Microvilli and related actin-based protrusions permit multiple interactions between cells and their environment. How the shape, length and arrangement of microvilli are determined remains largely unclear. To address this issue and explore the cooperation of the two main components of a microvillus, the central F-actin bundle and the enveloping plasma membrane, we investigated the expression and function of Myosin VIIA (Myo7A), which is encoded by crinkled (ck), and its interaction with cadherin Cad99C in the microvilli of the Drosophila follicular epithelium. Myo7A is present in the microvilli and terminal web of follicle cells, and associates with several other F-actin-rich structures in the ovary. Loss of Myo7A caused brush border defects and a reduction in the amount of the microvillus regulator Cad99C. We show that Myo7A and Cad99C form a molecular complex and that the cytoplasmic tail of Cad99C recruits Myo7A to microvilli. Our data indicate that Myo7A regulates the structure and spacing of microvilli, and interacts with Cad99C in vivo. A comparison of the mutant phenotypes suggests that Myo7A and Cad99C have co-dependent and independent functions in microvilli.
Actin-filament-based protrusions of the plasma membrane have important functions in cell structure, behavior, cell–cell or cell–environment interactions and physiology. Finger-like protrusions called microvilli are often found on the apical surface of epithelial cells. The microvilli of the ovarian follicular epithelium in Drosophila are a prominent example (Mahowald and Kambysellis, 1980; Trougakos et al., 2001; D'Alterio et al., 2005; Schlichting et al., 2006). Furthermore, microvilli can give rise to complex structures in sensory organs, such as rhabdomeres and bristles in insects and hair cell stereocilia in the vertebrate ear, which transduce light and mechanosensory stimuli, respectively (reviewed by DeRosier and Tilney, 2000; Frolenkov et al., 2004).
A microvillus contains a bundle of parallel actin filaments enveloped by plasma membrane. Plus ends of actin filaments point distally and minus ends are anchored in the terminal web, a meshwork of actin filaments and associated proteins in the apical cytocortex (reviewed by Frolenkov et al., 2004). The dynamics of actin polymerization and depolymerization control the length of microvilli. A number of factors have been identified that contribute to the length regulation of microvilli or microvillus-derived stereocilia and their organization and arrangement on the apical surface of epithelial cells. Many of these factors directly bind to actin in protrusions, either cross-linking actin filaments or linking them to the plasma membrane (reviewed by Lin et al., 2005; Sekerková et al., 2006; Manor and Kachar, 2008; Tepass, 2009; Fehon et al., 2010). In particular, myosin motor proteins, which bind to and travel along actin filaments, control multiple processes in microvilli and stereocilia. Myosins have been implicated in transporting, anchoring and concentrating protein complexes and membrane vesicles, in exerting tension force on the plasma membrane and in influencing the structure of actin networks (reviewed by Krendel and Mooseker, 2005; Nambiar et al., 2010; Schwander et al., 2010; Hartman et al., 2011).
One of the myosins that regulates actin-rich membrane protrusions in Drosophila and vertebrates is Myosin VIIA (Myo7A) (reviewed by El-Amraoui et al., 2008). This molecule is highly conserved in structure and sequence between flies and mammals. Myo7A consists of a motor head domain, a neck with IQ (isoleucine-glutamine) motifs and a tail containing MyTH4 (myosin tail homology) and FERM (Four-point-one, Ezrin, Radixin, Moesin)-MyTh7 domains and an SH3 (src homology F3) sequence (Kiehart et al., 2004). The head domain provides ATPase-dependent plus-end-directed motor activity along actin filaments (Udovichenko et al., 2002; Inoue and Ikebe, 2003). The activity of Myo7A seems to be controlled by the density of the actin network, as binding to actin is required for both the release of the autoinhibitory interaction between the head motor domain and the second FERM-MyTH7 domain and the subsequent activation of the ATPase (Yang et al., 2009; Umeki et al., 2009; Inoue and Ikebe, 2003). Once activated, Myo7A appears to be predominantly locked in an ADP-bound state, in which it strongly binds to actin. Its high duty ratio allows Myo7A to act as a processive motor when dimerized (Watanabe et al., 2006; Yang et al., 2006; Haithcock et al., 2011). Myo7A is thought to dimerize with the help of factors binding to its tail region (Yang et al., 2009; Umeki et al., 2009; Haithcock et al., 2011). The high-duty processive motor activity of Myo7A is consistent with a function in transporting cargo or in anchoring plasma membrane components to the actin bundle to generate tension forces (reviewed by El-Amraoui et al., 2008).
Mutations in the gene crinkled (ck), which encodes Drosophila Myo7A, cause pronounced structural defects in apical actin-based protrusions of the epidermis, such as mechanosensory bristles and non-innervated denticles and hairs (trichomes) (Nüsslein-Volhard et al., 1984; Gubb et al., 1984; Turner and Adler, 1998; Kiehart et al., 2004). Myo7A collaborates with the membrane-bound extracellular protein Miniature and the actin cross-linking factor Forked in denticle formation (Bejsovec and Chao, 2012). Moreover, Myo7A-deficient flies are deaf, owing to defects in the extracellular dendritic caps that attach the auditory sensory cells to the epidermis (Todi et al., 2005; Todi et al., 2008).
In vertebrates, Myo7A belongs to a group of molecules that are linked to Usher type I syndrome (USH1; reviewed by El-Amraoui et al., 2008). A hallmark trait of Myo7A mutants in vertebrates is deafness, which is caused, at least in part, by defects in the stereocilia of hair cells, including an abnormal arrangement of stereocilia, loss of cohesion between stereocilia, defects in length regulation of stereocilia and disruption of links between stereocilia (Self et al., 1998; Ernest et al., 2000; Küssel-Andermann et al., 2000; Michalski et al., 2007; Prosser et al., 2008; Lefèvre et al., 2008; Rzadzinska et al., 2009). Myo7A has been found to interact with several transmembrane proteins and scaffold proteins, promoting their presence and possibly transport in stereocilia (reviewed by El-Amraoui et al., 2008). Particularly well documented is the association of Myo7A with other USH1-linked proteins, cadherin 23 (CDH23), PDZ domain protein harmonin, scaffold protein SANS and the Ca2+ and integrin-binding protein CIB2 (Boëda et al., 2002; Adato et al., 2005; Bahloul et al., 2010; Grati and Kachar, 2011; Sahly et al., 2012; Riazuddin et al., 2012). Data corroborate the model that Myo7A forms a tension-generating link between the F-actin core of a stereocilium and CDH23 in the plasma membrane that is crucial for the function of the tip link, a connection between stereocilia that is thought to be involved in the mechanotransduction of hair cells (Grati and Kachar, 2011; Bahloul et al., 2010). Furthermore, in vitro data indicate that Myo7A also interacts with protocadherin 15 (PCDH15) (Senften et al., 2006), another USH1-associated protein that appears to interact heterophilically with CDH23, thereby connecting two neighboring stereocilia through a tip link (Kazmierczak et al., 2007). Similar to Myo7A mutants, Pcdh15 mutants show a severe disorganization of stereocilia bundles (Alagramam et al., 2000; Alagramam et al., 2001; Raphael et al., 2001; Seiler et al., 2005; Lefèvre et al., 2008). A recent study suggests that Myosin VIIB, a close relative of Myo7A, interacts with protocadherin-24 and mucin-like protocadherin, two heterophilically interacting cadherins that are involved in the proper formation of the brush border in enterocytes (Crawley et al., 2014).
Previous work by us and others has revealed that the fly ortholog of PCDH15, Cad99C, is a regulator of the integrity and length of follicle cell microvilli in Drosophila. In the absence of Cad99C, the brush border develops severe defects, including shortened or collapsed and irregularly distributed microvilli. By contrast, overexpression of Cad99C led to highly elongated microvilli (D'Alterio et al., 2005; Schlichting et al., 2006). To better understand how this cadherin regulates microvillus structure and brush border organization, we looked for interaction partners. Myo7A was an attractive candidate as it is known to affect various actin-based protrusions in flies (Kiehart et al., 2004), and has been shown to interact with PCDH15 in vitro (Senften et al., 2006). Here, we investigated the expression and function of Myo7A in oogenesis and undertook an in vivo analysis of the interactions between Myo7A and Cad99C.
Myo7A associates with parallel bundles of actin filaments during Drosophila oogenesis
Myo7A distribution was analyzed with a new antibody, the specificity of which was confirmed by western blot analysis (supplementary material Fig. S1A) and immunostaining (supplementary material Fig. S1B,C). Myo7A was found in the germline and associated somatic cells throughout oogenesis, showing developmentally regulated changes in expression level and subcellular distribution (Fig. 1). The main characteristics of the subcellular localization of Myo7A are a punctate distribution throughout the cytoplasm and association with specific F-actin-rich cell structures.
In the germline, the amount of Myo7A protein increased during follicle formation in the germarium and again at mid-oogenesis (stage 9) and was prominent until late oogenesis (Fig. 1A–J). Myo7A accumulated in the F-actin-rich cytocortex of the oocyte (stage 3 onwards, peaking at stages 9–10a) and, to a lesser degree, in the cortex of nurse cells (stage 6 onwards) (Fig. 1B–E). Myo7A was highly concentrated at two prominent F-actin structures. It colocalized with nurse cell struts (Fig. 1G,I; Fig. 2A), which consist of a series of overlapping bundles of parallel-oriented unidirectional actin filaments that anchor the nurse cell nuclei (Tilney et al., 1996). Myo7A was also found at ring canals (Fig. 1C; Fig. 2B), cytoplasmic bridges that connect the cells of a germline cyst. Ring canals consist of a dense inner ring of bipolar actin filaments and an associated crown and fishing-weir-like basket of needle-like actin bundles (Nicolas et al., 2009; reviewed by Hudson and Cooley, 2002). Myo7A was strongly enriched at the crown (Fig. 2B) and basket (not shown), but little was seen at the inner ring (Fig. 2B).
Myo7A was also expressed by somatic cells of the ovary, including all follicle cells (Fig. 1A–L). Compared with the germline, somatic expression was relatively low in early stages but increased after stage 8 when follicle cells undergo morphogenetic changes. Examples are the migrating border cells, which show high amounts of Myo7A in their actin-rich cellular protrusions (Fig. 1D,E; Fig. 2C), and the developing tubular dorsal appendages (Fig. 1H,K) and micropyle (Fig. 1L). In the follicular epithelium surrounding the oocyte, Myo7A was concentrated at the apical and basal pole and present at low levels throughout the cytoplasm (Fig. 1E). Similar to the germline, Myo7A shows a clear association with F-actin structures in columnar follicle cells. On the basal side, where bundles of parallel actin filaments span the cell floor similar to stress fibers (reviewed by Bilder and Haigo, 2012), Myo7A appeared to be particularly concentrated at their lateral attachment points, but was also found in a dotted pattern along the actin fibers (Fig. 2D). In the apical region, Myo7A associated with the microvilli that project from the surface of the follicle cells towards the oocyte (Fig. 2E,F). In summary, subcellular enrichment of Myo7A showed a strong association with areas rich in parallel bundles of actin filaments in both germline and somatic cells.
Myo7A is a component of the apical microvilli of follicle cells
Our analysis of Myo7A distribution in microvilli focused on stage 10, when a mature brush border has developed (D'Alterio et al., 2005). F-actin staining highlights the microvilli and the apical cytocortex (Fig. 2E). Reconstructed Z-stacks indicated that the actin scaffold that supports a microvillus consists of the F-actin bundle, which projects outward into the microvillus, a bulbous structure at the base of the protrusion, and a rootlet projecting basally into the cytocortex (Fig. 2F). A prominent overlap between Myo7A and F-actin was only observed at the base of microvilli (Fig. 2E,F). Here, the bulk of Myo7A appeared to be inserted into the bulbous F-actin structure and lower amounts were detected apically and basally to it (Fig. 2F).
To determine whether Myo7A is located inside microvilli, we compared its distribution to that of Cad99C, which is restricted to microvilli and found along their entire length (D'Alterio et al., 2005; Schlichting et al., 2006). Both Myo7A and Cad99C were seen at the interface between oocyte and follicle cells where the brush border is located (Fig. 3A). The spiky pattern of Myo7A signal and colocalization with Cad99C revealed the presence of Myo7A in follicle cell microvilli (Fig. 3B,C). High amounts of Myo7A were observed in the basal region of microvilli, and lower amounts were observed in the apical region (Fig. 3B; supplementary material Fig. S2A,B). However, the strongest signal of Myo7A resided below the Cad99C-positive zone of microvilli (Fig. 3B; supplementary material Fig. S2A,A′). Reconstructed Z-stacks indicate that each Cad99C-positive microvillus sits on a Myo7A-rich pedestal (Fig. 3C,D), a structure previously suggested to associate with the F-actin bulb (Fig. 2F). Expression of Myo7A::GFP in follicle cells yielded the same distribution as seen with the Myo7A antibody (supplementary material Fig. S2C).
The presence of Myo7A in the cytocortex was confirmed by colocalization with the apical marker Patj (Tanentzapf et al., 2000) (Fig. 3E). Moreover, Myo7A showed a largely alternating distribution with beta(Heavy)-spectrin (βH-spectrin) and Myosin II (Fig. 3F–I), two proteins of the terminal web that associate with the actin filament network between microvillar rootlets (Hirokawa et al., 1982; Mooseker et al., 1984; Phillips and Thomas, 2006). βH-spectrin (Fig. 3F,G) and Zipper (Zip), the heavy chain of Myosin II (Fig. 3H), filled the spaces between Myo7A foci in the apical cytocortex. In contrast to Myo7A, however, Myosin II was excluded from the region directly beneath the apical plasma membrane (Fig. 3I), as has also been reported for other epithelia (Mooseker et al., 1984; Thomas and Kiehart, 1994).
Taken together, our data indicate that Myo7A is strongly enriched in the apical cytocortex, where it associates with the microvillar rootlets and shows the highest concentration just below the apical membrane and in the basal region of follicle cell microvilli. Myo7A is also present, albeit at lower concentrations, in the apical parts of microvilli, indicating that Myo7A and Cad99C colocalize throughout the microvillus.
Myo7A is required for normal brush border morphogenesis
To determine whether Myo7A has a function in microvilli, we induced ck mutant cell clones in the follicular epithelium and studied them in stage-10 follicles where the brush border is fully developed. ck13 mutant cells were negative for Myo7A (Fig. 4A; Kiehart et al., 2004; Todi et al., 2005). We evaluated brush border integrity by looking at the pattern of Cad99C and actin filaments that mark the microvilli and by analyzing the arrangement of vitelline bodies with Nomarski optics. Vitelline bodies are precursors of the vitelline membrane that are secreted by follicle cells, and they are located between the microvilli during stage 10 (Mahowald, 1972; Trougakos et al., 2001). Vitelline bodies appeared in a regular stripe pattern on the apical surface of wild-type follicle cells but displayed an irregular distribution in ck mutant cells (Fig. 4B,B′), suggesting that microvilli might not project straight from the follicle cell surface to the oocyte as in wild-type follicle cells. In addition, the dome-shaped appearance of the brush border was more pronounced in wild-type than in ck mutant follicle cells. Microvilli of ck mutant cells, visualized with F-actin or Cad99C, looked shorter than normal microvilli (Fig. 4C–D′; supplementary material Fig. S3A), which correlated with a decreased height of vitelline bodies (supplementary material Fig. S3B). Moreover, a face-on view of a mosaic brush border suggested that ck mutant microvilli form a denser network than wild-type microvilli (Fig. 4E,E′).
Overexpression of Myo7A, using Myo7A::GFP that can rescue the ck mutant phenotype (Todi et al., 2005), caused a slightly wider spacing of microvilli in some follicle cell clones (supplementary material Fig. S3C), but most cell clones were indistinguishable from their wild-type neighbors. No obvious change in microvillus length was observed (data not shown). The level of Myo7A overexpression seemed relatively low (even at 29°C, when the Gal4 expression system is more active) and might not have been high enough to elicit consistent defects.
To gain more insight into the structural defects of the brush border of ck mutant follicle cells, we used transmission electron microscopy (TEM) images of early to mid stage 10b follicles for further analysis (Fig. 4G–R). In this case, all cells of a follicle, including germline and follicle cells, were either mutant or wild type. Prominent vitelline bodies are present at the oocyte–follicle cell boundary that are sandwiched between follicle cell microvilli and covered by oocyte microvilli in wild-type (Mahowald, 1972; Trougakos et al., 2001; Fig. 4G,I,K) and ck mutant follicles (Fig. 4H,J,L). Between vitelline bodies, oocyte microvilli and follicle cell microvilli contacted each other (Fig. 4K,L). In contrast to wild-type follicles, however, where vitelline bodies, although variable in shape, formed broad solid blocks (Fig. 4G,I), vitelline bodies formed narrower columns in ck mutants (Fig. 4H,J), consistent with differences in the spacing pattern of microvilli. Although some ck mutant follicle cell microvilli seemed to project straight towards the oocyte (Fig. 4L), as regularly observed for wild-type microvilli (Fig. 4K), mutant follicles frequently showed a crisscross pattern of vitelline bodies and microvilli that suggested irregularities in the orientation of microvilli in ck mutant follicle cells (Fig. 4L). Oocyte microvilli, which were not obscured by vitelline bodies and were therefore easier to analyze, showed clear differences. In wild-type follicles, oocyte microvilli formed a rather ordered brush border and converged onto vitelline bodies (Fig. 4M), whereas the ck mutant brush border was highly irregular and the underlying surface of vitelline bodies was jagged even at mid stage 10b when vitelline bodies begin to fuse into a continuous sheet (Fig. 4N). Defects in oocyte microvilli are consistent with an enrichment of Myo7A in the cell cortex of the oocyte (see Fig. 2E). In summary, our phenotypic analysis suggests that Myo7A is involved in regulating the structure and arrangement of follicle cell and oocyte microvilli.
TEM sections in the plane of the oocyte–follicle cell interface provided interesting information about the arrangement of follicle cell microvilli. Packed side-by-side, microvilli formed an extensive labyrinth of rows between vitelline bodies, which were usually only one, sometimes two microvilli wide (Fig. 4O). Such rows, which tended to encircle vitelline bodies, were also found in ck mutant follicles (Fig. 4P). The curved rows of microvilli explain the grid-like pattern seen in confocal images (Fig. 4E′; supplementary material Fig. S3C). Adjacent follicle cell microvilli were in close proximity (side view in Fig. 4Q and face-on view in Fig. 4R) and ‘kissing points’ were observed between their plasma membranes (Fig. 4Q). Follicle cell microvilli also made tight contact with vitelline bodies, as indicated by their imprints on the surface of vitelline bodies (Fig. 4R).
Interdependent localization of Myo7A and Cad99C in follicle cell microvilli
Cad99C was still present in the microvilli of ck mutant follicle cells, but the signal was noticeably weaker than in wild-type microvilli (Fig. 4D,D′; supplementary material Fig. S3A). Quantification of Cad99C signal intensities confirmed that the amount of this cadherin is significantly reduced in ck mutant microvilli compared to neighboring wild-type microvilli (Fig. 4F). As microvilli appeared to be shorter in ck mutant cells, the measured reduction of Cad99C is presumably an underestimate. Overexpression of Myo7A caused a higher than normal concentration of Cad99C at the base of microvilli where Myo7A is heavily enriched (supplementary material Fig. S3D). Myo7A seems not to be essential for normal subcellular localization of Cad99C but modifies the amount of this cadherin in microvilli.
We next analyzed the subcellular distribution of Myo7A in follicle cell clones that lacked Cad99C. The spiky pattern of Myo7A at the apical surface that represents its location in microvilli was missing (Fig. 5A,B). Myo7A still accumulated in the apical cytocortex, as indicated by colocalization with Patj (Fig. 5B), and appeared to be unaffected in the remaining cell body. The reduced apical Myo7A signal could be due to the strongly impaired structure and reduced size of Cad99C-depleted microvilli or could be a direct effect of the loss of Cad99C.
The cytoplasmic tail of Cad99C recruits Myo7A to follicle cell microvilli
To determine whether Myo7A localization in follicle cell microvilli is dependent on Cad99C, we tested the effects of various transgenic Cad99C isoforms. Overexpression of full-length Cad99C (Cad99C-FL) causes excessively long microvilli (D'Alterio et al., 2005). A dramatic change in the distribution of Myo7A was seen in response to Cad99C-FL expression. Myo7A appeared to be homogeneously distributed along the entire length of the microvilli, in contrast to wild type (Fig. 5C–D′). Moreover, Myo7A was present in abnormally high amounts in the overlong microvilli, an effect that became more pronounced during progression from stage 10a (Fig. 5D,D′) to 10b (Fig. 5C–C″). The strongly increased Myo7A signal in the microvilli was accompanied by a depletion of Myo7A in the remaining cell body (Fig. 5C′,D′). This included the apical cytocortex, which showed reduced colocalization of Myo7A with Patj (Fig. 5C″), and the central and basal region of follicle cells that displayed a drastic reduction in Myo7A signal intensity (Fig. 5I,I′). This implies that Myo7A is recruited to the overlong microvilli at the expense of Myo7A in the rest of the cell.
This excessive recruitment of Myo7A to microvilli might be directly induced by increased Cad99C expression or could be due to an association of Myo7A with the much longer actin filament bundles of the lengthened microvilli (supplementary material Fig. S3E). To distinguish between these possibilities, we studied the distribution of Myo7A in cells that expressed Cad99CΔcyt::GFP, in which the majority of the cytoplasmic tail is deleted and replaced by GFP. Nevertheless, similar to Cad99C-FL, this isoform induced abnormally long microvilli (D'Alterio et al., 2005), including a lengthened F-actin core (supplementary material Fig. S3F). Notably, no increase in Myo7A concentration was detected in microvilli containing Cad99CΔcyt::GFP (Fig. 5E–F″). Instead, the amount of Myo7A seemed to be somewhat reduced in microvilli and apical cytocortex at stage 10a (Fig. 5E′), although it appeared to be normal at stage 10b (Fig. 5F′,F″). The distribution of Myo7A in the remaining cell body was comparable to that of wild type (Fig. 5E′,F′,I′). We conclude that the Cad99C cytoplasmic tail, rather than the lengthening of the microvilli, is responsible for Myo7A recruitment to microvilli.
To further test this hypothesis, we expressed Cad99CΔex::GFP, in which all cadherin domains in the extracellular region are deleted and replaced with GFP. The basal region of microvilli showed an abnormally strong enrichment of Myo7A (Fig. 5G,G′), which corresponds to an accumulation of Cad99CΔex::GFP in this zone (Fig. 5H,H′). Similar to Cad99C-FL, recruitment of Myo7A to microvilli by Cad99CΔex::GFP was associated with a significant decrease in the amount of Myo7A in the remaining cell body (Fig. 5G′,I′). This result strongly corroborates the conclusion that the Cad99C cytoplasmic tail causes recruitment of Myo7A to microvilli in follicle cells.
Myo7A and Cad99C form a protein complex
The Cad99C-dependent recruitment of Myo7A to microvilli suggested that these two proteins interact physically. To test for in vivo interactions, co-immunoprecipitation (co-IP) experiments were carried out using tissue lysates from ovaries that expressed GFP-tagged transgenes encoding either full-length Myo7A (Myo7A::GFP) or the membrane-bound cytoplasmic tail of Cad99C (Cad99CΔex::GFP), respectively. The transgenic proteins were specifically expressed in follicle cells, where they localized to microvilli (see Fig. 5H′; supplementary material Fig. S2C). GFP-specific antibodies immunoprecipitated Cad99CΔex::GFP, Myo7A::GFP and the control protein mCD8::GFP, respectively (Fig. 6A,B, left panels). These precipitates were analyzed for the presence of Myo7A or Cad99C, using antibodies that were specific for these proteins (Myo7A, supplementary material Fig. S1A; Cad99C, supplementary material Fig. S4A,B). We found that Cad99CΔex::GFP but not mCD8::GFP interacted with Myo7A (Fig. 6A, center panel). In a reciprocal experiment, Myo7A::GFP but not mCD8::GFP precipitated Cad99C (Fig. 6B center panel; supplementary material Fig. S4C). In summary, our biochemical analysis indicates that Myo7A and Cad99C form a protein complex, with the Cad99C cytoplasmic region binding to Myo7A either directly or indirectly.
Protrusions on the apical surface of epithelial cells can dramatically differ in terms of their structure and organization, reflecting a large functional diversity. Here, we found that Myo7A affected several structural features of the apical brush border of the secretory follicular epithelium in the Drosophila ovary. Follicle cells lacking Myo7A produced microvilli that appeared to be shorter and were often misoriented. Moreover, the dome-like appearance of the brush border, which results from the increasing length of microvilli towards the center of the apical surface of individual follicle cells, was less pronounced in the absence of Myo7A. We discovered that follicle cell microvilli are organized in rows that meander between and surround vitelline bodies. Similar whorl patterns have been described for microridges on the surface of certain epithelia in vertebrates (e.g. Sperry and Wassersug, 1976), but whether the rows of follicle cell microvilli originate from microridges remains to be determined. Similar rows of microvilli were also observed in Myo7A-depleted follicle cells but were separated by narrower islands of vitelline bodies than in wild-type follicles, and showed occasionally a wider spacing pattern when Myo7A was overexpressed. In addition to defects in follicle cell microvilli, lack of Myo7A also caused defects in the oocyte brush border. Taken together, these results indicate that Myo7A is involved in regulating the structure and organization of microvilli.
Mutual co-immunoprecipitation of Cad99C and Myo7A indicates that these proteins form a complex in vivo in ovaries. Previous in vitro experiments showed that a tail portion of vertebrate Myo7A enables direct interaction with PCDH15 (Senften et al., 2006). The SH3 domain of Myo7A seems to be sufficient to bind to the cytoplasmic region of PCDH15. In contrast to PCDH15, the cytoplasmic region of Cad99C is rather poor in proline residues, and contains only one PxxP motif (PTGP at position 1596), fitting the minimal consensus for an SH3-binding site (Kaneko et al., 2011). This motif, however, is not conserved in Cad99C homologs of other insects (Anopheles gambiae, Apis mellifera, Tribolium castaneum) and is therefore an unlikely anchor for an SH3 domain. However, a direct interaction between the SH3 domain of vertebrate Myo7A and the Keap1 protein that also has no apparent SH3-binding site suggests that the SH3 domain of Myo7A might have a different binding specificity (Velichkova et al., 2002). Cad99C also has no apparent FERM domain-binding site that could interact with one of the two FERM domains of Myo7A. Another possibility is that the interaction between Cad99C and Myo7A is mediated by a linker, such as a harmonin-like molecule that could interact with the predicted PDZ-domain-binding sites of Cad99C (Schlichting et al., 2005, D'Alterio et al., 2005). In contrast to the co-dependency of Myo7A and Cad99C at the apical surface of follicle cells, Cad99C molecules that ectopically localize to the lateral plasma membrane owing to heavy overexpression did not cause an obvious recruitment of Myo7A. One possibility is that Myo7A failed to bind to Cad99C at the lateral membrane because F-actin organization is not favorable for releasing the self-inhibition of Myo7A at this position. Alternatively, an additional factor could be present in microvilli, acting to permit recruitment of Myo7A by Cad99C. Taken together, we conclude that Myo7A and Cad99C are part of a protein complex, in which the cytoplasmic region of Cad99C interacts, directly or indirectly, with Myo7A.
The data from our genetic interaction studies strongly corroborate the close tie between Myo7A and the cytoplasmic tail of Cad99C. Overexpression of full-length Cad99C, as well as expression of the membrane-anchored cytoplasmic portion of Cad99C, led to a strong recruitment of Myo7A into microvilli at the expense of Myo7A in other regions of the cell. By contrast, expression of the membrane-bound extracellular cadherin domains of Cad99C, which, similar to full-length Cad99C, caused overlong microvilli (D'Alterio et al., 2005), did not have this effect. These findings, together with an apical loss of Myo7A in Cad99C mutants, suggest that Cad99C is crucial for recruiting Myo7A to microvilli, and that both proteins are co-dependent in maintaining normal levels of one another in microvilli.
Much of the Myo7A that was recruited to microvilli when Cad99C was overexpressed derived from the large pool of Myo7A in the non-cortical cytoplasm of follicle cells. Such a pool has also been reported for hair cells (Hasson et al., 1995), and might represent monomeric (or autoinhibited) Myo7A that is not associated with F-actin. It will be interesting to investigate whether the binding of Cad99C to the tail of Myo7A contributes to the release of the self-inhibition of this molecule, which would initiate ATPase activity and consequently strong binding to actin filaments (Yang et al., 2009; Umeki et al., 2009). Clustering of activated Myo7A molecules at the microvillus membrane through interaction with Cad99C might promote activation and dimer formation so that Myo7A can move along the actin core in a processive manner, possibly transporting cargo. A similar mechanism has been proposed for the interaction of Myo7A with the MyRip–Rab27a complex (Sakai et al., 2011). Taken together, we propose that the interaction with Cad99C facilitates increased affinity of Myo7A to the actin core of microvilli.
Interestingly, in the absence of Myo7A, we found a marked reduction in the amount of Cad99C in microvilli, raising the possibility that Myo7A is involved in the transport of Cad99C. Albeit reduced, Cad99C was still present in microvilli lacking Myo7A and seemed to be normally distributed. This indicates that Myo7A is not necessary for Cad99C transport per se, although it might increase its efficiency. Alternatively, by binding to Cad99C and fixing it to the microvillus actin core, Myo7A might stabilize Cad99C in the plasma membrane, slowing its turnover rate. Notably, cytoplasmic vesicles in the subapical region of follicle cells that contained Cad99C were not positive for Myo7A (Fig. 5D,G), arguing against co-transport of these two molecules to or from the membrane. It is possible that dissociation from Myo7A occurs before Cad99C is endocytosed. Although the mechanism is not understood yet, our findings indicate that Myo7A modulates the amount of Cad99C in microvilli. Considering the function of Cad99C in supporting the length and stability of microvilli (D'Alterio et al., 2005; Schlichting et al., 2006), the reduced amount of Cad99C could be directly responsible for the reduced height and misalignment of Myo7A-depleted microvilli.
We previously reported that Cad99C protein that is missing most of its cytoplasmic tail (Cad99CΔcyt) was able to induce abnormally long microvilli even in the absence of full-length Cad99C (D'Alterio et al., 2005), suggesting that the extracellular cadherin domains of Cad99C can control structural integrity and outgrowth of microvilli independently of cytoplasmic interactions. We speculate, however, that the abnormally high expression level of the truncated Cad99C protein might have masked any modulatory effect of the cytoplasmic region. Here, we show that Myo7A is important to achieve a normal amount of Cad99C in microvilli and to produce microvilli of normal structure. This, together with the observed complex formation between Cad99C and Myo7A, indicates that the cytoplasmic tail of Cad99C contributes to the function of this cadherin in microvilli.
A comparison of the distribution and function of Myo7A and Cad99C in follicle cells reveals similarities and differences in the requirement of these two proteins for microvillus morphogenesis. It is curious, for example, that the distributions of Cad99C and Myo7A within microvilli are not identical, considering their co-dependence. In contrast to Cad99C, which appears to be homogenously distributed in microvilli (D'Alterio et al., 2005; Schlichting et al., 2006), Myo7A showed a much higher concentration in the basal portion than in the apical portion of a protrusion. It is possible that Myo7A interacts with additional factors in the basal microvillus region similar to mammalian Myo7A, which seems to interact with several transmembrane receptors at the base of stereocilia (Michalski et al., 2007). Alternatively, the uneven Myo7A distribution could be related to the actin treadmilling process. Notably, the phenotypes of Cad99C and ck mutants show differences, suggesting at least some independent requirements for these two genes. The brush border defects in Cad99C mutant follicle cells were considerably more severe than in ck mutant follicle cells. Consistent with this is the finding that the defective brush border in Cad99C mutants interfered with eggshell formation, causing eggs to dehydrate and collapse (D'Alterio et al., 2005; Schlichting et al., 2006; Elalayli et al., 2008). By contrast, eggs from ck mutant females did not collapse (n = 390), indicating that loss of Myo7A does not interfere with the formation of a functional eggshell. Neither multiple wing hairs nor larval denticle defects that are characteristic for ck mutants (Nüsslein-Volhard et al., 1984; Kiehart et al., 2004; Bejsovec and Chao, 2012) were found in Cad99C mutants (Schlichting et al., 2005; data not shown). We conclude that Myo7A and Cad99C have co-dependent and independent functions in the regulation of apical protrusions.
MATERIALS AND METHODS
Drosophila strains and genetics
The strains used in this study were as follows: null alleles Cad99C21-5 and Cad99C21-8 (D'Alterio et al., 2005) and ck13 (Kiehart et al., 2004), and the ck uncovering deletions Df(2L)ED3 (Ryder et al., 2004) and Df(2L)BSC299 (Cook et al., 2012). To generate ck mutant flies, ck13/Df(2L)ED3 and ck13/Df(2L)BSC299 first instar larvae were selected based on the absence of a GFP-balancer and raised on yeast-supplemented apple juice agar plates at 21–22°C and low humidity, which increased survival rates during first and second instar.
The cell clones were as follows: (1) ck13 and Cad99C mutant cell clones in the follicular epithelium were induced by FLP/FRT-mediated mitotic recombination, using hsFLP1 and FRT40A or FRT82B (Xu and Rubin, 1993). ubi-nlsGFP was employed for negative labeling of cell clones (Xu and Rubin, 1993), TubP-Gal4, TubP-Gal80 and UAS-CD8::GFP for positive marking (Lee and Luo, 1999). Recombination was induced by two 2-hour heat shocks at 37°C in adult females. (2) Expression of the transgenic constructs UAS-Cad99C (#T2 and T6) and UAS-Cad99CΔcyt::GFP (#18-10 and 18-127) (D'Alterio et al., 2005), UAS-Cad99CΔex::GFP (#13 and 17; see below), UAS-ck::GFP (Todi et al., 2005) and UAS-mCD8::GFP (Lee and Luo, 1999) in follicle cells was induced by the Act5c>mCD2>Gal4 FLPout cassette (Pignoni and Zipursky, 1997). The FLPout cassette was activated by an 11–35-minute heat shock at 37°C. Females were 0–1-days old at the time of heat shock and ovaries were dissected 2 days later. Control genotypes were Oregon R and w1118. Flies were raised at 25°C.
Cad99CΔex::GFP lacks all extracellular cadherin domains of Cad99C. Cad99CΔex::GFP consists of: (1) a partial 5′UTR plus sequence for the N-terminal 102 amino acids of Cad99C, including the signal peptide – NotI–PstI fragment of a Cad99C cDNA construct (D'Alterio et al., 2005); (2) eGFP – PstI–PvuI fragment, PCR-amplified by using the primers 5′-GAATTCTGCAGTCGACGGTAC-3′ and 5′-CGATCGCTTGTACAGCTCGTCC-3′ from pEGFP-N2 (Clontech Laboratories, Mountain View, CA); (3) sequence for the C-terminal 354 amino acids of Cad99C (transmembrane and cytoplasmic regions) and the 3′UTR (PvuI–KpnI fragment of the Cad99C cDNA). pUAST was used as a P element transformation vector (Brand and Perrimon, 1993). The generation of transgenic flies followed standard procedures.
Generation of Myo7A and Cad99C-specific antibodies
The peptide used for the production of antibodies against Myo7A consisted of amino acids 873–1065 of Myo7A (portion of the IQ and adjacent Myth4 domain), and the peptide used to produce antibodies against Cad99C consisted of amino acids 505–751 of Cad99C (cadherin domains 5–6). These peptides were expressed and extracted as GST fusion proteins and used to immunize guinea pigs and rabbits. Specificity of antisera (anti-Cad99C-Rb1, anti-Cad99C-Rb3, anti-Myo7A-GP6) was confirmed by western blot analysis and tissue immunostaining (Fig. 4A; supplementary material Fig. S1A–C; Fig. S4A,B).
Cellular markers and imaging
Primary antibodies for tissue immunostaining were as follows: polyclonal guinea pig anti-Cad99C (GP5, 1∶3000; D'Alterio et al., 2005), rabbit anti-Cad99C (Rb1, 1∶1000), guinea pig anti-Myo7A (GP6, 1∶2000), mouse anti-GFP (1∶500; Abcam, Toronto, ON, Canada), rabbit anti-Patj (1∶250; Tanentzapf et al., 2000) and rabbit anti-βH-spectrin (1∶1000; Thomas and Kiehart, 1994). Secondary antibodies (1∶400) were conjugated to Cy3, Cy5 (Jackson ImmunoResearch Laboratories, Westgrove, PA), Alexa Fluor 488, or Alexa Fluor 555 (Life Technologies, Burlington, ON, Canada). The GFP gene trap zip::GFP was used to detect Myosin II (David et al., 2010). F-actin was visualized with phalloidin conjugated to Alexa Fluor 488 or Rhodamine (1∶20; Life Technologies). Potato lectin was conjugated to FITC (1∶10; Vector Laboratories, Burlington, ON, Canada). Fluorescent images from preparations mounted in Vectashield (Vector Laboratories) were generated with a Zeiss LSM510 scanning laser confocal microscope using 40×/1.4, and 63×/1.4 Plan-Apo objectives (Carl Zeiss, Toronto, ON, Canada). All imaging was performed at room temperature. Figures were processed using Adobe Photoshop-CS6 and Illustrator-CS6 (Adobe Systems, Ottawa, ON, Canada).
Co-immunoprecipitation and western blot analysis
For co-immunoprecipitation experiments, for each genotype, 160 pairs of ovaries from 2-day-old well-fed females were dissected in PBS on ice and homogenized in 300 µl of lysis buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 0.3 µM aprotinin, 10 µM leupeptin, 1 mM PMSF). The supernatant of the tissue lysate was recovered after each of three consecutive centrifugation steps (5 minutes at 19,000 g at 4°C in a microcentrifuge) and the volume was adjusted to 2 ml. The total protein yield was ∼6 mg. The tissue lysate supernatant was incubated with 50 µl of anti-GFP antibodies conjugated to magnetic beads (µMACS; Miltenyi Biotec, San Diego, CA) for 30 minutes on ice. The immunocomplexes were isolated according to the instructions of the manufacturer (Miltenyi Biotec).
For western blot analysis, proteins from ovarian tissue lysates (3 µl, undiluted lysate) and from immunoprecipitations (25 µl) were separated by SDS-PAGE. Primary antibodies were polyclonal rabbit anti-Cad99C (Rb3, 1∶3000), guinea pig anti-Myosin VIIA (GP6, 1∶4000) and rabbit anti-GFP (1∶4000; Clontech Laboratories). HRP-conjugated secondary antibodies (1∶1500; Jackson ImmunoResearch Laboratories) and the ECL visualization system (GE Healthcare Life Sciences, Baie d'Urfe QC, Canada) were used for signal detection.
Measurements of immunosignal intensity
To compare the relative intensity of immunosignals between genotypically different cells, the pixel value was determined for a specified subcellular area of mutant and control cells using ImageJ (NIH). (1) To determine Cad99C signal in ck mutant cells, 13 ck mutant cell clones in stage-10 follicles were analyzed. For each clone, Cad99C signal was measured in two to five ck−/− cells and an equal number of neighboring ck+/+ or ck+/− control cells. An area of equal size was analyzed in all cells, which corresponded to the area of the microvillus region in a ck mutant cell. (2) To determine Myo7A signal in cells expressing Cad99C transgenes, seven to eight cell clones were evaluated per genotype. In most cases, eight cells were measured per mutant clone plus an equal number of neighboring wild-type cells. For each mutant cell, the ratio in signal intensity to a neighboring wild-type cell was determined (as the percentage of wild-type value). Prism 4 (GraphPad Software, La Jolla, CA) was used for statistical analysis (paired two-tailed t-test) and generation of graphs.
Transmission electron microscopy
Ovaries were dissected in cold Schneider's medium, fixed on ice in 2.5% glutaraldehyde, 2% paraformaldehyde and 2% acrolein in 75 mM cacodylate buffer pH 7.4 (Caco-buffer) for 1 hour, in 2.5% glutaraldehyde in Caco-buffer overnight and in 1% OsO4 and 0.1% sorbitol in Caco-buffer for 1 hour (EM grade fixatives and chemicals from Polysciences, Inc., Warrington, PA and Sigma-Aldrich, Oakville, ON, Canada). Samples were then dehydrated in an ethanol series, treated with 100% ethanol and propylene oxide for two 15-minute periods each and incubated in a 1∶1 mixture of propylene oxide and araldite-502/Embed-812 embedding medium (AEEM; Cedarlane, Burlington, ON, Canada) overnight. After evaporation of propylene oxide for 4 hours, samples were incubated in fresh AEEM overnight, mounted in fresh AEEM and hardened at 65°C for 12 hours.
100-nm-thick sections (Leica EM UC6 ultramicrotome, Leica Microsystems, Concord, ON, Canada), parallel or perpendicular to the length axis of microvilli (longitudinal sections through the center of the follicle or the follicle-cell–oocyte interface, respectively) were contrasted with uranyl acetate and lead citrate, and analyzed with a Hitachi HT-7700 TEM (Hitachi High-Technologies, Etobicoke, ON, Canada). For six control and eight ck mutant follicles (early to mid stage 10b), several sections were viewed and multiple (10–40) images were analyzed. Images were taken at ×14K-110K. Contrast and brightness were adjusted in Adobe Photoshop CS6. 1-µm-thick sections, stained with toluidine/methylene blue were used as a control for the age of the follicle and the position of ultra-thin sections within the follicle. Migration of centripetal cells and thickness of the follicular epithelium were used to stage follicles (King, 1970).
We are grateful to Maggie Au Yeung, Shang-Hsien Hwang and Mark Ng (University of Toronto, Canada) for technical assistance, and Roger Jacobs (McMaster University, Hamilton, Canada), Frank Laski (University of Los Angeles, Los Angeles, CA) and Ulrich Tepass (University of Toronto, Canada) for advice on TEM. We thank Ulrich Tepass and Tony Harris (University of Toronto, Canada) for helpful comments on the manuscript. We thank Daniel Eberl (The University of Iowa, Iowa City, IA), Graham Thomas (Penn State, State College, PA), Ulrich Tepass and the Bloomington Drosophila Stock Center (Indiana University, Bloomington, IN), Kyoto Drosophila Genetic Research Center (Kyoto, Japan) and FlyTrap (Yale University, New Haven, CT) for reagents, and our departmental imaging and animal facilities for technical support.
C.G., R.-H.S.L. and D.G. designed, performed and analyzed experiments. X.C. and A.D. planned and performed experiments. D.G. supervised the project and wrote the manuscript.
This work was funded by operating grants from the Canadian Institute of Health Research; and the Natural Sciences and Engineering Research Council of Canada (to D.G.).
The authors declare no competing interests.