Defective membrane repair machinery impairs survival of invasive cancer cells

Cancer cells are able to reach distant tissues by migration and invasion processes. Enhanced ability to cope with physical stresses leading to cell membrane damages may offer to cancer cells high survival rate during metastasis. Consequently, down-regulation of the membrane repair machinery may lead to metastasis inhibition. We show that migration of MDA-MB-231 cells on collagen I fibrils induces disruptions of plasma membrane and pullout of membrane fragments in the wake of cells. These cells are able to reseal membrane damages thanks to annexins (Anx) that are highly expressed in invasive cancer cells. In vitro membrane repair assays reveal that MDA-MB-231 cells respond heterogeneously to membrane injury and some of them possess a very efficient repair machinery. Finally, we show that silencing of AnxA5 and AnxA6 leads to the death of migrating MDA-MB-231 cells due to major defect of the membrane repair machinery. Disturbance of the membrane repair process may therefore provide a new avenue for inhibiting cancer metastasis.

Membrane rupture and repair assay. Membrane repair assay differed slightly from our previous studies 56 . MDA-MB-231 cells were cultured in complete growth medium on 18*18 mm glass coverslips (Nunc). A solution of 5 µg/mL FM1-43 (ThermoFisher Scientific, Waltham, MA, USA) in D-PBS containing 2 mM Ca 2+ was maintained over ice and subsequently added in a homemade coverslip cell chamber ( Supplementary Fig. 1), where cells containing coverslip was mounted. To induce membrane damage, cells were irradiated at 820 nm with a tunable pulsed depletion laser Mai Tai HP (Spectra-Physics, Irvine, USA) of an upright two-photon confocal scanning microscope (TCS SP5, Leica) equipped with an HCX PL APO CS 63.0 × 1.40 oil-objective lens. Irradiation consisted of 1 scan (1.3 s) of a 1 µm × 1 µm area with a power of 110 (± 5) mW. 512 × 512 images were acquired at 1.3 s intervals with pinhole set to 1 Airy unit. Membrane rupture and repair processes were monitored by measuring variations in fluorescence intensity of FM1-43 as previously described 56 . FM1-43 was excited by the 488-nm laser line (intensity set at 30% of maximal power) and fluorescence emission was measured between 520 and 650 nm. For quantitative analysis, the fluorescence intensity was integrated over the whole cell surface and corrected for the fluorescence value recorded before irradiation, using ImageJ software.

Transduction of AnxA5-targeting and/or
Lentiviral-based particles containing shRNAs were produced by Bordeaux University Lentiviral Vectorology Platform (US005, Bordeaux, France) by transient transfection of 293 T cells. 2·10 5 MDA-MB-231 cells were cultured in a 30 mm Petri Dish for 24 h and transduction was carried out by adding concentrated lentiviral particles to the cells at multiplicity of infection (MOI) of 10 in 2 mL Opti-MEM for 24 h. Transduced cells were cultured for 24 h in growth medium and then selected in selection medium composed by 2 µg/mL puromycin in DMEM for 48 h. Cells were passaged and subsequently cultured in 25 cm 2 cell culture flask in selection medium. At each passage, a fraction of cells was used for preparing protein extracts for western-blot analysis of the expression of endogenous Anx. Co-transduction of lentiviral-based particles containing AnxA5-and AnxA6-targeting

Localization of endogenous Anx in intact and damaged cells by immunofluorescence.
For subcellular localization of endogenous Anx in damaged MDA-MB-231 cells, cell membrane rupture was performed according to the protocol described above, but in the absence of FM1-43 to avoid fluorescence cross-talk. After laser irradiations, cells were fixed in 1% glutaraldehyde and permeabilized in 0.1% TritonX100 diluted in DPBS + Ca 2+ . All subsequent steps (saturation, antibody incubation and washes) were performed using 2% BSA in DPBS + Ca 2+ solution. Primary antibodies (1:100, except for anti-AnxA5 1:400), which were the same as used for western-blot, and secondary Alexa Fluor 488-coupled anti-mouse or anti-rabbit IgG goat antibodies (1:1000, ThermoFisher Scientific) were successively incubated with cells for 1 h at 37 °C. Finally, cells were washed in DPBS + Ca 2+ and nuclear counterstaining was performed with DAPI (Sigma). For each condition, about 30 cells from three independent experiments were analyzed. For the subcellular localization of endogenous Anx in intact MDA-MB-231 cells, cells were cultured in 8-well Lab-Tek chambered coverglasses. Cells were immunostained according to the protocol described above for the localization of Anx in damaged cells, from the step of glutaraldehyde fixation.

Results
Collagen I fibrils favor migration of MDA-MB-231 cells and formation of migrasome. First, the migration of MDA-MB-231 cells plated on glass coverslip coated or not with collagen I was compared through analysis by phase-contrast video-microscopy ( Fig. 1).
The fibrillar collagen substrate obtained in our experimental conditions has been previously characterized in depth 53,57 . Treatment with glutaraldehyde enables collagen I cross-linking, which increases the capacity of tumor cells to form invasive structures 58 . Tracking of cells was performed in both conditions by means of the MtrackJ plugin from the ImageJ software 55 and start-end distances were measured. Start-end distance is the Euclidian distance between positions of cell at t 0 and t end . It therefore positively correlates with the motility of the cell. In the presence of collagen I, cells exhibited elongated shape and migrated lengthwise ( Fig. 1 and Supplementary video 1), whereas in the absence, they presented mostly rounded shape and spinning movements ( Fig. 1 and Supplementary video 2). Mean start-end distance of cells seeded on collagen I was significantly (P value = 5.5E−11 for Student t test) higher (103 ± 33 µm) than in the absence of collagen I (13 ± 12 µm) after 2 h of migration. We have therefore concluded that collagen I fibrils favor the migration of MDA-MB-231 cells. www.nature.com/scientificreports/ In order to characterize the migrasome 59,60 of MDA-MB-231 cells moving on collagen I fibrils, cells were cultured on glass coverslip coated with collagen I for 24 h. Their migration was analyzed for 2 h by phase-contrast video-microscopy and at the end of the kinetics study cells were fixed and incubated with CellMask Orange. CellMask stains are lipophilic dyes that become fluorescent upon inserting into plasma membrane. The use of glass bottom dishes equipped with a square-patterned coverslip displaying an alphanumerical code in each square enabled cell tracking during different stages of the experiment and correlation of cell migration observed by phase-contrast video-microscopy and CellMask Orange staining analyzed by fluorescence microscopy. By means of fluorescence microscopy, we systematically observed (over 7 independent experiments) the presence of cell membrane fragments in the near periphery of approximately 70% of cells (Fig. 2a). At higher magnification, cell membrane fragments appeared as membrane-bound vesicular structures (MbVS, Fig. 2b), as previously described by Yu and collaborators 59 . By analyzing the immediate surrounding of a migrating cell followed by video-microscopy ( Fig. 2c and Supplementary video 3), we observed the presence of MbVS in the wake of the cell by fluorescence microscopy (Fig. 2d). We have concluded that MDA-MB-231 cells migrating on collagen I fibrils release MbVS in their wake. Formation and release of these cell structures may be induced by shearing forces existing between the extracellular collagen I fibrils and cell membrane.

MDA-MB-231 cells exhibit a highly efficient membrane repair machinery.
In order to investigate the membrane repair ability of MDA-MB-231 cells, plasma membrane was damaged by laser ablation and membrane repair was assayed by monitoring the kinetics of cell entry of membrane-impermeable FM1-43 dye, using a standard protocol 13,56 . Membrane resealing leads to a stop in the entry of FM1-43 molecules into the cytosol and therefore to a stop in the increase of the intracellular fluorescence intensity, whereas the absence of membrane resealing leads to a continuous entry of FM1-43 into the cytosol and a continuous and strong increase in the fluorescence intensity. Irradiation conditions were adjusted (110 mW) to cause mild membrane injury to cells, characterized by the entry of FM1-43 into the cytosol as early as a few seconds after laser ablation without the presence of a macroscopic tear at the site of membrane irradiation (Fig. 4a,b, frames 2, arrow). Two typical responses were observed within the 120 s following the membrane damage. 50% of cells exhibited an  Supplementary Fig. 2). The choriocarcinoma BeWo cell line is placental in origin and the placenta is expected to be the richest organ in Anx in Humans (www.prote inatl as.org). BeWo cells served therefore as a positive control of the expression of Anx. In order to standardize measurements, MDA-MB-231 and BeWo protein extracts were analyzed on the same PVDF membrane and the band intensity relative to the Anx was reported to GAPDH, used as loading control. As an example, a typical result obtained for the detection of AnxA5 is presented in Fig. 5b. We observed that AnxA4 is hardly, if at all, detected in both cell lines. A significant expression of AnxA1, A2, A5, and A6 is observed. For these Anx, the expression level is systematically as high for MDA-MB-231 as for BeWo cells. We even observed that the expression of AnxA2 is stronger in MDA-MB-231 compared to BeWo cells. We can therefore conclude that four Anx (A1, A2, A5 and A6), which are known to participate in membrane repair, are highly expressed in MDA-MB-231 cells.
We then investigated the subcellular distribution of endogenous Anx in MDA-MB-231 cells. Immunocytofluorescence experiments confirmed that AnxA1, AxnA2, AnxA5 and AnxA6 are significantly expressed in www.nature.com/scientificreports/ MDA-MB-231 cells whereas AnxA4 is weakly expressed (Fig. 5c). As observed in human myoblasts (our personal data) and murine neuroblasts 63 , AnxA1 localizes in the nucleus and in the cytoplasm of MDA-MB-231 cells. Instead AnxA2, AnxA5 and AnxA6 are exclusively cytoplasmic. These results are consistent with previous studies for AnxA2 63 and AnxA6 64 but different for AnxA5 30,63 , which has been reported to be present in nucleus and cytoplasm. The homogenous distribution of Anx within the cytoplasm supposes that it localizes in the cytosol. Strikingly, we observed that the expression level of AnxA5 or AnxA6 varies considerably from one MDA-MB-231 cell to another (Lower panels of Fig. 5c,d), whereas that of AnxA1 or AnxA2 is relatively constant (Higher panels of Fig. 5c,d). We observed that a cell expressing AnxA5 at a high level expresses also strongly AnxA6 (Supplementary Fig. 3). In addition, a low level of AnxA5 is systematically accompanied by a weak expression of AnxA6. This strong heterogeneity in the expression of AnxA5 and AnxA6 in MDA-MB-231 cells

AnxA5 and AnxA6 belong to the membrane repair machinery of MDA-MB-231 cells. Most,
if not all, proteins that are involved in the repair machinery, i.e. dysferlin 8 , AnxA1, AnxA2, AnxA6 12,14,65 , AnxA5 13,30 and MG-53 17 are recruited to the membrane disruption site immediately after membrane injury. In order to assess whether AnxA5 and AnxA6 may be involved in the membrane repair machinery of MDA-MB-231 cells, we investigated the subcellular localization of both Anx after laser ablation. For this purpose, MDA-MB-231 cells were cultured on gridded coverslips, thus enabling accurate tracking of irradiated cells. After laser irradiation, cells were fixed, permeabilized and immunostained for AnxA5 (Fig. 6a) or AnxA6 (Fig. 6b). After laser injury, endogenous AnxA5 and AnxA6 were found specifically accumulated at the disruption site of MDA-MB-231 cells. We have previously shown that AnxA5 promotes membrane repair mainly by interacting with the edges of the torn membrane, where it forms bi-dimensional arrays that strengthen the membrane and prevent the expansion of the tear 13,20 . In addition, it has been proposed that AnxA6 is recruited to the wound edges, where it induces a constriction force enabling wound closure 15 . Interaction between Anx and damaged plasma membrane requires only the presence of phosphatidylserine in the membrane and Ca 2+ at mM concentration. Therefore, it is likely that the recruitment of AnxA5 and AnxA6 at the disruption site is responsible for membrane resealing in MDA-MB-231 cells.
In order to confirm that AnxA5 and AnxA6 are required for membrane repair, both Anx were knocked-down in MDA-MB-231 cells using shRNA strategy. We estimated that the decrease of the expression of AnxA5 or AnxA6 in transduced cells was about 90% (Supplementary Fig. 4). AnxA5-deficient or AnxA6-deficient MDA-MB-231 cells were then submitted to membrane repair assay. After laser injury, we observed that AnxA5-deficient MDA-MB-231 cells exhibited a rapid and strong increase in the intracellular fluorescence intensity due to the entry of FM1-43 dye, indicating a major defect of membrane repair (Fig. 7a,c, middle panel). AnxA6-deficient MDA-MB-231 cells also exhibited a continuous entry of FM1-43 dye into the cytosol after membrane damage but in a lesser extent compared to AnxA5-deficient cells (Fig. 7b,c, right panel). The absence of plateau within 120 s after membrane damage indicates that these cells also present a defect of membrane repair. 120 s after membrane injury, imaging of FM1-43 by video-microscopy was stopped and gain and offset values were adjusted to obtain an unsaturated fluorescence image (Fig. 7d). We observed that most AnxA5-deficient MDA-MB-231 cells (n = 45 /49) exhibited a large membrane disruption at the wound site (Fig. 7d, left panel). We have proposed previously that AnxA5 molecules form 2D-arrays at the edges of the membrane disruption site in order to strengthen the membrane and to avoid the extension of the rupture 13,20,30 , which has been recently confirmed on model membrane 66 . The hole present at the irradiation site suggests that plasma membrane has been torn. Lipid material, which is marked by the lipophilic dye FM1-43, is accumulated near the disruption site but looks insufficiently dense for plugging it. Strikingly, in the case of AnxA6-deficient MDA-MB-231 cells, we often (n = 36/51) observed that lipid material is concentrated far away from the membrane injury (Fig. 7d, right panel). If the repair mechanism relies on the formation of a lipid "patch" that is supposed to clog the membrane disruption, AnxA6-deficient cells would suffer from a defect in the recruitment of this lipid "patch". Altogether, these results led to conclude that AnxA5 and AnxA6 are instrumental for membrane repair in MDA-MB-231 cells. www.nature.com/scientificreports/ plementary Fig. 6a,b). As observed for AnxA6-deficient MDA-MB-231 cells, lipid material appeared frequently concentrated far away from the membrane injury ( Supplementary Fig. 6c).
In order to assess the influence of a defect of the membrane repair machinery in cell life during migration, we seeded 2·10 5 control or Anx-deficient MDA-MB-231 cells on a dish coated with collagen I and cells were imaged and counted 24 h and 48 h after seeding. We hypothesized that migration of Anx-deficient MDA-MB-231 on collagen I may lead to the death of cells and their detachment from the coverslip, due to un-resealed membrane damages. 24 h after seeding, we observed that Anx-deficient cells exhibit a cell density and morphology similar to control cells indicating that the deficiency in Anx does not disturb cell adhesion on collagen I-coated coverslip (Fig. 8, 24 h). 48 h after seeding, phase contrast microscopy imaging showed that cell density was www.nature.com/scientificreports/ similar between control and AnxA5-AnxA6 deficient cells, though the percentage of rounded cells was strongly increased (Fig. 8, 48 h). Through video-microscopy analysis, we observed that elongated AnxA5-AnxA6 deficient MDA-MB-231 cells became rounded as soon as they were starting to migrate (Supplementary video 6). It was likely that rounded cells were dying cells, which were intended to be eliminated during washes prior to trypsinization. This was confirmed by cell counting, which showed that living adherent AnxA5-AnxA6 deficient cells represented in about half of the amount of control cells (Fig. 8). As control, we have analyzed the behavior of AnxA5-AnxA6 MDA seeded on gelatin, which corresponds to denaturated collagen I (Supplementary video 7). These cells migrated weakly and exhibited a number of rounded and living adherent cells similar to control cells. We observed that AnxA5 or AnxA6 deficiency had little effect on cell density and morphology, even the percentage of rounded cells appeared to be doubled for AnxA6-deficient MDA-MB-231 cells, without significant effect on the number of cells 48 h after seeding. A combined deficiency in AnxA5 and AnxA6 seems therefore to affect the migration of MDA-MB-231 cells on fibrillar collagen I.
To address the possibility that these AnxA5-AnxA6 deficient MDA-MB-231 cells suffer from a defect of membrane repair during their migration on collagen I, cells were loaded with Fluo-4-AM and kinetics study of cell migration on collagen I was performed by fluorescence microscopy. Images acquired in phase-contrast microscopy just before and after kinetics study ensured that cells have migrated (Fig. 9a). We observed important variation of Fluo-4-AM fluorescence in several cells during migration (Fig. 9b). The strong increase of fluorescence intensity revealed the presence of rupture(s) in cell membrane as observed in control cells (Fig. 3). However, in the case of AnxA5-AnxA6 deficient MDA-MB-231 cells we observed that over a large period the fluorescence oscillated between a high intensity and a lesser one (Fig. 9b,c). Through fluorescence videomicroscopy, variations of intensity appeared as a flashing light (Supplementary video 8). It is unlikely that these variations were due to a series of damage and repair events. It is instead more relevant to think that it results from a strong increase of the intracellular calcium concentration subsequent to a membrane damage and an attempt www.nature.com/scientificreports/ from the cell to counteract this Ca 2+ excess, through Ca 2+ absorption by mitochondria for example. Altogether these results suggest that AnxA5-AnxA6 deficient MDA-MB-231 cells are unable to reseal plasma membrane damages induced by their migration on collagen I.

Discussion
Migration on collagen I fibrils as a method for studying membrane repair. For reaching distant tissues, cancer cells are submitted to physical stresses during metastasis, due to their migration and invasion processes through dense extracellular matrix and due to intravasion and extravasion mechanisms for the entry into and the exit from the blood circulation, respectively. Shear stress induced by these processes may lead to cell membrane disruptions. This study aimed mainly at determining whether silencing of membrane repair machinery impairs survival of cancer cells during migration on extracellular matrix. It required first to find a method enabling to modelize shearing forces induced by the extracellular matrix. In this report, we have provided evidences that cell migration on collagen I fibrils leads to plasma membrane damages along with the release of membrane fragments in the wake of the cell. Membrane damage and release can be easily followed by fluorescence microscopy using labeled Ca 2+ indicator and plasma membrane stain. Beyond of its interest for research in cancer biology, this method is an outstanding tool for studying membrane repair in cells that exhibit the capacity of migration, notably epithelial and endothelial cells. It could replace current artificial methods such as laser irradiation, permeabilization using detergent, sprinkling with glass beads or scratching with a needle, that are far away from physiological injuries.
AnxA5 and A6 belong to the membrane repair machinery of MDA-MB-231 cells. We then needed to identify proteins belonging to the membrane repair machinery. Several Anx have been shown to participate in membrane repair processes in different cell types 6,67 , and particularly in cancer cells 15,48,49 . They constituted therefore prime targets for this achievement. Even though AnxA4 has been shown to play an instrumental role in membrane repair of MCF-7 cells 15 , we observed that it is weakly expressed in MDA-MB-231 cells, suggesting its participation in membrane resealing is unlikely. However, AnxA1, A2, A5 and A6 are highly expressed in MDA-MB-231 cells and could be expected to participate in membrane resealing. Our attention focused particularly on AnxA5 and AnxA6, which exhibit varying intracellular concentrations. This heterogeneity of Anx expression probably explains discrepancy in the ability of MDA-MB-231 cells to reseal damaged plasma membrane. Our results demonstrate that AnxA5 and AnxA6 are instrumental for membrane repair in MDA-MB-231 cells. Whether the description of mechanisms related to membrane resealing in cancer cells was outside the scope of this study, some observations that we have done gave interesting clues on the role played by AnxA5 and AnxA6. We have previously proposed that AnxA5 forms 2D-arrays at the edges of the torn membrane allowing to strengthen the membrane and prevent the expansion of the tear 13,20,30 . This hypothesis has been recently confirmed by a sophisticated study performed by Scheuring and collaborators, using notably highspeed atomic force microscopy 66 . They showed that AnxA5 self-assembly into lattices decreases lipid diffusivity, increases membrane thickness and increases lipid order. All together these results show that the formation of AnxA5 2D-arrays at the edges of the torn membrane induces a transition phase that stabilizes and structures the membrane into a gel phase. Here, we have shown that endogenous AnxA5 is recruited rapidly at the disruption site in MDA-MB-231 cells, where it probably self-assembles upon binding to the damaged membrane and promote membrane repair. Damage by laser irradiation in AnxA5-deficient MDA-MB-231 cells leads to the absence of membrane repair, which seems to be due to the expansion of the breach becoming too large to be plugged by the lipid "patch". This observation is totally in line with the expected function of AnxA5 in membrane repair that we have previously reported. AnxA6 is an unusual family member with two instead of one Anx core domains, each domain being composed by four 70-amino acid repeats. Together with other Anx, AnxA6 has been shown to assemble at the site of cell membrane damage and participate in membrane resealing in muscle 14,68,69 and cancer cells as well 15 . A specific role for AnxA6 in membrane resealing of damaged MCF-7 cells have been previously proposed 15 . From wound edges curved in out-of-plane by the action of AnxA4, AnxA6 induces constriction force responsible for the closure of the hole. It is important to note that MCF-7 cells were exposed to strong injuries by laser irradiation inducing a large wound diameter (up to 3 µm) in this study. As the membrane repair mechanism may vary depending on the cell type, the extent of the damage and the spatial position of the injury 6,15,70 , it is likely that the role played by AnxA6 for resealing MDA-MB-231 cells is slightly different in our study. Here, we show that AnxA6-deficient MDA-MB-231 cells suffer from a defect of membrane repair, which seems to be linked to the absence of the recruitment of cytoplasmic lipid material, ie the lipid "patch", to the site of membrane damage. The presence of two Anx core domains gives the ability to AnxA6 to bridge two adjacent membranes, such as plasma membrane and cytoplasmic lipid vesicle. In addition, phylogenetic analysis of Anx has shown that the N-terminal core domain of AnxA6 is similar to AnxA3 whereas the C-terminal one is similar to AnxA5 71 . It has been shown that Anx induce different and specific membrane morphologies upon interaction 72 . For instance, AnxA3 rolls the membrane in a fragmented manner from free edges, producing multiple thin rolls, whereas AnxA5 induces cooperative roll-up of the membrane. AnxA6 may therefore induce multiple membrane rearrangements at the disruption site at the interface between free edges of the damaged plasma membrane and the intracellular lipid "patch". We hypothesize that the absence of AnxA6 may be responsible for the lack of recruitment of the lipid "patch".

Defective membrane repair machinery impairs survival of invasive cancer cells. The migration
of MDA-MB-231 cells on fibrillar collagen I is accompanied by the release of membrane fragments responsible for the disruption of the cell membrane. In wild-type cells, in the presence of AnxA5 and AnxA6, these Scientific Reports | (2020) 10:21821 | https://doi.org/10.1038/s41598-020-77902-5 www.nature.com/scientificreports/ membrane injuries are resealed in less than a minute (Fig. 10a). As observed for damage by laser irradiation, it is likely that the increase of intracellular Ca 2+ concentration induces the recruitment of Anx at the disruption site, where they promote membrane resealing. Once membrane repair is done, Anx come back to the cytosol and the cell can continue to migrate. The combined inhibition of the expression of AnxA5 and AnxA6 leads to a major defect of the membrane repair machinery. In MDA-MB-231 cells, this silencing leads to the incapacity of cells to migrate on fibrillar collagen I (Fig. 10b). In this case, disruptions due to migration are indeed not repaired leading to cell death. Consequently, cell migration is finally discontinued. Strikingly, we have observed that the inhibition of only AnxA5 or AnxA6 was not sufficient to induce death of cells and the stop of migration. It is likely that shear stress induced by collagen fibrils on cell membrane was minor force leading to small ruptures (few hundreds nm), smaller than those created by laser irradiation (1 µm). In this case, the low residual concentration of AnxA5 or AnxA6 may be sufficient for the achievement of membrane repair. One can also envision that the lack of AnxA5 or AnxA6 may be compensated by other Anx. In a physiological context, we hypothesize that mechanical stress would be stronger than in our in vitro experimental conditions. Indeed, a cell that migrates in vivo through the extracellular matrix is submitted to shear forces on the whole surface of the plasma membrane, whereas in our in vitro experimental model shear forces affect only plasma membrane at the interface interacting with the substrate. In addition, we can expect that shear forces induced by other cells during intravasion and extravasion processes are likely to be forces with greater strength.
In conclusion, we show here that silencing of the membrane repair machinery induces a stop of migration of cancer cells and constitutes a promising approach for inhibiting cancer metastasis.
Received: 15 July 2020; Accepted: 17 November 2020 Figure 10. Model of migration of MDA-MB-231 cells equipped with a functional or defective membrane repair machinery. (a) In wild-type cells, migration on fibrillar collagen I induces the release of membrane fragments responsible for membrane injuries (Step 1). Anx, notably AnxA5 and AnxA6, are recruited to the disruption site, where they promote membrane repair (Step 2). Once the cell is repaired, Anx return to the cytosol (Step 3). (b) In cells presenting defective membrane repair, due for example to the absence of Anx, membrane injuries created by the migration on collagen I fibrils are not resealed (Step 2). This leads to the death of cells, which are released from the substrate (Step 3).