Structured RhoGEF recruitment drives myosin II organization on large exocytic vesicles

The Rho subfamily of small GTPases is a focal point for controlling cytoskeletal architecture and cell behavior in eukaryotic cells (Hodge and Ridley, 2016; Jaffe and Hall, 2005). Rho proteins are loosely anchored to the cell membrane via C-terminal lipid modifications. Both their upstream regulators and downstream effectors are cytoplasmic proteins recruited to the membrane by their transient association with Rho. Rho proteins shuttle between the active, GTP-bound, and inactive GDP-bound states. RhoGEF activators and RhoGAP inhibitors orchestrate the transition between these states. The spatial and temporal dynamics of these two classes of regulators, which vary according to the cell and tissue requirements, determine the resulting pattern of Rho activation (Bement et al., 2024; Bos et al., 2007; Denk-Lobnig and Martin, 2019). Activated Rho GTPases recruit downstream target proteins to the membrane, triggering a cascade of events that reorganize the cytoskeleton, modify cell adhesion and motility, and activate or repress gene expression (Spiering and Hodgson, 2011).

Upon activation, Rho-GTP promotes force generation through the actomyosin meshwork. This is accomplished by recruiting and activating both formin family proteins, which mediate nucleation and elongation of linear actin microfilaments, and myosin II motors, which generate force through binding and movement along anti-parallel actin cables. Formins are present in the cytoplasm in an inactive state due to the intramolecular association between the N- and C-termini of the protein (Higgs, 2005). Upon binding to membrane-associated Rho-GTP, formins transition from a closed to an open state and become active. Stable association with the membrane is mediated by binding to the phospholipid phosphatidylinositol (4,5)-bisphosphate (PIP2) via a basic domain in the formin protein (Rousso et al., 2013). The combined association with Rho-GTP and PIP2, termed ‘coincidence detection’, leads to a stable membrane interaction of active formin and the nucleation of linear actin microfilaments.

The recruitment and binding of non-muscle myosin II to the actin cables is driven by Rho-associated protein kinase (Rock) proteins. Like formins, Rock proteins are cytoplasmic proteins that are maintained in an inactive state due to the intramolecular association between their N- and C-termini. Following the initial recruitment by active Rho-GTP, a PH phospholipid-binding domain in Rock proteins provides tighter association with the membrane (Amano et al., 2000). Rock proteins phosphorylate the regulatory light chains of the myosin II hetero-oligomer, allowing the complex to bind the actin microfilaments. Thus, Rho-GTP is the common trigger for both actin polymerization and myosin II activation. However, how the same molecule temporally and spatially coordinates these two distinct events remains unclear.

Actomyosin plays a central role during exocrine secretion processes that employ large secretory vesicles (SVs) (Nightingale et al., 2012). A prominent example is the Drosophila larval salivary glands (LSGs), which utilize large SVs, of several micrometers in diameter, and display an orchestrated recruitment of actomyosin, which is required in order to expel their content into the lumen of the gland. Salivary gland SVs release large amounts of adhesive mucinous glycoproteins (‘glue’), which the fly pupa uses to attach to a solid surface. Secretion initiates shortly before pupation in response to a hormonal (Ecdysone) signal (Biyasheva et al., 2001), proceeds for ∼2 h, and involves exocytosis from hundreds of individual secretory vesicles from within each cell (Farkas and Sutakova, 1999). Exocytosis of the glue cargo from each fused vesicle via stabilized fusion pores takes 1–3 min (Biton et al., 2023; Rousso et al., 2016; Tran et al., 2015). Interestingly, actomyosin contractility mediates folding and ‘crumpling’ of the vesicle membrane, which drives a unique mode of exocytosis, wherein content release occurs while the vesicle membrane remains segregated from the apical cell membrane (Kamalesh et al., 2021). Activation of Rho1, the primary Drosophila Rho family homolog, triggers both arms of the actomyosin network in this setting – namely, actin polymerization, and myosin II recruitment and activation (Rousso et al., 2016). Although previous work identified RhoGAP71E as a RhoGAP protein recruited to the secretory vesicles to drive Rho1 inactivation and actin depolymerization (Segal et al., 2018), the corresponding RhoGEF proteins mediating Rho1 activation have not been identified. Moreover, the precise mechanism of actomyosin organization and force generation on the SVs has not been established.

In the current study, we examined the spatiotemporal dynamics of the Rho network during exocytosis of large secretory vesicles, by exploring the activation of Rho1 by RhoGEF proteins and its implication on actomyosin contractility. We show that upon vesicle fusion with the apical cell membrane, low levels of activated Rho1 appear on the vesicle membrane. These levels of activated Rho1 are sufficient to recruit the formin Diaphanous (Dia), generating a fairly uniform F-actin coat around the vesicle (Rousso et al., 2016; Segal et al., 2018), but appear incapable of recruiting sufficient levels of myosin II to mediate vesicle constriction. Subsequent recruitment of RhoGEF2 in an irregular structured pattern on the F-actin coat generates a corresponding activation pattern of Rho1, driving local myosin II recruitment and subsequent constriction. Vesicle membrane buckling occurs at sites of high myosin II concentration, leading to the crumpled shape of the vesicle membrane during content release. Our findings imply a paradigm of different thresholds of activated Rho1 for recruitment of Dia and Rock (also known as Rok, the single fly Rock homolog), enabling a biphasic mode of actomyosin recruitment that drives anisotropic membrane crumpling during large secretory vesicle exocytosis.

In the Drosophila LSGs, the fusion of large secretory vesicles with the apical cell membrane initiates an orchestrated process of actomyosin recruitment to the vesicle membrane that expels its content (Fig. 1A,B). To characterize actomyosin coat dynamics and organization by Rho1 signaling, we monitored the F-actin probe LifeAct–Ruby and the myosin II light-chain reporter Sqh–GFP, using Airyscan confocal live-imaging (Huff, 2015) of ex vivo cultured LSGs (Fig. 1C,D). F-actin and myosin II appeared almost simultaneously on the vesicles ∼40 s after fusion, but displayed distinct spatial assembly patterns (Fig. 1C; Movie 1). The actin coat appeared broadly spread over the vesicle, whereas myosin II displayed a distinct, punctate pattern (Fig. 1C). This visually prominent difference in the F-actin and myosin II distributions was quantitatively estimated by generating intensity line profiles for both these elements along the circumference of newly fused vesicles at ∼40 s post fusion (Fig. 1D; Fig. S1A). The variance in intensity of F-actin or myosin about the mean was used as a measure of dispersion for these elements and was found to be significantly lower for F-actin compared to myosin II (Fig. 1D and see figure legend).

To obtain a comprehensive 3D view of actomyosin distribution on the entire vesicle, we used Airyscan confocal super-resolution microscopy, deconvolution and 3D reconstruction to examine fixed LSGs, stained with both anti-GFP antibody and fluorescent phalloidin (Fig. 1E–K; Fig. S1B, Movie 2). Examining the vesicles at their equatorial planes showed myosin II localization as a punctate pattern (Fig. 1H), consistent with the live imaging (Fig. 1C,D). The 3D reconstructed surface view of the vesicle along the x-y imaging axis (Fig. 1G,I–K; Fig. S1B,C) revealed a more complete picture: a structured meshwork-like organization of myosin II filaments on the vesicle, alongside a broader, partially overlapping arrangement of F-actin cables (Fig. 1I–K). The observed differences suggest that although activated Rho1 is the common trigger, additional factors contribute to the distinct spatial patterning of F-actin and myosin II on SVs.

To examine the basis for the differential recruitment patterns of F-actin and myosin II to SVs, we searched for the RhoGEF protein(s) responsible for activating Rho1 in this setting. One prominent candidate was RhoGEF2, known for its role in activating Rho1 in a variety of Drosophila tissues (Fox and Peifer, 2007; Grosshans et al., 2005; Hacker and Perrimon, 1998; Mulinari et al., 2008; Nakamura et al., 2017; Padash Barmchi et al., 2005). To test the role of RhoGEF2 in this context, we monitored the distributions of the active Rho sensor AniRBD::GFP (Munjal et al., 2015) and F-actin following RNAi-mediated knockdown of RhoGEF2 in LSGs (Fig. 2A,B). In wild-type (WT) SVs, the Rho sensor initially appeared faint and spread out over the SV at 20±8.2 s after fusion, followed by F-actin appearance at 32.5±8.3 s (mean±s.d.; Fig. 2C; Fig. S2A). Notably, the Rho sensor appeared 12.5±4.3 s before F-actin (Fig. S2A; Movie 4), consistent with the notion that Rho activation precedes actin recruitment (n=12 SVs from three LSGs). Subsequently, the Rho levels intensified, became more punctate (Fig. 2C; Movie 3) and peaked at 56.7±13 s, followed by an F-actin peak at 84.2±15 s. Both signals persisted until exocytosis was complete.

In RhoGEF2 RNAi-expressing LSGs, the SVs exhibited prominent F-actin recruitment, with F-actin first appearing at 23.6±5 s and attaining peak levels comparable to that in WT (Fig. 2D,E; Fig. S2B). However, although the Rho sensor first appeared at 17.5±4.5 s, a timing comparable to WT, the signal was weak and remained at a markedly low level (Fig. 2D,F; Fig. S2B, Movies 5, 6). Quantitative analysis of Rho dynamics on multiple vesicles highlights the significantly reduced peak levels of the recruited Rho sensor on SVs in RhoGEF2 RNAi-expressing glands as compared to WT LSGs (Fig. 2F; Fig. S2D). Consistent with the reduced levels of active Rho, most of the fused vesicles stalled and failed to expel their content (75±8% in RhoGEF2-RNAi LSGs compared to 9±5% in WT, n=75 SVs from three LSGs in each, mean±s.d.; Fig. 2D; Fig. S2B). We have previously shown that approximately half (53%) of SVs in which secretion is halted by Rock inhibition, display assembly and disassembly cycles of their F-actin coats (Rousso et al., 2016; Segal et al., 2018). In contrast, either full retention of the coat or marked prolongation of these cycles was observed for 78% of the RhoGEF2-knockdown stalled vesicles. Our previous work has shown that RhoGAP71E activity on Rock-inhibited stalled vesicles is responsible for the cycling of F-actin levels (Segal et al., 2018). The disassembly of F-actin, mediated by RhoGAP71E, was thus also affected by knockdown of RhoGEF2 activity. These effects were specific to RhoGEF2, as expression of an RNAi construct targeting GEFmeso, another RhoGEF highly expressed in LSGs, displayed normal F-actin and Rho1 sensor recruitment, similar to WT (Fig. 2E,F; Fig. S2C,D, Movie 7).

Based on these observations, we conclude that RhoGEF2 is essential for enhancing the levels of active Rho1 on SVs, which in turn affects vesicular constriction. Furthermore, we find that F-actin recruitment to SVs is independent of the RhoGEF2-driven recruitment of Rho1 to SVs.

The stalling of vesicular constriction observed in the RhoGEF2-RNAi LSGs prompted us to examine the recruitment and distribution of myosin II under these conditions. To this end, we monitored the F-actin probe LifeAct–Ruby and the myosin II light-chain reporter Sqh–GFP in RhoGEF2 RNAi-expressing LSGs. We observed that myosin II recruitment and distribution were aberrant upon RhoGEF2 knockdown. The levels of recruited myosin II (Fig. 3A,B; Fig. S3A, Movies 8, 9) appeared significantly lower than those in WT (Fig. 1C; Movie 1) or with RhoGEFmeso-RNAi (Fig. S3B, Movie 10). Quantification of myosin II (Sqh–GFP) dynamics on multiple vesicles showed reduced myosin II peak levels in RhoGEF2 RNAi compared to the levels for WT or RhoGEFmeso-RNAi control LSGs (Fig. 3C–F), whereas F-actin levels were unaffected (Fig. 3F). The difference in myosin II distributions was quantitatively estimated by generating intensity line profiles for myosin along the circumference of newly fused RhoGEF2-RNAi SVs and compared to GEFmeso-RNAi control SVs (Fig. 3D). The variance in intensity of myosin about the mean was used as a measure of dispersion of the distribution, and was found to be significantly lower for the RhoGEF2-RNAi vesicles compared to the control SVs (Fig. 3E and legend).

Examination of the overall 3D organization of Sqh–GFP in fixed and immunostained LSGs revealed apparent differences in the myosin II pattern in RhoGEF2-RNAi SVs compared to WT (Fig. 3H). 3D rendering of myosin II organization from the surface of the vesicle revealed a considerably more dispersed distribution on RhoGEF2-RNAi SVs, when compared to the structured organization of myosin II foci observed in WT (Fig. 3Hiii,v versus Fig. 3Hi). Consistently, Sqh–GFP also appeared more uniform when viewed at the equatorial plane of the RhoGEF2 knockdown SVs (Fig. 3Hiv,vi versus Fig. 3Hii).

In conclusion, although compromising the levels of RhoGEF2 did not affect actin polymerization, a marked effect on myosin II recruitment to the vesicles was observed. We propose that fusion of the vesicle with the apical cell membrane leads to the hypothetical recruitment of low levels of active Rho by diffusion from the plasma membrane. In support of this notion, we observed the instantaneous diffusion of the transmembrane protein marker mCD8–GFP, from the apical membrane into the vesicle at fusion (Fig. S3C,D). However, diffusion becomes restricted by the time F-actin appears on the vesicle, which is around 20 s post fusion (Kamalesh et al., 2021). We propose that this initial level of active Rho1 is sufficient for the recruitment of Dia and induction of actin nucleation, but cannot drive proper myosin II recruitment and activation on its own, a matter we discuss below. The subsequent activity of RhoGEF2 is required to promote the recruitment of high levels of myosin II in a meshwork pattern.

To clarify the role of RhoGEF2 for appropriate myosin II recruitment, we first determined the spatiotemporal pattern of RhoGEF2 on the SVs. Live imaging showed that RhoGEF2–sfGFP was recruited to the vesicle from the cytoplasm in a punctate pattern, typically between 25 s and 30 s after fusion (Fig. 4A; Movie 11). RhoGEF2 recruitment preceded myosin II appearance (monitored by Sqh–RFP) by several seconds, suggesting a possible causal relationship (Fig. 4A). This notion is further supported by the observation that RhoGEF2 knockdown LSGs displayed only low levels of activated Rho1, lacking the typical WT punctate organization (Fig. 2C,D). Thus, we conclude that the prominent signal of activated Rho1 in WT vesicles can be attributed to its activation by RhoGEF2.

Next, we monitored the spatial distribution of RhoGEF2 and activated Rho1. The pattern of activated Rho1 showed a similar structured distribution to that of RhoGEF2 (Fig. 4Bii,v versus Fig. 4Dii,v). However, we could not compare them in the same vesicle, given that RhoGEF2 and Rho1 are only available as GFP fusion proteins. Because high levels of activated Rho1 are expected to recruit Rock and myosin II, we examined, as an alternative, the correspondence between the distribution of activated Rho1 and myosin II (Sqh–RFP) on the same vesicle. Indeed, a significant correlation between these distributions was observed and quantified, suggesting a causal relationship between punctate high levels of activated Rho1 and the corresponding recruitment of myosin II (Fig. 4B,C and legends).

If the structured RhoGEF2 distribution dictates the pattern of myosin II recruitment, we would also expect to see a correspondence between these patterns. By comparing the distribution of RhoGEF2 and myosin II on the same vesicle, we observed a similar distribution and a significant correspondence of the patterns at the equatorial sections that was quantitatively estimated (Fig. 4D,E and legends). It should be noted that although the spatial overlap between activated Rho1, RhoGEF2 and myosin is substantial, it is incomplete. This is to be expected, given that Rock activation generates a contractile actomyosin network, which can then change shape rapidly. These results suggest that local puncta of RhoGEF2 recruitment from the cytoplasm generate distinct foci of activated Rho1, which, in turn, lead to the corresponding structured pattern of myosin II recruitment.

Is the meshwork-like recruitment of myosin dictated exclusively by the pattern of RhoGEF2, or are there structural spatial asymmetries on the vesicle which might generate a pre-pattern for myosin II recruitment? To examine this possibility, we followed the distribution of a phospho-mimetic myosin protein (Sqh^EE), which is constitutively active and does not require phosphorylation by Rock (Munjal et al., 2015). The localization of Sqh^EE thus bypasses the normal recruitment signals. Sqh^EE localized in a fairly uniform pattern in the majority of the vesicles (Fig. 4F,G) and led to partial stalling of vesicle constriction (39±5%, n=75 SVs from three LSGs; mean±s.d.). This result implies that apart from the local patterning by RhoGEF2, there are no other inherent molecular or physical asymmetries on the SV membrane that impinge on the recruitment of activated myosin II. We conclude that the punctate distribution of RhoGEF2 dictates the spatial pattern of activated Rho and Rock on LSG SVs, which subsequently drive the structured pattern of myosin II recruitment.

Next, we examined what triggers the recruitment of cytoplasmic RhoGEF2 to the vesicles. To determine whether the recruitment of RhoGEF2 to the vesicle depends on the polymerization of F-actin, we followed its recruitment in LSGs treated with the actin polymerization inhibitor Latrunculin A (LatA; Fig. S4). Fused vesicles were marked and detected with the F-actin reporter LifeAct–Ruby or the membrane marker mCherry–CAAX. Prior to the addition of LatA, the normal punctate recruitment of RhoGEF2 was observed (Fig. S4A,B). However, following the addition of LatA, newly fused SVs were stalled and displayed an irregular F-actin pattern of a few puncta. No recruitment of RhoGEF2 to these SVs was observed (Fig. S4C,D). This experiment indicates that F-actin or an actin-associated protein is required to recruit RhoGEF2 to the fused SVs. Such a mechanism would ensure the pre-establishment of F-actin tracks for the myosin II motors to assemble.

Having established the basis for the punctate recruitment of myosin II, we next wanted to explore its role in vesicle constriction. Glue SVs do not undergo isotropic constriction, but rather demonstrate multiple local folds in a process we have termed ‘crumpling’ (Kamalesh et al., 2021). As a result, in this novel mode of secretion, when the vesicle content is effectively expelled, the vesicle membrane remains insulated from the apical cell membrane. Could the focal recruitment of myosin II direct the process of vesicle crumpling?

To examine this notion, we followed the spatial correlation between the distribution of myosin II and the topology of the constricting vesicle membrane. In cross sections of a squeezing vesicle, the sites of high myosin concentration indeed corresponded to the location of membrane folding (Fig. 5A; Movie 12), suggesting a causal relationship between local myosin recruitment and the sites of vesicle membrane folding. The actual squeezing of the vesicle content is a fairly rapid event that majorly occurs within a timeframe of 36±8 s (mean±s.d.) within the vesicle fusion-secretion window (Kamalesh et al., 2021). We propose that vesicle squeezing is executed by the concerted action of multiple confined actomyosin constriction events, which are dictated by the focal recruitment and distribution of myosin II. Local constrictions of the actomyosin mesh leads to buckling of the underlying vesicle membrane. The combination of multiple local events leads to the global squeezing of the vesicle, content release, and crumpling of the vesicle membrane (Fig. 5B).

We have monitored the mechanism of actomyosin recruitment to the large glue SVs in the Drosophila LSG, following their fusion with the apical cell membrane. This process lasts up to 3–4 min for each vesicle, culminating in the complete expulsion of vesicle cargo into the LSG lumen. Contraction of an actomyosin network that coats the vesicle plays an essential role in content release, leading us to explore in depth the molecular mechanisms underlying its formation and activation. In keeping with a standard and well-established pathway leading to actomyosin-based contractility, both F-actin assembly via the formin Dia and myosin II recruitment by Rock are triggered in this setting by activated Rho1. However, our observations reveal that the spatiotemporal characteristics of the two responses to active Rho1 are distinct.

Prior to fusion, little if any activated Rho1 can be detected on the secretory vesicle membrane. Fusion of the vesicle can allow diffusion of membrane-tethered active Rho1 from the apical membrane to the vesicle membrane. Diffusion of active Rho1 to the vesicle membrane is arrested by the time of formation of the actin coat (Kamalesh et al., 2021). This initial, low level of active Rho1, uniformly distributed on the vesicle membrane, leads to a corresponding pattern of recruitment of the formin Dia (Rousso et al., 2016) and the subsequent, fairly even formation of prominent actin cables, but provides for only low levels of myosin II recruitment, which are insufficient to drive vesicle constriction. Accumulation of F-actin on the vesicle, or of hitherto unknown actin-associated proteins, leads to the recruitment of RhoGEF2 from the cytoplasm to the vesicle membrane. This recruitment is uneven and marked by multiple local puncta of high RhoGEF2 levels, which drive local Rho activation and subsequent recruitment of myosin II (mediated by Rock) in a corresponding structured pattern. The distribution and activity of myosin II then drive the constriction of actomyosin at multiple locations, leading to buckling of the underlying vesicle membrane. These constrictions enable the unique secretion mode from such large vesicles (termed exocytosis by vesicle ‘crumpling’), which expels the vesicle content while keeping the vesicle membrane insulated from the apical membrane of the LSG (as shown schematically in Fig. 5B). We wish to note in this context the distinction between the actomyosin-rich apical cell membrane and the dynamic events leading to actomyosin-based contractility on the nascent vesicle membrane, which is devoid of the relevant signaling and cytoskeletal elements prior to fusion. Furthermore, and in support of this notion, we have recently observed that glue SVs which undergo a ‘full collapse’ mode of secretion and abnormally incorporate into the cell membrane immediately after fusion, still subsequently recruit myosin and even display contractile activity (Biton et al., 2023), underscoring the distinct compositions of vesicle and cell membranes.

Activated Rho has been universally demonstrated to be the crucial switch for the recruitment of both F-actin and myosin II (Hodge and Ridley, 2016; Jaffe and Hall, 2005; Spiering and Hodgson, 2011). In frog eggs, the appearance of F-actin precedes myosin II recruitment (Yu and Bement, 2007). We identified a similar biphasic recruitment mode in the Drosophila LSGs, where the activation of myosin II takes place after the formation of a filamentous actin coat and might actually rely on components generated by actin coat construction. The temporal basis for the biphasic recruitment in this system stems from an apparent difference in the levels of active Rho1 required to trigger each branch. Hypothetical low levels of activated Rho1, provided by diffusion from the apical membrane, appear sufficient for full-fledged recruitment of Dia and subsequent nucleation and polymerization of an F-actin coat. However, they are insufficient for complete myosin II recruitment, which relies on an enhancement of Rho1 activation, triggered by RhoGEF2 activity. In light of our observations, examining whether the two branches are activated simultaneously or consecutively in other systems of actomyosin recruitment, and if they have distinct Rho activation thresholds, is warranted. For example, such a mechanism might trigger the recently identified structured (and F-actin independent) recruitment of myosin II to secretory vesicles of mouse salivary glands (Ebrahim et al., 2019).

The biphasic mode of actomyosin recruitment is crucial for the proper secretory activity of the salivary gland. Without the second wave of myosin II recruitment, triggered by RhoGEF2, only low levels of myosin II accumulate on the vesicles, and they remain stalled. This regulatory mode serves two roles for the constricting vesicles. First, it allows harnessing of the high levels of inactive RhoGEF2 in the cytoplasm, by recruiting the protein and activating it at the vesicle membrane, presumably via G-protein-coupled receptors (GPCRs) (van Unen et al., 2015). Second, the recruitment and distribution of RhoGEF2 to the vesicles rely on elements that accumulated in the first wave of activation, namely F-actin or actin-associated proteins. These unevenly distributed local foci lead to a corresponding accumulation of RhoGEF2. The structured distribution of subsequent myosin II clusters dictates the constriction pattern of the vesicle, such that the underlying membrane buckles to release the vesicle content but does not integrate into the apical membrane of the cell. This mode of constriction is robust and efficient given that dozens of such myosin II foci drive the constriction of each vesicle, whereas their precise number and spacing can vary between vesicles. The resulting outcome of membrane crumpling maintains the unique composition of the apical membrane in the face of fusion to multiple large secretory vesicles (Kamalesh et al., 2021).

Drosophila fly lines used in this study include: lines obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537): fkh-GAL4(B-78060; Figs 1–5 and Figs S1–S4), UAS-LifeAct-Ruby (B-35545; Figs 1–5 and Figs S1–S4), sqh–GFP.RLC (B-57145; Figs 1,3,5 and Fig. S1,S3), UAS-RhoGEF2 RNAi (B-34643; Figs 2,3 and Figs S2,S3, and B-31239 data not shown), UAS-GEFmeso RNAi (B-42545; Figs 2,3 and Figs S2,S3), sqh-mCherry (B-59024; Fig. 4), UAS-mCherry-CAAX (B-59021; Fig. 5), and sqh-sfGFP-RhoGEF2 (B-76260, Fig. 4). Lines from other resources: ubi-AniRBD::GFP (Rho1 sensor; Munjal et al., 2015) kindly provided by Thomas Lecuit, IBDM, France (Figs 2,4); sqhE20E2–GFP (phosphomimetic Sqh; Royou et al., 2002); kindly provided by Andrea Brand, University of Cambridge, UK (Fig. 4).

All fly stocks were reared on standard cornmeal, molasses and yeast medium at 21°C in a temperature-controlled room. Crosses and flies used for imaging experiments were grown in 25°C incubators without internal illumination. Live imaging experiments were performed on ex vivo cultures of third-instar Drosophila LSGs. Larvae from crosses were used without distinguishing between sexes, as no obvious sex-specific differences in SG secretion were observed.

SG culturing was performed as previously described (Kamalesh et al., 2021). In brief, SGs from third-instar larvae were dissected out in Schneider's Drosophila medium (SDM) and up to six glands were 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. LSGs were typically imaged within an hour of dissection.

LSGs were imaged for at least 5 min before drug treatment, to ensure overall health and the commencement of secretion. With image acquisition stopped, LatA (1 µM; Sigma-Aldrich), was added to the medium with a micropipette, directed near but without touching the sample, and mixed gently with a pipette tip. Typically, the effect of LatA was clearly observable after 15 mins of its addition. Image acquisition was recommenced on the same region of the LSG to directly compare the state of F-actin (visualized using the Lifeact–Ruby reporter) before and after LatA treatment. Cultures maintained for up to 2 h after treatment did not display any overt abnormalities in tissue integrity.

Live imaging of the LSGs was performed on an LSM900 Airyscan 2 confocal microscope (Zeiss), using a 63×/1.4 NA (oil) objective, 1.5–2× digital zoom, in super-resolution mode. The high-sensitivity Airyscan GaAsP-PMT detector was used, as it significantly improved the signal-to-noise ratio, and thus spatial and temporal resolution. Special care was taken to monitor the alignment of the Airyscan detector throughout the live imaging session, as the movements inside the live specimen could be a source of fluctuations in the detector alignment. The Airyscan detector adjustments were set to activate alignments automictically during live and continuous scans, and during time-series acquisitions. Time series were acquired from a region close to the lumen of the LSG, where the vesicles fuse. To achieve the highest temporal resolutions, image acquisitions were made from single optical sections with no time intervals and for short durations of 5–10 mins only, due to profuse bleaching. For quantifying vesicular dynamics (see section below), two optical sections 2 μm apart were typically acquired with a 5 s time interval for a duration of 20 mins. For all live Airyscan imaging, scan speeds of ∼1 µs/pixel were optimal to maintain proper detector alignment while achieving the best temporal and spatial resolution.

Raw stacks were processed using the Airyscan processing utility on Zen 3.1 software (Blue edition) to obtain the Airyscan time series. Fiji software and Adobe Photoshop CC 2021 were used for cropping and adjustment of brightness and contrast of the Airyscan images for visualization purposes. The time series shown in the figures represents the typical dynamics observed in vesicles across more than six LSGs.

From the Airyscan time series acquisitions, vesicles that could be clearly followed from the onset of fusion (which is associated with vesicular expansion) were cropped out for analysis. The entire time-lapse focus series was background subtracted using Fiji software, before cropping out the individual vesicles for analysis. Time lapse analysis was performed on movies taken at 20 s intervals. A region of interest (ROI) was drawn around the perimeter of the vesicle for the probe being analyzed, on each time frame, starting from a frame before fusion and until the end of secretion, or as long as possible in the case of contraction-halted vesicles observed in fkh>RhoGEF2i LSGs. An auto-thresholding was then applied to the individual channels of the cropped-out vesicles. For analysis of the LifeAct–Ruby probe for F-actin, the Isodata type of thresholding was used, whereas for analysis of the Rho sensor and Sqh intensity, the mean type of thresholding was used. The integrated intensity of the probes post thresholding and within the drawn ROIs were measured. The measured intensity of each time frame was normalized over the average cytosolic intensity of the probe and plotted over time (Figs 2E, 3C). The peak intensities of the different probes were compared across genotypes.

Individual third-instar LSGs were dissected out in SDM, then transferred to another drop of SDM supplemented with 1% of freshly prepared PLPS fixative (4% paraformaldehyde, 0.1% glutaraldehyde, 0.01 M sodium meta-periodate, 0.075 M L-lysine, 0.035 M phosphate buffer and 0.1% saponin, pH 7.4), where they were swiftly bisected along their length using a pair of fine tweezers, thus exposing the lumen of the glands. The PLPS fixative contributes considerably to antigen unmasking but is only partially compatible with phalloidin-based visualization of F-actin. The bisected LSGs were immediately transferred to ice-cold PLPS fixative and kept on a rotating shaker for 30 mins at room temperature (RT). The bisected and fixed LSGs were then given a quick wash with Permeabilization neutralization (PN) solution (0.035 M phosphate buffer, 0.01 M sodium meta-periodate, 0.075 M L-Lysine, 0.35% Triton-X 100, 0.2% sodium deoxycholate and 0.2% glycine, pH 7.4), followed by another 15 min wash on a rotating shaker. Fixed LSGs were pooled together in PLPS buffer (0.035 M phosphate buffer, 0.01 M sodium meta-periodate, 0.075 M L-lysine, pH 7.4) until enough glands were obtained, and permeabilized with PN solution for 30 mins at RT followed by blocking in PN solution with 5% normal goat serum (NGS), for 1 h at RT. Primary antibody incubation was performed for 16–24 h at 4°C in PBST (PBS plus 0.3% Triton X-100) with 5% NGS, followed by three washes with PBST for 10 mins each. Secondary antibody incubation was performed for ∼6 h at RT or 12–16 h at 4°C, followed by three washes in PBST. Following an additional wash in PBS, the tissue was transferred to 90% glycerol (in PBS) for a couple of hours. The LSGs were then mounted between 2 glass coverslips with spacers, using 90% glycerol as the mounting medium for Airyscan imaging.

The following primary antibodies were used: rabbit anti-GFP (abcam cat. no. ab290) diluted 1:200 was used throughout, except when co-staining for mCherry, where chicken anti-GFP (abcam cat. no. ab13970) was used at 1:800; rat anti-RFP 1:1200 (ChromeTek cat. no. 5F8); rabbit anti-dsRed (Clonetech cat. no. 632496). Anti-rabbit-IgG, anti-chicken-IgY and anti-rat-IgG secondary antibodies conjugated to Alexa-Fluor-488, Alexa- Fluor-568 or Alexa-Fluor-633 were purchased from Molecular Probes (Invitrogen) and used as recommended by the manufacturer. Phalloidin–Atto647 (Fluka cat. no. 65906) or phalloidin–TRITC (Sigma cat. no. P-1951) were used along with primary and secondary antibody incubation when required.

LSGs with their lumens facing close to the coverslip were imaged on the LSM900 Airyscan 2 confocal microscope (Zeiss), using a 63×/1.4 NA (oil) objective, 2–4× digital zoom, in super-resolution mode. Scan speeds >2 µs/pixel were optimal for the fixed samples. ∼10–12 μm z-stacks were typically acquired using a 0.15 μm interval between optical sections. Multiple channels were acquired in sequential frame acquisition mode. The raw z-stacks were processed using the 3D Airyscan processing utility of Zen 3.1 software (Blue edition) to obtain Airyscan super-resolution stacks. Fiji software and Adobe Photoshop CC 2021 were used for cropping and adjustment of brightness and contrast of the Airyscan images for visualization purposes.

For 3D reconstruction and analysis of vesicles, all processing was done using the Zen 2 (blue edition) service pack. Full Z-stacks of individual vesicles were cropped out from Airyscan Z-stacks of a portion of the gland using the Create image subset utility. The images were further sharpened using the unsharp mask utility, followed by applying a Nearest Neighbor (NN) deconvolution algorithm (adjustable). The contrast was further improved by applying the Enhance contour utility. The 3D reconstruction of individual vesicles obtained with Zen 2 was used to analyze the patterns of myosin II, F-actin, Rho and RhoGEF2 and their overlap in co-immunostaining experiments. Staining of the first six sections along the x-y imaging axis from the surface of the vesicles (irrespective of the orientation of the vesicle) was sum projected, to depict the surface view defined in Fig. 1E. The surface view of myosin II staining in WT is shown in Fig. 1G and Fig. 3Hi and in fkh>RhoGEF2i in Fig. 3Hiii,v. Similarly, surface views of active Rho sensor and RhoGEF2 staining are shown in Fig. 4Bii and Dii, respectively. The surface view gives the impression of the overall 3D pattern of the molecules on the surface of the vesicles.

The equatorial plane view of the vesicles was used to analyze and obtain a quantitative measure of the distribution of myosin II, active Rho sensor and RhoGEF2 and their colocalization. The equatorial plane is the middle section of the vesicle from its z-stack that is defined in Fig. 1E. The equatorial plane view for myosin II staining on a WT vesicle shows a punctate distribution as in Figs 1H and 3Hii and in fkh>RhoGEF2i a more uniform distribution as in Fig. 3Hiv,vi. To obtain a quantitative measure of this distribution, line scans depicting the fluorescence intensity along the circumference of the vesicle using the equatorial plane view were plotted using Fiji software. For this analysis, the background signal (measured from the center of the vesicle outline) was subtracted from the equatorial plane view, a segmented line of 6-pixel width was drawn along the circumference of the vesicle that encompasses most of the fluorescence signal, and the Fiji plot profile function was used to plot the intensity of the molecules along the vesicle circumference. An example of the line scan plot showing the comparative distribution of F-actin and myosin II in WT is shown in Fig. 1D. This plot shows large fluctuations in fluorescence intensity distribution for myosin II, characteristic of a punctate or patterned distribution, whereas F-actin intensity is more uniform along the circumference.

To analyze the overlap between active Rho sensor or RhoGEF2 and myosin II, a line scan depicting the fluorescence intensity of both channels along the circumference of the vesicle from the equatorial plane view was used.

All experiments for each genetic background and drug treatment were repeated at least three times with glands from different organisms, and representative images or videos are shown. The statistical tests used for each experiment and P-values are mentioned in the figure legends. The GraphPad Prism software was used for all statistical analysis and plotting.

We thank Dr R'ada Massarwa and Vandana Bharadia for help at the initial stages of this work. O.A. is an incumbent of the Miriam Berman presidential development chair. B.-Z.S. is an incumbent of the Hilda and Cecil Lewis Professorial Chair in Molecular Genetics. This paper is dedicated to the memory of the late Dr R'ada Massarwa, a groundbreaking scientist with an unlimited passion to visualize and uncover the wonders of biology.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261944.reviewer-comments.pdf