Cell-intrinsic and -extrinsic roles of the ESCRT-III subunit Shrub in abscission of Drosophila sensory organ precursors Free

Cell junctions are essential for chemical and mechanical functions of epithelia (Higashi et al., 2016). Adherens junctions (AJs) and tight junctions (TJs) in vertebrates/septate junctions (SJs) in invertebrates ensure the mechanical and permeability barriers, respectively (Banerjee et al., 2006; Harris and Tepass, 2010; Shin et al., 2006; Tepass et al., 2001; Tsukita et al., 2001). Drosophila SJs are composed of a stable core complex containing over 20 proteins with interdependent localization, which include cytosolic proteins such as Coracle (Cora) and Discs large (Dlg; Dlg1); Claudin-like proteins including Kune-Kune; cell adhesion molecules such as Fasciclin III (Fas3); and ion channel transporters such as Na⁺/K⁺ ATPase alpha (ATPα) and beta [Nervana2 (Nrv2)] subunits (Faivre-Sarrailh, 2020; Genova and Fehon, 2003; Izumi and Furuse, 2014; Kaplan, 2002; Nelson et al., 2010; Snow et al., 1989; Ward et al., 1998; Oshima and Fehon, 2011).

Although cells need to establish and maintain a proper permeability barrier, SJs remain highly plastic, especially during cell division, a fundamental process for the development and function of all organs. Cytokinesis is initiated in anaphase with the contractile activity of an actomyosin ring, which drives cleavage furrow ingression and results in midbody formation, a docking platform connecting daughter cells, required to recruit essential proteins in later steps of cell division (Fededa and Gerlich, 2012; Glotzer, 2005; Green et al., 2012). Abscission, the final step in cytokinesis, results in the physical separation of the two daughter cells and has been proposed to help regulate cell fate acquisition (Chaigne et al., 2020; Ettinger et al., 2011). In Drosophila, the centralspindlin protein complex recruits ALiX, which in turn recruits the endosomal sorting complex required for the transport-III (ESCRT-III) component Shrub (the CHMP4B ortholog). Shrub/CHMP4B has the property to form polymers that, together with other ESCRT-III components, are proposed to drive abscission (Eikenes et al., 2015; Guizetti et al., 2011; Lie-Jensen et al., 2019; Matias et al., 2015). In addition, CHMP4B exerts different functions, including viral budding and plasma and nuclear membrane resealing (Agromayor and Martin-Serrano, 2013; Lie-Jensen et al., 2019; Vietri et al., 2020). Although there is a great deal of knowledge on abscission of isolated cells, abscission of tightly packed polarized epithelial cells with junctional complexes remains largely underexplored.

The formation of new AJs is coordinated with the early steps of cytokinesis, leading to the positioning of the midbody basal to the forming adhesive interface (Firmino et al., 2016; Founounou et al., 2013; Guillot and Lecuit, 2013; Herszterg et al., 2013; Higashi et al., 2016; Lau et al., 2015; Morais-de-Sá and Sunkel, 2013). Moreover, neighboring cells maintain SJ contact with the dividing cell via finger-like protrusions connected to the midbody (Daniel et al., 2018; Wang et al., 2018). Novel SJs assemble below AJs and spread basally, concomitantly with the basal displacement of the midbody. The midbody leaves the SJ plane ∼1.5 h after anaphase onset, and abscission occurs over 5 h after anaphase onset (Daniel et al., 2018; Wang et al., 2018). The sequence of events leading to abscission and the concomitant junctional remodeling are proposed to play a role in maintaining epithelial barrier functions in proliferative epithelia, yet remain poorly characterized in tissues composed of cells with distinct identities.

Drosophila notum consists of a single-layered epithelium containing two distinct types of cells, epidermal cells (ECs) and sensory organs (SOs) (Fig. 1A). Sensory organ precursors (SOPs) divide asymmetrically and generate two daughter cells, pIIa and pIIb, that undergo subsequent cell divisions, ultimately giving rise to adult SOs (Fig. 1A′) (Fichelson and Gho, 2003; Gho et al., 1999; Hartenstein and Posakony, 1989). At each division, cell fate determinants Numb and Neuralized (Neur) are unequally segregated to control Notch-dependent binary fate acquisition (Langevin et al., 2005; Le Borgne and Schweisguth, 2003; Rhyu et al., 1994). In contrast to ECs, SOPs and daughters are fast-cycling cells (Audibert et al., 2005) (Fig. 1A′), raising the question as to how barrier integrity and abscission occur during SOP cytokinesis to ensure subsequent cell divisions while preserving tissue homeostasis. In this study, we characterized the coordination between the midbody basal displacement and SJ remodeling during SOP cytokinesis, and investigated the functions of Shrub and SJ components in abscission. We report that Shrub contributes to regulate daughter cell physical cytoplasmic isolation and SJ remodeling, exerting two complementary functions to orchestrate SO abscission.

To monitor cytokinesis progression across time and throughout the three-dimensional cellular space, we live imaged the regulatory light chain of nonmuscle type 2 Myosin tagged with RFP (MyoIIRFP) tagged with RFP (MyoIIRFP) to label both the AJ plane and the actomyosin contractile ring, and ATPα tagged with GFP (ATPαGFP) to monitor the SJs (Fig. 1B,B′; Fig. S1A,A′). SOP and pIIa/pIIb daughter cells were identified using Histone 2B::IRFP670 (H2BIR), expressed under the minimal promoter of neur (Fig. 1B,B′). Throughout, time (t) is expressed in minutes, except when indicated otherwise, with t0 corresponding to the onset of anaphase. In both ECs and SOPs, the constriction of the actomyosin contractile ring gives rise to the midbody located basal to the AJ, within the plane of SJs (Fig. 1B,B′; Fig. S1A,A′ at t10) (Daniel et al., 2018; Founounou et al., 2013; Guillot and Lecuit, 2013; Herszterg et al., 2013; Morais-de-Sá and Sunkel, 2013). As previously described for dividing ECs (Fig. 1 and Fig. S1A, t10; Daniel et al., 2018; Wang et al., 2018), actomyosin ring pulling results in membrane ingression in the dividing SOPs while neighboring cells remain tightly associated with the latter via finger-like protrusions, as observed with ATPαGFP and Cora (Fig. 1B, t10, SJ plane; Fig. 1C,C′, t10, SJ plane). We found another commonality between SOPs and ECs, the basal displacement of the midbody (Fig. 1B′; Fig. S1A′, arrows), concomitant with a striking disparity between both cell types, with midbody basal displacement occurring 1.75 times faster in SOPs (2.8±0.1 µm/h) than in ECs (1.6±0.1 μm/h) in the 80 min following anaphase onset (Fig. 1D,E). From this timepoint onwards, we observed a sudden inflection in the curve, with the SOP midbody being displaced faster towards the basal side (6.9±0.2 µm/h) (Fig. 1E) as it leaves the SJ domain (Fig. 1B, t90-t115). Taking into account how newly assembled SJs act as a conveyor belt, triggering the basal displacement of the midbody in ECs (Daniel et al., 2018), the difference in midbody displacement kinetics between SOPs and ECs predicts faster SJ remodeling in the former, which we decided to investigate next.

To discriminate between SJ components present at finger-like protrusions and coming from the dividing cell or neighboring cells, we imaged EC or SOP dividing clones, expressing MyoIIRFP and endogenous untagged ATPα, adjacent to cells expressing MyoIIRFP and ATPαGFP. At t10, the presence of ATPαGFP signal in the finger-like protrusions pointing towards the midbody originated from the neighbor cell, and confirmed that SJ barrier integrity between the SOP and its neighboring EC is preserved during cytokinesis (Fig. 2A; Fig. S2A; Daniel et al., 2018). In SOPs, as the midbody underwent basal displacement, concomitantly at t70±7, lateral extremities within a finger-like protrusions drew closer until the edges could no longer be resolved by light microscopy (Fig. 2A, t65, SJ plane, n=4). At t97±4, ATPαGFP signal at finger-like protrusion was barely detectable (Fig. 2A, t100 SJ plane, n=5), and, shortly after, the midbody was displaced from the SJ plane (Fig. 2A, t110; Fig. 1C, t115). Finally, the midbody appeared displaced at distance, basal to the pIIa nucleus, described here as SOP midbody release (Fig. 2A, t120; Fig. 1C,C′). In striking contrast to SOP cytokinesis, during EC division, finger-like protrusions persisted after t120 (Fig. S1A; Daniel et al., 2018), confirming that finger-like protrusion disengagement and midbody release occur faster in SOPs than in ECs.

We next analyzed the converse situation and monitored the de novo assembly of SJs at the new SOP daughter interface by live imaging SOPs expressing ATPαGFP and MyoIIRFP surrounded by EC neighbors expressing endogenous untagged ATPα and MyoIIRFP. During SOP cytokinesis, the ATPαGFP signal appeared apical and above the midbody, forming new SJs between SOP daughter cells at t45 (Fig. 2B), much sooner than between ECs, in which this occurred at t80 (Daniel et al., 2018). These results suggest that SJ assembly occurs faster in SOP daughter cells than in ECs. To confirm this, we performed fluorescence recovery after photobleaching (FRAP) experiments (Fig. 2C) of ATPαGFP signal above the midbody. We observed that, 40 min after photobleaching one of the two finger-like protrusions of dividing ECs, ATPαGFP signal was barely recovered, presenting mainly an immobile fraction [ymax, 17%; t1/2, 7.4 min; in agreement with previous observations (Fig. 2D; Daniel et al., 2018). In contrast, following FRAP of the SOP daughter interface, ∼50% of the ATPαGFP signal was recovered at the new daughter interface (ymax, 49%; t1/2, 10.2 min; Fig. 2C,D).

Altogether, these data show that the formation of the new SJ, the disengagement of the finger-like protrusions and midbody release occur earlier in SOP daughters than in EC daughters.

We next studied when physical separation of pIIa and pIIb cytoplasm occurred. To do so, we generated cell-specific labeling using the photoconvertible probe KAEDE (Daniel et al., 2018) under the control of pannier promoter (pnr-Gal4) in combination with UAS-driven gene silencing. Upon directed UV excitation in the pIIa cell, the green-to-red photoconverted KAEDE probe was able to diffuse freely into the pIIb cell until cytoplasmic isolation was completed (Fig. S3A). For each SOP, identified using GAP43IR expressed under the minimal promoter neur, KAEDE photoconversion was conducted at a precise time point after anaphase onset. When photoconversion was performed in the pIIa 61 min after anaphase, the photoconverted probe diffused into the pIIb cell (Fig. 3A,A′). By contrast, when photoconversion was performed 63 min post anaphase, KAEDE remained restricted in the pIIa cell, indicating that cytoplasmic isolation had occurred (Fig. 3B,B′). The experiment was conducted at different time points after anaphase, between t46 and t71, and results show that the proportion of dividing SOP with completed cytoplasmic isolation increases over time (Fig. 3C; Fig. S3B, n=62). Owing to lethality with the pnr-Gal4 driver in one of our experimental conditions, scabrous promoter (sca-GAL4; Mlodzik et al., 1990) was used instead to drive gene silencing. Here also, the proportion of cells exhibiting cytoplasmic isolation increased as a function of time, but a delay was observed (Fig. S3B,C, Table S1A). We then determined the time at which 50% of the pIIa/pIIb cells had cytoplasmic isolation (referred to as t1/2), and obtained a t1/2 of 59±1 min and 66.5±1.3 min for pnr-Gal4 and sca-Gal4, respectively. The reason for the t1/2 difference between the two conditions is unknown. For the sake of simplicity in the comparison between different gene silencing conditions, and because both Gal4 driver lines were used as control values, we arbitrarily set the t1/2 of control pnr-Gal4 and control sca-Gal4 to a relative time of 0.

These data indicate that cytoplasmic isolation occurs at a time when the midbody is still connected to the finger-like protrusions (t85-t100, Fig. 2A), suggesting that midbody release and cytoplasmic separation are spatiotemporally decoupled. To determine the relative timing of both processes, we monitored the position of the midbody by imaging MyoIIGFP in regards to the plasma membrane marker GAP43IR (Fig. 3D) and observed that, until t100, the midbody colocalized with GAP43IR at the pIIa-pIIb interface (Fig. 3D). At t105, the midbody leaves the pIIa-pIIb interface (Fig. 3D, t111±13), just before or concomitant with pIIb cell division (t123±7, n=13). After leaving the pIIa-pIIb interface, the midbody moves onto the plasma membrane of the pIIa cell (Fig. 3D-D″, t105, t110, n=10/13; in three cases, the midbody disappears between two time frames, making it impossible to determine whether it localizes on the surface of pIIa before being released), until is no longer detected at t115 (Fig. 3D, t122±14, n=13). In isolated cells, midbodies have the ability to be internalized in acidic intracellular compartments following abscission (Crowell et al., 2014). By co-imaging MyoIIRFP and MyoIIGFP with GAP43IR, we found that the midbody labeled with RFP and GFP detaches from the plasma membrane of the pIIa cell, and is found at distance from the pIIa cell before both fluorescent signals were lost simultaneously (Fig. S3D,D′). A similar observation was made while imaging the kinesin-like protein Pavarotti (PavGFP), a member of the centralsplindin complex, together with MyoIIRFP and GAP43IR (Fig. S3E,E′). In both experimental conditions, the midbody found outside the pIIa cell was still positive for GFP, RFP and GAP43IR signals, arguing that the midbody remnant is in the extracellular space between ECs rather than in endocytic compartments in which the GFP signal should be quenched (Couturier et al., 2014).

Taken together, our quantitative analyses allow us to define a meticulous timeline encompassing SOP asymmetric cell cytokinesis events in a stepwise manner, starting with formation of the new SJs, followed by cytoplasmic isolation and the disengagement of the finger-like protrusions, and finishing with the release of the midbody (Fig. 3E). We next proceeded to elucidate what underlying molecular mechanisms govern SOPs abscission.

To further elucidate Shrub/CHMP4B function in proliferative epithelia and based on the evolutionarily conserved function of ESCRT-III/CHMP4B in abscission, spanning from archea to vertebrates, we next investigated Shrub/CHMP4B function in SOP abscission and first monitored Shrub localization. Because available anti-Shrub antibodies give no obvious membrane-associated signal upon chemical fixation (Pannen et al., 2020), we used a CRISPR/Cas9-engineered Shrub::GFP (ShrubGFP; Mathieu et al., 2022). By imaging MyoIIRFP and ShrubGFP together with GAP43IR, two distinct pools of ShrubGFP were detected (Fig. 4A). From t30 to t60, a first pool of ShrubGFP was detected in the plane of the midbody, along finger-like protrusions (Fig. 4A′,A″), and, at t70±10, the ShrubGFP signal along the finger-like protrusions was no longer detected. From t70 to t120, when the cell was poised for abscission, a second pool of ShrubGFP was detected on each side of the midbody (Fig. 4A′,A‴, n=16).

With regards to the formulated hypothesis on the potential role for Shrub during abscission, it was surprising to find that its recruitment at the midbody occurred at t70±10 (Fig. 4A′), significantly later than the time of cytoplasmic isolation, t59±1, previously reported (Fig. 3C). We found and here report that the addition of a GFP tag alters Shrub activity (see below). However, despite the fact that ShrubGFP activity is perturbed, its localization at both sides of the midbody is in agreement with previous studies (Addi et al., 2020; Green et al., 2013; Mierzwa et al., 2017). We further confirmed Shrub localization at finger-like protrusions (Fig. S4A-A″, n=3) and on both sides of the midbody (Fig. S4B-B″, n=5) using a Hemagglutinin (HA)-tagged version of Shrub (ShrubHA), a probe previously reported to be a bona fide reporter of Shrub localization (Mathieu et al., 2022). In addition, CHMP2BGFP (charged multi vesicular body protein 2B tagged with GFP), another ESCRT-III component recruited in a Shrub-dependent manner (Babst et al., 2002), was also detected at the finger-like protrusions (Fig. S4C,C‴, t17) and on both sides of the midbody (Fig.S4C,C″,C‴, t82). Altogether, these data indicate the presence of two spatiotemporally distinct pools of Shrub during abscission.

To determine the origin of the pool of ShrubGFP located along the finger-like protrusions, we imaged SOP or EC dividing clones, expressing MyoIIRFP and devoid of ShrubGFP, adjacent to cells expressing ShrubGFP (Fig. 4B; Fig. S4D). A ShrubGFP punctum emanating from the non-dividing neighboring cell was detected at the tip of the finger-like protrusion pointing toward the midbody of the dividing SOP (from t18 to t72, Fig. 4B; t36, Fig. 4B′; n=5) or that of the dividing EC (Fig. S4D; t36, Fig. S4D′; n=6). Conversely, a first punctum of ShrubGFP appeared at t26 at the midbody before the appearance of a second punctum at t60 in the SOP (Fig. 4C,C′, n=2). At t90, puncta were still present on both sides of the midbody in the SOP (Fig. 4C). In ECs, a punctum was detected on one side of the midbody from t28 to t90 (Fig. S4E; t60, Fig. S4E′; n=4). We conclude that Shrub is recruited at finger-like protrusions in a cell-non-autonomous manner and on both sides of the SOP midbody in a cell-autonomous manner, suggesting two distinct functions of Shrub that we next investigated.

Based on its localization on both sides of the midbody, we first explored the cell-autonomous function of Shrub in abscission. However, Shrub is required in several processes, making it difficult to analyze its function. Indeed, clones of epithelial cells homozygous mutant for a null allele of Shrub^G5 delaminate even when apoptosis was suppressed by p35 expression (Hay et al., 1994). We therefore opted for an UAS/Gal4-based tissue inducible-based RNA interference (RNAi) approach using sca-GAL4. Shrub depletion was attested by the presence of enlarged Hrs-positive endosomes (Fig. S5A) and Crumbs intracellular accumulation, as reported in the Drosophila trachea (Fig. 5A) (Dong et al., 2014). Shrub depletion also led to accumulation of Kune-Kune and Sinuous in intracellular compartments (Fig. 5A; Fig. S5B) as reported for other Claudin-like proteins (Pannen et al., 2020), and intracellular accumulation of NrxIV (Nrx-IV) (Fig. S5C). We next sought to determine the impact Shrub depletion had on cytoplasmic isolation kinetics and focused on sca-defined tissue regions in which KAEDE expression was limited to SO lineage. We found that SOP-specific depletion of Shrub delayed cytoplasmic isolation (Fig. 5B, with a t1/2 delay of 6±1.5 min compared with the sca control; Table S1B), as well as the timing of midbody release (Fig. S5D). As mentioned, addition of a GFP tag on Shrub alters its activity. In ShrubGFP/+ heterozygous flies, bearing a wild-type copy of Shrub and a copy of the shrubGFP allele, a delay in cytoplasmic isolation was observed in SOPs (Fig. 5C, t1/2+10±1 min compared with pnr control; Table S1C) and in the female germline stem cyst (67/181 germline stem cells with delayed abscission, Fig. S5E). These data further support the pivotal cell-autonomous function of Shrub during cytoplasmic isolation.

The fact that Shrub localizes at the finger-like protrusions raises the question of whether Shrub exerts a function there. As previously reported in the wing imaginal disc epithelium (Pannen et al., 2020), in the pupal notum, depletion of Shrub results in defective localization of SJ components, namely Claudins (Kune-Kune and Sinuous) (Fig. 5A; Fig. S5C), suggesting that Shrub is also required for SJ integrity in the pupal notum. This phenotype led us to wonder whether the effect of Shrub depletion on the kinetics of cytoplasmic isolation could also be due, in part, to its effect on SJs. If this were the case, the alteration of SJ integrity upon depletion of one of its components is expected to impact cytoplasmic isolation timing, a prediction that we next tested. We observed that SOPs showcased a premature cytoplasmic isolation upon Cora depletion in an SOP and its neighboring cells (Fig. 6A; Fig. S6A, with a t1/2−6±2 min compared with t1/2 pnr reference control; Table S1C). Following Cora depletion, as observed in the control, the midbody colocalized with GAP43IR at the interface between the two daughter cells up to 65 min after anaphase onset (Fig. 6B, from t10 to t65, midbody level). However, the midbody release from the pIIa-pIIb interface (Fig. 6C, t91±14, midbody level, n=7) and its detachment from the pIIa cell plasma membrane (Fig. 6D, t96±12, n=6) took place earlier than in control situations (Fig. 6C,D). This premature midbody release is unlikely to be caused by an acceleration of the cell cycle, as anaphase entry in pIIa and pIIb cells occurred later in Cora-depleted cells than in control counterparts (Fig. S6B).

In seeking to understand how SJs impact the timing of cytoplasm isolation, we wondered whether SJs could exert a mechanical effect on the midbody. In vertebrate isolated cells grown on glass coverslips, tensile forces on the intercellular bridge negatively regulate abscission (Lafaurie-Janvore et al., 2013). By analogy, could the finger-like protrusions exert mechanical tension on the midbody, and if so, could the removal of these tensile forces upon depletion of SJ components explain the premature abscission? To probe for putative tensile forces at the finger-like protrusions, we performed laser-based nanosurgery at these structures. No recoil was detected following ablation at one of the finger-like protrusions (Fig. 6E-E″; Movie 1, n=5). Importantly, the efficiency of the laser beam cut was validated by the fact that the cut was repaired within a couple of minutes, with the recruitment of MyoIIRFP and the re-appearance of ATPαGFP attesting the membrane repair (t1-t3, Fig. S6C; Movie 2, n=5) in a manner akin to the Rho flares described in Stephenson et al. (2019). Thus, acceleration of cytoplasmic isolation upon loss of Cora is unlikely to be due to a release of mechanical stress by finger-like protrusions. Together, these data indicate that SJ remodeling kinetics at finger-like protrusions delay the cytoplasmic isolation in SOPs.

As depletion of Shrub leads to altered localization of SJ components (Pannen et al., 2020; Fig. 5A; Fig. S5B,C), our data imply that loss of Shrub has two opposite outcomes. The first is that depletion of Shrub by affecting SJ components in the finger-like protrusions is expected to cause premature cytoplasmic isolation. The second outcome, in the dividing cell, is that depletion of Shrub is expected to delay cytoplasmic isolation (Fig. 5B). These proposed dual functions of Shrub led us to further study the relationship between Shrub and Cora. Silencing of Cora caused increased recruitment of ShrubGFP at the finger-like protrusions in one-third of the cases (Fig. 6F,F′, from t20 to t70, n=6; Fig. 6G), further suggesting a function for Shrub at the site of local SJ remodeling (see Discussion). Because shrubGFP is a mutant allele that delays abscission (Fig. 5C), we reasoned that the presence of ShrubGFP may modify the timing of abscission measured upon depletion of Cora in SOPs. Our prediction was confirmed, and Cora depletion in the presence of the shrubGFP mutant allele caused a premature cytoplasmic isolation compared with the shrubGFP allele alone (Fig. 6A, t1/2 difference of 12 min). Plotting results also suggest a tendency for Cora-depleted cells to exhibit a higher proportion of premature cytoplasmic isolation than when shrubGFP allele is present in Cora-depleted cells (Fig. 6A, t1/2 difference of 4 min).

Overall, our data support the notion that Shrub exerts two functions during SOP cytokinesis: a first cell-autonomous function, at the level of the SOP midbody, to control the timing of cytoplasmic isolation; and a second cell-non-autonomous function at finger-like protrusions during local SJ remodeling (Fig. 6H).

In this study, we characterized the coordination between abscission and the maintenance of epithelial permeability barrier functions during SOP asymmetric cell division. In the dividing SOP, below the plane of newly formed AJs, midbody assembly takes place while adhesiveness to neighboring cells is preserved via the formation of finger-like protrusions from pre-existing SJs. The basal displacement of the midbody and the de novo assembly of SJs occur more rapidly in SOPs than in ECs, resulting in faster disentanglement of finger-like protrusions in the former, prior to midbody release. At the molecular level, SJ components negatively regulate cytoplasmic isolation, suggesting a regulatory role of SJs in abscission. Finally, we report that Shrub exerts two complementary functions in SOP abscission, a cell-autonomous control of cytoplasmic isolation and a non-autonomous function at finger-like protrusions required to remodel SJs and prevent premature abscission.

Our FRAP analyses revealed that the assembly of SJs at the new pIIa-pIIb interface occurred faster in SOPs than in ECs. As a consequence, the disentanglement of the finger-like protrusions, corresponding to the disassembly of old SJs mediating contact between SOPs and neighboring cells occurs faster in SOPs than in ECs. This also causes a faster basal displacement of the SOP midbody compared with the EC midbody (Daniel et al., 2018). As the endosomal system contributes to transport and turnover of SJ complexes (Nilton et al., 2010; Pannen et al., 2020; Tempesta et al., 2017; Tiklová et al., 2010), our data suggest differences in trafficking in SOPs versus ECs. As SOP daughters enter into mitosis ∼2 h after SOP anaphase, membrane trafficking dynamics might be under cell cycle regulation. Furthermore, the SOP transcriptional program could include membrane trafficking regulators or SJ components impacting SJ assembly kinetics. Our data also suggest that the kinetics of SJ assembly and disassembly in the SOPs and daughters dictates that of SJ components in the neighboring ECs.

Our data raise the question as to how SJ components negatively regulate abscission. Our laser nanosurgery experiments suggest that SJs are unlikely to exert mechanical constraints on the midbody of SOPs (Fig. S6A,B), which is embedded in pre-existing SJs, assembled between SOPs and neighboring cells prior to mitosis. This topology ensures that the permeability barrier function is maintained in the plane of the midbody at the finger-like protrusions. Novel SJs between SOP daughters subsequently assemble above the midbody to build the permeability barrier. We envision the topology of finger-like protrusions as a means of preserving the mother-daughter permeability barrier throughout cytokinesis. The disentanglement of finger-like protrusions corresponds to the dismantlement of the old SJs, leading to the release of the midbody. Thus, apical to basal displacement of the finger-like protrusions involves intense membrane remodeling and possibly Shrub-dependent endosomal sorting and/or membrane repair functions. Upon depletion of Cora, Shrub signal becomes stronger at finger-like protrusions, and midbody release, hence abscission, is accelerated. The molecular trigger by which SJ components negatively regulate Shrub recruitment at the level of the midbody and in the finger-like protrusions has yet to be elucidated.

In isolated vertebrate cells, cytoplasmic isolation is concomitant with the physical separation of the divided cells (Guizetti et al., 2011; Steigemann et al., 2009). In SOPs, as in the Caenorhabditis elegans embryo (Green et al., 2013), cytoplasmic isolation and midbody release are temporally decoupled (Fig. 3E). A possible caveat in our study resides in the use of KAEDE (Ando et al., 2002), and we cannot exclude that a smaller probe would still be able to equilibrate after the time we have determined for cytoplasmic isolation. Nonetheless, the fact that Shrub is recruited to the midbody and regulates the time at which the KAEDE no longer diffuses indicates that cytoplasmic isolation takes place several minutes prior to midbody release. As the midbody release coincides with the disentanglement of the finger-like protrusions, we propose that the midbody is released only upon SJ disassembly. Consequently, the midbody is moved to the pIIa cell surface until detaching and moving away from the pIIa cell. Here, the midbody is still labeled with MyoIIGFP, PavGFP, MyoIIRFP and GAP43IR, suggesting that it is located in the extracellular space instead of being taken up by an adjacent epidermal cell.

A link between the regulation of cell division and fate decision was reported, and abscission is thought to regulate fate acquisition (Chaigne et al., 2020; Ettinger et al., 2011). The abscission of SOP is asymmetric with the midbody remnant residing at the pIIa plasma membrane for several minutes prior to its release outside the SOP daughter. Whether SOP midbody could signal and/or impact on proliferation, differentiation or cell fate as in mammalian cells (Ettinger et al., 2011; Kuo et al., 2011; Peterman et al., 2019; Pohl and Jentsch, 2009) awaits further monitoring of its behavior after its release.

Our results show that Shrub positively and cell-autonomously regulates abscission in SOPs, as observed in mammalian cultured cells and Drosophila germline stem cells (Eikenes et al., 2015; Matias et al., 2015). Shrub appears as two puncta recruited to the midbody long before midbody release. This is also observed in female Drosophila germline stem cells in which Shrub is recruited from G1/S prior to abscission occurring in G2 phase of the following cell cycle (Eikenes et al., 2015; Matias et al., 2015). Interestingly, recruitment of Shrub occurs around the S to G2 transition of the daughter cells and ∼1 h after the entry of SOPs into mitosis (Audibert et al., 2005). Then, the midbody release takes place just before the G2/M transition of pIIb. Thus, abscission in SOPs appears to follow a similar cell-cycle-dependent regulation as in the germline stem cells. Whether this cell-cycle dependence is specific to stem cells and progenitors or also applies to epidermal cells awaits further investigation.

A second pool of Shrub is recruited at finger-like protrusions and proposed to act in a cell-non-autonomous manner. What could be the function of this pool of Shrub? Based on the fact that Shrub regulates SJ steady state distribution and dynamics (Pannen et al., 2020), and is also known to regulate endosomal sorting, Shrub could contribute to local SJ remodeling. Alternatively, based on the geometry of the finger-like protrusions and the movement of plasma membrane concomitant to the basal displacement of the midbody, Shrub could be recruited there to promote membrane repair (Jimenez et al., 2014; Scheffer et al., 2014) and/or locally compensate for leakage in permeability barrier function. Consistent with this model, when the permeability barrier function is challenged by depletion of Cora, ShrubGFP accumulates at the finger-like protrusions.

A link between the regulation of cell division and fate decision is reported, and abscission is thought to regulate fate acquisition (Chaigne et al., 2020; Ettinger et al., 2011). The fact that abscission occurs 1 h after anaphase whereas Notch-dependent fate acquisition is initiated ∼15 min after anaphase (Bellec et al., 2018; Couturier et al., 2012) raises an intriguing question. Indeed, our photoconversion experiment with KAEDE shows that pIIa and pIIb share their cytoplasm while Notch intracellular domain (NICD; Notch) is translocated exclusively in the nucleus of pIIa. The cell fate determinants Numb and Neur are also unequally inherited by the pIIb cell long before cytoplasmic isolation. Thus, how can NICD, and Numb and Neur, remain confined to pIIa and pIIb cells, respectively, while at the same time the KAEDE probe freely diffuses bidirectionally? This situation is reminiscent of that of phospho-Mad, which remains restricted to the most anterior of Drosophila germline stem cells in stem cysts (Eikenes et al., 2015; Mathieu et al., 2013; Matias et al., 2015). These data suggest that passage must be regulated/selective at the intercellular bridge (Mullins and Biesele, 1977; Norden et al., 2006; Steigemann et al., 2009), as reported for proteins and organelles in yeast (Lengefeld and Barral, 2018).

Overall, our study sheds light on two complementary functions of Shrub in the coordination between cytoplasmic isolation and maintenance of epithelial permeability barrier functions in progenitors. Based on the apicobasal topology of mechanical and permeability barriers, apical positioning of the midbody within the TJs in vertebrates (Dubreuil et al., 2007; Higashi et al., 2016; Jinguji and Ishikawa, 1992) and the role of ESCRT components in TJ protein trafficking (Dukes et al., 2011; Raiborg and Stenmark, 2009), it is tempting to speculate that the cell-autonomous and cell-non-autonomous effects we described in this study may also be at play during epithelial abscission in vertebrates.

Key resources are listed in Table S2.

Drosophila melanogaster stocks were maintained and crossed at 25°C. Somatic clones were induced using hs-FLP with two heat shocks (60 min at 37°C) in second- and third-instar larvae. pnr-Gal4 was used to drive the expression of UAS-CoraRNAi and the photoconvertible probe UAS-KAEDE. Ay-Gal4 was used to drive the expression of UAS-ShrubGFP. sca-Gal4 was used to drive the expression of UAS-ShrubRNAi, UAS-KAEDE and UAS-ShrubHA. neur^PGAL4 was used to drive the expression of UAS-ShrubHA.

Declaration of contained use of genetically modified organisms (GMOs) of containment class no. 2898 was made to the French Ministère de l'Enseignement Supérieur, de la Recherche et de l'Innovation.

Drosophila genotypes in figures are as follows:

(B) MyoII::RFP; neur-H2B::IR/+; ATPα::GFP/+; (C,C′) neur-H2B::IR; (E) MyoII::RFP; neur-H2B::IR/+ (ECs and SOPs) and MyoII::RFP; neur-H2B::IR/+; ATPα::GFP/+ (ECs and SOPs).

(A,B) hs-FLP/MyoII::RFP; neur-H2B::IR/+; FRT82B ATPα::GFP/FRT82B nls::RFP; (C,D) MyoII::RFP;; ATPα::GFP/neur-GAP43::IR.

(A-C) UAS::KAEDE; neur-GAP43::IR, pnr-Gal4/+; (D,D′) MyoII::GFP, neur-H2B;; neur-GAP43::IR, pnr-Gal4/+.

(A,A′) MyoII::RFP; Shrub::GFP/+; neur-GAP43, pnr-Gal4::IR/+; (B,C) hs-FLP; UAS-Shrub::GFP/Ay-Gal4; MyoII-MyoII::mcherry.

(A) sca-Gal4/UAS-ShrubRNAi; UAS-KAEDE/+; (B) sca-Gal4/UAS-ShrubRNAi; UAS-KAEDE/+, sca-Gal4/UAS-ShrubRNAi; UAS-KAEDE/neur-GAP43::IR and sca-Gal4/+; UAS-KAEDE/+; (C) UAS::KAEDE; neur-GAP43::IR, pnr-Gal4/+ and UAS::KAEDE/ShrubGFP; neur-GAP43::IR, pnr-Gal4/+.

(A) UAS-KAEDE/+; neur-GAP43::IR, pnr-Gal4/+ and UAS-KAEDE/+; neur-GAP43::IR, pnr-Gal4/UAS-CoraRNAi and UAS-KAEDE/Shrub::GFP; neur-GAP43::IR, pnr-Gal4/+ and UAS-KAEDE/Shrub::GFP; neur-GAP43::IR, pnr-Gal4/UAS-CoraRNAi; (B) MyoII::GFP, neur-H2B;; neur-GAP43::IR, pnr-Gal4/UAS::CoraRNAi; (C,D) MyoII::GFP, neur-H2B;; neur-GAP43::IR, pnr-Gal4/+ and MyoII::GFP, neur-H2B;; neur-GAP43::IR, pnr-Gal4/UAS::CoraRNAi; (E-E″) MyoII::RFP;; ATPα::GFP/+; (F,F′) MyoII::RFP; ShrubGFP/+; neur-GAP43::IR, pnr-Gal4/UAS::CoraRNAi.

(A) MyoII::RFP; neur-H2B::IR/+; ATPα::GFP/+.

(A) hs-FLP/MyoII::RFP; neur-H2B::IR/+; FRT82B ATPα::GFP/FRT82B nls::RFP.

(B) UAS::KAEDE; neur-GAP43::IR, pnr-Gal4/+ and sca-Gal4/+; UAS-KAEDE/+; (C) UAS::KAEDE; neur-GAP43::IR, pnr-Gal4/+ and sca-Gal4/+; UAS-KAEDE/+; (E) MyoII::RFP/MyoII::GFP;; GAP43::IR/+; (F) MyoII::RFP;; GAP43::IR/Pav::GFP.

(A,A′) sca-Gal4/UAS-Shrub::HA; (B,B′) neur^PGAL4/UAS-Shrub::HA; (C-C″) MyoII::RFP; CHMP2BGFP/+; pnr-Gal4, neur-GAP43/+; (D,E) hs-FLP; UAS-Shrub::GFP/Ay-Gal4; MyoII-MyoII::mcherry.

(A-C) sca-Gal4/UAS-ShrubRNAi; UAS-KAEDE/+; (D) MyoII::RFP; sca-Gal4; UAS-KAEDE/+ and MyoII::RFP; sca-Gal4/UAS-ShrubRNAi; UAS-KAEDE/+; (E) w¹¹¹⁸ and Shrub::GFP/+.

(A) UAS-KAEDE/+; neur-GAP43::IR, pnr-Gal4/UAS-CoraRNAi; (B) MyoII::GFP, neur-H2B;; neur-GAP43::IR, pnr-Gal4/+ and MyoII::GFP, neur-H2B;; neur-GAP43::IR, pnr-Gal4/UAS::CoraRNAi; (C) MyoII::RFP;; ATPα::GFP/+.

Pupae aged from 16 to 24 h after puparium formation (APF) were dissected using Cannas microscissors in 1× phosphate-buffered saline (1× PBS, pH 7.4) and fixed for 15 min in 4% paraformaldehyde solution at room temperature (RT) (Gho et al., 1996). Following fixation, dissected nota were permeabilized using 0.1% Triton X-100 in 1× PBS (PBT) and incubated with primary antibodies (for details, see Table S2) diluted in PBT for 2 h at RT. After three washes for 5 min in PBT, nota were incubated with secondary antibodies diluted in PBT for 1 h at RT. Following incubation, nota were washed three times in PBT and once in PBS prior to mounting in 0.5% N-propylgallate with 1,4-diazabicyclo[2.2.2]octane (DABCO) dissolved in 1× PBS/90% glycerol.

For Drosophila germline stem cell identification, antibody staining and Hoechst staining were performed according to standard protocols. Briefly, ovaries or testes were dissected in PBS, fixed in 4% paraformaldehyde, rinsed and permeabilized in PBT (PBS-0.2% Triton X-100) for 30 min, left overnight with primary antibodies in PBT at 4°C, washed for 2 h in PBT, left with secondary antibodies in PBT for 2 h at RT, washed for 1 h in PBT and mounted in Citifluor (Eikenes et al., 2015; Matias et al., 2015).

Germline stem cells were identified with the fusome staining (round or linking the cystoblast (CB) and counted. Stem cysts (more than two cells anchored to the niche and linked by a fusome) and polyploid germline stem cells (higher DNA and larger fusome than control) were counted. Stacks were acquired every 0.7 μm.

Live imaging was performed on pupae aged for 16-22 h APF at 20-25°C (Gho et al., 1999). Pupae were stuck on a glass slide with a double-sided tape, and the brown pupal case was removed over the head and dorsal thorax using microdissection forceps. Spacers made of four to five glass coverslips were positioned at the anterior and posterior side of the pupae, respectively. A glass coverslip covered with a thin film of Voltalef 10S oil was then placed on top of the pillars such that a meniscus was formed between the dorsal thorax of the pupae and the glass coverslip. Images were acquired with a confocal microscope Leica SP5, SP8 or SPE equipped with a 63×/1.4 NA PlanApo objective and controlled by LAS AF software. Confocal sections were taken every 0.5 μm unless otherwise specified. All images were processed and assembled using ImageJ/FIJI software and Photoshop CS4.

FRAP experiments were performed in pupae expressing MyoIIRFP with ATPαGFP and neur-GAP43IR. Regions of interest corresponding to one of the two ATPα::GFP finger-like protrusions pointing to the midbody were bleached (488 nm at 100% laser, one iteration of 100 ms) using a LSM Leica SP8 equipped with a 63×/1.4 NA PlanApo objective. Confocal stacks were acquired every 2 min and 30 s after photobleaching.

Photoconversion assays were performed in pupae expressing the green to red photoconvertible probe KAEDE. KAEDE was photoconverted (405 nm laser at 0.5-2% power, point bleach, one to two iterations of 300 ms each) using a LSM Leica SP8 equipped with a 63×/1.4 NA PlanApo objective. Confocal stacks were acquired every 2 min after photoconversion and imaged at 22-25°C.

Photoablation experiments were performed on 16 h APF live pupae using an Airyscan Zeiss LSM800 confocal microscope equipped with a 63×/1.4 NA PlanApo objective. Ablation was carried out on pIIa/pIIb cell finger-like protrusions at the SJ level with a two-photon laser-type Mai-Tai HP from Spectra Physics set to 800 nm and a laser power of 2.9 W. Timing of ablation was set between 35 and 45 min after anaphase onset. Ablation parameters were laser trans 40% and two iterations of 1.89 s.

For each FRAP experiment, three measurements were performed: the photobleached junction, the control junction (the finger-like protrusion opposite to the finger-like protrusion after FRAP) and the background.

All information concerning the statistical tests are provided in the main text and in the figure legends, including the number of samples analyzed in each experiment. GraphPad Prism 8 software or R 4.2.1 (https://cran.rproject.org/bin/windows/base/old/4.2.1/) was used to perform the analyses.

Line plots use the following standards: thick lines indicate the means, bar plots represent the means, and errors bars represent s.d.

Shapiro–Wilk normality test was used to confirm the normality of the data, and the F-test was used to verify the equality of s.d. The statistical differences of Gaussian datasets were analyzed using unpaired two-tailed Student's t-test.

For the analysis of the midbody displacement, we performed an ANCOVA to test the effect of the interaction between time (between 5 and 120 min) and the ‘ condition ’ parameter (SOP versus EC). For the midbody displacement and FRAP, the colored areas show s.d.

Time to cytoplasmic isolation was modeled via logistic regression, using the GLM function in R, with a binomial error distribution and a logistic link function. We began with a fully parameterized model in which cytoplasmic isolation varied as a function of time, condition, and the interaction between time and condition (i.e. both the slopes and intercepts of the model describing temporal effects differed among condition groups). Statistical significance of each model term was evaluated via analysis of deviance, with forward model selection of terms and goodness-of-fit assessed against a Chi-squared distribution. Additionally, we ran two additional nested models: one excluding the interaction term (i.e. a common model slope with differences among condition groups in their intercept terms; equivalent to differences in the means amongst groups), and a second in which cytoplasmic isolation state varied only as a function of time (i.e. no effective differences amongst condition groups). We compared overall fit for each model using Akaike's information criterion (AIC), selecting the parsimony model as the one with the lowest AIC score. Results of analysis of deviance and model comparisons corroborated each other, and so the parsimony model was used to predict the mean percentage of cells demonstrating cytoplasmic isolation, as well as standard errors of the estimates, and to visualize differences among groups. In some lines, the interaction between time and treatment was not significant; however, the interaction between sca> and pnr> was slightly significant (Table S1). These results were corroborated by model comparison via AIC, with the selection process favoring models including only time and treatment effects, i.e. although proportions of cells in cytoplasmic isolation differed among conditions at a given time point, the rate of relative increase over time did not differ among groups. This lack of significant time-treatment interaction facilitated the comparison of treatment effect size. To do so, we used model results to estimate the time at which 50% of the pIIa/pIIb cell indicated cytoplasmic isolation (herein termed t1/2). Each t1/2 was estimated using the two models [(sca> and sca>ShrubRNAi) and (pnr>, pnr>+ShrubGFP/+, pnr>+CoraRNAi, pnr>CoraRNAi+ShrubGFP/+)]. Statistical significance was represented as follows: not significant (ns), P>0.05; *P≤0.05; **P≤0.01; and ***P≤0.001.

We thank the Bloomington Drosophila Stock Center (Indiana University, USA) the Vienna Drosophila RNAi Center and the National Institute of Genetics Fly Stock Center for providing fly stocks. We also thank the Microscopy Rennes Imaging Center-BIOSIT (France). The monoclonal antibodies against Elav, Cut, Cora, E-Cad and α-Spectrin were obtained from the Developmental Studies Hybridoma Bank, generated under the auspices of the National Institute of Child Health and Human Development, and maintained by the University of Iowa Department of Biological Sciences. We thank Dr Echard and Marta Mira Osuna for critical reading of the manuscript, and Dr McCairns for help with the GLMs.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201409.reviewer-comments.pdf