Endophilin-A2 functions in membrane scission in clathrin-independent endocytosis

Abstract

During endocytosis, energy is invested to narrow the necks of cargo-containing plasma membrane invaginations to radii at which the opposing segments spontaneously coalesce, thereby leading to the detachment by scission of endocytic uptake carriers. In the clathrin pathway, dynamin uses mechanical energy from GTP hydrolysis to this effect, assisted by the BIN/amphiphysin/Rvs (BAR) domain-containing protein endophilin. Clathrin-independent endocytic events are often less reliant on dynamin, and whether in these cases BAR domain proteins such as endophilin contribute to scission has remained unexplored. Here we show, in human and other mammalian cell lines, that endophilin-A2 (endoA2) specifically and functionally associates with very early uptake structures that are induced by the bacterial Shiga and cholera toxins, which are both clathrin-independent endocytic cargoes. In controlled in vitro systems, endoA2 reshapes membranes before scission. Furthermore, we demonstrate that endoA2, dynamin and actin contribute in parallel to the scission of Shiga-toxin-induced tubules. Our results establish a novel function of endoA2 in clathrin-independent endocytosis. They document that distinct scission factors operate in an additive manner, and predict that specificity within a given uptake process arises from defined combinations of universal modules. Our findings highlight a previously unnoticed link between membrane scaffolding by endoA2 and pulling-force-driven dynamic scission.

Extended Data Figure 1. Screening of BAR domain protein library

a, Listing of BAR domain proteins that were tested in our localization screen. Briefly, 24 hours after transfection of HeLa cells with fluorescent protein-tagged constructs, cells were ATP-depleted for 20 min, incubated for 20 min with 1 µM STxB-Cy3 or STxB-A488, fixed at 37°C, mounted, and viewed by confocal microscopy. The initial phenotype that was scored was BAR domain protein localization on STxB-induced plasma membrane invaginations. TOCA1, TOCA3, and amphiphysin-2 (yellow underlay) colocalized with STxB on tubular structures (not shown). See below for endoA2 (green underlay). b, At variance with the expected phenotype, expression of GFP-tagged endoA2 (green) led to the disappearance of long STxB-induced plasma membrane invaginations (red). Tubule length was quantified in non transfected control cells (n=50), and endoA2-GFP expressing cells (n=59). Quantifications show means ± SEM of 2 independent experiments. *** p<0.001, two-tailed Mann-Whitney U test. c, STxB-Cy3 (50 nM, red) was incubated for 45 min at 37°C with cells expressing GFP (i), endoA2-GFP (ii), or Rab5-Q79L-GFP (iii). Expression of endoA2-GFP did not affect STxB trafficking to the Golgi/TGN membranes. n=48 for GFP expressing cells, n=43 for endoA2-GFP expressing cells, and n=46 for Rab5-Q79L–GFP expressing cells (2 independent experiments). Quantifications show means ± SEM. *** p<0.001, ns = non significant, Bonferroni’s multiple comparison test. Scale bars: b,c=10 µm.

Extended Data Figure 2. Intracellular localization analysis of endoA2 and STxB

a–c, STxB is internalized in endoA2-GFP positive vesicles and induces the recruitment of endoA2-GFP to the plasma membrane. HeLaM cells stably expressing endoA2-GFP-FKBP (a,c), or HeLa cells transiently expressing endoA2-GFP (b) were incubated continuously at 37°C respectively with 0.2 µM STxB-Cy5 or 0.5 µM STxB-Cy3. Observation by live cell imaging using a spinning disk microscope (a,c) or TIRFM (b). a, STxB endocytosis. Image series of 170 sec (for the complete sequence, see Supplementary information, Video V1). Arrows show the formation in the cell periphery of STxB-Cy5 and endoA2-GFP positive vesicles that move into cells. b, Kymographs of HeLa cells transiently expressing endoA2-GFP in the absence (−) or presence (+) of STxB-Cy3. Quantification from TIRFM recordings of endoA2-GFP spot lifetime (−STxB, n=1768 events; +STxB, n=144 events; 3 independent experiments; *** p<0.001, two-tailed Mann-Whitney U test). c, Plasma membrane recruitment of endoA2. STxB-Cy5 was added 15 sec after the beginning of image acquisition. Two time points are shown (for the complete sequence, see Supplementary information, Video V2): before STxB addition (− STxB, 0 sec) and after STxB addition (+ STxB, + 80 sec). A striking recruitment of endoA2-GFP to the plasma membrane was observed (white arrowheads). Fluorescence intensities along plasma membrane segments of endoA2-GFP and STxB-Cy5 were followed over time (means ± SEM, n=6 cells, 3 independent experiments). Note that both rose in a similar manner. d–i, EndoA2 and Shiga toxin poorly colocalize with markers of clathrin-mediated endocytosis. All images show live cell TIRF microscopy recordings. d, HeLa cells were transfected with plasmids expressing endoA2-GFP and µ2-mCherry. The overlap between both markers was very small. e, The genome-edited cell line SK-MEL-2 expressing LCa-mRFP was transfected transiently with endoA2-GFP plasmid. Again, both markers showed very little overlap. f, HeLa cells were transfected with plasmids expressing endoA2-GFP and DNM2-mRFP. A substantial overlap was observed between both markers. For d–f, quantifications are reported in Fig. 1d. g–i, HeLa cells transiently expressing mRFP-LCa (g) or µ2 -mCherry (h), and the genome-edited cell line SK-MEL-2 expressing LCa-mRFP (i) were continuously incubated with 0.5 µM STxB-A488 for 5 min at 37°C. Note the weak overlap between STxB and the other markers. j, Quantification of colocalization for g–i. Means ± SEM of the following numbers of cells: mRFP-Lca, n=11; µ2-mCherry, n=6; LCa-mRFP, n=19. 2 independent experiments. *** p<0.001, Bonferroni’s multiple comparison test. Scale bars: a,c-i=2 µm.

Extended Data Figure 3. Effect of endoA2 depletion on Shiga toxin endocytosis at early times of uptake

a, After binding on ice for 30 min, STxB-Cy3 (50 nM, red) was incubated for 5 min at 37°C with HeLa cells that had been transfected with negative control siRNAs (18 images), or siRNAs against endoA2 (23 images). After fixation, Golgi membranes were labeled with antibodies against giantin (green). Note the presence of short STxB-containing tubules (arrowheads) in the endoA2 depletion condition. The Western blot documents the efficiency of endoA2 depletion. Clathin heavy chain (CHC) was used as a loading control. b–c, Characterization of STxB conjugates with monofunctionalized nanogold. b, HeLa cells were incubated for 45 min at 37°C in the indicated conditions, fixed, treated for silver enhancement, and viewed by transmission light microscopy. Note the strong perinuclear signal (arrowheads) that was visible only when cells were incubated with STxB-nanogold (113 images). c, HeLa cells were incubated for 45 min at 37°C with STxB-nanogold, fixed, treated for silver enhancement, and viewed by electron microscopy after sectioning (32 images). Note the strong signal in the Golgi region (G), which validated the functionality of the conjugate. Some STxB-nanogold could still be seen at the plasma membrane (PM) and in endosomes (E). N, nucleus. d, HeLa cells were transfected with negative control siRNAs, or siRNAs against endoA2, incubated for 30 min on ice with nanogold-coupled STxB, shifted for 5 min to 37°C, and then fixed. Silver enhancement was used to enlarge nanogold particles. Note that STxB-containing invaginations (arrowheads) were much longer in endoA2-depleted cells than in negative control siRNA-transfected cells. Quantification in shown in Fig. 2b (see legend for cell numbers). Scale bars: a,b=10 µm, c=200 nm, d=100 nm.

Extended Data Figure 4. Effect of endoA2 depletion on STxB and CTxB uptake, and rescue with endoA2 mutants

a, Efficiency of endoA2 depletion with different siRNA sequences. HeLa cells were transfected with 3 different siRNA sequences against human endoA2 (#5, #7 and #8). The efficiency of endoA2 depletion was monitored by Western blotting with antibodies against endoA2. Negative control siRNA-transfected cells were used for comparison. The Western signal for clathrin heavy chain (CHC) served as loading control. Quantification shows means ± SEM (n=3 independent experiments). ** p<0.01, *** p<0.001, ns = non significant (Bonferroni’s multiple comparison test). Trafficking studies on these cells are reported in Fig. 2c. b, Effect of endoA2 depletion on STxB binding to cells. Hela cells were transfected with siRNAs as in a, detached, incubated for 30 min on ice with 1 µM STxB-A488 and 10 µg/ml Tf-A647, washed, and analyzed by FACS. Cells treated with the glycosylceramide synthase inhibitor PPMP were used as a control for signal specificity. Means ± SEM of 3 independent experiments are shown (except for PPMP which was done twice). The increased STxB binding with siRNA #8 is not explained at this stage. These cell surface binding data serve as controls for the sulfation experiment of Fig. 2c. c, EndoA2 functions in CTxB uptake. CTxB (5 nM) uptake assay (n=3 independent experiments) after 10 min at 37°C in conditions of endoA2 depletion (* p<0.05, two-tailed t-test). As opposed to STxB, CTxB could be removed from the plasma membrane by acid washes, and endocytosis could therefore be measured directly using a plate reader assay (see Extended Methods in Supplementary information). d, STxB trafficking rescue experiment under endoA2 depletion conditions. HeLa cells were transfected with negative control siRNAs, or siRNAs against endoA2. After 48 hours, endoA2-depleted cells were transfected for 24 hours with siRNA-resistant expression vectors coding for: endoA2ΔH0-GFP (ΔH0), endoA2ΔSH3-GFP (ΔSH3), or full-length wild-type endoA2-GFP (FL). STxB-Cy3 (50 nM, red) was incubated with these cells for 45 min at 37°C (for quantification and statistical data, see Fig. 2d). The H0 helix deletion mutant did not rescue STxB transport to perinuclear Golgi/TGN membranes, as opposed to wild-type endoA2 and endoA2ΔSH3. The Western blot documents the efficiency of endoA2 depletion in siRNA-transfected cells. Clathin heavy chain (CHC) was used as a loading control. e, Effect of endoA2 mutants on STxB-induced membrane invaginations. Non transfected HeLa cells (Ctrl) or cells expressing GFP-tagged full-length wild-type endoA2 (FL), H0 helix deletion mutant (ΔH0), or SH3 domain deletion mutant (ΔSH3) were ATP-depleted and incubated for 10 min at 37°C with 0.5 µM STxB-Cy3. Tubule length was quantified (Ctrl, n=277 in 102 cells; FL, n=90 in 15 cells; ΔH0, n=183 in 48 cells; ΔSH3, n=164 in 36 cells). 2 independent experiments. Quantifications show means ± SEM. *** p<0.001, ns = non significant, Bonferroni’s multiple comparison test. Scale bars: d,e=10 µm.

Extended Data Figure 5. Intracellular trafficking analysis

a–c, EndoA2 depletion does not affect general trafficking processes. All experiments were performed on negative control siRNA transfected cells, and on cells that were depleted for endoA2. a, Transferrin uptake (left, n=3) and recycling (right, n=4). EndoA2 depletion did not affect any of these processes. Quantifications show means ± SEM of the indicated numbers of independent experiments. b, Steady-state localization of TGN46 and CI-MPR (red), as analyzed by immunofluorescence. Golgi membranes were labeled with antibodies against giantin (green). EndoA2 depletion did not affect the steady-state localization of these markers (TGN46 in siCtrl or siEndoA2 cells: 12 images; CI-MPR in siCtrl or siEndoA2 cells: 10 images; 2 independent experiments). c, Anterograde trafficking of E-cadherin. After release from endoplasmic reticulum, SBP-mCherry-E-cadherin protein was detected at the cell surface with anti-mCherry antibodies (0, 20 and 60 min time points). After 60 min, the relative means (± SEM) of anti-mCherry fluorescence per unit area were quantified for control (6 images, 87 cells) and endoA2-depleted cells (6 images, 81 cells). 3 independent experiments. No significant difference was observed using a two-tailed t-test (p>0.05). d–f, Depletion of α-adaptin does not affect Shiga toxin trafficking. All experiments were performed on negative control siRNA transfected cells, and on cells that were depleted for the indicated proteins. Quantifications show means ± SEM. d, Sulfation analysis of retrograde STxB transport (3 independent experiments). HeLa cells in the indicated conditions were incubated for 20 min at 37°C with STxB-Sulf₂ in the presence of radioactive sulfate, and sulfated STxB-Sulf₂ was measured by autoradiography. Note that α-adaptin depletion did not affect sulfation of STxB-Sulf₂, while depletion of endoA2 or syntaxin-16 (STX16) had a strong effect. ** p<0.01, ns = non significant (Bonferroni’s multiple comparison test). e, Immunofluorescence analysis. HeLa cells in the indicated conditions were incubated for 45 min at 37°C with 0.05 µM STxB-Cy3 (red). During the last 10 min, 10 µg/ml Tf-A488 (green) were added in the growth medium. Cells were placed on ice, and cell surface exposed Tf was removed by acid washes. After fixation, cells were labeled for giantin (blue). Note that α-adaptin depletion strongly inhibited Tf uptake, but not retrograde transport of STxB to TGN/Golgi membranes. Tf uptake was quantified for control (5 images, 102 cells) and α-adaptin-depleted cells (5 images, 108 cells). 2 independent experiments. *** p<0.001 (two-tailed t-test). f, siRNA-mediated depletion of endoA2 and of syntaxin-16 was analyzed by Western blotting. Clathrin heavy chain (CHC) was used as loading control. Scale bars: b,c,e=10 µm.

Extended Data Figure 6. Cell and model membrane experiments

a, Knocksideways. HeLaM cells stably expressing Mito-YFP-FRB and rat endoA2-GFP-FKBP (green) were transfected with negative control siRNAs (siCtrl), or siRNAs against human endoA2 (siEndoA2) that did not cross with the rat sequence. STxB-Cy3 (0.5 µM, red) was incubated for 15 min at 37°C with ATP-depleted cells. The cells were then fixed at 37°C, and viewed by confocal microscopy. Quantification of tubule formation and cell numbers are shown in Fig. 3a. Note that STxB-induced tubule length reversibly increased after endoA2-GFP-FKBP sequestration. The depletion of endogenous human endoA2, and the expression of GFP-FKBP-tagged rat endoA2 were assessed by Western blotting with anti-endoA2 and anti-GFP antibodies, respectively. Western blotting against clathrin heavy chain (CHC) was used as loading control. b–c, Interfering with microtubules or dynein motors strongly affects STxB-induced tubule length. b, HeLa cells were incubated for 1 hour at 37°C with DMSO (+ DMSO) or 10 µM nocodazole (+ nocodazole), ATP-depleted for 20 min, and then incubated for 10 min at 37°C with 0.5 µM STxB-Cy3 (red) in the same conditions. Labeling with an antibody against a-tubulin (green) was used to visualize the efficiency of nocodazole treatment. Long tubular structures containing STxB could not be detected after incubation with nocodazole (−ATP/+DMSO: 18 images; -ATP/+Nocodazole: 16 images; 3 independent experiments). c, Heavy chains of cytoplasmic dyneins (DYNC1H1 and DYNC2H1) were depleted from HeLa cells with siRNAs. Cells were ATP-depleted and then incubated for 10 min at 37°C with 0.5 µM STxB-Cy3 (red). The presence of long STxB-induced tubules was strongly decreased under these conditions. Tubule length was quantified for negative control siRNA treated cells (siCtrl, n=188) and for DYNC2H1 depleted cells (siDYNC2H1, n=165). 2 independent experiments. Quantifications show means ± SEM. *** p<0.001, two-tailed Mann-Whitney U test. The Western blot documents the efficiency of DYNC1H1 depletion. a-tubulin was used as a loading control. d–e, Model membranes experiments. d, Measurement of tube pulling force over time in the absence (− endoA2ΔH0) or presence of endoA2ΔH0 mutant (+ endoA2ΔH0, 7 µM in injection pipette). e, Scission experiments with tethers that were coated with endoA2ΔH0. Scale bars: a,c=10 µm, b,e=5 µm.

Extended Data Figure 7. Electron microscopy of STxB-nanogold on ATP-depleted cells following treatment with nocodazole

HeLa cells were treated for 1 hour at 37°C with DMSO (top panel) or 10 µM nocodazole (bottom panel). During the last 20 min, ATP was depleted. Cells were then incubated for 10 min with STxB-nanogold in the continued presence of inhibitors, fixed, and prepared for electron microscopy. In the DMSO condition (top panel), long tubular structures connected to cells surface were observed (magnified views in right insets, red arrowheads in C), as expected from light microscopy experiments. These structures were in close proximity with microtubules, as indicated with blue arrowheads in magnifications A and B. In the nocodazole condition (lower panel), STxB-induced plasma membrane invaginations were still present (arrowheads), but much shorter (mean length of 118.2 ± 7.0 nm, n=109 invaginations; 0.90 ± 0.12 invaginations/µm of plasma membrane, n=28 images; 3 independent experiments) than in the absence of the compound. Scale bar sizes are indicated on each micrograph.

Extended Data Figure 8. Actin and endoA2

a–c, EndoA2 codistribution with actin. a, HeLa cells transiently co-expressing endoA2-GFP and mCherry-LifeAct were observed by time-resolved TIRF microscopy. The panel shows acquisitions at the plasma membrane that were taken at 5 sec intervals. Arrowheads point out examples of structures on which endoA2 and actin colocalize in a dynamic manner. Quantification of colocalization of endoA2-positive structures with LifeAct is presented as mean ± SEM (n=7 cells, 2 independent experiments). b, EndoA2-depleted HeLa cells were incubated continuously for 5 min at 37°C with 0.5 µM STxB-Cy3. After fixation, actin filaments were stained with Phalloidin-FITC. Arrowheads indicate STxB-induced tubules that are decorated by actin. c, HeLa cells transfected with negative control or endoA2 siRNAs (siCtrl or siEndoA2, respectively) and transiently expressing mCherry-LifeAct were observed by TIRF microscopy after addition of 0.5 µM STxB-Cy3 at 37°C. Arrowheads point out examples of structures on which STxB and actin colocalize. Quantification of colocalization of STxB-positive structures with LifeAct is presented as means ± SEM (n=6 cells, 2 independent experiments). ns = non significant (two-tailed t-test). d–e, Analysis of STxB-induced plasma membrane invaginations in function of endoA2 depletion and/or actin perturbation. d, HeLa cells were transfected with negative control siRNAs or with siRNAs against endoA2, and treated or not with 0.5 µM latrunculin-A. The cells were then incubated continuously for 5 min at 37°C in the presence of 0.5 µM STxB-Cy3 (red), fixed, and labeled with phalloïdin (green). The quantification of STxB tubule length is shown in Fig. 4a. Note that tubule length increased with combined treatments (siCtrl+DMSO: 22 images; siEndoA2+DMSO: 11 images; siCtrl+LatA: 14 images; siEndoA2+LatA: 14 images). e, Depletion of endoA2 was analyzed by Western blotting. Clathrin heavy chain (CHC) was used as loading control. f, Efficiency of ARPC2 depletion. HeLa cells were transfected for 72 hours with a smartpool of 4 siRNA sequences against ARPC2. The efficiency of ARPC2 depletion was monitored by Western blotting with antibodies against ARPC2. The Western signal for clathrin heavy chain (CHC) served as loading control. Corresponds to experiments in Fig. 4a. Scale bars: a,c=2 µm, b=5 µm, d=10 µm.

Extended Figure 9. Combined effects of interference with endoA2, dynamin and actin on STxB-induced membrane invaginations

a–c, Endogenous endoA2 and dynamin are found on STxB-induced plasma membrane invaginations. a, Dynamin-2 was depleted from cells (siDNM2), which were then incubated continuously for 5 min at 37°C in the presence or absence of 0.5 µM STxB-Cy3 (red), fixed, labeled for the indicated markers, and analyzed by confocal microscopy. Note that endoA2-containing tubules (green) were seen only in the presence of STxB. No overlap was observed with clathrin (blue). –STxB: representative of 15 images; +STxB: representative of 35 images; 2 independent experiments. b, Experiment as in (a) on wild-type cells that were analyzed by wide field microscopy. As above, endoA2 (green) was found on tubular structures only in the presence of STxB (red). The quantification shows means ± SEM of 3 independent experiments on 234 cells without STxB (− STxB) and 921 cells with STxB (+ STxB). ** p<0.01, two-tailed t-test. c, Negative control siRNA transfected HeLa cells and endoA2-depleted cells were incubated for 5 min at 37°C in the presence of 0.5 µM STxB-Cy3 (red). Endogenous endoA2 (blue) and dynamin (green, arrowheads) were labeled with specific antibodies, detected by immunofluorescence, and viewed by structured illumination microscopy. Note that dynamin localized in spots on STxB-induced invaginations, while endoA2 distributed in a continuous manner. d, Analysis of STxB-induced plasma membrane invaginations in function of endoA2 and dynamin-2 depletion, and actin perturbation. HeLa cells were depleted for dynamin-2 alone, dynamin-2 in combination with endoA2, or both depletions in combination with 0.25 µM latrunculin-A treatment, as indicated. These cells were then incubated continuously for 5 min at 37°C with 0.5 µM STxB-Cy3 (red), fixed at 37°C, and viewed by confocal microscopy. Note that the tubulation phenotype increased with each additional interference modality (siDNM2: 21 images; siEndoA2+siDNM2: 15 images; siEndoA2+siDNM2+LatA: 16 images; 2 independent experiments). The quantification of tubule length in the different experimental conditions of this figure is shown in Fig. 4a. The depletion of dynamin-2 and endoA2 was validated by immunoblotting. Clathrin heavy chain (CHC) was used as loading control. Scale bars: a,b=5 µm, c=0.5 µm, d=10 µm.

Extended Data Figure 10. Intoxication curves

a–d, HeLa cells were depleted for endoA2 (a), dynamin-2 (b), incubated with 0.25 µM latrunculin-A (c), or submitted concomitantly to all three perturbations (d). These cells were then further incubated for 1 hour in the presence of increasing concentrations of STx-1, at the end of which protein biosynthesis was measured. Note that only the triple treatment condition had a strong effect on cell intoxication. The protection factors determined on 4–13 independent experiments are shown in Fig. 4c. Error bars show the SEM.

Figure 1. EndoA2 localization to endocytic pathways

Quantifications show means ± SEM. All conditions: incubation for 5 min at 37°C (except if stated otherwise). a, BSC-1 cells with 50 nM STxB-Cy3 (n=20 cells) or 5 nM CTxB-Alexa555 (n=50 cells) for 3 min at 37°C, and labeling for endoA2 (3 independent experiments). b, Cryoelectron microscopy on HeLa cells incubated with 0.5 µM STxB: STxB 15 nm, endoA2 (arrows) 10 nm (representative of 50 images). c, HeLa cells transiently expressing endoA2-GFP incubated with 0.5 µM STxB-Cy3 and analyzed by TIRFM (n=25 cells, 3 independent experiments). d, Colocalization analysis by TIRFM of endoA2 with the indicated markers (transient expression in HeLa cells: DNM2-mRFP, n=10; mRFP-LCa, n=8; µ2-mCherry, n=12; genome-edited SK-MEL-2: LCa-mRFP, n=8; 2 independent experiments; *** p<0.001, Bonferroni’s multiple comparison test). Scale bars: a,c,d=2 µm, b=100 nm.

Figure 2. EndoA2 functions in Shiga toxin uptake

Quantifications show means ± SEM. a, 5 min incubation at 37°C of control (siCtrl) or endoA2-depleted (siEndoA2) HeLa cells with nanogold-coupled STxB (arrowheads indicate invaginations; representative images of numbers of cells as shown in b). b, Frequency (siCtrl, n=25 cells; siEndoA2, n=27 cells) and length (siCtrl, n=149 tubules; siEndoA2, n=138 tubules) of STxB-containing invaginations on experiments as in (a) (ns = non significant, *** p<0.001, two-tailed Mann-Whitney U test). c, Sulfation analysis (n=3 independent experiments) on HeLa cells (ns = non significant, * p<0.05, ** p<0.01, *** p<0.001, Bonferroni’s multiple comparison test). d, Rescue experiment on endoA2-depleted HeLa cells. Incubation for 45 min at 37°C with 50 nM STxB-Cy3. One a.u. corresponds to 78.7 ± 1.2% of STxB in Golgi area. Numbers of cells: siCtrl, n=145; siEndoA2, n=190; siEndoA2+EA2ΔH0, n=46; siEndoA2+EA2ΔSH3, n=31; siEndoA2+EA2 FL, n=56; 2 independent experiments; ns = non significant, *** p<0.001, Bonferroni’s multiple comparison test. Scale bars: a=100 nm, d=10 µm.

Figure 3. Model membrane experiments

a, Knocksideways on ATP-depleted HeLaM cells in control (siCtrl) or endoA2-depletion conditions (siEndoA2). Incubations for 15 min at 37°C with STxB (0.5 µM). Means ± SEM of the following numbers of tubules: siCtrl, n=65 (30 cells); –sequestration, n=101 (52 cells); +sequestration, n=313 (131 cells); release, n=626 (187 cells). 2 independent experiments. *** p<0.001, Dunn’s multiple comparison test. b, Inverse emulsion technique in conditions i to iv, as indicated. Number of vesicles: i, n=46; ii, n=39; iii, n=17; iv, n=32; 3 independent experiments. Arrows point to tubular structures. c–e Nanotube tethers. c, Tube retraction force in presence (1 µM ) or absence of endoA2. d,e, Measurement of retraction force over time upon stepwise elongation at 0.5 µm/s of an endoA2-scaffolded tube (d), and pulling force-driven scission (e). Scale bars: b=10 µm, c,e=5 µm.

Figure 4. Additive effects of scission factors

Quantifications show means ± SEM. a, Incubation of HeLa cells for 5 min at 37°C with STxB-Cy3 (0.5 µM). Determination of tube length on fixed cells (number of tubules per condition: A, n=254; B, n=79; C, n=94; D, n=146; E, n=118; F, n=338; G, n=131; H, n=115; 2 independent experiments; ns = non significant, * p<0.05, *** p<0.001, Dunn’s multiple comparison test). b, Frequency distribution of tubules according to length (conditions as in a). c, Intoxication analysis. Protection factors on 4 to 13 independent experiments per condition (*** p<0.001, Bonferroni’s multiple comparison test).