zeiss LSM 510. Super-resolution images were obtained with Zeiss ELYRA SR-SIM.

In situ hybridization was performed as described previously (Lehmann and Tautz, 1994). LD23052 cDNA (DGRC) was used to generate an anti-sense digoxigenin-labeled Cad99C RNA probe. An anti-sense RNA probe for ndg was made using a PCR fragment (nucleotides 1383-3046) obtained by reverse transcriptase PCR (RT-PCR). cDNAs were made from the total RNAs isolated with TRIzol (Invitrogen) from wild-type embryos.

Stage 16 SGs were used for analyses in all cases, except for the SG luminal length measurement shown in Fig. 2B, where SGs of both stage 15 and 16 were measured. Cad99C^120B/Cad99C^57A (M/Z) and Cad99C^57A/Cad99C^120B (M/Z) mutants are the progeny from crossing homozygous Cad99C^120B females and homozygous Cad99C^57A males, and from crossing homozygous Cad99C^57A females and homozygous Cad99C^120B males, respectively. Maternal mutants Cad99C^120B/+ (M) and Cad99C^57A/+ (M) are the progeny from crossing homozygous Cad99C^120B and homozygous Cad99C^57A females to wild-type males, respectively. Cad99C^57A/Cad99C^120B and Cad99C^120B/Cad99C^57A zygotic mutants are the progeny from crossing heterozygous parents. The images of stage 15 and stage 16 embryos (Crb staining) from perfect ventral views were taken with ProgRes C14plus camera (Jenoptik). The middle of the SG lumen was traced using the Image J program (NIH). For quantification shown in Fig. 4F, lateral views of the projected confocal images of wild-type, Cad99C-FL- and Cad99CΔCyt-overexpressing SGs (SAS staining) were used.

To measure the luminal area, confocal images of Crb or SAS staining were projected to visualize the entire lumen. The luminal area was traced and calculated using ImageJ. To measure the apical domain size of individual SG cells, confocal images of E-Cad staining were traced and measured using ImageJ. The elongation ratio of the same cells was determined by calculating the ratio of apical domain length oriented along the proximal-distal (PD) axis to the apical domain length along the dorsal-ventral (DV) axis passing the centeroid of each cell. Statistical significance was calculated using Student's t-test (two-tailed).

Wild-type, Cad99C^120B/Cad99C^57A (M/Z), fkh>Cad99C-FL and fkh>Cad99CΔCyt embryos were stained for membrane markers and with DAPI. Cross-sections of confocal images of SGs were obtained with the Zeiss LSM 510 program. Nuclei were counted in three independent cross-sections per SG and averaged. Statistical significance was calculated using Student's t-test (two-tailed).

Dechorionated embryos were high pressure frozen with the Leica EM HPM 100, in 200 μm aluminum specimen carriers filled with yeast paste containing 10% methanol (McDonald and Morphew, 1993). Samples were then freeze substituted in a fixing cocktail containing 1% osmium, 0.1% uranyl acetate, 95% acetone and 5% water for better preservation of membranes (Walther and Ziegler, 2002). The freeze substitution was performed in a Leica EM AFS 2/FSP with a modified schedule (correspondence Rick Fetter, Janelia Farms HHMI). Samples were gradually infiltrated with 33%/50%/75% increasing Eponate 12 (Pella) mixture in acetone, with a pure resin infiltration overnight, and cured at 60°C for 2 days. Sections (60 nm) were stained with 1% tannic acid followed by 2% uranyl acetate and 0.3% lead citrate. Grids were viewed on a Phillips CM120 Transmission Electron Microscope and digital images captured with an ER-80 AMT 8 megapixel CCD camera. Sections from at least four independent embryos were analyzed for wild type and fkh-Gal4>Cad99C-FL, and from two embryos for Cad99C^M/Z, fkh-Gal4>Cad99CΔCyt and fkh-Gal4>SAS-FL.

We thank C. Dahmann, D. Godt, D. Cavener, R. Fehon, H. Bellen, A. Holz, G. Thomas, the Bloomington Stock Center and Drosophila Genomics Resource Center for providing fly stocks, antibodies or cDNAs used in this study. We thank FlyBase for information on the genes used in this study (St Pierre et al., 2014). We also thank M. Delannoy and B. Smith of the microscope facility in the Johns Hopkins University School of Medicine for technical assistance for TEM analysis. In addition, we thank members of the Andrew laboratory, and A. Ewald, A. Hubbard, G. Seydoux and three anonymous reviewers for their comments on this manuscript.