Drosophila E-cadherin is required for the maintenance of ring canals anchoring to mechanically withstand tissue growth

Significance

This work addresses the interplay among membrane trafficking, cell adhesion, and tissue integrity maintenance in the Drosophila female germline. The clathrin adaptor protein 1 (AP-1) complex is shown to regulate the trafficking of E-cadherin to ring canals (RCs), a structure resulting from incomplete cytokinesis and allowing intercellular communication. E-cadherin assembles adhesive clusters that, as revealed by EM analyses, organize a dense microvilli meshwork wrapping around RCs. Although dispensable for RC biogenesis and maturation, AP-1 and E-cadherin are required to maintain RCs’ anchoring to the plasma membrane at the onset of vitellogenesis, when cells experience exponential growth and increased mechanical stress. Our study unravels a previously unidentified function for E-cadherin in maintaining RC anchoring to the plasma membrane.

Abstract

Intercellular bridges called “ring canals” (RCs) resulting from incomplete cytokinesis play an essential role in intercellular communication in somatic and germinal tissues. During Drosophila oogenesis, RCs connect the maturing oocyte to nurse cells supporting its growth. Despite numerous genetic screens aimed at identifying genes involved in RC biogenesis and maturation, how RCs anchor to the plasma membrane (PM) throughout development remains unexplained. In this study, we report that the clathrin adaptor protein 1 (AP-1) complex, although dispensable for the biogenesis of RCs, is required for the maintenance of the anchorage of RCs to the PM to withstand the increased membrane tension associated with the exponential tissue growth at the onset of vitellogenesis. Here we unravel the mechanisms by which AP-1 enables the maintenance of RCs’ anchoring to the PM during size expansion. We show that AP-1 regulates the localization of the intercellular adhesion molecule E-cadherin and that loss of AP-1 causes the disappearance of the E-cadherin–containing adhesive clusters surrounding the RCs. E-cadherin itself is shown to be required for the maintenance of the RCs’ anchorage, a function previously unrecognized because of functional compensation by N-cadherin. Scanning block-face EM combined with transmission EM analyses reveals the presence of interdigitated, actin- and Moesin-positive, microvilli-like structures wrapping the RCs. Thus, by modulating E-cadherin trafficking, we show that the sustained E-cadherin–dependent adhesion organizes the microvilli meshwork and ensures the proper attachment of RCs to the PM, thereby counteracting the increasing membrane tension induced by exponential tissue growth.

E-cadherin (E-Cad) is a core component of intercellular adhesion complexes in cohesive metazoan tissues. E-Cad assembles into clusters that are stabilized by actin filaments via β- and α-catenin at the level of adherens junctions and form an adhesive belt mechanically linking cells together. A key feature of adherens junctions is their plasticity, which enables tissue remodeling, sustained by a constant endocytosis- and exocytosis-regulated E-Cad turnover (1) that is critical for various morphogenetic processes in epithelia (2–5).

Drosophila oogenesis is a rich, multifaceted developmental process during which E-Cad function is not limited to epithelia, because it also regulates intercellular collective migration (6, 7) and the adhesion of stem cells to their niche (8). Cells derived from two different stem cell populations initially assemble into egg chambers composed of a follicular epithelium surrounding a 16-cell germline cyst (GC), itself composed of one oocyte and 15 nurse cells. During the next 64 h, GC cells grow to hundreds of times their initial volume. Oocyte growth is supported by cytoplasmic connections with nurse cells through ring canals (RCs) (Fig. 1 A and B), intercellular bridges that, instead of undergoing abscission, are stabilized on arrested cleavage furrows (9, 10). Recent findings revealed that RCs play a vital role in germline as well as in somatic tissues (10). RCs are composed of a noncontracting subcortical actin ring (11), the inner rim, attached to an electron-dense plasma membrane (PM) (12), the outer rim (Fig. 1A). RCs have been studied mainly in Drosophila female GCs (9) where genetic screens uncovered a variety of actin regulators controlling their establishment at the onset of oogenesis and their growth throughout the entire process (13–17). However, the molecular machinery involved in anchoring the PM to the RC remains unknown. Mutations in several membrane-traffic regulators affect the integrity of nurse cells’ PM, causing multinucleation and giving rise to remnants of detached RCs (18–24); these observations suggest that an unidentified membrane cargo is required for anchoring RCs to the PM.

Fig. 1.

Nurse cells’ multinucleation in AP-1 mutant female GCs. (A) Schematic representation of the GC consisting of a single oocyte (Oo, nucleus) connected to 15 nurse cells (blue) via RCs (red) and a surrounding monolayer of about 650 somatic follicle cells (green). (Inset) Schematic representation of a transverse section through the RCs composed of an inner rim [red, containing the Adducin-like Hu-li tai shao (Hts) (13, 15) and the filamin Cheerio (16)] contacting an electron-dense PM (outer rim, black) that itself is connected to the rest of the nurse cell PM (gray). (B) Stereotyped organization of the female GC before and after detachment of nurse cells’ RCs. The oocyte has a gray nucleus; nurse cells have colored nuclei. (C) Stage 8 wild-type and AP-1 mutant [identified by the loss of nuclear localization signal (NLS)::GFP, blue] GCs stained for actin (green) and DAPI (red). Arrows indicate RCs connecting nurse cells in control GCs. Arrowheads indicate RCs floating in the cytoplasm of multinucleated nurse cells in AP-1 mutant GCs (at least one floating ring was observed in 29 of 34 mutant stage 8 or older GCs). (C′) Quantitation of multinucleated AP-1 mutant GCs at stage 7 to stage 9 or older. (D) Maximal projections of 5 µm of anchored and clustered floating RCs in control and AP-1 mutant GCs.

Here we describe an RC detachment phenotype in mutants of the clathrin adaptor protein 1 (AP-1), a protein complex regulating polarized membrane protein sorting from the trans-Golgi network and endosomal compartments (25), and provide direct evidence that polarized membrane trafficking to RCs allows an E-Cad–mediated mechanical strengthening of RC anchoring necessary to resist the membrane tension generated by cellular growth.

Results

Loss of AP-1 Induces Multinucleation of Nurse Cells in Female GCs.

In this study we generated homozygous AP-47^SHE11 mutant GCs in the female germline (the μ subunit of AP-1, hereafter referred to as “AP-1 mutants”). Actin staining revealed that nurse cells of AP-1 mutant GCs progressively became multinucleated, exhibiting floating RCs organized in clusters (Fig. 1 C and D). Multinucleation, indicating a defect in membrane stability, was first observed at the onset of vitellogenesis (stage 8). Loss of AP-1 never caused loss of oocyte membrane integrity, suggesting that this membrane is more robust than that of nurse cells, presumably because of its differential organization and composition (26).

A Faster Growth Rate Correlates with Higher PM Tension.

Multinucleation begins at a stage during which the oocyte accumulates yolk and GCs grow 4.6 (stage 8) to 34 (stage 10a) times faster than at previous stages (Fig. 2 A and B). Because a faster growth rate could affect mechanical membrane properties, providing a rational for this stage-dependent multinucleation phenotype, we probed PM tension by making 5-µm-wide holes in nurse cells’ PM using laser ablation. Such holes did not heal but instead propagated until they reached the PM of neighboring nurse cells, leading to multinucleation (Fig. 2C and Movie S1). We measured the retraction of PM extremities and vertices after ablation and found that membrane recoil was about four times slower (1 μm/min) than in epidermal cells and did not differ significantly in slow- and fast-growing GCs (Fig. 2D and Fig. S1 B–E). However, cutting the sheet-like nurse cell–nurse cell interface may not release tension as efficiently as cutting the string-like belt of adherens junctions in epithelia. Furthermore, tensions may not be released as efficiently at later stages as at the earlier ones, because we made holes of the same size in nurse cell–nurse cell interfaces although the surfaces of these interfaces differ by 4.5-fold between stages 5 and 9.

Fig. 2.

A faster growth rate correlates with an increase in PM tension. (A) The change in GC volume after exit from the germarium. Oogenesis stages are indicated above the curve. The arrow indicates the onset of vitellogenesis. (B) GC growth rates from stages 3–10a. The arrow indicates the onset of vitellogenesis. (C) Nano-ablation of a PM from the nurse cell of a stage 7 GC expressing PH::GFP. The PM regresses progressively after a 5-µm-long wound was made. (Upper) Confocal planes where the wound was made. (Lower) Orthogonal views to visualize the entire targeted portion of the PM. (D) Displacement of the free PM extremities (arrows in the Inset) generated by PM laser ablation in stage 5 and stage 9 nurse cells. (E) Representative cases of ablation of nurse cells’ PM from stage 6 (Upper) and stage 9 (Lower) GCs. PM deformations were observed before the cut (arrows) in stage 6 GCs but not in stage 9 GCs, and more deformations appear after the cut in the PM contiguous to ablated PM (arrowheads) in stage 6 GCs than in stage 9 GCs. (F) Density of deformations in the PM contiguous to the ablated PM before and after the cut (stages 5 and 6, n = 14; stage 7, n = 26; stage 8, n = 23; and stages 9 and 10, n = 23).

Fig. S1.

PM fluctuations following laser ablation in WT and AP-1 mutant cells. (A) Color-coded time projections of stage 5 and stage 9 PH::GFP-expressing egg chambers following laser ablation of a segment of nurse cells’ PM. The arrows indicate the site of laser ablation. Nurse cells neighboring the targeted nurse cells undergo more fluctuations in stage 5 egg chambers than in stage 9 egg chambers, as illustrated by the enlarged peak on an signal intensity plot profile in A′ along the line marked by the double arrow in A. (B) Retraction of vertices (arrows) in PH::GFP-expressing nurse cells of a stage 7 egg chamber (Upper) and in E-Cad::GFP-expressing epithelial cells of the pupal notum of Drosophila (Lower) following laser ablation of a segment of nurse cells’ PM (arrowhead). (B′) Velocities of vertex retraction. (C) The positions of the two PM extremities (arrowheads) generated by laser ablation and vertices (arrows) were tracked over time. (D and E) Examples of the displacement of vertices (D) and extremities (E) from their original positions following laser ablation. The red slope shows the measurement of their maximal velocities. (F and G) Recoil velocity of vertices (F) and extremities (G) following laser ablation in control and AP-47^SHE11 GCs. For every stage, no significant difference was observed between controls and mutants. (H) Density of tubular deformations before and after laser ablation in control and AP-47^SHE11 GCs.

Although these two experimental biases prevented us from assessing PM tension directly, we noticed that adjacent nurse cells were subjected to more fluctuations in cell shape when ablation was performed at early stages rather than later stages (Fig. S1 A and A′). This observation indicates that PMs are more prone to deformations at early stages, possibly because of lower PM tension. Accordingly, we observed that after ablation, tubular pleckstrin homology domain::GFP⁺ (PH::GFP⁺) deformations appeared on the PM contacting adjacent nurse cells (Fig. 2E, arrowheads). Because such tubular deformations are reminiscent of those observed in vitro at the surface of giant unilamellar vesicles and in vivo at the PM of cells upon the reduction of PM tension (27, 28), we reasoned that tension release induced by ablation is causal to the appearance of deformations. We found that tubular deformations were frequent in slow-growing GCs but were hardly detectable in fast-growing ones (Fig. 2 E and F, Fig. S1H, and Movies S2 and S3). Strikingly, PM tubular deformations already were present before ablation in slow-growing GCs but not in fast-growing ones (Fig. 2 E and F) and therefore (because their presence does not rely on laser ablation) could be used as a reliable readout for PM tension. Thus, we concluded that the PM tension is higher in stages 8–10a GCs than at earlier stages. Finally, similar recoil velocities and tubular deformations following laser ablation were obtained upon loss of AP-1, indicating that AP-1 does not significantly regulate PM tension (Fig. S1 F–H). PM tension is the sum of the in-plane lipid bilayer tension and the protein-dependent membrane-to-cortex attachment (29). We did not further assess the respective contributions of these factors to changes in PM tension during oogenesis, but we propose that the exponential growth that begins at stage 8 affects the mechanical membrane properties, eventually causing multinucleation in AP-1 mutants.

AP-1 and Rab11 Control the Maintenance of RC Anchoring to the PM and E-Cad Localization in Nurse Cells.

To follow the dynamics of the disappearance of the PM, we monitored the distribution of the PM marker E-Cad using an E-Cad::GFP knockin line (30) in AP-1 mutant GCs. Live imaging revealed that multinucleation was caused by the detachment of the PM from RCs, immediately followed by PM fragmentation (Fig. 3A). We never observed fragmentation of portions of PM devoid of RCs, suggesting that multinucleation was caused exclusively by the detachment of the PM from the RCs. This notion was supported further by our analysis of fixed tissue: According to the stereotyped organization of the GC, loss of all RCs connecting nurse cells to nurse cells but not of RCs connecting nurse cells to the oocyte should lead to the formation of three syncytia containing two, four, and eight nurse cell nuclei. This exact configuration was observed in AP-1 mutants (Figs. 1B and 3B), further indicating that nurse cell–nurse cell interfaces devoid of RCs remain stable in the AP-1 mutant and that nurse cells’ multinucleation is caused by RC detachment. Thus, although dispensable for RC establishment, AP-1 activity is required to maintain RCs’ anchoring to the PM beginning at stage 8.

Fig. 3.

AP-1 and Rab11 control the maintenance of RCs’ anchoring and E-Cad localization in nurse cells. (A) Time-lapse imaging of a stage 8 AP-1 mutant GC. The RC is seen as a hole (arrow) in the otherwise continuous E-Cad::GFP⁺ PM. (Upper) Single focal planes. (Lower) Orthogonal sections. PM detachment starting from the RC (t = 5 s) is followed by PM fragmentation (t = 10 s, n = 5). (B) A stage 9 AP-1 mutant GC stained for actin (green) and Rab11 (red). The two panels correspond to projections of different focal planes. With the exception of the oocyte (gray asterisk), every nurse cell PM initially bearing an RC collapsed, resulting in three syncytia containing 8, 4, and 2 nurse cell nuclei (as indicated by colored asterisks and the corresponding cartoon in Fig. 1B) and clusters of detached RCs (arrowheads). (C) RCs in stage 9 GCs stained for E-Cad (green) and actin (red) (maximal projections of 5–7 µm). E-Cad is enriched around RCs, and this enrichment can be lost in stage 8 and older AP-1 mutants. (C′) Quantification of data in C. Total numbers of RCs and GCs examined are indicated above the columns. (D) Stage 8 control (NLS::GFP⁺) and AP-1 mutant (NLS::GFP⁻) GCs stained for Rab11. Enlarged endosomes were observed in 17 of 18 AP-1 mutant GCs (from stages 4–10). (E) Stage 8 AP-1 mutant nurse cell stained for E-Cad and Rab11. An entire GC is shown at low magnification in the left panel. The other three panels show magnified views of the boxed area in the left panel. Arrowheads indicate enlarged endosomes positive for Rab11 and E-Cad, which partially colocalized in 12 of 16 stage 4–10 GCs. In three stage 8 GCs, 94% of E-Cad⁺ endosomes (145/155) were positive for Rab11, and 63% of Rab11⁺ endosomes (145/222) were positive for E-Cad. (F) RCs of stage 9 GCs overexpressing WT Rab11 or Rab11^S25N were stained for E-Cad (green) and actin (red) (maximal projections of 5–7 µm). Loss of E-Cad enrichment was seen in 35 of 83 RCs from 10 stage 9 or older Rab11^S25N-expressing GCs. (G) A stage 10 GC overexpressing Rab11^S25N. The arrow indicates floating RCs (seen in 9 of 13 stage 8 or older GCs). The right panel shows a magnified view of the boxed area in the left panel.

Live imaging of E-Cad::GFP and immunostaining of endogenous untagged E-Cad revealed that the whole surface of nurse cells’ PM is decorated by E-Cad⁺ clusters visibly enriched around RCs (Fig. 3C). In AP-1 mutant GCs, this enrichment started to disappear at stage 8 (Fig. 3 C and C′). In mutant GCs, E-Cad also localized to cytoplasmic puncta that were absent from control cells and already were present in non-multinucleated GCs at stage 8 (Fig. 3E), indicating that cytoplasmic mislocalization of E-Cad in AP-1 mutant GCs precedes multinucleation. In mammalian cells, AP-1 controls the subcellular localization and function of the Rab11⁺ recycling endosome compartment (31, 32), and E-Cad transits through Rab11⁺ compartments (33–35). This function raises the possibility that E-Cad mislocalization in AP-1 mutants involves a defect in Rab11-dependent trafficking. Consistent with this proposition, Rab11 localization changed from small endosomes distributed throughout the entire cytoplasm in control nurse cells to enlarged endosomes in AP-1 mutant nurse cells (Fig. 3D), and the majority of E-Cad cytoplasmic puncta localized to Rab11⁺ compartments (Fig. 3E). To assess the effect of Rab11 on E-Cad trafficking, we overexpressed a dominant-negative form of Rab11 (Rab11^S25N) that was reported to block entry into recycling endosomes in mammalian cells (36). Overexpression of Rab11^S25N phenocopied AP-1 mutants with loss of E-Cad enrichment around RCs (Fig. 3F) and multinucleation of stage 8 and older nurse cells (Fig. 3G). Thus, in both AP-1 and Rab11 mutant backgrounds, the presence of fewer E-Cad clusters surrounding RCs correlates with RC detachment leading to multinucleation.

E-Cad Controls RCs Anchoring to the PM.

This correlation raises the possibility that E-Cad/Shotgun (Shg) is necessary to anchor RCs. Consistent with this suggestion, GCs mutant for the shg^IG29 loss-of-function allele (37) and null β-catenin (β-Cat) alleles arm^XP33 (Fig. 4A), arm^YD35, or arm^XK22 (6, 37, 38) display nurse cell multinucleation. However, the amorphic shg^IH and the null shg^R69 mutant alleles do not cause nurse cell multinucleation (6, 39). We reasoned that this apparent discrepancy could be explained by functional compensation by the classical cadherin N-cadherin (N-Cad) in E-Cad–null mutant GCs, as reported in other tissues in ref. 40. We found that N-Cad was not detected in control GCs, but in E-Cad-null mutant GCs N-Cad was expressed ectopically and was localized to the PM (Fig. 4B). We propose that E-Cad somehow negatively regulates N-Cad transcription and/or translation, although we cannot rule out the possibility that N-Cad is translated in control GCs but is targeted to degradation and is below our detection threshold. Nevertheless, in the absence of E-Cad, β-Cat still localized to the PM of nurse cells, albeit at lower levels than in controls (Fig. 4B), as is consistent with functional compensation. This observation prompted us to prevent N-Cad ectopic expression by using N-Cad^RNAi in E-Cad–null mutant GCs. N-Cad silencing in shg⁺ GCs did not cause any detectable phenotype (n = 30), but N-Cad silencing in shg^R69 mutant GCs induced multinucleation (Fig. 4C) in addition to the oocyte mispositioning defects expected from the loss of E-Cad (39). Furthermore, we observed that shRNA-mediated E-Cad depletion also caused nurse cell multinucleation (Fig. 4D). In this situation, we speculate that incomplete E-Cad depletion is sufficient to disrupt E-Cad function in RC anchoring but not in repressing N-Cad expression. Accordingly, N-Cad was not ectopically expressed, and β-Cat was no longer recruited to the PM in E-Cad–depleted GCs (Fig. S2). Together, our results show that N-Cad is responsible for a functional compensation of E-Cad loss in RC anchoring, explaining why nurse cell multinucleation is observed in β-Cat– but not in E-Cad–null alleles, and enable us to conclude that E-Cad participates in RCs anchoring.

Fig. 4.

E-Cad controls RCs’ anchoring to the PM. (A) Stage 10 arm mutant GCs stained for Actin. Arrowheads indicate floating RCs (at least one floating RC was seen in three of four arm^XP33 mutant stage 8 or older GCs). (B) Stage 6 control and E-Cad mutant (shg^R69) GCs stained for N-Cad (green) and Armadillo (red). Ectopic expression of N-Cad was observed in 15 of 15 stage 4 and older shg^R69 GCs. (C) A stage 10 E-Cad mutant GC expressing N-Cad shRNA^HMS02380 stained for DAPI (red) and actin (green). Floating RCs (arrowhead) were observed in 5 of 10 stage 8 or older GCs. The right panel shows a magnified view of the boxed area in the left panel. (D) GCs expressing E-Cad shRNA^GL00646 in stage10 egg chambers stained for actin. The arrow indicates floating a RC. (At least one floating RC was seen in 14 of 22 E-Cad shRNA^GL00646 and in 10 of 13 E-Cad shRNA^HMS00693 stage 8 or older GCs.)

Fig. S2.

RNAi-mediated E-Cad depletion is not accompanied by ectopic N-Cad expression. (A) Control and MTD > E-Cad^shRNAGL00646 ovarioles stained for β-Cat. In E-Cad^shRNAGL00646, β-Cat localizes to small membrane patches in stage 6 or older GCs (n = 5) localizing or not to RCs and is absent from the PM in stage 7 and older GCs (n = 16). (B) MTD > E-Cad^shRNAGL00646 stage 9 egg chamber presenting an oocyte localization defect (asterisk), stained for DAPI (blue), Actin (red), and N-Cad (green). N-Cad is not expressed in E-Cad–depleted egg chambers (n = 4) except for a weak signal observed at the oocyte PM in three of four stage 7 or older GCs.

Disruption of a Microvillosities-Rich PM Around RCs in AP-1 Mutant GCs.

How exactly could the cadherin–catenin complex participate in the maintenance of RC anchorage? Transmission electron microscopy (TEM) analysis of multinucleated AP-1 mutant GCs revealed that the inner rim of detached RCs remained attached to the outer rim, which itself was still connected to portions of the PM surrounding the RC (Fig. 5A). Thus, RC detachment does not result from the detachment of the inner rim from the outer rim but rather from the disconnection of a portion of the PM surrounding RCs. TEM further revealed that in control GCs the PM surrounding RCs appeared highly convoluted (Fig. 5 A–B′). In striking contrast, this region appeared devoid of such convolutions in RCs still anchored in AP-1 mutant GCs, (Fig. 5 B and B′). We further examined the ultrastructural topology of the nurse cells’ PMs using scanning block-face EM (41). This analysis shows that the complex convolutions surrounding RCs are caused by tightly packed tubular extensions of PM, 65 ± 14 nm in diameter and 1,500 ± 400 nm in length, that protrude into the intercellular space between nurse cells (Fig. 5C and Movies S4 and S5). Such protrusions also were observed at lower density over the rest of the PM distant from RCs (Fig. 5 C and D and Movies S4 and S5). We further characterized these structures using light microscopy. We propose that actin-positive filaments at the PM at a distance from RCs (Fig. 6A) correspond to individual protrusions and that the high density of actin (Fig. 6A) and the presence of the actin crosslinker α-Actinin (Actn) (Fig. 6 C and D) (42), the actin regulator Enabled (43), and the microvilli marker phospho-Moesin (Fig. 6D) at the PM surrounding RCs is caused by the local abundance of the protrusions revealed by TEM (Fig. 5A). AP-1 mutant cells displayed lower levels of Actn around RCs (Fig. S3A), consistent with the loss of PM convolutions around RCs (Fig. 5 B and B′ ), further indicating that AP-1 is necessary for protrusions organization around RCs.

Fig. 5.

Disruption of a microvillosity-rich PM around RCs in AP-1 mutant GCs. (A) TEM image of anchored and detached RCs in control and AP-1 mutant stage 9 GCs. (Left) Low magnifications; nuclei are cyan; cytosol is yellow. (Center) High-magnification views of the boxed areas in left panels. The RC actin-rich inner rim is red. (Right) Interpretative drawings of the center panels. The inner rim and parallel fibers inside protrusions are red. The inner rim (red) is attached to the outer rim (arrows, thick line) in anchored and detached RCs. The outer rim of the detached RC is itself still connected to a portion of PM (arrowheads). (B and B′) TEM image of control and AP-1 mutant anchored RCs (stage 8). Boxed areas in B show a portion of the PM surrounding RCs. High magnifications of these areas in B′ show complex PM convolutions in the control and the absence of such convolutions in the mutant. (C) Projection over 3 µm (30 sections) of consecutive scanning block-face EM images of an RC (inner rim in red) and the neighboring PM in a control stage 8 GC. The RC is surrounded by microvilli-like protrusions also present at lower densities over the rest of the PM. (D) Consecutive projections over 1.7 µm (17 sections) of TEM images after PM segmentation through the volume of a RC in a control stage 8 GC and over 8.5 µm (85 sections) through its whole volume.

Fig. 6.

E-Cad organizes microvillosity-like structures. (A and A′) RC from a control stage 8 fixed GC stained for E-Cad (green) and actin (red). Actin staining is displayed under two different brightness/contrast settings to illustrate properly both the faint actin-positive filaments all over the cortex (seen better in A′) and the intense actin signal at the periphery of RCs (seen better in A). (B) RCs from control and E-Cad shRNA^GL00646 live stage 8 GCs expressing the PH::GFP probe. (C) Control and E-Cad shRNA^GL00646 fixed stage 8 GCs stained for Actn (green) and the inner rim marker Hts-RC (red). (D) Control fixed stage 9 GC stained for E-Cad (red), P-Moesin (green), and Actn (blue). E-Cad clusters are more peripheral than the portion of P-Moesin/Actn⁺ PM surrounding the RCs.

Fig. S3.

Localization of Actn, Par-3 (Baz), and Dlg around RCs. (A) AP-47^SHE11mosaic GC in a stage 10 egg chamber stained for Actin (red) and for Actn (green). The nls::GFP clonal marker (blue) is lost from some nurse cells, indicating that these cells are AP-47^SHE11 mutants and that the clonal marker still produced by control cells does not diffuse from the control cells to a great extent. The center and right panels show magnified views of the boxed area in the left panel, where three controls (arrows) and one AP-47^SHE11 mutant (arrowheads) are in the same focal plane. The Actn signal in RCs was two to three times higher in control cells (n = 5 RCs) than in AP-47^SHE11 mutant cells (n = 4 RCs). (B) RC of a control stage 8 egg chamber stained for Par3 (red), Dlg (green), and E-Cad (blue). Projections of 2 µm. (C and D) PM of control nurse cells from stage 7 egg chambers stained for Actin (green) and Par-3 (red in C) or Dlg (red in D). Single focal planes are shown.

E-Cad Organizes Microvillosity-Like Structures in Nurse Cells.

Loss of E-Cad enrichment and loss of protrusions around RCs in AP-1 mutants prompted us to analyze the direct requirement for E-Cad in protrusion organization. In E-Cad–depleted GCs, lower PH::GFP signals (Fig. 6B) and an almost complete loss of Actn signals around RCs indicate that protrusions surrounding RCs are severely affected (Fig. 6C). Furthermore, protrusions distributed all over the PM of nurse cells were visibly affected (Fig. S4 A and A′). Although this effect indicates that E-Cad controls the organization of protrusions, the E-Cad⁺ clusters enriched around RCs do not localize to the protrusion-dense region but rather to its immediate periphery (Fig. 6 A and D), and clusters distributed all over the rest of the PM do not colocalize with actin-positive linear structures (Fig. 6A). Finally, we found that the polarity markers Par-3 and Discs large 1 (Dlg) that are enriched around RCs (Fig. S3B) also localized to the rest of the nurse cells’ PM, but neither localized to microvilli (Fig. S3 C and D). Thus, we propose that AP-1–dependent E-Cad clusters organize protrusions independently of a polarized distribution of Par3 and Dlg.

Fig. S4.

E-Cad organizes microvillosity-like structures in nurse cells. (A) Control and E-Cad shRNA^GL00646 fixed stage 9 egg chambers stained for Actin. (A′) Magnified views of boxed areas in A. In controls, most of the actin-positive structures are linear and are clearly separated by actin-negative regions. In contrast, in addition to smaller, fainter linear structures, the cortex of E-Cad–depleted nurse cells presents punctate actin-positive structures with fewer actin-negative regions between them.

Discussion

In this article, we report that AP-1/Rab11 regulate the polarized trafficking of E-Cad and that E-Cad assembles adhesive clusters that are needed to maintain the anchoring of RCs to the PM at the time of exponential GC growth; this growth is associated with a change in mechanical membrane properties that probably is caused by increased membrane tension.

We show that in Drosophila nurse cells, defects in AP-1/Rab11 function lead to the progressive disappearance of E-Cad⁺ clusters surrounding RCs, suggesting that AP-1/Rab11 ensure the polarized delivery of E-Cad to RCs. Several studies in Drosophila, Caenorhabditis elegans, and mammals have implicated AP-1/Rab11 in E-Cad trafficking (33–35, 44, 45). Mammalian E-Cad carries a tyrosine-based AP-1 sorting signal, but the absence of such a motif in Drosophila E-Cad argues against a direct recognition of E-Cad by AP-1. Instead, interactions between E-Cad and membrane-trafficking regulators can be mediated by adaptors such as β-Cat (46) and the type I γ phosphatidylinositol phosphate kinase PIPKIγ (34). Alternatively, because AP-1 also controls the position and morphology of recycling endosomes (31, 32), the E-Cad trafficking defect we describe could result from malfunctioning recycling endosomes. Nurse cells’ multinucleation also has been described for Rab6 (18), Rab11 (24), PI4KIIIa (23), and components of the Exocyst (20, 21) and ESCRT (19) complexes. Although the involvement of PI4KIIIa and ESCRT in E-Cad trafficking is unknown, E-Cad trafficking requires the activity of Rab6, Rab11, and the exocyst complex (46, 47). We therefore anticipate that defective intracellular trafficking of E-Cad toward adhesive clusters contributes to the multinucleation phenotypes in these trafficking regulators.

Whether AP-1 acts directly or not, this study unravels previously unidentified E-Cad functions. We show that E-Cad is required for the maintenance of RCs’ anchorage. Intriguingly, E-Cad also organizes microvillosities at the surface of nurse cells, even though it does not localize to these microvillosities but rather to adhesive clusters interspersed between them. How could E-Cad organize microvilli remotely? One could speculate that close apposition of membranes through E-Cad–dependent adhesion somehow stabilizes protrusions, possibly by allowing specific contacts between protrusions. In epithelia, intermicrovillar adhesion is assured by protocadherins (48–50), and although any requirement for microvilli remains to be demonstrated, we envisage that they reinforce RCs anchorage. This remote action of E-Cad is somewhat reminiscent of another E-Cad function during oogenesis: E-Cad clusters at the nurse cells’ PM control the orientation of the filopodia-like actin cables that position nuclei during later stages of oogenesis. In a similar fashion, these E-Cad clusters are interspersed between the membrane-originating tips of filopodia (51).

E-Cad functions in cell adhesion, migration, and stem cell maintenance during Drosophila oogenesis have been studied extensively. However, as illustrated by this study and ref. 49, additional unsuspected roles for E-Cad remain to be identified. One of the reasons these roles have not been identified previously is that, as shown in Drosophila follicular epithelium (40), in mammals (52, 53), and, as we report, in GCs, N-Cad compensates for E-Cad function. Our observations also provide a rationale for the previous discrepancy between E-Cad (no multinucleation) and β-Cat (multinucleation) phenotypes in the GCs (6, 38). We further propose that similar functional intercompensation between classical cadherins is likely to occur in a number of other tissues, developmental stages, and organisms and is likely to bias observations similarly, depending on the methods used to affect E-Cad function.

AP-1/Rab11/E-Cad are required for maintenance of RCs’ attachment throughout vitellogenesis, during which faster cellular growth is accompanied by an increase in PM tension. We therefore propose that RCs’ anchoring must be reinforced through the AP-1–mediated delivery of E-Cad to withstand the increased membrane tension generated by exponential growth during vitellogenesis. Without reinforcement, these forces would be sufficient to tear the PM surrounding the RCs physically. Why would the PM tear only at this location and not anywhere else? We can only suppose that dynamic rearrangements of the PM surrounding RCs that are required either to organize the microvilli-rich region or to accommodate the growth of RCs somehow destabilize the PM. Based on the conservation of the RC functions in the germlines of invertebrates and vertebrates, we anticipate that this function of E-Cad is evolutionarily conserved.

Materials and Methods

Materials and methods are briefly described here. Further details are given in SI Materials and Methods.

Drosophila Stocks and Genetics.

AP-47^SHE11, shg^R69, and arm^XP33 mutant germline clones were generated using the FLP/FRT system. The MTD-GAL4 line was used to drive RNAi and Rab11 dominant-negative expression in the germ line.

Immunofluorescence and Antibodies.

Ovaries from adult flies were fixed in 4% (wt/vol) paraformaldehyde and stained with primary antibodies: rat anti–DE-Cad [1:100; DCAD2; Developmental Studies Hybridoma Bank (DSHB)], rat anti–N-Cad (1:500; DSHB), mouse anti-Rab11 (1:100; DSHB), mouse anti-Armadillo (1:200; N27A1; DSHB), rabbit anti–γ-Adaptin (1:1,000) (54), rat anti–α-Actn (1:50; DSHB), mouse anti–Hts-RC (1:5; Creative Diagnostics), and rabbit anti–P-Moesin (1:100) (55). We then used Cy2-, Cy3-, or Cy5-coupled secondary antibodies (Jackson Laboratories) diluted 1:250 and/or Phalloidin–Alexa-647 (Life Technology) diluted 1:500.

Imaging and Laser Ablation.

Live ovarioles were dissected and maintained in Schneider medium adjusted to pH 7.0 after supplementation with 15% (vol/vol) FCS and 200 µg/mL bovine insulin, as described in ref. 56. Fixed specimens and movies were acquired using LSM Leica SP5 and SP8 microscopes equipped with a 63× plan Apo-NA 1.4 lens or using a spinning-disk confocal microscopy equipped with a CSU-X1 disk, a Cool-SNAP-HQ2 camera, a Piezo stage, and a 100/3 plan Apo-NA 1.4 lens under the control of the MetaMorph Software. All images were processed using ImageJ.

Laser ablation was performed on live GCs using a Leica SP5 confocal microscope. Ablation was carried out on nurse cell membranes 16.5 ± 0.5 µm below the coverslip with a two-photon laser-type Mai-Tai HP from Spectra Physics set to 800 nm.

EM.

Ovaries were fixed in 2% (wt/vol) paraformaldehyde plus 2.5% (wt/vol) glutaraldehyde in 0.1 M cacodylate buffer for 2 h at room temperature and were processed for uranyl acetate contrast and embedded in Epon-Araldite mix (57, 58). Samples were observed directly either after ultrathin sectioning using a JEOL JEM-1400 electron microscope (Jeol) operated at 80 kV, equipped with a Gatan ORIUS SC1000 camera, or with a Gatan 3View microtome within an FEI Quanta 250 field emission gun scanning electron microscope as described in ref. 57.

SI Materials and Methods

Drosophila Stocks and Genetics.

AP-47^SHE11 clones were recovered from

• •hs-FLP/+ ; + ; FRT82B, AP-47^SHE11/FRT82B, Ubi-GFP(S65T)nls for immunofluorescence (Figs. 1 C and D and 3 B–E and Fig. S3A)
• •hs-FLP/+ ; E-Cad::GFP/+ ; FRT82B, AP-47^SHE11/FRT82B, P(Ubi-mRFP.nls)3R for live imaging of egg chambers
• •hs-FLP/+ ; E-Cad::GFP /+ ; FRT82B, AP-47^SHE11/FRT82B, P(ovoD1-18)3R for live imaging of egg chambers (Fig. 3A)
• •hs-FLP/+ ; + ; FRT82B, AP-47^SHE11/FRT82B, P(ovoD1-18)3R for electron microscopy (Fig. 5 A and B)

Rab11 dominant-negative–expressing GCs were recovered from

• •Otu-GAL4/+ ; Nos-GAL4/+ ; Nos-GAL4/ P(UASp-YFP.Rab11.S25N)06 (Fig. 3 F and G)

shg^R69 clones were recovered from

• •hs-FLP /+; FRT42B, shg^R69/FRT42B, P(Ubi-GFP.nls)2R1 P(Ubi-GFP.nls)2R2 (Fig. 4B)

N-Cad shRNA-expressing GCs were recovered from

• •Otu-GAL4/+ ; Nos-GAL4/+; Nos-GAL4/ P(TRiP.HMS02380)

shg^R69 N-Cad shRNA-expressing GCs were recovered from

• •hs-FLP/Otu-GAL4; FRT42B, shg^R69/FRT42B, P(Ubi-GFP.nls)2R1 P(Ubi-GFP.nls)2R2; Nos-GAL4/P(TRiP.HMS02380) (Fig. 4C)

E-Cad shRNA-expressing GCs were recovered from

• •Otu-GAL4/+ ; Nos-GAL4/+; Nos-GAL4/P(TRiP.HMS00693)
• •Otu-GAL4/+ ; Nos-GAL4/P(TRiP.GL00646); Nos-GAL4/+ (Figs. 4D and 6C and Figs. S2 and S4)

PH::GFP-expressing GCs were recovered from

• •Ubi-PH::GFP/CyO (Fig. 2 C–F, Fig. S1 A and B, and Movies S1–S3)

PH::RFP-expressing AP-1 mutant GCs were recovered from

• •hs-FLP/+ ; Ubi-PH::RFP /+ ; FRT82B, AP-47^SHE11/FRT82B, P(Ubi-mRFP.nls) (Fig. S1 F–H)

PH::GFP, E-Cad shRNA-expressing GCs were recovered from

• •Otu-GAL4/+ ; Nos-GAL4/Ubi-PH::GFP; Nos-GAL4/+ (Fig. 6B)

arm^XP33 clones were recovered from

• •arm^XP33, FRT19A/ Ovo^D1-18, hs-FLP, FRT19A ; + ; + (Fig. 4A)

Unless stated otherwise, fly stocks were obtained from the Bloomington Drosophila Stock Center. Other fly stocks used were knockin E-Cad::GFP allele (30), FRT42B, shg^R69 (39), FRT82B, AP-47^SHE11 (54), and Ubi-PH::GFP (59). The MTD-GAL4 line was used to drive GAL4 expression in GCs. Flippase expression to generate germinal clones was induced by two 37 °C heat shocks for 1 h at the second- and third-instar larval stages. Egg chamber staging was performed using King’s standard stages of oogenesis (60).

Immunofluorescence and Antibodies.

Ovaries from 1- to 3-d-old female flies were dissected in PBS to isolate individual ovarioles, which were fixed immediately for 20 min in 4% formaldehyde (with the exception of Actn and P-Moesin staining, which required 5-min fixation with 1% trichloroacetic acid (TCA) in PBS (61). Fixator (TCA or formaldehyde) was rinsed three times in PBS, and ovarioles were permeabilized for 2 h in PBS and 0.1% Triton X-100 (PBT). Ovarioles then were incubated in primary antibodies diluted in PBT for 2–3 h, washed three times for 15 min in PBT, incubated for 45 min in Cy2-, Cy3-, or Cy5-coupled secondary antibodies (Jackson Laboratories) and/or Phalloidin-Alexa-647 (Life Technology), washed three times for 10 min in PBT, incubated for 5 min in DAPI, rinsed three times in PBS and once in 50% glycerol, and finally mounted in mounting medium (90% glycerol, 2.5% n-propyl gallate) (Fluka) in 1× PBS) with two bands of stretched parafilm used as 70- to 90-µm spacers between slide and coverslips. (The fixed samples shown in Fig. S3B were not mounted between glass and coverslip ensure that the peripheral location of E-Cad/β-Cat clusters as compared with α-Actn and Phospho-Moesin was not caused by the egg chamber being squeezed; instead, egg chambers were simply placed on the coverslip in a drop of mounting medium.) Every step was performed at room temperature.

Primary antibodies were rat anti–DE-Cad (1:100; DCAD2; DSHB), rat anti–N-Cad (1:500; DSHB), mouse anti-Rab11 (1:100; BD Biosciences), mouse anti-Armadillo (1:200; N27A1; DSHB), rabbit anti–γ-Adaptin (1:1,000) (54), rat anti–α-Actn (1:50; ab50599; AbCam), mouse anti–Hts-RC (1:5; Creative Diagnostics), and rabbit anti–P-Moesin (1:100) (55).

Egg Chamber Growth Measure.

Egg chamber dimensions were determined following King (60) for stages 2–7. Stages 8–10a were identified using morphological criteria described by King (60) and were measured without being compressed between a glass slide and coverslip. Egg chamber volume was calculated as follows: V = 4/3 × π × length × width². Growth rates were calculated using the stages durations described by King (60).

Culture of Egg Chambers for Live Imaging.

Live ovarioles were dissected and maintained in Schneider medium (Gibco) adjusted to pH 7.0 after supplementation with 15% FCS and 200 µg/mL bovine insulin (Sigma Aldrich), as described in ref. 56. Ovarioles were immersed in a minimal volume of medium covered by mineral oil (Sigma Aldrich) to prevent evaporation while allowing proper oxygenation. Live imaging (Z stacking of 1.5 µm with one stack of 8–10 images acquired every 5–7 min to avoid phototoxicity) was performed using a spinning-disk confocal microscope.

Laser Ablation.

Laser ablation was performed on live egg chambers using a Leica SP5 confocal microscope. Ablation was carried out on nurse cell membranes, 16.5 ± 0.5 µm below the coverslip, with a two-photon laser-type Mai-Tai HP from Spectra Physics set to 800 nm using the fluorescence recovery after photobleaching (FRAP) module of the Leica LAS software. Settings used were as follows: Trans 100%; gain 80%; offset 60%; numerical zoom 4×; lens, 63× 1.4 NA; confocal series; scanning speed 400 Hz; resolution 512 × 512 pixels; seven iterations in a 5 × 3 µm area. Images were acquired every 1.3 s after ablation (Fig. 2D) or every 25 s to acquire confocal stacks of the whole targeted membrane after exiting the FRAP module (Fig. 2C).

Fluorescence Imaging.

Fixed and live specimens were imaged using Leica SP5 and Leica SP8 microscopes laser-scanning microscopes equipped with 63× plan Apo-NA 1.4 and 100× plan Apo-NA 1.4 lenses or using spinning-disk confocal microscopy on a Nikon Ti-E microscope equipped with CSU-X1 disk, a Cool-SNAP-HQ2 camera (Roper Scientific), a Piezo stage, a 100/3 plan Apo-NA 1.4 lens, and 491 (50 mW) and 561 (50 mW) lasers, under the control of the MetaMorph Software.

Image Processing.

All images were processed using ImageJ. We applied Gaussian blurs with a radius of 0.8 pixels to increase the signal-to-noise ratio for better visualization. We applied intensity projections of several focal planes as shown in figure legends. To project signals only from the PM and not from the surrounding cytosol (for example from E-Cad⁺ vesicles), we first sliced confocal stacks orthogonally with an output spacing of one pixel. Next, we deleted the signal outside the plane of the PM on the resulting resliced stack. Third, we sliced the cleared resliced stack orthogonally to return to the initial view, and, fourth, we applied the projection on the resulting resliced stack. We used minimum instead of maximum intensity projections in Fig. 3B on the Rab11 channel to accentuate the negative staining of nuclei and in Fig. 5C to visualize electron-dense material on TEM pictures. For movies of live imaging of egg chambers (Fig. 3A), we used the Subtract Background filter with a rolling ball radius of 50 pixels after applying the Gaussian blur. We applied a gamma filter on the green channel in Fig. 3B to accentuate the actin staining of remaining nurse cell membranes. The TEM pictures projections in Fig. 5D and Movie S5 were made after PM segmentation using the pixel classification workflow of the Ilastik software. Images were mounted on Adobe Photoshop Elements 10.0. For analysis of vertices and cut extremities displacement after photoablation (Fig. 2D and Fig. S1 B–G), the signal-to-noise ratio at the PM was enhanced using the SteerableJ plugin of ImageJ, and movies then were thresholded and skeletonized using ImageJ. Using a custom-made macro, vertices and free extremities were tracked automatically on skeletonized movies by tracking three and one neighboring pixels, respectively.

EM.

For chemical fixation Drosophila ovaries were dissected from adult females in PBS and were fixed immediately in a mix of 1% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M cacodylate buffer for 2 h at room temperature. To enhance the contrast, samples were incubated in 1% OsO₄ and 1.5% K₄[Fe(CN)₆] followed by a 1-h incubation with freshly made 1% tannic acid. Further incubation in 2% OsO₄ was followed by a 1-h incubation in 1% uranyl acetate at ambient temperature for 40 min (57, 58). After the dehydration cycles, samples were embedded in Epon-Araldite mix (EMS hard formula). Samples were observed directly either after ultrathin sectioning using a JEM-1400 electron microscope (JEOL) operated at 80 kV and equipped with a Gatan Orius SC 1000 camera or with a Gatan 3View microtome within an FEI Quanta 250 field emission gun scanning electron microscope as in ref. 57.