dEHBP1 controls exocytosis and recycling of Delta during asymmetric divisions

Notch signaling is an evolutionarily conserved, intercellular signaling pathway that plays a seminal role in numerous biological processes, including cell fate acquisition and differentiation (Artavanis-Tsakonas et al., 1999; Bray, 2006; Fortini, 2009; Fortini and Bilder, 2009; Kopan and Ilagan, 2009; Tien et al., 2009). The versatile role of Notch signaling during development and adult tissue homeostasis relies upon the context-dependent function of different regulators and downstream effectors (Bray, 2006; Yamamoto et al., 2010). Given the importance of Notch signaling in development, cancer, and human diseases (Gridley, 2003, 2007; Weng and Aster, 2004; Roy et al., 2007; Watt et al., 2008; Bolós et al., 2009), the identification of new regulators of Notch (Berdnik et al., 2002; Sasamura et al., 2003; Hutterer and Knoblich, 2005; Jafar-Nejad et al., 2005; Vaccari and Bilder, 2005; Gallagher and Knoblich, 2006; Acar et al., 2008; Tien et al., 2008; Rajan et al., 2009; Saj et al., 2010; Vaccari et al., 2010) has played an important role in advancing our understanding of the molecular and cellular basis of development and disease. To understand the mechanisms of activation and identify novel regulators of Notch signaling, we performed forward genetic screens to identify genes that affect the asymmetric divisions of cells of the external sensory organs (ESOs), in which cell fate decisions depend on Notch signaling (Lai, 2004; Le Borgne et al., 2005; Gönczy, 2008).

The ESO lineages give rise to micro- and macrochaetae, which develop on the thoraces and appendages of adult flies in a highly organized pattern (Gho et al., 1999; Reddy and Rodrigues, 1999; Bellaïche and Schweisguth, 2001; Lai, 2004; Lai and Orgogozo, 2004; Le Borgne et al., 2005). Each ESO consists of four cells that develop from a single precursor, hereafter named the pI cell, through consecutive rounds of asymmetric divisions (Fig. 1 a). In the microchaetae lineages, the pI cell divides into a posterior pIIa and an anterior pIIb cell. The pIIa cell gives rise to the trichogen (shaft) cell and its surrounding tormogen (socket) cell, both visible on the exterior surface of the thoracic cuticle. The pIIb cell divides into a pIIIb and a glial cell, which migrates away and eventually dies. The pIIIb cell produces the neuron and the thecogen (sheath) cells.

The efficacy and directionality of Notch signaling during asymmetric divisions is accomplished at multiple levels by asymmetric endocytosis (Fürthauer and González-Gaitán, 2009a,b). Endosomes that are positive for SMAD anchor for receptor activation (SARA) are segregated asymmetrically, but loss of function of SARA does not lead to cell fate transformation defects in the ESO lineage (Coumailleau et al., 2009). The cell fate determinants Numb and Neuralized (Neur) form a crescent at the anterior cell cortex of pI in a Par complex–dependent manner (Betschinger et al., 2003; Langevin et al., 2005; Roegiers et al., 2005; Wirtz-Peitz et al., 2008) and segregate into the anterior pIIb signal–sending cell, where they function as regulators of vesicular trafficking. In the pIIa signal–receiving cell, which does not inherit Numb, Sanpodo (Spdo) localizes at the plasma membrane together with the Notch receptor where it positively regulates Notch function (O’Connor-Giles and Skeath, 2003; Hutterer and Knoblich, 2005; Langevin et al., 2005). In the pIIb cell, Numb inhibits the plasma membrane localization of Spdo and converts Spdo into a negative regulator of Notch (Babaoglan et al., 2009). In addition to Numb, Neur, an E3 ubiquitin ligase, controls the ubiquitination and endocytosis of Delta (Lai and Rubin, 2001; Pavlopoulos et al., 2001) in the signal-sending pIIb cell (Le Borgne and Schweisguth, 2003).

Delta endocytosis in the pIIb signal–sending cell may serve the purpose of “pulling” the Notch receptor via their physical interaction away from the cell receiving the Notch signal, possibly rendering the target sites of various proteases in Notch accessible to proteolytic cleavages. These cleavages are necessary for Notch activation (Parks et al., 2000; Nichols et al., 2007a,b). A nonexclusive model is the “recycling” model (Fig. 1 b). In this model, Delta is first inserted into the plasma membrane, but it is unable to signal. Delta is then endocytosed in a vesicular compartment. These Delta-bearing vesicles return back to a specialized plasma membrane domain at the interface of pIIa and pIIb cells, where the actin-rich structure (ARS) resides (Rajan et al., 2009). This process is mediated via a Rab11-positive recycling endosomal compartment (Emery et al., 2005; Benhra et al., 2011). Furthermore, loss of Sec15, a member of the exocyst complex and an effector of Rab11, leads to a basal accumulation of Delta and Spdo (Jafar-Nejad et al., 2005).

In the present study, we report the identification of CG15609, the Drosophila homologue of Eps15 homology domain containing protein-binding protein 1 (dEHBP1), as a novel component of Notch signaling during asymmetric cell divisions of the ESO lineages. We show that dEHBP1 regulates the levels and localization of Spdo as well as the trafficking of Delta at the signaling interface of the pIIa/pIIb. These data provide critical links between the key players required for the trafficking of Delta.

To identify novel genes in the Notch signaling pathway we performed a forward genetic mosaic screen on chromosome arm 2R to isolate mutations that disrupt the stereotypic pattern of the mechanosensory bristles on the thorax of the adult flies (Andrews et al., 2009). Mutations in the 2R11 complementation group are homozygous lethal and clones of these mutations on the adult thorax bear no bristles (Fig. 1, c and d). The 2R11 group consists of four alleles: A28, O4, N8, and T14.

To assess if the lack of bristles in the 2R11 mutant clones is due to loss of sensory precursor pI cells, we stained pupal thoraces at 15 h after puparium formation (APF) with anti-Senseless (Sens) antibody, which marks the pI and its progeny pII cells (Nolo et al., 2000). The emergence of Sens-positive pI cells is not affected within mutant regions of pupal thoraces (unpublished data). To examine whether the cell fates in the ESO lineages are affected, we assessed the expression of appropriate cell fate markers at 24–26 h APF pupal thoraces, when the cell types in the ESO lineage have been determined. We observed that more than one internal cell is positive for Elav, a neuronal marker (Fig. 1, e–e′′′) and Prospero, a sheath cell marker (Fig. 1, f–f′′′) in mutant clones. In addition, we observe loss of Su(H), a marker for socket cells (Fig. 1, g–g′′′). Hence, the emergence of supernumerary neurons and sheath cells occurs at the expense of external cell types, i.e., socket and shaft cells. The transformation of cells within the mutant clones exhibits incomplete penetrance, i.e., in 2R11^O4 homozygous mutant clones, ∼70% of the ESO cell clusters exhibit no Su(H) socket cells, whereas ∼40% of all Cut-positive ESO cells express Elav, as opposed to 25% in wild-type lineages. The incomplete penetrance of cell fate transformations has also been observed in other mutants that regulate Notch signaling, such as sec15 and spdo (Hutterer and Knoblich, 2005; Jafar-Nejad et al., 2005).

To establish if Notch signaling is indeed impaired, we assessed the expression pattern of the transcription factor Tramtrack (Ttk). Ttk is expressed in the pIIa cell (signal receiving), but not in the pIIb (signal sending) cell (Guo et al., 1995), as a result of Notch signaling (Okabe et al., 2001; Pi et al., 2001). In mutant 2R11 clones, posterior cells do not express Ttk (Fig. 1, h–h′′′) similarly to their anterior pIIb sibling cells, indicating that Notch signaling is indeed affected and that pIIa cells have transformed into pIIb cells. Hence, 2R11 mutations affect Notch signaling during asymmetric divisions of ESO cells.

The 2R11 mutations were mapped by meiotic recombination using P elements (Zhai et al., 2003) followed by complementation tests with deficiencies and lethal alleles in the region of interest. All 2R11 alleles failed to complement Df(2R)ED2751, Df(2R)Exel6065, Df(2R)ED3181 molecularly mapped deficiencies (Parks et al., 2004; Ryder et al., 2007; Cook et al., 2010). In addition, 2R11 alleles do not complement P{lacW}l(2)k09837, a lethal insertion of a P transposable element in the 5′ upstream regulatory region of the CG15609 gene (Spradling et al., 1999), the “P” allele (Fig. 2 a). The lethality of P{lacW}l(2)k09837 is due to the P element insertion because it is reversed upon precise excision of the P element. We also generated a deletion, CG15609^ΔEx24, by imprecise excision of the P element (Fig. 2 a). Flies homozygous for CG15609^ΔEx24 are lethal and fail to complement all other alleles (Fig. 2 d). The molecular lesions in the 2R11 mutants are either stop codons, deletions, or alterations in splice acceptor or splice donor sequences (Fig. 2 a), which are all predicted to cause truncated protein products of CG15609 (Fig. 2 c). Finally, a genomic rescue construct spanning CG15609 rescues all hetero-allelic combinations (Fig. 2, a and d). Similarly, ubiquitous overexpression of the N-terminally FLAG-tagged cDNA isoform B rescued the lethality of A28 and O4 mutations over Df(2R)Exel6065 (Fig. 2 d). These data demonstrate that 2R11 corresponds to CG15609.

CG15609 encodes the fly homologue of the human EHD protein-binding protein 1, an interacting partner of the Eps15 homology domain (EHD) 1 and 2 proteins (Guilherme et al., 2004a,b). In adipose cells and in the absence of insulin signaling, EHBP1 functions along with EHD2 in endocytosis of the GLUT4 glucose transporter, whereas upon insulin stimulation, EHBP1 and EHD1 mediate the recycling of GLUT4 back to the plasma membrane. In Caenorhabditis elegans, EHBP1 interacts with constitutively active Rab GTPase variants and is involved in the endocytic recycling of clathrin-independent cargoes (Shi et al., 2010), but it has not been implicated in Notch signaling in worms or vertebrates.

EHBP1 proteins contain multiple evolutionarily conserved domains (Fig. 2 b; Friedberg, 2010), i.e., an ∼200-aa-long, N-terminal C2-like lipid-binding domain (Zhang and Aravind, 2010), a middle calponin homology (CH) actin-binding domain (Gimona and Mital, 1998; Gimona et al., 2002; Korenbaum and Rivero, 2002), and a C-terminal, 200-aa-long, coiled coil region, which serves as a protein–protein interaction platform (Shi et al., 2010). In addition, the fly and mammalian EHBP-1 homologues have a putative CAAX box that may be involved in membrane anchoring, but this motif is not conserved in C. elegans (Shi et al., 2010).

To determine the expression pattern and subcellular localization of CG15609, we generated an anti-EHBP1 antibody against the full-length Drosophila protein. Anti-dEHBP1 fails to recognize the endogenous protein in either thoracic pupal clones homozygous for dEHBP1^O4 (Fig. 2, e–e′′′) or embryos that lack dEHBP1 (dEHBP1^ΔEx24; Fig. 2, f–g′′). dEHBP1 is expressed ubiquitously, but it is enriched in the embryonic central nervous system (CNS; Fig. 2, f–f′′). Moreover, in both epithelial thoracic cells (Fig. 2, e–e′′′) and embryonic CNS (Fig. 2, f–g′), dEHBP1 is mainly localized at the plasma membrane.

In summary, we have isolated loss-of-function mutations in CG15609, the fly homologue of EHBP1. Loss of dEHBP1 disrupts Notch signaling during asymmetric divisions in the ESO lineage, and dEHBP1 is ubiquitously expressed and localized mostly at or near the plasma membrane.

To determine the subcellular localization of dEHBP1 we localized the protein with respect to an array of markers that delineate distinct plasma membrane subdomains along the apical–basal axis of epithelial cells (Humbert et al., 2003). Patj determines the apical region, Armadillo (Arm) and E-cadherin (Ecad) define the adherens junction, whereas Fasciclin III (FasIII) and Discs large (Dlg) mostly mark the lateral area of the plasma membrane (Fig. 3 h). dEHBP1 is present all around the plasma membrane of both pIIa and pIIb cells, but is especially enriched basally as well as laterally, at the interface of pIIa/pIIb cells, where it mostly colocalizes with FasIII and Dlg (Fig. 3, a–e′′′, g–j′′; and Fig. S1). The pIIa/pIIb interface is rich in actin filaments that assemble along the apico-basal axis and in apical microvilli in both cells, forming the ARS (Rajan et al., 2009). Because dEHBP1 is a putative actin-binding protein, we determined if dEHBP1 colocalizes with F-actin. Indeed, dEHBP1 colocalizes with F-actin at the pIIa/pIIb interface (Fig. 3, f–f′′ and i–j′′; and Fig. S1).

dEHBP1 exhibits a discontinuous, punctate pattern that likely reflects a dynamic, vesicular mode of trafficking (Fig. 3, g–g′′). To follow the trafficking of dEHBP1 in real time, we tagged dEHBP1 at its N terminus with mCherry. Expression of mCherry-dEHBP1 does not cause any cell fate transformations and is able to rescue the loss-of-function dEHBP1 phenotypes (unpublished data). Overexpression of mCherry-dEHBP1 in ESO lineages results in its localization at the interphase of pII cells and in intracellular punctae, recognized also by the anti-dEHBP1 antibody (Fig. 3, i–j′′). Live imaging of ESO clusters at the one- and two-cell stage revealed that mCherry-dEHBP1 punctae are present in both pIIa/pIIb cells to an equal extent (Fig. 4, a–d; and Video 1). Moreover, mCherry-dEHBP1 concentrates transiently at the interface of pIIa/pIIb cells shortly after the division of the pI cell (Fig. 4, g and n).

To determine the nature of dEHBP1 punctae, we analyzed their colocalization with key Notch-signaling components, such as Spdo and Delta, which are also enriched at the interface of the pIIa/pIIb cells (Rajan et al., 2009; Tong et al., 2010; Benhra et al., 2011), as well as with a variety of subcellular markers including multiple Rab proteins, each identifying a discrete intracellular compartment (Fig. 4, Fig. S1; Stenmark, 2009). Hence, we assessed the colocalization of mCherry-dEHBP1 and Spdo-GFP in live imaging studies as well as endogenous dEHBP1 and Delta in immunofluorescent stainings. Although Spdo GFP is predominantly apical, whereas mCherry–dEHBP1 is enriched in subapical domains, we were able to detect colocalization of Spdo and dEHBP1 at the interface of the ESO cells (Fig. 4, g and i–k; and Video 2). However, we did not observe any colocalization of Spdo and dEHBP1 punctae in the cytoplasm (Fig. 4, l–r; Video 3). Similarly, we failed to see any colocalization between dEHBP1 and Delta in immunofluorescent stainings (Fig. 4, s–s′′′). Therefore, we conclude that dEHBP1 traffics to the interface of pII cells separately of Spdo and Delta vesicles. In addition, our live imaging analysis revealed only a rare and transient colocalization between dEHBP1 and Rab5-GFP or Rab11–GFP (unpublished data). dEHBP1 does not colocalize with Golgi markers, such as p120 kD, Syntaxin 16, and GM130, with the early endosome markers Avalanche and Rab5YFP, the late endosome markers Rab9YFP and Hook, the recycling endosome markers Rab11 and Rab11YFP, and the exocytic marker Rab10YFP (Fig. S1). Interestingly, we discovered that there is significant colocalization of dEHBP1 with Rab8YFP, a marker of exocytic vesicles that travel from trans-Golgi and through the recycling endosome to the plasma membrane (Huber et al., 1993; Ang et al., 2003; Henry and Sheff, 2008; Shi et al., 2010; Das and Guo, 2011). Intriguingly, in nematodes, EHBP1 colocalizes with Rab10 and Rab8 GTPases, (Shi et al., 2010), two very similar GTPases in the secretory-recycling pathway (Pereira-Leal and Seabra, 2001; Das and Guo, 2011), and loss of EHBP1 phenocopies loss of rab10 in endosomal recycling (Shi et al., 2010).

Even though we do not detect significant colocalization between dEHBP1 and Rab10 (Fig. S1), we addressed the role of rab10 in asymmetric cell divisions by overexpressing dominant-negative Rab10YFP in ESO lineages, as there are no available alleles for rab10. We did not observe any defects in cell fate acquisition under these experimental conditions. Thus, loss of Rab10 function does not mediate the effect of dEHBP1 loss-of-function mutations. To address whether loss of rab8 phenocopies loss of dEHBP1, we tested a rab8¹ allele. The molecular lesion in rab8¹ is a point mutation in the GTPase domain that alters the evolutionarily conserved serine at position 17 to phenylalanine (S17F). When rab8¹ is placed over a deficiency that uncovers the rab8 locus, it results in pupal lethality. To perform clonal analysis, we recombined rab8¹ on a FRT80B chromosome. Loss of rab8 in thoracic clones or overexpression of dominant-negative Rab8YFP in ESO lineages does not confer any cell fate phenotype. Therefore, loss of Rab8 function does not mediate the effect of dEHBP1 loss-of-function mutations.

The prominent accumulation of dEHBP1 at the interface of the pIIa/pIIb cells, where Spdo and Delta are targeted, strongly suggests that dEHBP1 may play a role in the localization of Spdo, Delta, or Notch. Indeed, loss of dEHBP1 causes a strong up-regulation of Spdo in the pI cell and its progeny (Fig. 5, Fig. S2, b–c′; and unpublished data). We did not observe the expansion of any of the other subcellular compartments, including ER, Golgi, and endosomal compartments (Fig. S2, a–e′). We were therefore unable to determine the specific subcellular compartment that contains Spdo. Importantly, we did not observe any aberrant localization of Delta or Notch using standard immunofluorescence assays (Fig. 5, b–e′). Neither did we observe any defects in the endocytosis of Notch (Fig. 5, f–f′′′). Furthermore, we did not observe any alterations in the apical–basal cellular polarity, as judged by the subcellular localization of Arm (unpublished data). In addition, the localization of Bazooka (Baz), a member of the Par complex that controls the anterior localization of the cell fate determinants Numb and Neur (Betschinger et al., 2003; Langevin et al., 2005; Roegiers et al., 2005; Wirtz-Peitz et al., 2008), occurs properly at the posterior cortex of the dividing pI cells (unpublished data). Finally, the cell fate determinants Numb and Neur form the proper crescent at the anterior cortex of the dividing pI cell, indicating that their asymmetric segregation occurs correctly in the absence of dEHBP1 (Fig. S3, i–l′). In summary, the asymmetric segregation of key determinants occurs properly.

The aberrant up-regulation of Spdo may reflect its retention in the cytoplasm and a failure to reach the plasma membrane, as previously observed with the loss of sec15 (Tong et al., 2010). We observed that, even in the absence of dEHBP1, Spdo is colocalized with FasIII at the plasma membrane, suggesting that it is not exclusively cytoplasmic (Fig. 5, g–j). It is however possible that certain parameters of Spdo trafficking may have been altered. For example, at this level of analysis, we cannot conclude whether Spdo is properly inserted in the plasma membrane. In addition, Spdo may linger in subcortical pools for a longer period of time and its access to Notch receptor may have been reduced, thus rendering its activation more difficult. Alternatively, the aberrant up-regulation of Spdo may negatively impact Notch signaling (Babaoglan et al., 2009). However, we and others have not observed any effects of Spdo overexpression upon cell fate acquisition in the ESO lineage (this paper; Jafar-Nejad et al., 2005; Tong et al., 2010).

To further dissect the function of dEHBP1 in ESO lineage specification, we overexpressed a constitutively active form of atypical protein kinase C (DaPKC^ΔN) in the wild-type and dEHBP1^−/− thoracic clones. Overexpression of DaPKC^ΔN has been shown to result in retention of Spdo at the plasma membrane, conferring a partial Notch gain-of-function phenotype, i.e., generation of extra socket cells (Langevin et al., 2005; Roegiers et al., 2005). However, when we overexpressed DaPKC^ΔN in thoracic clones lacking dEHBP1 we observed the same phenotype as in the cells that only lack dEHBP1, i.e., a Notch loss of function (Fig. S3, a, b, d, and e). Because constitutively active DaPKC results in Spdo localizing at the plasma membrane, we examined whether loss of dEHBP1 suppresses the overexpression phenotype of DaPKC^ΔN by altering the subcellular localization of Spdo. We observe that Spdo still colocalizes with the membrane marker FasIII in dEHBP1^−/− ESO lineages that overexpress constitutively active DaPKC, similarly to neighboring wild-type ones (Fig. S3, g–h′′). Therefore, dEHBP1 may affect other aspects of Notch signaling, such as the activity of the ligand Delta.

To corroborate our hypothesis, we examined whether loss of dEHBP1 suppresses the Notch gain-of-function phenotype of the overexpressed activated Notch intracellular domain (Notch^NEXT). Notch^NEXT functions in a ligand-independent, presenilin-dependent manner. We observe that loss of dEHBP1 cannot suppress the supernumerary sockets that the overexpression of Notch^NEXT causes, as judged by the macroscopic expression of enlarged socket structures within the thoracic clones and the supernumerary Su(H) socket cells in clones in pupal nota 24 h APF (Fig. S3, a, c, d, f, and m–n′′). Thus, dEHBP1 functions upstream of the presenilin-mediated S3 cleavage of Notch. In conclusion, our data suggest that dEHBP1 may not affect the activity of the Notch receptor, but most likely regulates other aspects of Notch signaling, such as the function of Spdo and/or the activity of the ligand Delta.

If dEHBP1 positively regulates the ability of Delta to signal, then loss of dEHBP1 would be epistatic to gain of function of Notch signaling, achieved by ectopic expression of Delta. Indeed, loss of dEHBP1 results in the development of extra neurons in clones of cells despite the ectopic expression of an “activated” variant of Delta ligand, which is called Delta^R+ and is also dependent on ubiquitination and epsin-mediated recycling, similar to Delta (Fig. S3, o–o′′′; Wang and Struhl, 2004). These results suggest that dEHBP1 affects the output of Notch signaling even upon overexpression of Delta^R+, because it may affect the function of Spdo and/or it affects the trafficking and signaling capacity of Delta^R+.

The localization of Delta is unaffected in the absence of dEHBP1 in immunofluorescent stainings of pupal thoraces at 17 h APF using standard immunofluorescence stainings (Fig. 6, a–b). These stainings are based on permeabilizing conditions that reveal the localization of both extracellular and intracellular Delta (Fig. 6, a–b). It has been proposed that Delta is endocytosed mainly at the basal side of ESO lineages in a neur-dependent fashion (Benhra et al., 2011). Upon endocytosis, it is then transcytosed to the apical side, where it will interact with Notch (Benhra et al., 2011). To assess whether Delta endocytosis is affected, we performed Delta endocytosis assays (Le Borgne and Schweisguth, 2003). In brief, extracellular Delta molecules are conjugated with anti-Delta antibodies under nonpermeabilizing conditions over a 20-min period at room temperature, allowing uptake in the cells. The internalized complexes of Delta antibodies are then detected by fluorescent secondary antibodies (Fig. 6 c). Interestingly, uptake of Delta antibody complexes is dramatically impaired in dEHBP1^−/− clusters (Fig. 6, d–e), suggesting that either endocytosis is impaired or that the abundance of Delta molecules present on the cell surface is much decreased.

The total pool of Delta present at the membrane consists of a pool of Delta that is initially delivered to the plasma membrane, as well as a pool of Delta that is recycling back to the ARS at the interface of pIIa and pIIb cells upon endocytosis (Emery et al., 2005; Jafar-Nejad et al., 2005; Rajan et al., 2009; Benhra et al., 2011). To assess Delta trafficking, we performed Delta pulse-chase assays (Fig. 7 a). In brief, pupal thoraces were incubated under nonpermeabilizing conditions with anti-Delta antibodies over a 30-min period on ice, to block endocytosis. This allows anti-Delta-antibodies to bind to Delta at the cell surface. If the tissue is fixed immediately upon incubation with the antibodies, we can estimate the total amount of Delta that is accessible by the antibodies and therefore, the total membrane-associated pool of Delta. We define this time point as 0 min. If the tissue is fixed after chase for 60 min at room temperature we can estimate the rate of uptake. We define this as time point 60 min. We find that in dEHBP1^−/− cells, the amount of Delta is reduced both at time point 0 and 60. These data indicate that exocytosis of Delta is impaired in the absence of dEHBP1 (Fig. 7, b–c′, g).

To corroborate our conclusions, we analyzed the distribution of exocytosed Delta by detecting extracellular Delta in ESO clusters under nonpermeabilizing conditions: the tissue is fixed in the absence of detergent and incubated overnight with anti-Delta. The data show that in dEHBP1^−/−, extracellular Delta is severely decreased, suggesting that exocytosis of Delta is indeed impaired. Some of the remaining Delta is localized basally (Fig. 7, d–f′′′ and h). These data suggest that Delta exocytosis is impaired and that trafficking from the basal to the apical membrane is severely disrupted.

The enrichment of dEHBP1 at the signaling interface of pIIa/pIIb cells correlates with the failure of Delta to localize at the interface in dEHBP1^−/− cells. Therefore, dEHBP1 may control the abundance and/or subcellular localization of components that affect Delta localization at the ARS, such as Rab11 and its effector Sec15 (Zhang et al., 2004). We found that dEHBP1 does not affect the subcellular localization of Rab11 or Sec15 (Fig. S2, e–f′). Alternatively, dEHBP1 may affect the integrity of ARS. However, the umbrella-shaped morphology of the ARS is undisturbed in the absence of dEHBP1 (Fig. S2, g–i).

Conversely, we asked whether any of the components that regulate Delta recycling to the ARS are important for the localization dEHBP1, such as Arp3 and sec15. Because dEHBP1 is a putative actin-binding protein, we first examined whether disruption of the ARS results in mislocalization of dEHBP1 and consequently whether it confers accumulation of Spdo. We found that in the absence of Arp3 activity (Rajan et al., 2009), dEHBP1 is correctly localized and Spdo fails to accumulate within the ESO cluster (unpublished data). Thus, loss of function of Arp3 does not affect the subcellular localization and function of dEHBP1. As loss of rab11 is cell lethal (Emery et al., 2005; Banks et al., 2011), we asked whether loss of sec15 leads to mislocalization of dEHBP1. As shown in Fig. 8 a, the subcellular localization of dEHBP1 is severely disrupted in the absence of sec15 because dEHBP1 accumulates at the basal side in both epithelial cells as well as within ESO lineages, where it colocalizes with Delta in the same subcellular compartment (Fig. 8, a–c′′). Thus, both dEHBP1 and Delta depend on Sec15 for their proper intracellular trafficking.

The dependence of dEHBP1 subcellular distribution on sec15 prompted us to assess if dEHBP1 and Sec15 interact physically. Indeed, dEHBP1 coimmunoprecipitates with Sec15 from whole-cell extracts of transiently transfected S2 cells (Fig. 8 f). The physical interaction between dEHBP1 and Sec15 may reflect their colocalization in vivo. Indeed, Sec15-GFP colocalizes with dEHBP1 at the interface of pII cells in ESO clusters (Fig. 8, d–d′′) and within punctate structures (Fig. 8, e–e′′). Taken together, these results suggest that Sec15 regulates the recycling of Delta and the subcellular distribution of dEHBP1, with which it physically interacts and colocalizes in vivo.

We attempted to map the domain of dEHBP1 that is necessary for interaction with Sec15 by in vitro GST pull-down assays unsuccessfully, suggesting that dEHBP1 may interact with Sec15 via other protein intermediates. One possibility is that dEHBP1 may interact with Sec15 via its interaction with Rab11 GTPase, whose effector is Sec15. However, we were unable to observe any direct interaction of dEHBP1 and Rab11 in in vitro GST pull-down assays (Fig. 8 g) or in yeast two-hybrid experiments (not depicted). To further examine the possibility of physical interactions between dEHBP1 and Rab11, we performed coIP from S2 whole-cell extracts that were transiently transfected with wild-type or constitutively active or dominant-negative HA-tagged variants of Rab11 and FLAG-dEHBP1. As shown in Fig. 8 h, dEHBP1 interacts strongly with the dominant-negative variant of Rab11 and very weakly with the wild-type or constitutively active variant of Rab11. As dominant-negative forms of Rab proteins mimic the GDP-bound inactive state, these data suggest the possibility that dEHBP1 may act at a step just before the activation of Rab11. In conclusion, dEHBP1 depends on Sec15 for its subcellular localization, and it can physically interact with Rab11 and Sec15. We envision that dEHBP1 may act along with Rab11 and Sec15 during the trafficking of Delta-bearing vesicles toward the ARS.

In the present study, we describe the identification of dEHBP1 as a novel, positive regulator of Notch signaling in asymmetrically dividing cells in the ESO lineage in Drosophila. In the absence of dEHBP1, external cell types, such as socket and shaft cells, are transformed into internal cell types, i.e., neuron and sheath cells, one of the hallmarks of loss of Notch signaling. EHBP1 has been previously studied in mammalian cell culture systems and in vivo in C. elegans. In mammalian adipocytes, EHBP1 affects endocytosis and recycling of the glucose transporter GLUT4 in the context of insulin signaling, depending on its interaction via the NPF motifs present in its N-terminal region with EHD2 or EHD1, respectively (Guilherme et al., 2004a,b). However, the fly and worm EHBP1 lack the NPF motifs (unpublished data and Shi et al., 2010), suggesting that the EHD–EHBP1 interaction may have emerged later in evolution. In C. elegans, EHBP1 was shown to impair rab10-mediated endocytic recycling of clathrin-independent endocytosed cargoes, such GLR-1 glutamate receptor (Shi et al., 2010). Here, we show for the first time that dEHBP1 is required in the exocytosis and recycling of Delta, a ligand of the Notch receptor. Notch signaling defects were not reported in C. elegans ehbp1 mutants. Therefore, it would be interesting to investigate whether EHBP1 and its homologues play an evolutionarily conserved role of EHBP1 in Notch signaling.

dEHBP1 is a ubiquitous protein that is associated with the plasma membrane, enriched at the lateral and basal surface of pII cells, where it colocalizes with F-actin. Our live imaging with mCherry-dEHBP1 and immunofluorescent stainings with anti-dEHBP1 antisera also reveal dEHBP1-positive, punctate, intracellular structures within ESO lineages. An extensive analysis with a diverse array of intracellular markers revealed that these punctae colocalize with Rab8, indicating their exocytic nature (Huber et al., 1993; Ang et al., 2003). Importantly, in C. elegans, EHBP1 physically interacts and colocalizes with Rab8 and Rab10, and controls the recruitment of Rab10 in recycling endosomal structures (Shi et al., 2010). However, in our studies, overexpression of dominant-negative forms of Rab10 or Rab8 in the ESO lineages as well as thoracic clones of a newly identified Rab8 loss-of-function allele do not confer any cell fate phenotypes. Furthermore, we have not detected any interaction between dEHBP1 and Rab8 or Rab10 in a yeast two-hybrid analysis. Therefore, we believe that loss of either Rab8 or Rab10 function does not underlie the dEHBP1 mutant phenotypes we describe.

Notably, many key players that affect cell polarity or mark subcellular compartments, including Arm, Rab11, Sec15, and F-actin, are not affected by the loss of dEHBP1. In addition, cell fate determinants Numb and Neuralized are correctly segregated upon asymmetric cell division in dEHBP1 mutant cells. However, loss of dEHBP1 specifically affects the abundance and localization of Spdo, a regulator of Notch signaling in asymmetrically dividing ESO cells, and the exocytosis and trafficking of Delta.

Spdo facilitates reception of Notch signal at the plasma membrane of the signal-receiving cell (Dye et al., 1998; O’Connor-Giles and Skeath, 2003; Benhra et al., 2010; Tong et al., 2010). Therefore, accumulation of Spdo in dEHBP1^−/− ESO clusters and its presence in the plasma membrane should result in a Notch gain of function, instead of the loss-of-function phenotype we observe. We and others have not observed any effects of Spdo overexpression upon cell fate acquisition in the ESO lineage (this paper; Jafar-Nejad et al., 2005; Tong et al., 2010). Alternatively, the accumulation of Spdo in the absence of dEHBP1 in these cells may reflect defects in its trafficking and membrane localization, which render the activation of Notch signaling more difficult.

dEHBP1 mutations cannot suppress the gain of function phenotype of overexpressed ligand-independent, activated Notch intracellular domain. In addition, dEHBP1 does not affect the steady-state levels of Notch protein, as well as its endocytosis. Therefore, we conclude that dEHBP1 functions at a level upstream of presenilin-mediated S3 cleavage of Notch during reception of the signal. Although we cannot exclude that dEHBP1 functions in the signal-receiving cell, where it may control the trafficking and localization of Spdo, we conclude that dEHBP1 also functions in the sending of the signal. Our conclusion is based on the fact that dEHBP1 mutations are able to suppress the gain of function of Notch phenotype conferred by the overexpression of DaPKC^ΔN. Overexpressed constitutively active DaPKC^ΔN places Spdo at the plasma membrane, enabling the activation of Notch signaling (Langevin et al., 2005; Roegiers et al., 2005). We find that upon loss of dEHBP1, Spdo is still found at the plasma membrane under conditions of overexpression of DaPKC^ΔN. Therefore, the suppression of the overexpression phenotype of DaPKC^ΔN by loss of dEHBP1 may be because of other defects, such as loss of the ability of Delta to signal (Fig. S3). Furthermore, loss of dEHBP1 leads to development of additional neurons despite the concomitant ectopic expression of Delta^R+, a variant of Delta, in clones within pupal nota at 36 h APF. Because the steady-state levels of Delta are not affected in dEHBP1^−/− ESO lineages, we examined whether dEHBP1 affects Delta trafficking in the signal-sending cell. Upon loss of dEHBP1, the abundance of Delta at the cell surface is significantly reduced, suggesting that exocytosis is defective. Importantly, most of the remaining extracellular Delta protein localizes at the basal side of the signal-sending cell. This suggests that in addition to affecting exocytosis of Delta, dEHBP1 may also play a role in basal-to-apical trafficking of Delta. This leads to a reduced level of Delta at the signaling interface, which interferes with proper Notch signaling in the cell receiving the signal. Although our results do not exclude a possible role of dEHBP1 in other aspects of Delta trafficking, such as endocytosis, reduced exocytosis of Delta should mask an endocytic defect in our assays. The enrichment of dEHBP1 in the basal and lateral area of the plasma membrane, its colocalization with F-actin at the ARS at the interface of the pIIa and pIIb cells, the reduction of Delta exocytosis in mutant cells, and the absence of Delta at the interface and the apical surface of the ESO cluster in mutant cells indicate a role of dEHBP1 in the Sec15/Rab11 recycling pathway. Indeed, the colocalization of dEHBP1 and Delta in sec15^−/− ESO lineages implies that the exocyst component, Sec15, controls exocytosis of Delta, Spdo, and dEHBP1 to the apical plasma membrane through a common compartment. Because loss of dEHBP1 does not affect the localization of either Rab11 or Sec15, we conclude that sec15 lies more upstream in the trafficking pathway regulating the localization of multiple components, while dEHBP1 functions during the later stages of intracellular trafficking. Furthermore, the physical interaction between dEHBP1 and Sec15 as well as Rab11 suggest a mechanism how dEHBP1 may regulate the membrane localization of Delta via its interaction with Sec15 and Rab11 at the pII cells interface, even though such interaction was detected under transient overexpression conditions. We propose that dEHBP1 is an adaptor of the Rab11/Sec15-positive, Delta-bearing vesicles required for exocytosis (Fig. 9).

The identification of dEHBP1 provides further compelling evidence that the exocytosis and recycling pathway of Delta during asymmetric divisions is tightly regulated (Emery et al., 2005; Jafar-Nejad et al., 2005; Rajan et al., 2009). The recycling pathway of Delta appears to be context dependent, i.e., it is not required in all cells that use Notch signaling (Windler and Bilder, 2010; Yamamoto et al., 2010; Banks et al., 2011). Still, the discovery of dEHBP1 as a novel player in Notch signaling provides the opportunity to test its role in Notch-related neurobiological behaviors, such as sleep and addiction (Kaun et al., 2011; Seugnet et al., 2011), as well as in Notch-related diseases, as for example in Wiskott-Aldrich syndrome, an immunodeficiency characterized by abnormal differentiation and function of T cell lineages (Cotta-de-Almeida et al., 2007; Radtke et al., 2010; Thrasher and Burns, 2010). Furthermore, because the anthrax toxins lethal factor (LF) and edema factor (EF) inhibit the Sec15/Rab11-dependent Delta-recycling pathway in flies and endothelial cells (Guichard et al., 2010), it would be interesting to hypothesize whether they target dEHBP1 to mediate their toxicity.

The following stocks were used: yw;FRT42D, yw;FRT42DdEHBP1^A28/CyO, yw;FRT42DdEHBP1^O4/CyO, yw;FRT42DdEHBP1^N08/CyO, yw;FRT42DdEHBP1^T14/CyO, dEHBP1^ΔEx24/CyO, yw; CyO, H{w[+mC]=PDelta2-3}HoP2.1/Bc¹; P{lacW}l(2)k09837/CyO, Df(2R)ED2751/CyO, Df(2R)Exel6065/CyO, Df(2R)ED3181/CyO, ywUbx-FLP; FRT42D UbiGFP, ywUbx-FLP; FRT42D UbiGFP PcDNA, hsFLP tub-GAL4 UAS-GFP;; FRT82B y+ tub-GAL80/TM6, UAS-mCherry-dEHBP1, UAS-Flag-dEHBP1, P{lArB}neurA101 ry506/TM3, ry^RK Sb¹ Ser¹, P{GawB}neurGAL4-A101 Kg^V /TM3, Sb¹ (BL #6393), UAS-cd8GFP, UAS-Rab5-YFP, UAS-Rab7-YFP, UAS-Rab8-YFP, UAS-Rab9-YFP, UAS-Rab10-YFP, UAS-Rab11-YFP, UAS-Sec15-GFP, UAS-Spdo-GFP, FRT82Bsec15³/TM6B, yw;P{tubP-GAL4}LL7/TM3, Sb¹. All crosses were performed on standard media at 25°C unless otherwise indicated. Fly strains bearing the docking sites VK00033 and/or VK00020 were used for injections of the cDNA and genomic rescue constructs, respectively. Fly transformation was mediated by phiC31 transgenesis according to standard protocols.

cDNA constructs were constructed by PCR amplification of CG15609 isoform B, rab11, and sec15 with appropriately designed primers and subsequent cloning into suitable vectors. Cloning protocols and DNA purification were performed according to standard protocols. Primer sequences are available upon request.

The genomic rescue construct was constructed by recombineering the 9.1-kb region from gene CG6805 to gene CG8963, residing on either side of CG15609, into the F25 P[acman] vector (Venken et al., 2006). All constructs were verified by sequencing before injection or transfection.

The full-length cDNA of dEHBP1-B was cloned in pET28a (EMD) as a NotI PCR fragment. The plasmid was transformed into BL21 Rosetta pLys (EMD). Recombinant protein production was induced by addition of IPTG at a final concentration of 1 mM, in two liters of bacterial culture, previously grown at 37°C, OD₆₀₀ nm = 0.7. Induction of recombinant protein production was performed for 3 h at 37°C. Bacteria were then harvested and used for purification of recombinant dEHBP1 under denaturing conditions using metal affinity chromatography with Ni-NTA His-Bind Resins (EMD), according to the manufacturer’s instructions.

Dissections, stainings, Delta endocytosis, and recycling assays were performed as described previously (Le Borgne and Schweisguth, 2003; Jafar-Nejad et al., 2005; Rajan et al., 2009). In brief, for assaying Notch endocytosis, Delta endocytosis, and Delta recycling, pupal thoracic epithelia were dissected at the selected time points after puparium formation (APF) in Schneider’s medium, followed by incubation with the appropriate primary antibody (anti-Notch or anti-Delta). In Notch and Delta endocytosis assays, dissections and subsequent incubations were performed at room temperature (RT). The samples were then washed three times in Schneider’s medium and fixed in 4% paraformaldehyde in Schneider’s medium for 20 min at RT. After washing with Schneider’s medium, thoraxes were permeabilized in 1x PBS, 0.2% Triton X-100 (1x PBT) and incubated with other primary antibodies in 1x PBT, 5% normal goat serum (NGS) at 4°C overnight. The following day, samples were washed three times in 1x PBT for 5 min at RT and incubated with appropriate secondary antibodies. For Delta recycling, thoracic epithelia were dissected in Schneider’s medium and then incubated with primary anti-Delta monoclonal antibody in Schneider’s medium for 1 h on ice, to label the extracellularly accessible pool of Delta molecules while the same time preventing their endocytosis. After this step of incubation, excess antibody was washed away with Schneider’s medium and thoraxes were fixed in 4% paraformaldehyde in Schneider’s medium for 20 min at RT either immediately (0-min time point, exocytosed Delta) or after allowing the internalization and trafficking of Delta at RT for 1 h in Schneider’s medium (60-min time point, recycled Delta). After washing, samples were permeabilized and incubated with other primary antibodies as already mentioned. For the exocytosis of Delta, thoracic epithelia were dissected in Schneider’s medium and immediately fixed with 4% PFA in Schneider’s medium. Then, they were incubated with primary anti-Delta monoclonal antibody in Schneider’s medium at 4°C overnight, to label all extracellularly accessible Delta. After washing, samples were permeabilized and incubated with other primary antibodies as already mentioned.

Primary antibodies used in this study are: mouse α-Arm 1:100 (N2 7A1, Developmental Studies Hybridoma Bank [DSHB], Iowa City, IA; Peifer et al., 1994), rabbit α-Avl (Lu and Bilder, 2005), rabbit α-Baz 1:1,000 (Wodarz et al., 1999), mouse α-Cut 1:500 (2B10, DSHB; Blochlinger et al., 1990, 1993), mouse α-Delta 1:1,000 (C54.9B, DSHB; Klueg et al., 1998; Qi et al., 1999), mouse α-Dlg 1:100 (4F3, DSHB; Parnas et al., 2001), rat a-E-Cad 1:100 (DCAD2, DSHB; Oda et al., 1994), mouse α-FasIII (7G10, DSHB; Patel et al., 1987), rat α-Elav 1:500 (7E8A10, DSHB), guinea pig α-Hsc3 1:500 (Ryoo et al., 2007), rabbit anti-Hook (Krämer and Phistry, 1996), guinea pig α-HRS 1:600 (Lloyd et al., 2002), rabbit (O’Neill et al., 1994) α-Neur 1:600 (Lai and Rubin, 2001), mouse α-NICD 1:100 (C458.2H, DSHB), rabbit α-Numb 1:1,000 (Rhyu et al., 1994), rabbit α-Patj 1:500 (Bhat et al., 1999), rabbit α-Pros 1:1,000 (Y.N. Jan, HHMI, University of California, San Francisco, San Francisco, CA), mouse anti-Rab1 (Satoh et al., 1997), mouse α-Rab11 antibody 1:200 (BD; Khodosh et al., 2006), guinea pig α-Sec15 1:2,000 (Mehta et al., 2005), guinea pig α-Sens (Nolo et al., 2000), rat α-Spdo 1:200 (O’Connor-Giles and Skeath, 2003), rat α-Su(H) (a kind gift from F. Schweisguth, Institut Pasteur, Paris, France), rat α-Ttk69 1:500 (P. Badenhorst, MRC Laboratory of Molecular Biology, Cambridge, UK; Guo et al., 1995). Alexa 488–conjugated (Invitrogen) and Cy2-, Cy3-, and Cy5-conjugated (Jackson ImmunoResearch Laboratories) secondary antibodies were used at 1:200–1:500.

Live imaging of pupae were performed as described previously (Zitserman and Roegiers, 2011) using an inverted microscope (model TE2000U; Nikon) equipped with a C1 confocal imaging system (488-, 543-, and 633-nm lasers; Nikon) and using a 60x 1.45 NA objective. In brief, pupae at the selected developmental time points were mounted on double-stick tape on a slide dorsal side up. The pupal case was removed carefully so that the thorax and anterior portion of the abdomen were revealed without puncturing them. The pupae were then freed from the pupal case and transferred by a soft brush to be subsequently mounted between a glass slide and coverslip: The pupae were surrounded by a moistened frame of Whatman paper. A continuous bead of silicone vacuum grease extruded from a 5-cc syringe functioned to seal the slide–coverslip combination, protecting the pupa from desiccation and elevating the coverslip so it rested gently on the pupal thorax. A small drop of water (1 µl) at the interface between the coverslip and thorax cuticle was then applied to improve image quality when using immersion objectives. Frames were taken every minute during a time period that initiated at the pI stage, just before cytokinesis, and lasted well into the pII cell stage.

All other confocal scans were acquired with a confocal microscope (model LSM510; Carl Zeiss) with its accompanying software using Plan Apochromat 40× NA 1.4 and Plan Apochromat 63× NA 1.4 objectives (Carl Zeiss). Samples were mounted in Vectashield (Vector Laboratories). Images of fly adult thoraces (treated by boiling in 10% KOH, dissected and mounted in 70% glycerol) were captured with a camera (MicroFire; Olympus) mounted to a stereomicroscope (model MZ16; Leica) using ImagePro Plus 5.0 acquisition software (Media Cybernetics).

Image processing (Z-reslicing and projections) and quantifications were performed using ImageJ software (National Institutes of Health). Fluorescence of Delta was quantified by measuring the mean gray value of pixel intensity in the corresponding channel in XY maximal projections of stacks of confocal sections of ESO lineages and within areas delineated by Sens nuclear staining and expanded by 20% to include Delta punctae toward the periphery of the cells. Stacks were typically spanning the area from apical to basal side of ESO clusters, containing more than 70 optical sections. A similar calculation was performed in areas of same size, but consisting of only epithelial cells and devoid of any ESO clusters, so that the background fluorescence would be estimated and subtracted from the fluorescence intensity value of Delta within the ESO lineages. Images were assembled in Canvas X software (ACDSee Products).

S2 cells were maintained in Schneider’s medium, supplemented with 10% heat-inactivated, filter-sterilized fetal bovine serum (FBS) (Sigma-Aldrich) and 1× penicillin and streptomycin antibiotic mix (Invitrogen). Handling of S2 cells was performed according to manufacturer’s instructions (Invitrogen).

Transfections of S2 cells were performed using Effectene (QIAGEN) using 0.4 µg of total DNA/1.5 ×10⁶ cells according to manufacturer’s instructions. For each type of coimmunoprecipitation (coIP), three wells of ∼5 × 10⁶ cells in total were seeded for transfection. Cells were harvested 48 h after transfection, into 15-ml Falcon polypropylene tubes by pipetting. After centrifugation at RT at 1,500 rpm for 5 min, the cells were washed in 1× PBS (equilibrated at RT) and centrifuged as before. The cells were then suspended in 0.2 volumes of ice-cold lysis buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 5% glycerol, and 0.2% Triton X-100), supplemented with EDTA-free cocktail of protease inhibitors (Roche) and placed on ice for 60 min with occasional mild agitation. The lysate was then cleared by centrifugation at 13,000 rpm at 4°C for 10 min. After removing one tenth of the volume for input control, 30 µl of 50% EZ red anti FLAG M2 or EZ red anti-HA Agarose beads slurry (Sigma-Aldrich) was added to each coIP reaction used. The beads had been previously equilibrated in cold lysis buffer including EDTA-free protease inhibitors (Roche). Binding was performed for at least 4 h at 4°C. After centrifugation at 3,000 rpm at 4°C, the supernatant was discarded and the beads were washed with ice-cold lysis buffer, including EDTA-free protease inhibitors. The procedure was repeated two more times. The beads were finally suspended in 1x Laemmli buffer and boiled for 10 min before electrophoresis and Western blot analysis.

pGEX4T-1 and pGEX4T1-Rab11 were transformed into BL21 pLys (Invitrogen) and single colonies were kept as glycerol stocks after being tested for successful induction of corresponding proteins. Overnight cultures were then prepared from the glycerol stocks and used the following morning for inoculation of larger volumes of LB medium at a final dilution of 1:50. The cultures were then allowed to grow at 37°C for ∼2 h. GST and GST-Rab11 proteins were induced at 37°C for two additional hours by addition of IPTG at a final concentration of 0.1 mM. Bacteria were lysed in 1x Bugbuster (EMD). Purification of GST proteins from cleared lysates was achieved using GST beads (GE Healthcare) following the manufacturer’s instructions. Before binding to in vitro–translated products, beads and proteins were equilibrated in ice-cold lysis buffer containing EDTA-free cocktail of protease inhibitors (Roche), as described above. GST proteins were incubated with in vitro–translated dEHBP1 in ice-cold lysis buffer for at least 5 h at 4°C with rocking.

In vitro translation of CG15609 isoform B was performed using the TnT T7 Quick Coupled Transcription/Translation System (Promega) combined with the Transcend t-RNA chemiluminescent nonradioactive detection method (Promega). As a template, we used the T7 promoter containing pET28a-dEHBP1 clone, which was originally used for bacterial expression of dEHBP1 in our antibody production experiments.

Fig. S1 shows that dEHBP1 does not colocalize with markers of Golgi and early, late, or recycling endosomes (related to Figs. 3 and 4). Fig. S2 shows that the distribution of different subcellular compartments and the integrity of ARS are not affected in the absence of dEHBP1 (related to Fig. 5). Fig. S3 shows that loss of dEHBP1 is epistatic to gain of function of Notch signaling, achieved by ectopic expression of constitutively active DaPKC^ΔN or Delta^R+, but not by activated Notch^NEXT. Video 1 shows that mCherry-dEHBP1 is localized within intracellular vesicles and at the interface of pIIa and pIIb cells. Video 2 shows that mCherry-dEHBP1 is localized partially with Spdo-GFP at the apical side of the interface of pIIa and pIIb cells. Video 3 shows that mCherry-dEHBP1 is localized partially with Spdo-GFP at the medial side of the interface of pIIa and pIIb cells.

We thank the Bloomington Stock Center, and the Developmental Studies Hybridoma Bank for reagents. We are grateful to Dr. Francois Schweisguth for the anti-Su(H), Dr. Marcos Gonzalez-Gaitan for the anti-Rab5 and Dr. Jim Skeath for anti-Spdo antibodies, respectively. The rat α-120kDa antibody was a kind gift from Dr. Satoshi Goto. We thank Yuchun He and Hongling Pan for help with injections. We thank Dr. Manish Jaiswal, Dr. Hamed Jafar-Nejad, Dr. Timothy Mahoney and Dr. Chao Tong and Vafa Bayat for helpful comments on the manuscript.

Confocal microscopy was supported by the National Institutes of Health grant P30HD024064 to the BCM Intellectual and Developmental Disabilities Research Center. N.G. has been supported by European Molecular Biology Organization long term fellowship and Howard Hughes Medical Institute. S.Y. is supported by the Nakajima Foundation. F.R and D.Z. are supported by NIH RO1 NS059971 grant and the Pennsylvania Tobacco Settlement fund. H.B. is an HHMI Investigator.