Exocytosis by vesicle crumpling maintains apical membrane homeostasis during exocrine secretion.

Exocrine secretion commonly employs micron-scale vesicles that fuse to a limited apical surface, presenting an extreme challenge for maintaining membrane homeostasis. Using Drosophila melanogaster larval salivary glands, we show that the membranes of fused vesicles undergo actomyosin-mediated folding and retention, which prevents them from incorporating into the apical surface. In addition, the diffusion of proteins and lipids between the fused vesicle and the apical surface is limited. Actomyosin contraction and membrane crumpling are essential for recruiting clathrin-mediated endocytosis to clear the retained vesicular membrane. Finally, we also observe membrane crumpling in secretory vesicles of the mouse exocrine pancreas. We conclude that membrane sequestration by crumpling followed by targeted endocytosis of the vesicular membrane, represents a general mechanism of exocytosis that maintains membrane homeostasis in exocrine tissues that employ large secretory vesicles.

Introduction

Diverse exocrine tissues secrete their cargos via large vesicles that reach up to 10 μm in diameter. Prominent settings include lung alveolar cells secreting surfactant proteins, cells in the endothelium secreting pro-hemostatic proteins, lacrimal epithelial acinar cells secreting tear proteins, and exocrine cells along the digestive tract, such as salivary glands, the intestine, and the pancreas, which secrete enzymes and additional molecules (Nightingale et al., 2012). Compared to a synaptic vesicle, each large secretory vesicle (SV) would add up to ten-thousand times more membrane to the cell surface, underscoring the formidable challenge these tissues face in maintaining membrane homeostasis during exocrine secretion.

Plasma membrane homeostasis in secretory cells has been extensively studied in systems utilizing smaller vesicles such as chromaffin granules (∼300 nm in diameter) (Steyer et al., 1997) or synaptic vesicles (∼40 nm in diameter) where two major mechanisms have been described. In the first, the vesicular membrane collapses and integrates into the cell surface, while the excess membrane is subsequently retrieved by endocytosis at the exocytic site or in its vicinity (full collapse model, Figure S1A). In the second, the vesicle fuses transiently, such that the content is partially released and the vesicle detaches, not adding membrane to the cell surface (kiss-and-run model, Figure S1B) (Harata et al., 2006; Wu et al., 2014). While similar mechanisms have been proposed for large SVs such as cortical granules in sea urchin and Xenopus oocytes (Yu and Bement, 2007a; Zimmerberg et al., 1999) as well as in exocrine tissues (Porat-Shliom et al., 2013; Shitara and Weigert, 2015; Stevenson et al., 2017), the fate of the vesicular membrane in micron-sized vesicles remains incompletely resolved.

Here, we used Drosophila melanogaster larval salivary glands (LSGs) as a model to elucidate the mechanisms that allow cells to maintain apical membrane homeostasis during extended periods of secretion. The LSG is a monolayered tubular epithelium comprised of large, highly polarized columnar secretory cells (Figure 1A). The primary role of these cells is to secrete adhesive mucinous glycoproteins (SGS or glue proteins), which are subsequently used by the fly pupa to attach to a solid surface during metamorphosis (Figure 1B). Prior to onset of secretion, glue proteins are packaged into large SVs that range between 2 to 10 μm in diameter (Figures 1C and 1D) (Burgess et al., 2012, 2011; Rousso et al., 2016; Tran and Ten Hagen, 2017). Shortly prior to pupation, an ecdysone hormonal signal initiates secretion, which proceeds over a period of approximately 2 h (Biyasheva et al., 2001). Fusion of each vesicle to the apical surface is followed by a highly orchestrated process of actomyosin assembly and disassembly on the vesicle surface, which facilitates the expulsion of vesicle content to the lumen within 3 to 4 min (Figures 1B, 1D, and 1E) (Rousso et al., 2016; Segal et al., 2018; Tran et al., 2015). The formation of such an actomyosin meshwork around large vesicles is a prominent mechanistic feature in several vertebrate exocrine tissues, establishing it as a conserved attribute of exocrine exocytosis (Jerdeva et al., 2005; Masedunskas et al., 2011; Miklavc et al., 2009; Nemoto et al., 2004; Nightingale et al., 2012, 2011; Porat-Shliom et al., 2013; Valentijn et al., 2000; Yu and Bement, 2007a, 2007b).

Figure 1

The membrane of large secretory vesicles does not integrate into the apical surface during exocytosis

(A) Schematic of D. melanogaster third-instar LSGs, composed of a single-layered polarized epithelium. The LSG is filled with glue-protein-containing large SVs (5.78 ± 3.82 μm in diameter, n = 60 vesicles, 3 LSGs; cyan), which fuse to the apical surface and release their content into the lumen (L).

(B) Schematic of a single fused SV (SV membrane, green; apical membrane, gray). An F-actin coat (magenta) assembles on the vesicles after fusion. SVs can be imaged in the orthogonal or parallel planes with respect to the apical membrane (dashed lines).

(C and D) Confocal slice from a representative LSG cultured ex vivo, expressing salivary gland secretary protein-3fused to GFP (Sgs3-GFP; white) and the F-actin probe, LifeAct-Ruby (magenta); apical membrane (dashed line).

(E) Confocal images of a representative LSG expressing the membrane marker mCD8-GFP (green); F-actin (magenta). Insets in (D and E) show a magnified view of F-actin-positive fused SVs.

(F) Time series of a representative secreting vesicle (orthogonal view) from cultured LSGs, showing the content release that is marked by Sgs3-GFP (gray) and F-actin (magenta).

(G) Time series of a representative vesicle during secretion (orthogonal view) showing the vesicle membrane dynamics as marked by mCD8-GFP (i, green and iii, gray) and F-actin (i, magenta and ii, gray). Vesicle fusion is considered t = 0.

(H) Normalized intensity profiles of mCD8-GFP (green) and F-actin (magenta) of the vesicle shown in (G). Total membrane intensity remained constant throughout secretion, implying that the vesicle does not collapse into the apical surface. Vesicle diameter (gray) is normalized to the pre-fusion diameter.

(I) Total membrane intensities of six additional vesicles during the period of F-actin localization (in different colors). Membrane intensities were normalized to their respective peak intensities. (J) Time series of a representative SV membrane (parallel view) during secretion. At t = 45 s post-fusion, the vesicle membrane starts folding. Scale bars: 50 μm (C–E) and 5 μm (F, G, and J). Also see Figure S1.

We found that after fusion of glue-protein-containing SVs to the surface, the vesicular membrane neither collapses into the surface nor detaches back into the cytoplasm as suggested by previous models of exocytosis. Instead, the vesicular membrane becomes progressively folded by contraction of the actomyosin meshwork, which squeezes the content out of the vesicle while retaining and sequestering the membrane, giving it a crumpled appearance. Moreover, we show that transmembrane and lipid-anchored proteins do not freely diffuse between the cell surface and the fused SV membrane. Finally, actomyosin-mediated membrane crumpling is essential for triggering clathrin-mediated endocytosis (CME) directly from the highly folded vesicular membrane, which in turn retrieves it over a period that is up to six-times longer than the duration of content release. We also show that SVs of the mouse exocrine pancreas undergo membrane crumpling. Thus, we conclude that plasma membrane homeostasis in exocrine cells is maintained by mechanochemical sequestration of the vesicular membrane, enabling efficient content release and removal of the excess membrane by CME over a prolonged period of time.

Results

The membranes of large secretory vesicles do not incorporate into the apical plasma membrane during exocytosis

To follow the dynamic morphological changes and fate of the SVs membrane during exocytosis, we monitored the transmembrane marker, mCD8-GFP, and the F-actin probe, LifeAct-Ruby (Figure 1E). Once vesicles had fused to the apical surface their spherical shape was maintained for a significant, yet variable, amount of time (59 ± 42 s). Within this time window, F-actin assembled on the vesicle membrane, first appearing within 45.2 ± 20 s after fusion. This was followed by a phase of content release, during which the vesicle size rapidly decreased within a relatively constant time frame (36 ± 8 s), irrespective of the variability in the initial size of the vesicles (n = 30 events; 3 LSGs, Figures 1F and 1G and Video S1). As vesicle size decreased, the mean fluorescence intensity of mCD8-GFP on the vesicular membrane increased but the total fluorescence remained constant (Figures 1G–1I). Strikingly, monitoring the vesicle membrane from a view that is parallel to the apical surface, we observed that it underwent anisotropic folding and crumpling (Figures 1B, 1J, and S1C; Video S2). Consequently, once the content was released, the vesicular membrane appeared as a persistent, bright diffraction-limited spot (Figures 1F and 1G, arrowhead). Taken together, these observations imply that the total amount of vesicular membrane does not appreciably change during content release and that it does not integrate into the apical surface after fusion.

Membrane crumpling couples content release and membrane sequestration

To resolve the changes in vesicular membrane ultrastructure during exocytosis, we used correlative light and electron microscopy (CLEM), combining on-section fluorescence microscopy (FM) and electron tomography (ET) (Figure 2A; Kukulski et al., 2011). The fluorescence of LifeAct-Ruby surrounding fused vesicles correlated with the F-actin surrounding the vesicles in the tomographic reconstructions and it was subsequently used to assign vesicles to progressive phases of secretion (Figures 2B–2I). F-actin surrounding the vesicles was visible in the ET data as an apparent exclusion zone, devoid of ribosomes, as previously described (Kukulski et al., 2012; Maupin and Pollard, 1983). We estimated the diameter of the F-actin coat based on these exclusion zones and found that it grew in thickness during secretion (from 236 ± 56 nm in early vesicles to 441 ± 136 nm in late vesicles; n = 12; Figures 2B–2I), consistent with the increase in total F-actin fluorescence intensity observed by live imaging (Figure 1). In parallel, the vesicular membrane appeared progressively folded (Figure 2B–2I; Videos S3 and S4) and occasionally projected thin membrane protrusions approximately 20 nm in diameter (Figures 2G and S2A–S2D). Late-stage vesicles appeared as crumpled, compact structures that were still connected to the apical surface through a narrow pore (∼100 nm internal pore diameter; Figures 2I and S2D). These observations suggest that the vesicular membrane is sequestered from the apical surface and imply that the actomyosin meshwork is responsible for folding the vesicular membrane to extrude the content but without integrating the membrane into the apical surface.

Figure 2

CLEM reveals the stages of SV membrane folding

(A) On-section FM, TEM, and overlay. F-actin (magenta) highlights the apical membrane (dashed line) and fused SVs (arrowheads); L, lumen. The white box indicates the region of the ET acquisition of the vesicle shown in (B).

(B–I) 2D TEM images of SVs at progressive stages of secretion. Orthogonal (B and I) and parallel (C–H) views of the vesicles. Insets show the correlated FM. F-actin exclusion zone is pseudo-colored magenta in (B); early-fused vesicles (B and C). While the average F-actin coat thickness increased during secretion, it was lower where protusions emanate from the vesicle membrane. The average F-actin coat thickness is 190 ± 50 nm in (B), 168 ± 33 nm in (C), 263 ± 48 nm in (D), 280 ± 46 nm in (E), 345 ± 98 nm in (F), 376 ± 166 nm in (G), and 456 ± 146 nm in (H).

(D and E) SVs started showing mild folding.

(F and G) In mid secretion, SVs exhibited highly convoluted membranes.

(H and I) Late SVs exhibited crumpled and compacted membranes. Insets, lower left in (G–I) are slices from the ETs acquired in the boxed region (white dots) showing thin membrane protrusions (G, arrowhead) or tubes (H) and a narrow fusion pore in a late-crumpled SV (the dotted box with the arrowhead). Scale bars: 5 μm (A), 1 μm (B–I), and 100 nm (insets). Also see Figure S2.

If the vesicular membrane does not incorporate into the apical membrane while vesicle volume decreases, then the surface area of the SV membrane should remain constant. To obtain quantitative estimates of the total surface area and volume of large SVs during different phases of secretion, we used a CLEM workflow combining block-face confocal microscopy and focused ion beam milling combined with scanning electron microscopy (FIB-SEM; Figures 3 and S3). F-actin fluorescence within resin-embedded glands was used to identify regions rich in secreting vesicles, and the 3D ultrastructural data was acquired with an isotropic voxel (Figures 3A and 3B; Videos S5 and S6). To estimate the changes in surface area and volume, we segmented and compared unfused, early fused (i.e., after fusion but before membrane crumpling), and crumpled vesicles. While the volume of the crumpled vesicles was significantly smaller than the unfused and early-fused vesicles, the surface area was not significantly different (Figures S3E and S3F; Table S1). To accommodate for the variability in vesicle size, we calculated the surface-area-to-volume ratio (SA/V) for each vesicle. This value should reflect the extent of membrane retention irrespective of vesicle size. We found that the crumpled vesicles had a significantly higher SA/V compared to unfused and early-fused vesicles (Figure 3C). Hence, these measurements provide an independent, quantitative validation of membrane sequestration, demonstrating that the vesicular membrane does not appreciably integrate into the apical surface during exocytosis.

Figure 3

3D-ultrastructure of secretory vesicles

(A and B) FIB-SEM serial surface imaging of complete vesicles. Representative examples of segmented early-fused and crumpled vesicles, respectively. Insets: slices through the FIB-SEM stack; L, lumen. Fusion pore (red dashed box in inset).

(C) Surface-area-to-volume ratio of SVs at different stages. Unfused (n = 8), early fused (n = 5), and crumpled (n = 9). Scale bar: 1 μm (inset). The surface-area-to-volume ratio of crumpled vesicles was much higher than the expected ratio for vesicles simulated to be undergoing full collapse (inset on bottom right, see STAR Methods for details). p value < 0.01; using the non-parametric Kolmogorov Smirnov test. Also see Figure S3 and Table S1.

Limited diffusion between the vesicular and apical cell membranes after fusion

Physical sequestration of the vesicle membrane would maintain the size and biophysical properties of the apical surface during secretion but its composition could still be altered by diffusion of molecules into or out of the vesicular membrane. Therefore, we wondered whether diffusion between the apical and vesicular membranes might be limited. To quantify the diffusion between the apical and vesicular membranes, we measured fluorescence recovery after photobleaching (FRAP) of mCD8-GFP. Photobleaching of mCD8-GFP on half of an unfused vesicle resulted in rapid redistribution of the marker, as expected from freely diffusing molecules in a closed system where there is no addition of molecules (n = 16 events, Figures 4A and 4B). In contrast, after an initial increase in mCD8-GFP fluorescence immediately after fusion, no recovery of its fluorescence was observed after photobleaching on fused vesicles. These vesicles secreted their content normally, as determined by following F-actin fluorescence (n = 24 events; Figures 4C and 4D; Table S2). Moreover, the fluorescently labeled V-SNARE, Sec22-GFP, which localized exclusively to the SV membrane before fusion, was maintained on the vesicular membranes throughout secretion, without observably diffusing into the apical surface (Figures 4E and 4F). Importantly, photobleaching of mCD8-GFP on the apical surface resulted in only partial recovery to a maximum of ∼50%, indicating that only a fraction of the molecules is mobile and can diffuse freely (n = 10 events, Figures 4G and 4H). Similar distribution and diffusion characteristics were also observed for the lipidated membrane probe, myristoylated GFP (myr-GFP; Figure S4 and Table S2). These results demonstrate that the inherent properties of the apical surface contribute to limiting the diffusion of both proteins and lipids between the vesicular and apical membranes (Trimble and Grinstein, 2015).

Figure 4

Restricted diffusion between the vesicle and apical membrane

(A–J) FRAP of mCD8-GFP on SVs and apical surface. Photobleaching time is indicated by a red arrowhead, the red dashed line shows the region of interest (ROI) used for photobleaching. Mean intensity (black dashed line), standard deviation (gray area), and fitted curve (red curve). (A and B) Representative time-lapse showing mCD8-GFP FRAP on half of an unfused vesicle (A) and the corresponding mean fluorescence recovery profile (B). Mobile fraction (FM) 0.203 ± 0.002 (avg. ±SEM, n = 5 events), consistent with the redistribution of freely diffusing molecules within a closed system.

(C) Representative time-lapse showing mCD8-GFP FRAP of an entire fused vesicle undergoing exocytosis, showing no recovery of mCD8-GFP (top) and LifeAct-Ruby (bottom), showing that the vesicle continued to secret normally.

(D) The corresponding mean fluorescence recovery profile (n = 5 events) showing limited mCD8-GFP recovery.

(E and F) Sec22-GFP expressed from the endogenous promoter (top, cyan; bottom, gray), localized to the vesicle membrane before fusion, and did not appear to diffuse into the apical membrane during exocytosis as quantified in (F, avg. ± SD, n = 7 vesicles). F-actin (LifeAct-Ruby; top, magenta).

(G and H) Representative time-lapse showing mCD8-GFP FRAP on the apical surface (G) and the corresponding mean fluorescence recovery profile (H). FM 0.54 ± 0.02 (avg. ±SEM, n = 5 events), indicating that only a fraction of the molecules is mobile and can diffuse freely.

(I and J) FRAP analysis on vesicles that cannot crumple. (I) time-lapse showing mCD8-GFP FRAP on stalled vesicles from LSGs treated with the ROCK inhibitor (Y27632), and (J) corresponding FRAP curve showing negligible recovery (avg. ±SEM, n = 5 events). Scale bars: 5 μm. Also see Figure S4 and Table S2.

Since folding of the membrane could potentially influence diffusion rates (Huttner and Zimmerberg, 2001), we asked whether crumpling has a role in limiting diffusion into the vesicle. To this end, LSGs were treated with the Rho-associated protein kinase (ROCK) inhibitor Y-27632, which blocks actomyosin contractility, resulting in fused but spherical vesicles that do not squeeze out their content (Segal et al., 2018). FRAP under these conditions showed no recovery of mCD8-GFP fluorescence on the vesicle membrane (n = 12 events) (Figures 4I and 4J; Table S2), demonstrating that the limited diffusion is independent of actomyosin-driven membrane folding and is maintained for a long time.

These results, together with the observation that the total fluorescence intensity of mCD8-GFP on the vesicular membrane remained largely constant after vesicle fusion (Figures 1G–1I), are consistent with a scenario in which the vesicular membrane remains chemically sequestered from the apical surface as it extrudes its content, thus maintaining apical membrane homeostasis.

Extended sequestration of the crumpled vesicle membrane allows retrieval by clathrin-mediated endocytosis

Next, we followed the fate of the empty and crumpled vesicular membrane. Since it is well documented that various secretory cells efficiently employ endocytosis to retrieve SV membranes and their components for reuse after cargo release (Gundelfinger et al., 2003; Smith et al., 2000; Stevenson et al., 2017; Wu et al., 2014; Xie et al., 2017), we sought to determine whether this is also the case for LSGs. To monitor endocytic uptake specifically from the apical membrane, we stained this domain by injecting the membrane dye CellMask into the lumen of secreting LSGs. CellMask accumulated over time in multiple intracellular locations, visible as bright fluorescent puncta (Figure 5A, left panel). To monitor uptake from the basolateral domain, the dye was added to the surrounding medium. This treatment resulted in incorporation of CellMask specifically into the basolateral surfaces but no subsequent accumulation in intracellular puncta was observed (Figure 5A, right panel). These results demonstrate that cell membrane uptake, as marked by CellMask, occurs primarily from the apical surface of secreting LSGs.

Figure 5

Clathrin-mediated endocytosis is upregulated in secreting glands

(A and B) Uptake of CellMask. (A, left) CellMask (gray) injected into the lumen of a representative secreting LSG, stained the apical surface of the cells, and accumulated as puncta inside the cells. The white dashed box highlights a region inside the cytosol that is magnified in the inset to show the intracellular CellMask puncta. (A, Right) A representative LSG incubated with CellMask added to the bath solution. The basolateral cell membranes were stained with CellMask, but no intracellular internalization of the dye was observed.

(B) CellMask injected into the lumen of secreting LSGs expressing the temperature-sensitive allele of shi [fkh>UAS-shits1]. (Left) LSGs maintained and imaged under the restrictive temperature of 29°C, showed minimal intracellular CellMask puncta, while (right) LSGs grown under permissive temperature of 18°C showed marked CellMask accumulation. This suggested that the intracellular CellMask puncta originated due to endocytosis from the apical surface. L, lumen; C, Cytosol.

(C) myr-tdTomato expressed in LSGs [fkh>UAS-myr-tdTomato] (gray) was used to quantify endocytosis over time. myr-tdTomato marked all the cell surfaces and faintly marked the SV membranes. (Left) before the onset of secretion, LSGs had a narrow lumen (L). (Right) The secreting LSGs had a wide lumen and showed an increased number of intracellular myr-tdTomato puncta compared to the pre-secretion phase, suggesting that upregulation of internalization is coupled to secretion. The apical surface of the LSGs in (A–C) is highlighted by yellow dashed lines.

(D) Representative Z-projections from live secreting LSGs showing an increase in the number and size of myr-tdTomato puncta (arrowheads) over time.

(E) Representative images of AP2 knockdown [fkh>UAS-AP2σ RNAi] showing a decrease in myr-tdTomato puncta internalization, quantified in (G; see STAR methods for details).

(F) Representative images from Shi knockdown LSGs [fkh>UAS-shiRNAi] showing complete absence of endocytic puncta. Scale bar: 10 μm (A–C), and 25 μm (D–F).

To determine whether the puncta observed following apical surface staining were indeed a consequence of endocytosis, we inhibited endocytosis using a temperature-sensitive allele of shibire (shi); shi is the Drosophila homolog of Dynamin, a protein essential for the scission of endocytic vesicles (De Camilli et al., 1995; Koenig and Ikeda, 1989). Injecting CellMask into the lumen of the temperature-sensitive shi-mutant LSGs (fkh>UAS-shi) at the restrictive temperature (29°C), resulted in a significant reduction in the number of stained intracellular puncta compared to LSGs at the permissive temperature (18°C) (Figure 5B). These results suggest that the observed intracellular puncta represent membranes that are internalized via endocytosis from the apical surface.

In order to quantify endocytic uptake in secreting LSGs over time, we utilized the myr-tdTomato membrane marker. The number of intracellular myr-tdTomato puncta in secreting LSGs was markedly higher than in non-secreting LSGs (Figure 5C). Knocking down the Drosophila dynamin homolog Shi, which is essential for CME, or the Drosophila homolog of adaptor protein 2 (AP2) (Mettlen et al., 2018), led to a complete loss or significant reduction of internalized myr-tdTomato puncta, respectively (Figures 5D–5G). Taken together, these observations suggest that endocytosis from the apical surface of secreting LSGs is mediated by the clathrin pathway.

To determine whether the enhanced rate of endocytosis observed in secreting LSGs is attributable to fused SVs, we followed the dynamic localization of fluorescently labeled Shi (Shi-YFP). Shi-YFP first appeared on late-stage vesicles and remained visible for 13 ± 7.8 min (n=10 events), during which its intensity increased several fold and subsequently declined (Figures 6A and S5A; Video S7). A similar persistence on late-stage vesicles was observed for other components of the CME pathway, such as Liquid facets (Lqf)—the Drosophila homolog of Epsin 1 (Figure S5B; Ford et al., 2002; Wang and Struhl, 2004).

Figure 6

Prolonged sequestration allows retrieval of the vesicular membrane by CME

(A) Time series of a representative vesicle showing the dynamics of Shi-YFP (green in merge) with respect to LifeAct-Ruby (F-actin; magenta in merge). Shi-YFP was recruited to late-crumpled SVs (arrowhead).

(B and C) CLEM using Shi-YFP showed that the number of CCPs on late-crumpled SVs was higher than on early SVs.

(B) On-section FM showing Shi-YFP clusters on the apical surface (white box).

(C) A slice through the corresponding ET, showing that the Shi-YFP-labeled vesicle is a late-crumpled vesicle exhibiting several clathrin-coated pits (CCPs, arrowheads). The red dashed box highlights the fusion pore by which such late-stage vesicles remain attached to the apical surface.

(D) Quantification of the number of CCPs on early (LifeAct positive, Shi-YFP negative) and late (Shi-YFP positive) SVs. p values were compared using a non-parametric Kolmogorov Smirnov test.

(E) Z-projections from LSGs showing myr-tdTomato internalization in zip (Myo II heavy chain homolog) knockdown [fkh>UAS-zip RNAi]. Very few myr-tdTomato intracellular puncta were observed in knockdown LSGs.

(F) Quantification of the myr-tdTomato puncta (endosomal volume) over time. zip knockdown showed no significant increase in endosome volume compared to driver-only controls.

(G) ROCK inhibitor (Y27632) treatment of LSGs led to stalled vesicles (arrowheads), which showed no recruitment of Shi-YFP. Scale bars: 5 μm (A, B, and G), 0.5 μm (C), and 25 μm (E). Also see Figure S5.

Persistence of the CME machinery (Figures 6A and S5A) and the V-SNARE Sec22 (Figure 4G) on late-crumpled vesicles suggested that the vesicle membrane remains sequestered for a prolonged time after content release, during which it might be retrieved by CME. To test this directly we used CLEM, with Shi-YFP as a marker for late-stage vesicles. Indeed, Shi-positive vesicles were highly crumpled and decorated with multiple clathrin-coated pits (CCPs) ∼100 nm in diameter (Figures 6B–6D and S5C–S5F; Video S8). CCPs were identified based on the presence of the typical clathrin coat (Figures S5E and S5F; Avinoam et al., 2015). The average number of CCPs observed in a 300-nm thick section of late-stage crumpled SVs was significantly higher than on fused vesicles at the early stages of content release, defined by the absence of Shi-YFP on LifeAct-Ruby-positive vesicles, (mean = 6.4 ± 2.6, n = 12 and mean = 0.4 ± 0.6, n = 21, respectively; Figure 6D), indicating that CME is upregulated specifically on crumpled vesicles.

To test whether membrane crumpling is essential for recruitment of CME to the sequestered vesicular membrane, we inhibited actomyosin contractility using RNAi-mediated knockdown of zipper (zip), the Drosophila homolog of myosin II heavy chain, or of RhoGAP71E, which mediates actomyosin disassembly (Rousso et al., 2016; Segal et al., 2018). Under these conditions, the SVs fused to the apical membrane but retained their initial size and spherical shape. Strikingly, the endocytic accumulation of myr-tdTomato puncta was markedly reduced (Figures 6E, 6F, S5G, and S5H). In addition, LSGs treated with the ROCK inhibitor, Y-27632, which inhibits myosin II assembly, failed to recruit Shi-YFP to SVs (Figure 6G). Knock down of RhoGAP71E, where myosin II is retained on the stalled spherical vesicles also led to loss of Shi-YFP, indicating that the reduction in endocytic uptake is not due to loss of myosin II recruitment (Figure S5I; Segal et al., 2018). Collectively, these observations indicate that actomyosin-driven membrane crumpling is essential for the recruitment of CME, which retrieves the folded and sequestered vesicular membrane into the cell after the content has been released.

Exocytosis by membrane crumpling in the mouse exocrine pancreas

Finally, we explored whether SV membrane crumpling, as observed in D. melanogaster LSGs, represents a general mechanism employed by secretory glands, which utilize large vesicles for exocytosis. The mouse exocrine pancreas secretes enzymes that are important for digestion via large actin-coated SVs that range between 0.5 to 3 μm in diameter (Ebrahim et al., 2019; Geron et al., 2013). We isolated pancreatic acini from a mouse strain ubiquitously expressing a membrane-targeted tdTomato (mT) and transduced them with an adenovirus to express LifeAct-GFP (Geron et al., 2013). To visualize membrane dynamics during exocytosis, we performed live imaging of the acini after treatment with cholecystokinin (CCK) or Carbachol, which stimulate secretion (Figures S6A–S6D). We observed two modes of exocytosis: compound exocytosis, where vesicles fuse to an already secreting vesicle; as well as conventional exocytosis, where individual vesicles fuse to the apical surface (Geron et al., 2013). The duration of conventional exocytosis was 1–4 min. The vesicular membranes appeared to crumple irrespective of the mode of exocytosis (Figures 7A, 7B, and S6A–S6C; Video S9).

Figure 7

Secretory vesicles of the mouse exocrine pancreas undergo membrane crumpling

(A) FM image of pancreatic acini expressing the mT probe (magenta) and LifeAct-GFP (F-actin; green). LifeAct-GFP marked the vesicles undergoing secretion after induction with carbachol (arrowheads).

(B) Vesicular membrane and F-actin dynamics during secretion. Time-lapse of a representative single-vesicle fusion event. Immediately after fusion the vesicle was still spherical. Subsequently, both mT (top, gray; bottom, magenta) and LifeAct-GFP (middle, gray; bottom, green) showed a crumpled appearance during secretion. Note the persistence of the SV long after content release (arrowhead).

(C) CLEM of SVs in mouse pancreatic acini. (Left) On-section FM overlay image of an exocrine pancreas sample prepared for EM. LifeAct-GFP (green) marked the cortical actin at the apical membrane and secreting SVs (arrowheads). mT (magenta) marked the cell outlines as well as all the vesicular membranes. (Middle) A TEM image of the same region. (Right) Overlay of the FM and TEM images. White box marks the region where the ET of the vesicle shown in (D) was acquired. Insets show the entire captured field of view of the acinar sample; L, lumen.

(D–F) Slices from the ETs showing progressive vesicle membrane folding during secretion. Insets show the correlated FM images. The boxes mark the regions where the ETs were acquired. Inset in (F) highlights the outline of the folded vesicular membrane. CCPs (orange arrowheads) were only observed on folded vesicles. Scale bars: 5 μm (A–C), 0.5 μm (D–F).

(G) Model for exocytosis by membrane crumpling. Also see Figure S6.

To definitively resolve SV membrane architecture during secretion in the acinar pancreas we employed CLEM, using the LifeAct marker that pinpoints fused vesicles (Figures 7C–7F). Indeed, F-actin-coated SVs showed a folded membrane ultrastructure (Figures 7E, 7F, and S6B). In addition, CCPs were observed on the folded vesicles, suggesting that CME may occur directly from the vesicular membrane, as in D. melanogaster LSGs (Figures 7F and S6D, arrowheads). We also observed persistence of the SV membrane after crumpling of the vesicular membrane (Figure 7B, arrowhead), consistent with the observations made by Thorn and colleagues, who had previously proposed the notion of vesicle-membrane retention and restricted diffusion in this tissue (Thorn et al., 2004; Thorn and Parker, 2005). Taken together, we conclude that exocytosis by membrane crumpling and restricted diffusion is conserved and propose that they are essential to maintaining cell membrane homeostasis.

Discussion

Preserving the integrity and homeostatic state of the apical surface of epithelial cells, during prolonged periods of secretion via large SVs that continuously add membrane, presents a formidable challenge. The machineries that operate in secretory cells employing more conventional vesicles (<300 nm) may not suffice to balance the addition of membrane by large vesicles, which range up to 10 μm in diameter. Using the well-characterized D. melanogaster LSG model, together with a dynamic, morphological, and ultrastructural characterization, we identified an exocytic mechanism that mechanochemically insulates the fused vesicular membrane from the plasma membrane, thus providing a longer time window for membrane removal by conventional endocytosis. Apparent crumpling of the vesicle membrane and endocytosis during secretion is also observed in the exocrine pancreas, where vesicle exocytosis is actomyosin dependent. Moreover, multiple examples of crumpled SVs can be found in published EM data from exocrine tissues such as the mammalian alveolar type II lung cells (Kliewer et al., 1985; Tsilibary and Williams, 1983), lacrimal grands (Meneray et al., 1994), salivary glands (Segawa et al., 1998), and pancreatic acini (Reinecke, 1967; Valentijn et al., 1999). In type II lung cells, a thick layer of “actin-like material” is shown to surround the crumpled vesicular membrane (Tsilibary and Williams, 1983). Thus, we conclude that membrane crumpling is a previously overlooked mode of exocytosis that results in mechanochemical sequestration of the vesicular membrane, which maintains homeostasis during exocrine secretion.

This process is executed in several consecutive stages (Figure 7G). First, the vesicle fuses to the apical surface, forming a pore that initially expands but then quickly stabilizes at a maximal diameter. Meanwhile, an actomyosin coat assembles on the vesicle, leading to dynamic inward folding of the vesicle membrane. The vesicle membrane folds progressively until the entire membrane is packed into a small dense structure, driving complete expulsion of the content. This process, which we refer to as “crumpling,” mechanically sequesters the vesicular membrane and, coupled to the limited diffusion we observed, should maintain the size, composition, and biophysical properties of the cell surface. Finally, the empty, sequestered, and crumpled vesicle remnant persists for a prolonged period of time, during which CME is specifically recruited to remove it. We speculate that the high curvature at the tips of the fine membrane tubes that make up the folded vesicular structure facilitate the specific recruitment of the clathrin machinery (Zeno et al., 2020; Henne et al. 2010), thereby coupling endocytosis to exocytosis. We have shown previously that myosin II assembles around vesicles in a striped organization (Rousso et al., 2016). We now understand that the function of myosin is not only to constrict the vesicle but also to crumple the SV membrane, in order to extrude the content without integrating the vesicle membrane into the apical surface. It is likely that beyond actomyosin, proteins and lipids that sense, generate, and stabilize curvature, have a role in sustaining the highly curved tubes and sheets of the crumpled membrane. While secretion from the LSGs occurs only once in the lifetime of the organism and the glands are subsequently histolyzed, most other secretory glands of this type need to maintain membrane homeostasis throughout the lifetime of the organism, during repeated phases of secretion. It would be interesting to test whether the retrieved SV membrane is recycled for de novo vesicle biogenesis in these vertebrate models.

It would also be important to explore whether the limited diffusion of proteins and lipids between the vesicular and apical membranes is governed by the nature of the apical surface or by an active diffusion barrier (Trimble and Grinstein, 2015). For synaptic vesicles, evidence of mixing between vesicle and target membranes is varied (Chanaday and Kavalali, 2017). While over-expressed vesicular SNAREs show lateral diffusion after fusion (Granseth et al., 2006), vesicular proteins with lower copy numbers, such as neurotransmitter transporters, manifest very limited lateral diffusion (Balaji and Ryan, 2007; Leitz and Kavalali, 2011). In chromaffin and mast cells, the medium-sized granules appear to predominantly use kiss-and-run exocytosis, which likely avoids the need to limit diffusion (Breckenridge and Almers, 1987a, 1987b; Elhamdani et al., 2006; Henkel et al., 2001).

The utilization of large vesicles for secretion in numerous organs and organisms raises the question of the possible advantages of such vesicles over conventional-sized SVs that are up to two orders of magnitude smaller in diameter. It is possible that trafficking of secretory cargos in larger parcels will reduce clogging at the apical membrane during the process of secretion that can extend over hours. With respect to the challenge of membrane homeostasis, large vesicles provide the most economical way of packaging cargos. As vesicles increase in size, vesicle surface area increases by the square of the radius, while vesicle volume increases as a cube of the radius. Thus, large vesicles that have a radius that is two orders of magnitude larger than conventional vesicles, can package a given volume using only 1% of membrane surface that would be utilized to package the same volume by small vesicles. Conversely, when considering the smaller endocytic vesicles that remove the crumpled membrane, these vesicles are efficient in clearing large amounts of membrane while removing from the lumen only 1% of the original content that was deposited by the large vesicles.

In conclusion, while packaging secretory cargos in large vesicles minimizes the amount of membrane, this excess membrane must still be efficiently removed to maintain homeostasis. Exocytosis by actomyosin-mediated membrane crumpling achieves this goal by sequestering the vesicular membrane, which is then cleared by endocytosis over an extended period.

Limitations of the study

Our work demonstrates that the membrane of large exocrine vesicles undergoes crumpling and is sequestered from the apical membrane. The diffusion of membrane-associated proteins between the apical membrane and the retained vesicle membrane is restricted, preserving the composition of the apical membrane. Further studies should reveal whether this limited diffusion is a general feature of the apical membrane or alternatively, if a specific mechanism is employed to block the diffusion between the apical membrane and the vesicular membrane. The mechanism of exocrine vesicle crumpling and retention was found in the D. melanogaster salivary gland as well as in the mouse acinar pancreas. However, the mouse pancreatic vesicles are only up to 3 μm in diameter, while the D. melanogaster vesicles can be as large as 10 μm in diameter. It remains to be seen whether these size differences lead to subtle differences in the secretory process. The D. melanogaster salivary gland functions only once and will eventually undergo histolysis, in contrast to other exocrine tissues that maintain homeostasis for the lifetime of the organism. We do not know how the retrieval of membrane from the crumpled vesicle by CME contributes to apical membrane homeostasis of such tissues.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for reagents and resources should be directed to and will be fulfilled by the Lead Contact Ori Avinoam (ori.avinoam@weizmann.ac.il)

Materials availability

This study did not generate new unique reagents or transgenic lines.

Data and code availability

This study did not use any unpublished custom code, software, or algorithm.

Experimental model and subject details

Drosophila strains and rearing conditions

The following Drosophila lines were obtained from the Bloomington Drosophila Stock Center: UAS–LifeAct–Ruby (B-35545), Sgs3–GFP (B-5884), UAS-mCD8-GFP (B-5130), 10xUAS-IVS-myr-tdTomato (B-32221), UAS-shits1 (B-44222), UAS-lqf-GFP (B-57350), UAS–zip RNAi (B-38259). The following stocks were obtained from the Vienna Drosophila Resource Center: UAS-AP2-σ RNAi (v110725), UAS-shi RNAi (v10597) and sec22-GFP (2XTY1-SGFP-V5-preTEV-BLRP-3XFLAG, v318332). The UAS-CD63-EGFP line was a gift from Javier A. Sanchez-Lopez, UAS-shi-YFP was a gift from Maria Leptin (EMBL, Heidelberg) (Fabrowski et al., 2013), UAS-GPI-GFP was a gift from Isabel Guerrero (CSIC-UAM, Autonomous University of Madrid). fkh-Gal4 was used to drive UAS-based expression in salivary glands (Maybeck and Röper, 2009).

All fly stocks were reared in standard cornmeal, molasses and yeast media used by Bloomington Drosophila stock Centre, in 25°C incubators without internal illumination. Crosses with SGs expressing AP2 -σ RNAi were grown at 25°C and then shifted to 29°C from second instar larval stage onwards, for maximum effect of RNAi. Crosses expressing shi RNAi were grown at 18°C and then shifted to 25°C from the third instar larval stage onwards. shits1 is a temperature sensitive mutant allele. Crosses with UAS-shits1 were grown at 18°C (permissive temperature) and shifted to 29°C (restrictive temperature) for 1hr before dye injection and imaging. Imaging of these glands was performed on a 29°C heated stage (restrictive temperature). Gal4 driver-only controls were grown under the same regimens, showed normal secretion. All other crosses were grown at 25°C throughout. Larvae from crosses were used without distinguishing between sexes, as no obvious sex specific differences in secretion of LSG were observed. Figures 1A, 1E–1G, 4A, 4E–4G, 5A, 5E–5G, 6A, 6E–6G, S1A, S1B, S1F–S1H, S4A, S4B, S4G–S4I, S5A, S5B, and S5G–S5I contain live imaging experiments performed on ex-vivo cultures of 3rd instar Drosophila LSGs. Figures 2B–2D, 3B–3D, 6B–6D, S2C–S2F, S3C–S3F, and S5C–S5F contain CLEM experiments on cryo-fixed and resin embedded Drosophila salivary gland samples (see further details under section – Sample Preparation for CLEM).

Mouse lines

Mice were handled in accordance with Institutional Animal Care and Use Committee (IACUC) guidelines. 6–10-week-old, mT mice (membrane tdTomato strain) derived from Stock No: 007576, from the Jackson Laboratory were used for isolation of pancreatic acini. Both male and female mice were used for all experiments without discernable difference between sexes.

Method details

Culturing third instar LSGs for live imaging

LSG culturing was previously described (Rousso et al., 2016). In brief, SGs from third instar larvae were dissected out in Schneider’s medium and transferred to a 35-mm dish, with a 10 mm #1.5 glass bottom well (Cellvis D35-14-1.4-N), containing 200 μl of fresh medium for live imaging. SGs that were naturally secreting were identified by their expanded lumen, visible under a stereomicroscope before imaging. Alternatively, non-secreting SGs could be induced to secrete by treating them with Ecdysone (5 μM concentration of 20-hydroxyecdysone; Sigma-Aldrich) with mild shaking for 2-3 h at room temperature (RT) before imaging.

Drug treatment of LSGs

To stall vesicular secretion, the LSGs were treated with a ROCK inhibitor (Y-27632, 50 μM final concentration; Sigma-Aldrich) for 20 mins with mild shaking at RT, before imaging (Segal et al., 2018). Cultures maintained for up to 2 h after such treatment did not display any overt abnormalities in tissue integrity or structure of the labeled components of the cells.

Quantification of vesicle dynamics

Time series were acquired from a region close to the lumen of the LSG where the vesicles were fusing. While imaging, the vesicles can be captured in various orientations, ranging from orthogonal to parallel relative to the apical surface. Vesicles in an orthogonal orientation present a clear view through the fusion pore and apical surface. Vesicles for which the mid-plane through the pore was acquired while imaging were selected and cropped out for analysis. The entire time lapse focus series were bleach corrected (by histogram matching) and background subtracted (Rolling ball radius 20) in Fiji before cropping out the individual vesicles for analysis. A region of interest (ROI) was drawn around the perimeter of the mCD8-GFP probe on the vesicle membrane for each time frame, starting 5 frames before fusion and until the end of secretion (or longer). The same procedure was carried out for the LifeAct-Ruby (F-actin) probe imaged along with the membrane marker. The mean and integrated intensity of the mCD8 and LifeAct markers in the drawn ROIs were measured. The intensities of each time frame were normalized to the peak intensity and plotted over time (Figures 1H and 1I). The ROI around the mCD8-GFP signal was also used to estimate mean vesicle diameter at each time frame during secretion. In the plot in Figure 1H the vesicle diameter is normalized to the first time-frame. As representative of vesicular membrane dynamics in the secreting population of vesicles, the normalized total membrane intensity for 6 different vesicles was plotted over time for a duration of 100s starting with F-actin assembly (LifeAct appearance) to the end of vesicular constriction (Figure 1I). A similar strategy was adopted to quantify the normalized total Shi-YFP intensity over time in Figure S5A.

FRAP on LSGs

For Fluorescence recovery after photobleaching (FRAP) experiments, secreting LSGs expressing the mCD8-GFP (membrane) and LifeAct (F-actin) probes were set up for live imaging (pre-bleaching scan). Once the LifeAct marker appeared on the vesicle, acquisition was halted, an ROI corresponding to the LifeAct marker around the vesicle was drawn, and this region was targeted for photobleaching. The 405 laser line was used for photobleaching with a dwell time of at least 10 ms/pixel. The total photobleaching time was variable, depending on the size of the vesicle. Time lapse image acquisition was resumed immediately after photobleaching. The total intensity values for the mCD8 and LifeAct probes were estimated as described above for quantification of vesicle dynamics. The intensity at each time point after photobleaching was normalized to an average of 5 prebleach time frames. The fluorescence recovery traces were normalized using double normalization (Phair et al., 2004) and further normalized to full scale. FRAP recovery was fitted using a single exponential equation. Normalization, fitting, mean mobile fraction (FM) ± S.D were calculated and plotted using OriginPro2021. All vesicles analyzed in Figures 4 and S4, belong to LSGs in different larvae. Percentage of fluorescence recovery at ∼30 s (which is t ½ from the recovery plots), is included for more vesicles in Table S2 for both mCD8-GFP and myr-GFP membrane probes.

Endocytic uptake in LSGs

To perform dye uptake experiments, the apical surface of LSGs was stained by injecting CellMask Deep Red (ThermoFisher Scientific, diluted 1:1000 in Schneider’s medium) into the gland lumen during the glue-secretion phase, using a Femtojet express microinjector (Eppendorf) and custom-made capillaries. Imaging was started immediately after injection with an interval of 30mins. The basolateral membranes of the SGs were stained almost instantly after placing them in imaging media containing CellMask (1:1000 dilution).

To monitor endocytic internalization and accumulation of the genetically encoded myr-tdTomato marker, Z-stacks were acquired for secreting LSGs at intervals of 30 mins for at least 1.5 h. Three planes were chosen from the Z-stack that harbored a maximal number of intracellular myr-tdTomato puncta. These planes were usually between the apical surface and the nuclei, where most endosomes accumulated over time. The same 3 planes were chosen for each time interval for estimating the total endosomal volume over time. Quantification of endosome volume is intensity independent, and hence unaffected by variation in intensities along the depth of the tissue and variations due to photobleaching. For endosomal volume quantification, the Z-stack was background subtracted (Rolling ball radius 10). ROIs were drawn excluding the cell membrane of each cell and this region was cropped out for quantification. Intensity thresholding was then performed, such that only the myr-tdTomato puncta that displayed a higher intensity than the myr-tdTomato signal on the SVs were selected. The total volume as well the number of myr-tdTomato puncta were then estimated using the 3-D object counter plugin in Fiji. The total volume of all the vesicles inside a single cell was then pooled. This was done for at least three different cells in the imaging frames, across 3 intervals of 30 mins and from three different glands. The pooled endosomal volume of each cell at each time point was normalized to the pooled endosomal volume for the same cell at the first acquired time point. This was performed for SGs from AP2 σ, zip and RhoGAP71E knockdown larvae and compared to driver only controls. Endosomal volume quantification was not applied to SGs from shi knockdown larvae, since intracellular myr-tdTomato puncta were completely absent in this perturbation. The slopes of the endocytic uptake plots were compared using linear regression analysis.

Isolation and live imaging of mouse pancreatic acini

Mouse pancreatic acini were isolated as described previously (Geron et al., 2013). In brief, the mT mice (membrane tdTomato strain derived from Stock No: 007576, from the Jackson Laboratory) were euthanized according to institutional animal care guidelines, and their pancreases were excised. The excised tissues were minced to 1–3-mm pieces, digested for ∼15 min in oxygenized Krebs-Ringer buffer (KRB) medium supplemented with 0.5 mg/ml BSA and 0.1 mg/ml soybean trypsin inhibitor (STI) (resuspension medium) and with collagenase P (0.75 mg/ml). The digested tissues were washed with resuspension medium and filtered first through a coarse mesh, then through a 70-μm nylon mesh. The exocrine acini were allowed to settle on the bottom of 15-ml tubes; this step was repeated two to three times to remove floating, damaged acini. For overnight culturing, acini were resuspended in DMEM supplemented with FBS (5% volume), sodium pyruvate (1%), antibiotics (1%), l-glutamine (0.5%), BSA (10 mg/mL), and STI (0.2 mg/mL). Acini were plated at low density on plates coated with 5 μg/ml collagen IV for 5 h and were incubated at 37°C in 5% CO2 humidified air. Acini were transduced with 106 pfu/ml adenoviral LifeAct-GFP for 9–16 h before imaging. The acini were transferred to glass bottom coverslip dishes for live-imaging experiments that were carried out at 37 °C in resuspension medium after inducing secretion with Carbachol (1 μM, Sigma) or Cholecystokinin/CCK-8 (50 pM, Research plus). Cultured and transduced acini retained their in vivo polarity, and cell viability could be assessed by CellMask staining. Live cells showed cell membrane staining while damaged or punctured cells showed cytoplasmic CellMask staining. Vesicular secretion was imaged on a Yokogawa spinning disk confocal scanning unit (CSU-W1-T2) coupled to an inverted Olympus IX83 microscope (details under section: Microscopy and image processing).

Sample preparation for CLEM

For correlative light and electron microscopy, the dissected Drosophila LSGs or the isolated mouse pancreatic acini with appropriate fluorescent markers, were fixed using a high-pressure freezer (HPF) [Leica EM ICE (Leica Microsystems GmbH, Germany)]. For HPF, the glands or acini were placed inside aluminum planchets (Wohlwend GmbH, Switzerland; 0.3 mm/flat part #1314 and 0.15/0.15mm part #1315) filled with PBS supplemented with 10% BSA and 10% FBS as cryoprotectant. Freeze-substitution and resin embedding were performed using a temperature-controlled device, AFS2 (Leica Microsystems GmbH, Germany). The steps in freeze substitution were as follows: the HPF-fixed samples were placed in 0.1% (w/v) uranyl acetate in dry acetone at - 90°C for 70 h. The temperature was then raised to -45°C during the next 45 h, held at -45°C for 30 h followed by 3 washes with dry acetone for 10mins each. Infiltration with Lowicryl HM20 (Electron Microscopy Sciences, USA) was carried out with increasing concentrations of 10%, 25%, 50% and 75%, for 12hr each. Infiltration with 10% and 25% Lowicryl was performed at -45°C, and for 50% and 75% Lowicryl the temperature was raised to -35°C and -25°C, respectively. Next, temperature was held at -25°C with 3 exchanges of 100% Lowicryl HM20 every 12-15 hours, followed by polymerization of the resin under UV for an additional 48 hours at -25°C. The temperature was then raised to 20°C (2°C/hour), and further polymerization of the resin under UV for 48 hours.

On-section FM from resin embedded samples

For on-section fluorescence microscopy (FM), 300-nm-thick sections were cut using a Leica EM UC7 ultramicrotome (Leica Microsystems GmbH, Germany) mounted with a diamond knife (Diatome, Biel, Switzerland), and picked onto carbon-coated 200 mesh copper EM grids (Electron Microscopy Sciences, Hatfield, PA). The imaging of EM grids by FM prior to electron tomography acquisition was performed as described previously (Avinoam et al., 2015). Multispectral 50 nm TetraSpeck (TS) microspheres were allowed to adsorb to one side of the grid. The EM grids were then placed on a drop of PBS sandwiched between two glass coverslips (Menzel-Gläser, #1.5), sealed with grease, and imaged with the sections facing toward the objective of an Olympus IX83 fluorescence microscope (details under section: Microscopy and image processing).

Electron tomography

Electron tomography (ET) was performed as describe previously (Avinoam et al., 2015). Tomographic fiducial markers (15 nm protein A-coupled gold beads) were adsorbed to both sides of the EM grids. Transmission electron microscopy (TEM) maps were acquired for correlation with the FM. Either scanning transmission electron microscopy (STEM) or TEM tilt series were then acquired using a Tecnai F20 electron microscope (Thermo Fischer Scientific) at 200 kV or a Tecnai F30 electron microscope (Thermo Fischer Scientific) at 300 kV using SerialEM (Mastronarde, 2005). The TEM images for correlation were acquired at 4.47 nm pixel size using an UltraScan 4000 camera (Gatan). The STEM tilt series were acquired over a 60° to −60° tilt range (1° increment), at 1.67 nm pixel size in nanoprobe mode using the bright-field detector, with a 10-μm C2 aperture and 320 mm camera length. TEM tilt series were acquired using OneView camera (Gatan) or the K2 direct detector (Gatan). The tomograms were reconstructed using the IMOD software package (versions 4.9.4).

The multispectral 50 nm TetraSpeck (TS) microsphere beads visible in both FM and EM were used to correlate the 2 data sets. After we acquainted ourselves with the object of interest with the help of such correlations and were able to visibly identify crumpled vesicles in EM, we also utilized cellular features visible in both FM and EM for correlation.

Correlations of FM and TEM images were performed using the ec-CLEM plugin with the ICY software. The correlation precision using TS achieves correlation precision of 100nm or better (Avinoam et al., 2015; Kukulski et al., 2011).

Block-face FM of resin embedded samples

The blocks containing the resin embedded samples were trimmed on all sides to obtain a small block face ∼1 by 1.5mm. The face of the block was trimmed till it reached the lumen of the tissue (visible under a stereomicroscope) and then polished using a diamond knife (Diatom, Biel, Switzerland). For block face imaging, it was mounted with its polished surface facing down on the glass coverslip (Menzel-Gläser, #1.5) with a drop of PBS for hydration, and sealed on all sides with all-purpose plastic clay. The block was imaged through the cover glass on an Olympus spinning disc confocal IX83 microscope (details under section: Microscopy and image processing). Iterative block face imaging and trimming and polishing was performed till a region of interest with LifeAct-Ruby positive vesicles was identified, that was suitable for acquisition of FIB-SEM.

FIB-SEM

Prior to FIB-SEM imaging, blocks were mounted on SEM stubs (EMS cat# 75220) with conductive carbon adhesive tape (EMS cat#77816). Additional straps of conductive tape were applied to the blocks, to connect the stub and the block surface. An 8-10nm layer of Ir was deposited on the surface of the block using compact coating unit (CCU) - 010 (Safematic GmbH, Switzerland). Samples were then mounted onto a Helios NanoLab DualBeam 600 microscope (Thermofisher Scientific, USA). After finding ROIs in the SEM, 0.5-1μm of Pt was deposited on top of the ROI using either an ion beam (30kV,0.46-0.92nA), or with an additional electron beam deposition step (2kV, 5.5nA) prior to ion beam deposition. To expose the cross-section of the ROI, a frontal trench was milled with the ion beam (30kV, 21nA); trench dimensions varied with respect to ROI dimensions. Samples were subsequently imaged using a Crossbeam 550 system (Carl Zeiss GmbH, Germany) at a 54° tilt angle and a 5 mm working distance. Prior to serial surface imaging, cross-sections were polished with an ion beam (30kV, 1.5-7nA) and then milled in serial section mode (30kV, 0.7-1.5nA), at different dose factors (5-8) depending on the specific specimen. SEM micrographs were acquired under 2kV, 350pA. A combination of detectors, in-lens secondary electrons (inLens)/ energy-selective backscattered electrons (EsB) or EsB/secondary electrons type 2 (SE2), were used for image acquisition. Detection was determined per acquisition with respect to the image quality obtained in every specific stack.

Image processing: collected data was filtered using unsharp masking or local contrast enhancement (CLAHE) followed by smoothing (if needed) in Fiji. All other processing steps were performed using Amira version 2019.3 (Thermofisher Scientific, USA). Alignment was done with Align Slices module, allowing translation only, and refined manually to correct for misaligned slices. Segmentation of the vesicles was made semi-automatically using a brush tool with a threshold mask, followed by interpolation. The threshold was adjusted to include the volume inside the vesicles. The fusion pore area was segmented from the negative curvatures formed at the connection with the plasma membrane. Segmentation was performed every 4th slice of the stack unless new features appeared (especially in crumpled vesicles), then a smaller gap was used to include the new feature in the interpolation. The gaps were then interpolated, and the interpolation was corrected manually in cases of mis-interpolated features. A surface was generated from the segmented data using the Generate Surface module, followed by volume and surface area calculations using the Surface Area Volume module.

Targeted FIB-SEM and SA/V determination

Focus series were obtained by block-face imaging of Lowicryl resin embedded LSGs expressing LifeAct-Ruby (F-actin). Regions of the LSGs enriched in LifeAct positive vesicles were chosen for FIB-SEM acquisition. A surface SEM image of the same block was obtained and features on the LSGs close to the surface of the block were overlayed with LifeAct-Ruby fluorescence from block-face imaging. This allowed to target regions enriched with LifeAct-Ruby positive vesicles, that were subjected to FIB milling followed by serial surface acquisition of large volumes. FIB-SEM data was acquired with an isotropic voxel size. FIB-SEM data was processed as described above. The surface area (SA) and volume (V) were calculated for unfused, early fused (but still spherical) and late crumpled vesicles. Only the crumpled vesicles whose boundaries could be clearly identified at the apical surface (that is crowded with membrane extensions like filopodia) were chosen for segmentation, surface area and volume estimation.

Reason for using SA/V

The amount of surface area per unit volume (surface area to volume ratio - SA/V) is a characteristic of an object of a specific shape and size. The secretory vesicles are closest to a sphere in shape. Among all regular 3-dimensional shapes, a sphere has the lowest possible surface area for a given volume. Figure S3H is a plot showing the change in SA/V ratio with diameter (d) for a sphere. The arrowhead in this plot points to the average diameter of the vesicles seen in LSGs and the grey region highlights the range of vesicle diameters observed in LSGs. Under the scenario of the full collapse mode of exocytosis (schematic on the plot, Figure S3H), where the vesicle membrane is added to the surface as content volume is lost, the surface area and volume of the vesicle would both change as secretion progresses. During this mode of secretion, the SA/V (=6/d) can be expected to change according to the plot above, assuming that the shape of the vesicle remains close to a sphere.

We estimated the actual volume from the segmented vesicles at different stages- unfused, early fused and late stage crumpled vesicles, from the FIB-SEM stacks (Figure S3F and Table S1). Assuming a full-collapse exocytosis, we deduced the surface area from the measured volume (as per the plot above). For the deduction of surface area, the approximate diameter of the vesicle was estimated from the actual volume and this diameter was used in the calculation of the surface area using the equation π d2. A ratio between the deduced surface area to the actual volume is referred to as - the simulated SA/V ratio (Figure S3G).

We also quantified the actual surface area for each of the segmented vesicles (Figure S3E and Table S1). The ratio between the actual surface area and volume was calculated and is represented in the black box plots in Figure 3C. For the late stage crumpled vesicles, the actual SA/V ratio was significantly higher than the simulated SA/V ratio (grey box plot in Figure 3C) and hence represented for comparison. For the unfused and early fused vesicles, the simulated SA/V ratio and actual SA/V ration were not significantly different. Thus, the SA/V ratio is a way of normalization when the initial size of the vesicle is unknown and surface area and volume are obtained from the segmented FIB-SEM data.

Microscopy and image processing

Live imaging of Drosophila LSGs and pancreatic acini as well as block-face imaging of LSGs were performed using a Yokogawa automatic Spinning Disk confocal scanning unit (CSU-W1-T2) mounted on an inverted Olympus IX83 microscope. 60x 1.4 NA and 100x 1.49 NA oil immersion objectives were used for data acquisition. Images were captured by the back illuminated Prime 95B sCMOS camera (Photometrics), controlled by VisView software (Visitron Systems GmbH). The following fluorescence excitation and emission filter sets were used: 470/40 nm and 525/50 nm for GFP, 560/40 nm and 630/75 nm for Ruby or tdTomato, and 620/60 nm and 700/75 nm for CellMask Deep Red. The CellMask uptake experiments on LSGs were imaged on a Zeiss LSM 800 confocal microscope system, using 40x 1.2 NA water or 60x 1.4 NA oil objectives. On-section fluorescence microscopy was performed using Olympus IX83 microscope controlled via VisiView software (Visitron Systems GmbH) and equipped with a CoolLED pE-4000 light source (CoolLED Ltd., UK), 60x 1.4 NA and 100x 1.49 NA oil immersion objectives, and a Prime 95B sCMOS camera (Photometrics).

Post imaging processing was performed using Fiji and Adobe Photoshop CC 2020 for cropping and adjustment of brightness/contrast and visualization purposes only.

Quantification and statistical analysis

All experiments for each genetic background and/or drug treatment were repeated at least three times with different organisms, and representative images/videos are shown. The specific statistical tests used for each experiment and P values are mentioned in figure legends. The GraphPad Prism software was used for all statistical analysis other than FRAP experiments for which OriginPro2021 was used.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Adenovirus (Ad)-Lifeact-GFP	From Ben-Zion Shilo (Weizmann Institute of Science)Geron et al., 2013	Made by Erez Geron, Ben-Zion Shilo lab
Experimental models: Organisms/strains
D. melanogaster: y1 w∗; P{UAS-Lifeact-Ruby}VIE-19A	Bloomington Drosophila Stock Center	BDSC: 35545FlyBase:FBtp0064438
D. melanogaster: w∗; P{Sgs3-GFP}2	Bloomington Drosophila Stock Center	BDSC: 5884FBgn0003373
D. melanogaster: y1 w∗; PinYt/CyO; P{UAS-mCD8::GFP.L}LL6	Bloomington Drosophila Stock Center	BDSC: 5130FBst0005130
D. melanogaster: w∗; P{10XUAS-IVS-myr::tdTomato}attP2	Bloomington Drosophila Stock Center	BDSC: 32221FBti0131951
D. melanogaster: w∗; P{UAS-lqf.GFP.F}2	Bloomington Drosophila Stock Center	BDSC: 57350FBti0162749
D. melanogaster: w∗; P{UAS-shits1.K}3	Bloomington Drosophila Stock Center	BDSC: 44222FBti0151794
D. melanogaster: myr membrane probe w∗; P{UAS-gapGFP}AC1	Bloomington Drosophila Stock Center	BDSC: 4522FBti0010578
D. melanogaster: RNAi for Zipper y1 sc∗ v1 sev21;P{TRiP.HMS01703}attP40/CyO	Bloomington Drosophila Stock Center	BDSC: 38259FBti0149450
D. melanogaster: sec22-GFPPBac{fTRG01262.sfGFP-TVPTBF}	Vienna Drosophila Resource Center	v318332FBgn0260855
D. melanogaster: RNAi for shibireP{KK101444}VIE-260B	Vienna Drosophila Resource Center	v105971FBgn0003392
D. melanogaster: RNAi for AP-2 σP{KK102525}VIE-260B	Vienna Drosophila Resource Center	v110725FBgn0043012
D. melanogaster: y,w∗;P{UASp-shi::YFP}	Gift from Maria Leptin (EMBL, Heidelberg)Fabrowski et al., 2013	https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3753550/
D. melanogaster: w∗; P{UAS-GPI::GFP}	Gift from Isabel Guerrero (CSIC-UAM, Autonomous University of Madrid)	Made by Isabel Guerrero lab
Mouse: membrane tdTomato strain mT/mG, mTmG	Derived from Stock Gt(ROSA) 26Sortm4 (ACTB-tdTomato,-EGFP)Luo/J from the Jackson Laboratory	JAX: 007576
Chemicals, peptides and recombinant proteins
Ecdysone / 20-hydroxyecdysone	Sigma-Aldrich, Merck	Cat# H5142CAS: 5289-74-7
ROCK inhibitor, Y-27632	Sigma-Aldrich, Merck	Cat# Y0503
Carbachol/ Carbamoylcholine chloride	Sigma-Aldrich, Merck	Cat# C4383
Cholecystokinin/ 26-33 sulfated CCK-8	Research Plus Inc.	Cat# 01-0340-02
Collagenase P, from Clostridium histolyticum	Roche Diagnostics GmbH	Ref# 11213865001
Trypsin inhibitor, from Glycine max (soybean)	Sigma-Aldrich, Merck	Cat# T9003
Protein A- gold beads 15nm	UMC Utrecht	PAG 15 nm https://www.cellbiology-utrecht.nl/products.html
TetraSpeck beads 50nm	ThermoFisher Scientific	TetraSpeck Microsperes, 50nm customized
Lowicryl HM20 kit	Electron Microscopy Sciences	Cat# 14340 EMS
Software and algorithms
Fiji	(Schindelin et al., 2012)	https://imagej.nih.gov/ij
SerialEM	(Mastronarde, 2005)	https://bio3d.colorado.edu/SerialEM/
IMOD version 4.9.4	(Kremer et al., 1996)	https://bio3d.colorado.edu/imod/
ICY	(de Chaumont et al., 2012)	https://icy.bioimageanalysis.org/
eC-CLEM	(Paul-Gilloteaux et al., 2017)	http://icy.bioimageanalysis.org/plugin/ec-clem/
Amira version 2019.3	(Stalling et al., 2005)	https://doi.org/10.1016/B978-012387582-2/50040-X
GraphPad Prism version 8.3.0 for Mac	GraphPad Software	www.graphpad.com
OriginPro 2021	OriginLab Corporation	www.originlab.com
Others
200 mesh copper grids	Electron Microscopy Sciences	CF200-CU-50
Aluminum planchets for HPF	M. Wohlwend GmbH	Part # 1314 and 1315