Research Article

ESCRT Machinery Is Required for Plasma Membrane Repair

See allHide authors and affiliations

Science  28 Feb 2014:
Vol. 343, Issue 6174, 1247136
DOI: 10.1126/science.1247136

Structured Abstract

Introduction

Plasma membrane damage can result from numerous threats, including mechanical stress or biochemical agents such as pore-forming toxins. Different mechanisms for plasma membrane repair have been described in a variety of cellular models, including patching with endomembranes, endocytosis, and extracellular budding. We found that the endosomal sorting complex required for transport (ESCRT), which is implicated in numerous membrane fission events (such as during cytokinesis or for the budding of several viruses) was also required for the rapid closure of small wounds made at the plasma membrane.

Embedded Image

ESCRT recruitment mediates pinching out of wounded plasma membrane. (A) Cells expressing the ESCRT subunit CHMP4B-EGFP and wounded (arrow) in the presence of propidium iodide (PI) were observed by means of fluorescence imaging. (B) Model for ESCRT-mediated detection and shedding of wounded plasma membrane.

Methods

We used micropipettes, detergents, pore-forming toxins, and laser wounding to damage the plasma membrane of mammalian cells in tissue culture. Ultraviolet or two-photon lasers were used to induce small, localized wounds, and cell reactions were followed with time-lapse imaging. Propidium iodide (PI) entry in wounded cells was used to allow imaging of the plasma membrane opening and to quantify the rate of closure of single wounds. Mathematical fit of PI entry kinetics was used to estimate the diameter and the rate of closure of individual wounds. Characterization of PI fluorescence and diffusion gave us an estimation of wound sizes. Transfection of small interfering RNA or dominant-negative mutants of ESCRT subunits allowed us to assess their importance during plasma membrane repair. Last, using correlative-scanning electron microscopy we examined the ultrastructure of wounded plasma membranes.

Results

The various wounding methods used here revealed a systematic recruitment of ESCRTs to the plasma membrane. Wounding with a laser beam showed that ESCRTs—and in particular, ESCRT-III proteins—were specifically recruited to wound sites and were accumulated until wound closure. This recruitment depended on calcium, which is known to be a crucial signaling molecule for wound repair. The depletion of important ESCRT subunits such as CHMP4B, CHMP2A, or Vps4 was deleterious for a subpopulation of cells bearing small wounds (less than 100 nm in diameter). Correlative scanning electron microscopy and time-lapse imaging revealed that wounding was followed by ESCRT-positive membrane budding and shedding. Energy depletion did not prevent—and rather increased—ESCRT accumulation but prevented both membrane shedding and repair.

Discussion

These results show that ESCRT proteins play an important role in the detection and removal through the extracellular shedding of small wounds present at the plasma membrane. We propose that different mechanisms for membrane repair (patching, budding, or endocytosis) can be used by cells depending on the type and size of the wound. These mechanisms are stimulated by common early signaling events, such as calcium, but downstream events are likely to depend on the physiochemical characteristics of the wounds.

ESCRT-positive plasma membrane shedding has been observed in a variety of normal and pathological conditions. It remains unclear whether these phenomena are linked to local plasma membrane damage and whether ESCRT-III proteins are involved in these processes.

ESCRT Your Wound Away

The ESCRT (endosomal sorting complex required for transport) protein complex plays a role in budding into multivesicular bodies, in cytokinesis, and in HIV budding. Now, Jimenez et al. (p. 10.1126/science.1247136, published online 30 January) propose a role for ESCRT proteins in wound repair at the plasma membrane. In vivo imaging, modeling, and electron microscopy were used to reveal how the ESCRTs participate in a rapid energy-independent, calcium-dependent, membrane-shedding process at the plasma membrane that reseals small wounds caused by toxins or laser treatment.

Abstract

Plasma membrane damage can be triggered by numerous phenomena, and efficient repair is essential for cell survival. Endocytosis, membrane patching, or extracellular budding can be used for plasma membrane repair. We found that endosomal sorting complex required for transport (ESCRT), involved previously in membrane budding and fission, plays a critical role in plasma membrane repair. ESCRT proteins were recruited within seconds to plasma membrane wounds. Quantitative analysis of wound closure kinetics coupled to mathematical modeling suggested that ESCRTs are involved in the repair of small wounds. Real-time imaging and correlative scanning electron microscopy (SEM) identified extracellular buds and shedding at the site of ESCRT recruitment. Thus, the repair of certain wounds is ensured by ESCRT-mediated extracellular shedding of wounded portions.

Endosomal sorting complex required for transport (ESCRT) has been implicated in multivesicular body (MVB) biogenesis, viral budding, cytokinesis, and spontaneous budding of the plasma membrane (1). They are involved in local membrane deformation and scission. The topology of the bent membrane is very characteristic, with its concave side facing the cytoplasm. ESCRT subunits are classified in five complexes: ESCRT-0, ESCRT-I, ESCRT-II, ESCRT-III, and ESCRT disassembly subcomplex, which are recruited in a serial fashion during MVB biogenesis (24). ESCRT-III and ESCRT disassembly proteins appear to be involved in every ESCRT-associated function. ESCRT-0 and ESCRT-II have not been implicated in cytokinesis and are dispensable during viral budding (5, 6). It is likely that adaptor proteins such as ALG-2–interacting protein X (ALIX) can fulfill the role of ESCRT-II for the recruitment of the ESCRT-III machinery (5, 714).

Extracellular budding events have been described during cellular response to plasma membrane wounding by pore-forming toxins (15, 16). Because the topology of these budded membranes was reminiscent of ESCRT-dependent membrane deformation, we decided to explore the role of ESCRTs in plasma membrane repair.

We first analyzed the behavior of the ESCRT-III complex upon localized plasma membrane damage. We followed the dynamics of CHMP4B, an ESCRT-III protein necessary for all known ESCRT functions in mammals. The plasma membrane of HeLa cells was wounded using three different means. In the first assay, we incubated HeLa cells expressing CHMP4B-EGFP (enhanced green fluorescent protein) at endogenous levels (17), with two pore-forming molecules (digitonin at 250 μM or saponin at 0.05%) added to the cells for a short time. In both cases, we observed a redistribution of cytosolic CHMP4B-EGFP to localized spots at the cell membrane less than 3 min after the addition of the pore-forming molecules (Fig. 1A). Similar results were obtained with the pore-forming toxins streptolysin O and listeriolysin O (Fig. 1B). Second, we mechanically damaged the plasma membrane using a micropipette in a standard microinjection setup. Similarly, CHMP4B-EGFP was recruited to the wounded site in HeLa cells in about 25% of the cells analyzed (n = 20) (Fig. 1C). Finally, to standardize wounding conditions and to couple localized plasma membrane damage to fast imaging of the cellular response, we performed laser-based plasma membrane wounding. We used either a scanning confocal microscope equipped with a two-photon laser or a spinning disc confocal microscope equipped with an ultraviolet (UV) laser. These experiments revealed a very fast (less than 30 s) and localized recruitment of CHMP4B-EGFP to wounded domains (Figs. 1D and 2, A and C). CHMP4B was found on the damaged plasma membrane colocalizing with the plasma membrane marker pMyrPalmEGFP (Fig. 1D). ESCRT-III subunits have been observed before at the plasma membrane when overexpressed or when lacking autoinhibitory domains (18, 19) and during viral budding or uncharacterized nonviral budding events (1). Certain ESCRT-III subunits, such as CHMP3, bind with high affinity to phosphatidylinositol 3,4-bisphosphate, a lipid exclusively produced at the plasma membrane (20). Because lysosomes/late endosomes (LLEs) were found to be involved in plasma membrane repair (21, 22) and because ESCRT proteins are involved in LLE functions, we tested whether CHMP4 was transported by LLE to the wounded area of the plasma membrane. Cells expressing CHMP4B-EGFP were immunostained with endolysosomal markers [lyso-bis-phosphatidic acid (LBPA), lysosomal-associated membrane protein 1 (Lamp1), and early endosome antigen 1 (EEA1)] after localized laser wounding to look for approaching endolysosomal vesicles. No LBPA, Lamp1, or EEA1 vesicular structures were detectable close to the wound in conditions where CHMP4 was strongly recruited (fig. S1). As reported before (22), however, the presence of Lamp1 can be revealed by surface staining at the wound (fig. S2A). However, the recruitment of ESCRTs at the wound in cells with depolymerized microtubules suggests that it is mainly independent of vesicular transport (fig. S2B). This is further suggested by the fact that recruitment of CHMP4B on wound sites was energy-independent (see below).

Fig. 1 CHMP4B is recruited to the wounded plasma membrane.

(A to C) HeLa cells stably expressing CHMP4B-EGFP at endogenous levels (referred below as HeLa CHMP4B-EGFP) were used. Cells were treated with digitonin or saponin for 1 min and fixed for 3 min after the beginning of the treatment (A) or were treated with streptolysin O or listeriolysin O for 2.5 min before fixation (B). Cells were fixed and immunolabeled with antibodies to GFP (for better visualization of CHMP4B-EGFP recruitment) (A to C). Plasma membrane wounding with a glass needle (highlighted with dashed lines) was followed by differential interference contrast (DIC) microscopy imaging, and CHMP4B-EGFP recruitment was observed by confocal microscopy in 6 of 20 cells analyzed (C). (D) HeLa cells cotransfected with CHMP4B-mCherry and MyrPalm-EGFP constructs were wounded at the plasma membrane with a UV laser/spinning disc microscope system. Scale bars, 10 μm.

Fig. 2 CHMP4B-recruitment occurs before resealing.

(A to E) The plasma membranes of HeLa CHMP4B-EGFP cells were locally wounded (arrow) in the presence of PI at t = 0 (A, B, D, and E) or at several time points (C), using a UV laser. Time-lapse imaging was performed using a spinning disc microscope system to follow the recruitment of CHMP4B-EGFP and the entry of PI. (A) PI fluorescence is presented using a “royal” look-up table. (B) Corresponding graph for the cell in (A), which is representative of 15 cells analyzed. The red curve represents PI entry, and the green bars represent CHMP4B-EGFP recruitment measured around the wound. (C) Similar experiments were performed for cells subjected to successive wounding. Two localized wounding were performed at 0 and 10 min on two spatially separated spots of the membrane of cells, followed by a deadly wounding at 20 min. The cell presented is representative of 15 cells analyzed. The top panel illustrates CHMP4B recruitment at the wound sites at different time points. CHMP4B-EGFP recruitment measured in regions 1 and 2 is represented by the histograms in light green and dark green, respectively, and PI entry is represented by the red curve. Scales are shown on the right and left of the graph (see corresponding movie S1). (D) Curves of PI entry and CHMP4B-EGFP recruitment were averaged for 15 cells around the time of closure (Tclosure) (time when PI reaches a plateau), with Tclosure centered at 0. Error bars correspond to the SE of each mean. (E) The difference between Tclosure and the time of maximum CHMP4B-EGFP recruitment gives a ΔT for each cell analyzed. The distribution of all ΔT was plotted; average = −0.0227, t test not significantly different from 0, P = 0.9421. Arrows point at the wound site to which CHMP4B-EGFP is recruited. Scale bars, 10 μm.

These data suggest that soluble ESCRT-III proteins are directly targeted to physical holes present at the plasma membrane, where they may be involved in repair. The disruption of the membrane and the closure of the wound were monitored following the entry of the impermeant nucleic acid dye propidium iodide (PI) into wounded cells. Reversible physical damage upon laser-based wounding can be efficiently followed in real time, quantifying cellular PI fluorescence. CHMP4B recruitment was triggered by single or sequential and spatially distinct laser wounds (Fig. 2, A to C, and movie S1). The suitability of PI to monitor wound opening and repair was confirmed by comparing PI with other dyes (fig. S3). Using PI as a tracer, or other fluorescent dyes, we observed that wounds could stay open for a few hundreds of seconds (Fig. 2B and fig. S3). This was further confirmed by adding PI to the cells at different time points after wounding, which indicated that some wounds remained open up to 4 min (fig. S4). The recruitment of this ESCRT subunit always preceded the closure of the wound (Fig. 2D). Furthermore, we calculated the difference between the time of closure (time of reach of the PI plateau) and the time of maximum recruitment of CHMP4B, which was not statistically different from 0, strongly suggesting that these two events are correlated (Fig. 2E). Thus, as soon as the wound closes, ESCRT-III starts to dissociate from the wounded area.

We next investigated whether other subunits of the ESCRT-III subcomplex were involved in this targeted response. Immunostaining of digitonin-treated cells revealed that, upon damage, endogenous CHMP3, CHMP2A, and CHMP2B, but not CHMP6, colocalized with CHMP4B-EGFP at the plasma membrane (Fig. 3, A to C, E, and F, and fig. S5). Proteins from other ESCRT subcomplexes, such as hepatocyte growth factor–regulated tyrosine kinase substrate (HRS), were not detected (fig. S6A), although recruitment of overexpressed and endogenous TSG101 was observed (fig. S6, B and C). We investigated whether ALIX, a protein capable of bypassing ESCRT-0, ESCRT-I, and ESCRT-II complexes and recruiting ESCRT-III proteins directly, was involved in the recruitment of ESCRT-III proteins. ALIX colocalized with CHMP4B-EGFP at the plasma membrane of cells treated with digitonin (Fig. 3D) or wounded with a laser beam (Fig. 4C). Polyubiquitination of the plasma membrane wounded by a laser or by digitonin was observed by immunofluorescence (fig. S7). Its belated detection befogs the role of ubiquitin as a potential early signal for ESCRT recruitment via ubiquitin-binding domains (23, 24). Nevertheless, although we cannot exclude that the belated detection is due to a limited antibody sensitivity, the presence of ESCRTs at the wound sites in the absence of energy, as described below, argues in favor of ubiquitin-independent recruitment of ESCRTs. Calcium, an essential molecule for wound repair (25), is likely to be involved in early signaling of membrane damage and stimulate the local recruitment of ESCRTs. As expected, calcium depletion enhanced cell necrosis when cells were wounded with a laser (Fig. 4A and fig. S8). These conditions also strongly inhibited ESCRT recruitment (Fig. 4B). This argues for a role of calcium entry at wounded sites in the local recruitment of ESCRT. Accordingly, the recruitment of ALIX mutants lacking calcium-binding domain (26) (as well as lipid-binding mutants) was strongly impaired as compared to a control mutant or wild-type proteins (Fig. 4, C and D). Thus, calcium plays a major, and early, role in the direct recruitment of ESCRTs to the wounded plasma membrane, but the presence of both ALIX and TSG101 suggests that redundant pathways may be involved in the recruitment of ESCRTs at the wounded plasma membranes. The lipid-binding domain of ALIX has been described by Bissig et al. (26) as an LBPA-specific lipid-binding domain. Nevertheless, this study only considers lipids present on MVBs. Our data suggest that ALIX can bind to the plasma membrane through its LBPA-binding domain, suggesting that this domain can also bind other lipids present at the plasma membrane, or that LBPA might be transiently present at the wounded plasma membrane. This finding opens research perspectives concerning ALIX binding properties.

Fig. 3 CHMP2, CHMP3, and the ESCRT-associated protein ALIX are recruited to wounds.

(A to D) HeLa CHMP4B-EGFP cells were treated with digitonin and immunostained for CHMP2B (A), CHMP3 (B) (37), CHMP6 (C), and ALIX (D) (38). Intensity measurements of CHMP4B-EGFP and the other ESCRT subunits were performed along the lines highlighted at the plasma membrane edge and are represented as intensity spectrums. (E and F) HeLa CHMP4B-EGFP (E) or HeLa (F) cells were wounded and imaged with a spinning disc microscope equipped with a UV laser. Cells were fixed 2 min after wounding and immunostained for CHMP2A. EGFP signal was enhanced in all experiments as before. White arrows point at the wound where CHMP4B-EGFP is recruited. Scale bars, 10 μm.

Fig. 4 ESCRT-III proteins and the ESCRT- associated protein ALIX are recruited in a calcium-dependent manner.

HeLa CHMP4B-EGFP cells were incubated in media with or without calcium and supplemented with BAPTA [1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid] or EGTA. Both media were supplemented with PI at 160 μg/ml. Laser wounding was performed with a UV laser mounted on a spinning disc microscope. Cells were examined for necrosis after wounding on the basis of morphology by transmission light and levels of PI (see fig. S8). (A) Surviving and necrotic cells were quantified. Asterisks indicate differences tested by Fisher’s exact test: P > 0.0001 (for BAPTA); P > 0.0001 (for EGTA). (B) The amount of recruited CHMP4B-EGFP at the wound was quantified at 1 min after wounding to avoid measurement artifacts due to cell shape changes and blebbing. Asterisks indicate differences tested with the unpaired t test with Welch’s correction: P = 0.0013 (for BAPTA); P = 0.0219 (for EGTA). (C and D) HeLa cells transfected with different forms of split-YFP (yellow fluorescent protein) ALIX plasmids (26) were wounded and imaged. (C) Representative cells are presented for each mutant 2 min after wounding. (D) The recruitment of YFP-ALIX wild type (WT) or mutants was quantified at the wound site. Asterisks indicate differences tested with the unpaired t test with Welch’s correction: P = 0.0107 (D178A mutant); P = 0.6661 (D316A mutant); P = 0.0277 (QQ mutant). Error bars correspond to half 95% confidence intervals. Scale bar, 10 μm.

Finally, we analyzed the recruitment of the AAA+ adenosine triphosphatase Vps4 (responsible for the disassembly of ESCRT-III polymers) and also observed the rapid kinetics of recruitment (fig. S9 and movie S2). The timing for Vps4 recruitment and removal from the wounding site was very close to that of CHMP4B recruitment and removal. We did not detect other markers such as annexin A1 that have previously been detected in streptolysin O wounds (16).

We then asked whether the ESCRT machinery was recruited at the wounded area following the classical sequential order of events described for MVB formation. We knocked down CHMP2A in HeLa cells expressing CHMP4B-EGFP (17). CHMP2 proteins, in interaction with Vps4, are necessary to recruit the ESCRT disassembly complex (2, 3, 2730). In control cells, CHMP4B-EGFP local accumulation reached a plateau about 2 to 4 min after wounding and then started to decrease right after wound closure (see Fig. 2, B to D). In contrast, in CHMP2A knockdown cells, CHMP4B continued accumulating up to three times more at the wounded site and reached a plateau only 15 min after wounding (Fig. 5, A and B, fig. S10A, and movie S3). This suggests that ESCRT-III recruitment is the result of a dynamic equilibrium between the assembly and disassembly of the complex on the wound. Knockdown of CHMP3 had a similar, albeit milder, effect on CHMP4B accumulation (figs. S10B and S11A), suggesting a role for CHMP3 in the disassembly of ESCRT-III machinery.

Fig. 5 Perturbation of wound resealing after depletion of ESCRT-III and ESCRT disassembly subunits.

Laser wounding and time-lapse acquisition were performed with a UV laser/spinning disc microscope system on cells incubated with PI. (A and B) HeLa CHMP4B-EGFP cells were treated with siRNA luciferase or CHMP2A; see movie S3. (A) Quantification of CHMP4B recruitment at the wound site. (B) Representative cells for CHMP4B-EGFP recruitment in cells treated with siRNA luciferase or CHMP2A. Arrows point at the wound site. Scale bar, 10 μm. (C and D) HeLa cells transfected with different plasmids or siRNA were incubated in medium with PI and wounded using a UV laser. The total fluorescence of PI intensity was fitted using a mathematical model (see supplementary materials and methods) to obtain an initial effective radius (Reff) of the wound and an effective rate of closure (v) (see fig. S13 for raw data). The distribution of Reff for Reff < 1.5 × 106 was not different between siRNA luciferase and CHMP2A conditions; t test with Welch’s correction, P = 0.9537 (C). Distribution of v for Reff < 1.5 × 106 was significantly different between siRNA luciferase and CHMP2A conditions; t test with Welch’s correction, P = 3.849 × 10−6 (D). (E) Quantification of cell survival upon plasma membrane damage in HeLa cells transfected with Vps4B-EGFP WT or DN or treated with siRNA against luciferase or CHMP4B. Statistical analysis was performed using a two-sided χ2 test for proportion comparison (α = 5%). n.s., nonsignificant differences; *P < 0.0001. (F) PI entry was quantified and averaged for each time point in each condition. Error bars correspond to 95% confidence intervals.

To better understand the role of ESCRT-III proteins in plasma membrane repair processes, we followed the opening and closure of the wound by recording and quantifying the kinetics of entry of PI into wounded cells. A simple model of wound closure was established, and the curves of PI entry after cell damage were fitted using this model. Successive wounding of the same cell showed that the plateau of PI entry reached after each damage was caused by wound closure and not by saturation of PI labeling and thus could be used when fitting the curves (Fig. 2C and fig. S12). Quantitative information about the initial radius of the hole in the plasma membrane (Reff) and its rate of closure (v) could be estimated using the model (see details in supplementary materials and methods and fig. S12). Laser-based plasma membrane damage led to a diversity in wound size (see the cumulative distribution in fig. S13). When we plotted closing rate values against the relative wound size, closure rates increased with wound size. However, two modes were apparent: one for relatively smaller Reff (smaller holes), where rates increased rapidly in function of Reff, and one for relatively larger Reff (larger holes), where closing rate increased more moderately. This suggested that different mechanisms may be used depending on wound sizes. Indeed, inhibition of CHMP2A affected only the closure rate of relatively small holes with no impact on wound size distribution (Fig. 5, C and D, and fig. S13, A and B), whereas no effect was observed for larger holes (fig. S12B). Using quantitative microscopy and calculation, we could estimate that ESCRTs play a role in the repair of wounds of less than 100 nm (fig. S14). No effect on v could be detected upon treatment of cells with small interfering RNA (siRNA) against CHMP3 (figs. S10B and S11, C and D).

ESCRT-III function was further perturbed by transfecting the dominant-negative (DN) mutant Vps4BE223Q (31) or by treating cells with an siRNA targeting CHMP4B (32) (for siRNA efficiency, see fig. S10C). These treatments inhibit ESCRT-III activity: CHMP4B is the main component of ESCRT-III filaments (33), and Vps4B DN blocks depolymerization of ESCRT subunits and therefore their recycling (3). In these conditions, we did not detect a reduction in the rate of closure. However, when we plotted the distribution of wound size, we observed that perturbation of Vps4B or CHMP4B functions leads to a reduction in the number of cells displaying a relatively smaller wound (fig. S13, C and D). No such effect was seen upon inhibition of CHMP2A or CHMP3 (figs. S11B and S13A). This suggested that, upon perturbation of Vps4B or CHMP4B functions, cells with smaller wounds did not survive and, thus, could not be analyzed. Accordingly, we observed a significant reduction of cell survival upon laser wounding in these conditions, in comparison to cells treated with a control siRNA or overexpressing wild-type Vps4B (Fig. 5E). This was consistent with the higher levels of PI entry observed (Fig. 5F). This indicates that ESCRT-III activity is essential for cell survival after plasma membrane damage and suggests that the importance of ESCRT-III role in repair is related to the size of the wound.

Our observations indicate that ESCRT-III proteins are rapidly, specifically, and dynamically recruited to damaged areas of the plasma membrane to quickly repair the defect. Knowing the usual topology of ESCRT activity, this suggests a model whereby ESCRT proteins would be involved in extracellular budding of the damaged area and sealing of the plasma membrane.

To directly test this model by observing the wound site at high resolution, we set up a correlative scanning electron microscopy (SEM) assay to (i) get a spatiotemporal control of the wound driven by a laser, (ii) observe CHMP4B-EGFP dynamics and specific recruitment, and (iii) obtain a high-resolution image of the zone where CHMP4B-EGFP was recruited. The surface of the CHMP4B-EGFP HeLa cells grown on gridded coverslips was wounded using a UV laser (Fig. 6, A to F, and movie S4). The recruitment of CHMP4B-EGFP was followed by fluorescence time-lapse imaging (Fig. 6A). Lateral projection of image stacks showed that CHMP4B was recruited on plasma membrane protrusions (Fig. 6B). Cells imaged by time-lapse microscopy were then fixed and observed by SEM (n = 6). Low-magnification observation allowed the detection of wounded areas (Fig. 6C). At higher magnification, a cluster of extracellular buds of various size was detected on damaged zones and could be aligned with CHMP4B-EGFP–positive areas (Fig. 6, D to F). This suggested that ESCRT-III proteins were involved in extracellular budding at wounded sites.

Fig. 6 ESCRT-positive cell surface budding upon wounding.

HeLa CHMP4B-EGFP cells were used for these experiments. Wounds and time-lapse acquisition was performed on a spinning disc microscope equipped with a UV laser. (A) Time-lapse observation of the recruitment of CHMP4B-EGFP in wounded cells. (B) Lateral view of the cell 4 min after wounding. (C) Correlative image of the cell fixed and processed for SEM (representative of six cells analyzed). (D and F) Higher magnification of the areas pointed in (A) and (C). (E) Fluorescence to SEM alignment based on cell morphology criteria; see movie S4. (G) Recruitment of cytoplasmic CHMP4B-EGFP at blebs generated at the wounded site. The resorption of the bleb is accompanied by concentration of CHMP4B at the wounding site. Lateral views correspond to an orthogonal projection where the z images were separated by 5 pixels and intrapolated. White arrows point at the wound site where CHMP4B is recruited. Scale bars, 10 μm (A, C, and G); 1 μm (D to F).

We observed ESCRT-positive membrane shedding at the cell surface at wounded sites using time-lapse confocal microscopy (Fig. 7A, fig. S15, and movie S5). Accordingly, using spinning disc confocal time-lapse microscopy and SEM, we observed that CHMP4B-EGFP–positive buds were accompanied with shedding (fig. S16). Both membrane repair and ESCRT-positive shedding were adenosine 5′-triphosphate (ATP)–dependent mechanisms (Fig. 7, B to E, and movie S6). ESCRT recruitment itself still occurred in the absence of energy, but pinching of the membrane was only observed when energy production was restored (Fig. 7C and fig. S17). Thus, ESCRT shedding itself appears necessary for the wound closure, and ATP-dependent mechanisms, most likely the ESCRT disassembly machinery, may also be required to resolve the neck constriction, bud detachment, and wound closure.

Fig. 7 Energy-independent ESCRT recruitment and energy-dependent shedding of wounded membrane.

HeLa CHMP4B-EGFP cells were wounded and imaged on a spinning disc microscope equipped with a UV laser. (A) Shedding of ESCRT-positive particles from wound site pointed by the arrow. For corresponding lateral view, see movie S5. (B and C) Cells were incubated in phosphate-buffered saline (PBS) + PI supplemented with glucose, glucose + sodium azide, or deoxyglucose + sodium azide, as indicated. Cells were wounded as previously described. PBS containing deoxyglucose and sodium azide was washed out in certain assays [(C) and fig. S17] and replaced with PBS-glucose 10 min after wounding (WO). (D and E) PI entry (D) and CHMP4B (E) recruitment were quantified over time and averaged. Arrows point at the wound site where CHMP4B is recruited. Empty arrows point at CHMP4B-positive shedding particles. Error bars correspond to half 95% confidence intervals. Scale bars, 10 μm.

The precise mechanisms that allow cells to rapidly detect wounds at the plasma membrane will need further investigation, but our results indicate a key role played by calcium increase. In addition, time-lapse imaging of CHMP4B recruitment led to an intriguing observation: We often noticed that plasma membrane wounding was followed by the rapid formation of a localized bleb of a few micrometers in diameter. The bleb rapidly became positive for CHMP4B-EGFP (Fig. 6G). As the bleb regressed, the staining of CHMP4B-EGFP became denser. Previous studies have revealed that events lowering membrane tension (such as exocytosis and local removal of the actin cytoskeleton) occur upon wounding of the plasma membrane (34). On the basis of these studies and our observations, we thus propose a model whereby a lesion in the plasma membrane triggers a local drop in membrane tension reinforced by local membrane supply through exocytosis and/or cytoskeleton remodeling. This drop in tension together with calcium entry would promote ALIX and ESCRT-III local recruitment. Other signals such as ubiquitination of proteins at the wound site or loss of lipid asymmetry and the recruitment of lipid-binding proteins may reinforce this binding. The assembly of ESCRT-III into helix-shaped filaments would then trap the damaged portions of plasma membrane on buds, shed them from the plasma membrane, and ensure sealing of nondamaged areas. Rapid binding of ESCRT proteins may also prevent extension of the wound and enhance cell survival. Although these mechanisms would be sufficient to explain localized membrane repair, it may be associated with other mechanisms described previously, such as patching of membrane gaps with intracellular vesicles (fig. S2A). The size of the gap may dictate the choice of the mechanism involved for repair because the patching model has clearly been demonstrated for larger wounds and because we observed critical involvement of ESCRT-III proteins to repair wounds smaller than 100 nm. These wounds were rather long-lasting as compared to what has been observed for larger wounds.

In conclusion, plasma membrane integrity is essential for cell homeostasis and survival. We propose that ESCRT-III proteins play a central role in repairing local injuries by ensuring extracellular shedding of damaged areas. Plasma membrane shedding, sometimes positive for ESCRTs, has been observed in several normal and pathological conditions (35, 36), and it will now be important to test whether this is linked to local plasma membrane damage and whether ESCRT-III proteins are involved in these processes.

Supplementary Materials

www.sciencemag.org/content/343/6174/1247136/suppl/DC1

Materials and Methods

Figs. S1 to S17

Movies S1 to S8

References (3943)

  • * These authors contributed equally to this work.

References and Notes

  1. Acknowledgments: We are indebted to the electron microscopy platform of Institut Pasteur and in particular to P. Bomme for their services with correlative light to SEM experiments. We are also indebted to our colleagues who shared materials; I. Poser and A. Hyman for the CHMP4B-GFP cell line; C. Wunder and C. Bissig for ESCRT plasmids; M. Maki for the CHMP6-GFP plasmid; R. Tweten for the streptolysin O plasmid; R. Y. Tsien for the MyrPalm-EGFP plasmid; A. Draeger for the annexin plasmids; H. Stenmark, S. Urbé, and R. Sadoul for the antibodies against ESCRT subunits; G. van der Goot for the listeriolysin O toxin; A. El Marjou for the production of streptolysin O; V. Fraisier, L. Leconte, L. Sengmanivong, and F. Waharte of the Nikon Imaging Center and the PICT-IBiSA of the Institut Curie for technical support in microscopy; A. Pinheiro for quantitative polymerase chain reaction protocols and help; P. Vargas for advice and reagents for calcium experiments; J. B. Manneville, D. Guet, and M. Pinot for technical assistance with micropipette use; and F. Niedergang, G. Boncompain, A. Dimitrov, and F. Brochard for commenting on the results. Funding: This work was supported by the Institut Curie, CNRS, the French Agence Nationale de la Recherche (“ANR”) (to F.P.), a Ligue Contre le Cancer grant (to M.P.), a French ministerial fellowship (to J.L.-J.), Fondation pour la Recherche Medicale fellowships (to J.L.-J. and P.M.), and an Institut Curie fellowship (to A.J.J.). F.P.’s team is supported by the LaBex CelTisPhyBio (ANR-10-LBX-0038 part of the IDEX PSL no. ANR-10-IDEX-0001-02). Author contributions: A.J.J. and J.L.-J. began the project. A.J.J. designed and performed all the experiments under the supervision of F.P. and with the help of the electron microscopy platform of Institut Pasteur for the SEM experiments and performed most quantifications. A.J.J. wrote the paper together with F.P. J.L.-J. performed preliminary experiments under the supervision of M.P., was involved in discussions, and participated in the writing of the manuscript. P.M. performed experiments for determining hole size together with A.J.J., performed the computational modeling of PI entry and hole size estimation, performed several quantitative analysis in close interaction with A.J.J. and under the supervision of M.P. and F.P., was involved in discussions, and participated in the writing of the manuscript. S.D. helped A.J.J. with revision experiments. M.P. was involved in discussions, supervised the quantitative analysis, and participated in the writing of the manuscript. F.P. supervised this work and wrote the manuscript together with A.J.J.
View Abstract

Navigate This Article