Patch repair protects cells from the small pore-forming toxin aerolysin Free

The virulence of many lethal bacteria is mediated partly by the secretion of pore-forming toxins (PFTs). PFTs are the largest family of bacterial toxins and are classified into subfamilies based on structure. Two subfamilies include the cholesterol-dependent cytolysins (CDCs), secreted primarily by Gram-positive bacteria, and the aerolysin family, secreted by a wide range of bacteria. Archetypal toxins representing CDCs include streptolysin O (SLO) from Streptococcus pyogenes and intermedilysin (ILY) from S. intermedius. CDCs form large (20–30 nm) pores that require cholesterol for pore formation (Thapa et al., 2020). SLO also uses cholesterol for binding, whereas ILY binds to the glycosylphosphatidylinositol (GPI)-anchored protein human CD59 (Giddings et al., 2004). Aerolysin is the archetypal aerolysin family member, secreted by Aeromonas hydrophila. Aerolysin forms small (4 nm) pores and binds to GPI anchors (Abrami et al., 1998b). Aerolysin is produced as the inactive precursor pro-aerolysin, which requires cleavage of a C-terminal peptide for activation (Abrami et al., 1998a; Iacovache et al., 2011). Cleavage can be mediated by cellular furins, which are present on most mammalian cell membranes or in vitro using trypsin (Abrami et al., 1998a; Iacovache et al., 2011). After binding, both aerolysin and CDCs form pre-pores on the host plasma membrane that insert and cause cell death if not resisted by the host cell.

Cellular resistance mechanisms are best described for resisting CDCs. To resist CDCs, four Ca²⁺-dependent membrane repair mechanisms have been proposed – toxin removal by caveolar endocytosis, pore clogging by annexins, pore removal by microvesicle shedding, and membrane resealing by patch repair. Although caveolar endocytosis was proposed to remove CDCs (Corrotte et al., 2020), no work to date has demonstrated removal of intact pores by this mechanism. Instead, we previously showed that endocytosis acts after repair is complete to remove toxin oligomers and damaged membranes (Romero et al., 2017). In contrast, annexins form a crystalline lattice to physically blockade the pore (Jimenez and Perez, 2017). CDCs are also sequestered on microvesicles, which are subsequently shed (Keyel et al., 2011). Although shedding can occur independently of proteins, both the endosomal sorting complex required for transport (ESCRT) (Jimenez et al., 2014) and mitogen-activated kinase kinases (MAP2K1 and MAP2K2; hereafter denoted MEK) (Ray et al., 2022) catalyze this process. Shed vesicles contain annexin A1 and annexin A6, although neither is required for shedding (Ray et al., 2022; Ray et al., 2019). Finally, patch repair, the homotypic or heterotypic fusion of vesicles with the membrane (Cooper and McNeil, 2015), involves patch repair proteins including the muscle-specific dysferlin (Cai et al., 2009; Roostalu and Strähle, 2012) and synaptotagmin 7 (Cooper and McNeil, 2015). The contribution of patch repair to resisting CDCs is not known.

In contrast to repair mechanisms triggered by CDCs, no repair mechanisms for aerolysin are known. Understanding repair mechanisms triggered by aerolysin is important because it is expected to enable selective targeting of toxins in the future and provide new insights into the generality of repair mechanisms. Repair of cells following aerolysin challenge might take longer than seen with CDCs, potentially due to the small size of the pore (Cooper and McNeil, 2015; Gonzalez et al., 2011; von Hoven et al., 2017). Furthermore, it is controversial if Ca²⁺ flux is triggered by aerolysin, so it has been questioned whether Ca²⁺-dependent membrane repair responses are activated. Although early studies suggested aerolysin conducts Ca²⁺ (Krause et al., 1998; Nelson et al., 1999; Tschodrich-Rotter et al., 1996), a recent study reported minimal Ca²⁺ flux following aerolysin challenge (Larpin et al., 2021). However, it has also been suggested that Ca²⁺ release from the endoplasmic reticulum (ER) mediates repair (Bittel et al., 2020; Krause et al., 1998; Tschodrich-Rotter et al., 1996). Alternatively, aerolysin could trigger repair downstream of K⁺ efflux (Gonzalez et al., 2011). K⁺ efflux might trigger repair in response to the actinoporin sticholysin and CDC listeriolysin O (Cabezas et al., 2017). Thus, it is unclear what repair responses, if any, are activated by aerolysin.

Here, we tested the hypothesis that cells use similar repair mechanisms to resist both CDCs and aerolysin. Instead, we found a difference in repair responses. Nucleated cells were 2–4-fold more sensitive to aerolysin than to CDCs. Removal of extracellular Ca²⁺ protected cells from aerolysin, suggesting that Ca²⁺-dependent cell death signals outweigh repair benefits. However, when we removed intracellular Ca²⁺, cells became more sensitive to aerolysin. Live-cell imaging showed sustained extracellular Ca²⁺ influx when cells were challenged with aerolysin. This suggested that Ca²⁺-dependent repair was active following aerolysin challenge. We tested four proposed repair pathways – caveolar endocytosis, MEK-dependent shedding, annexin-mediated clogging, and patch repair. We found that caveolar endocytosis was not required for repair against aerolysin or CDCs. MEK-dependent repair was not required to resist aerolysin. Annexin A6 was recruited to the membrane more slowly following aerolysin challenge compared to that seen with CDCs. Importantly, cells transfected with dysferlin were more resistant to aerolysin than to CDCs. This suggests that cells utilize patch repair as the primary repair mechanism against aerolysin. Overall, these findings suggest that aerolysin and CDCs trigger distinct repair responses, which suggests repair responses are tailored to specific toxin threats.

To investigate cellular resistance to aerolysin cytotoxicity, and to compare cytotoxicity between aerolysin and CDCs, we normalized toxin activity using human erythrocyte lysis. One hemolytic unit (HU) was defined as the quantity of toxin required to lyse 50% of erythrocytes solution in 30 min at 37°C (Keyel et al., 2012a; Romero et al., 2017). This means 1 HU is the LC₅₀ for the toxin against human red blood cells. Using toxin activity in erythrocytes instead of toxin mass allows comparison of relative cellular resistance, and controls for variation in specific activity between toxin preparations.

We compared HeLa cell cytotoxicity to SLO, perfringolysin O (PFO), ILY, pro-aerolysin, and trypsin-activated aerolysin (aerolysin) by flow cytometry (Fig. 1A,B; Fig. S1A–C). We selected SLO because it is our best-characterized toxin (Haram et al., 2022; Ray et al., 2022; Romero et al., 2017), and PFO as a second CDC to compare to aerolysin. Finally, we include ILY because it engages the GPI-anchored protein human CD59, and aerolysin also engages GPI-anchored proteins. To compare the cytotoxic activity, we generated dose–response curves, calculated toxin specific lysis at each dose, and fit the dose–response curve to a logistic model to determine the toxin dose needed to kill 50% of the cells (LC₅₀). To control for impurities in toxin purification, we challenged cells with an equivalent mass of the non-hemolytic aerolysin Y221G (aerolysin^Y221G) (Tsitrin et al., 2002), whereas we controlled for trypsin activation using trypsin alone (Fig. S1B). Neither trypsin alone nor aerolysin^Y221G killed cells (Fig. S1B). Given that previous results suggest that aerolysin takes longer to kill cells than CDCs (Cabezas et al., 2017; Gonzalez et al., 2011; Husmann et al., 2009; Ray et al., 2019), we performed a time course to measure aerolysin cytotoxicity. In contrast to SLO, which kills within 5 min post-addition (Keyel et al., 2011; Ray et al., 2019), aerolysin required at least 30 min before robust cytotoxicity was measured (Fig. S1B,C). Cytotoxicity increased between 30 min and 1 h (Fig. S1B,C). As the flow cytometry assay is not suitable for toxin challenge times longer than 1 h, we used an MTT assay to measure viability in independent replicates at 3–24 h. Cytotoxicity was not different between samples in the first 12 h post-challenge (Fig. S1D,E). Consistent with previous results (Ray et al., 2019), both a propidium iodide (PI) uptake assay and MTT were comparable to each other (Fig. 1A,B; Fig. S1). Based on these results, we chose 30 min as the optimal time for comparison to CDCs. When we compared pro-aerolysin or aerolysin to CDCs, we found that HeLa cells were more sensitive (∼5-fold) to aerolysin, independent of the method of activation (Fig. 1A,B). These data indicate that aerolysin is more toxic than CDCs to HeLa cells.

To determine whether this was a cell-specific phenomenon, we tested two additional cell types, murine fibroblasts (3T3), a model cell line for membrane repair, and caspase 1/11^−/− [null for both Casp1 and Casp11 (also known as Casp4)] primary murine bone marrow-derived macrophages (Casp 1/11^−/− BMDMs). We used Casp 1/11^−/− cells to rule out cell death due to inflammasome-dependent Gasdermin D activation (Keyel et al., 2013). The advantage of testing these two cell lines is to give a broader overview of mouse–human differences in sensitivity, and a comparison between immune cells and non-immune cells. When we challenged 3T3 cells with the toxins, we found that 3T3 cells were ∼4- to 6-fold more sensitive to pro-aerolysin and aerolysin compared to PFO and SLO, similar to HeLa cells (Fig. 1C,D). In contrast to 3T3 or HeLa cells, Casp 1/11^−/− BMDMs were ∼100-fold more sensitive to aerolysin compared to SLO and PFO (Fig. 1E,F). The LC₅₀ for aerolysin was lower for Casp 1/11^−/− BMDMs (7.1±1.7 HU/ml) compared to that for HeLa (87.7±8.0 HU/ml) or 3T3 cells (108.7±24.5 HU/ml) (Fig. 1). This stands in stark contrast to what was seen with CDCs, where BMDMs are ∼5–10-fold more resistant than HeLa cells (Keyel et al., 2013). Based on these data, we conclude that nucleated cells are more sensitive to aerolysin than to CDCs.

The increased sensitivity of aerolysin could be due to reduced repair or activation of cell death pathways. It is controversial whether cells utilize Ca²⁺- or K⁺-dependent repair responses to resist aerolysin (Cabezas et al., 2017; Gonzalez et al., 2011; Larpin et al., 2021). We tested the impact of ion flux on aerolysin either by chelating extracellular Ca²⁺ with EGTA, or by preventing K⁺ efflux with addition of 150 mM KCl to the extracellular medium. We found no change in the LC₅₀ for aerolysin or SLO when K⁺ efflux was blocked (Fig. 2A; Fig. S2A). Consistent with previous results (Ray et al., 2019; Romero et al., 2017), lack of extracellular Ca²⁺ sensitized cells to SLO (Fig. 2B; Fig. S2B). In contrast, removal of extracellular Ca²⁺ protected HeLa cells from pro-aerolysin and aerolysin because the LC₅₀ increased ∼10-fold (Fig. 2B; Fig. S2B). To confirm that Ca²⁺ chelation prevented cell death instead of interfering with toxin binding, we measured the binding of Alexa Fluor 647-conjugated pro-aerolysin^K244C (pro-aero^K244C) (Iacovache et al., 2006). Pro-aero^K244C bound equally well to HeLa cells in a dose-dependent manner independently of extracellular Ca²⁺ (Fig. S2C). Based on these data, we conclude that Ca²⁺ influx, but not K⁺ efflux, is critical for aerolysin-mediated killing.

Given that Ca²⁺ is toxic to cells following aerolysin challenge, we determined the amount of extracellular Ca²⁺ necessary for cell death. The medium used in killing assays, RPMI, contains 0.4 mM Ca²⁺, so we used Tyrode's buffer to control extracellular Ca²⁺. We first compared HeLa cell cytotoxicity between RPMI and Tyrode's buffer, each supplemented with 2 mM CaCl₂. Aerolysin was equally potent in both buffers, but SLO and ILY were more potent in Tyrode's buffer compared to RPMI (Fig. 2C; Fig. S2D). To measure the impact of extracellular Ca²⁺, we measured cytotoxicity at a range of Ca²⁺ concentrations using a toxin dose that killed ∼70% of cells at 2 mM Ca²⁺. Reducing extracellular Ca²⁺ to ∼10 μM protected cells from aerolysin (Fig. 2D). Notably, this is consistent with prior results (Iacovache et al., 2016) showing that aerolysin can form pores at low/no Ca²⁺, and suggests EGTA does not interfere with oligomerization or pore formation. In contrast, CDC killing plateaued at ∼20 μM (Fig. 2D). We conclude that aerolysin cytotoxicity is caused by Ca²⁺ overload.

To better measure Ca²⁺ overload, we labeled cells with fluo-4 and measured Ca²⁺ influx in HeLa cells challenged with aerolysin or SLO. To analyze Ca²⁺ flux, we categorized cells into three subsets based on their survival time (Ray et al., 2022) – cells dying within 5 min of toxin challenge, cells dying within 5–45 min of toxin challenge, and cells surviving throughout the entire 45 min imaging experiment (Fig. 3A; Fig. S2E, Movie 1). We chose these categories to stratify cells that were immediately overwhelmed and killed by toxin, cells dying during the course of the imaging, and cells with functioning proximal repair mechanisms. We expect cell death occurring after 45 min to be the result of repair-independent mechanisms, so these processes were not examined. Although overall cell death was comparable between SLO and aerolysin, SLO killed more cells within 5 min of toxin challenge than aerolysin (Fig. 3A), consistent with our findings that aerolysin kills more slowly than SLO (Fig. S1B,C). In contrast, minimal cell death was observed in cells challenged with aerolysin plus EGTA, non-hemolytic aerolysin^Y221G or PBS alone (Fig. 3A). The maximal Ca²⁺ flux from control cells was lower compared to toxin-challenged cells (Fig. 3B,C; Movie 2). Notably, the maximal Ca²⁺ flux was close to saturating in our system (fMAX=1024) for toxin-challenged cells, except for aerolysin-challenged cells that died within 5 min of challenge (Fig. 3C; Movie 1). When we measured the time to reach the fMAX, all cells that died within 5 min had a rapid peak in Ca²⁺ concentration (∼4 min) (Fig. 3D,E). Although surviving cells had a longer time to reach fMAX compared to cells dying between 5 and 45 min, SLO challenged cells reached fMAX faster than aerolysin challenged cells (Fig. 3D,F,G). SLO challenged cells showed rapid Ca²⁺ oscillations throughout the imaging period (Movie 1). In contrast, aerolysin-challenged cells showed delayed, yet sustained levels of Ca²⁺, with reduced oscillations (Fig. 3D,F,G; Movie 1). The steady, sustained increase in Ca²⁺ flux caused by aerolysin required pore formation because we did not observe similar effects with cells challenged with PBS, aerolysin with EGTA or aerolysin^Y221G (Fig. 3H). We quantified Ca²⁺ oscillations by determining the realized volatility of the Ca²⁺ flux for each cell. SLO-challenged cells had greater volatility compared to aerolysin-challenged cells (Fig. 3I; Fig. S2F). This analysis showed that the non-hemolytic aerolysin^Y221G also triggered Ca²⁺ oscillations above background, although EGTA reduced Ca²⁺ oscillations to background levels (Fig. 3J; Fig. S2F). These data suggest that pores formed by aerolysin trigger a sustained Ca²⁺ flux that leads to cell death.

Although extracellular Ca²⁺ is toxic, intracellular Ca²⁺ could trigger membrane repair responses. Repair proteins, including annexins and dysferlin, bind Ca²⁺ during repair (Cooper and McNeil, 2015). A prior study suggested that intracellular Ca²⁺ can drive repair (Bittel et al., 2020). To test whether intracellular Ca²⁺ triggers repair responses to aerolysin, we chelated intracellular Ca²⁺ using BAPTA-AM and challenged cells with aerolysin with or without also chelating extracellular Ca²⁺. Removal of extracellular Ca²⁺ protected cells from aerolysin regardless of intracellular Ca²⁺ chelation (Fig. 4A; Fig. S3A). However, chelation of intracellular Ca²⁺ showed a trend to sensitizing cells to aerolysin (Fig. 4A; Fig. S3A). These data suggest that Ca²⁺-dependent repair pathways could be activated by aerolysin.

To examine Ca²⁺-dependent repair, we first examined MEK-dependent repair. We recently showed that ∼70% of Ca²⁺-dependent repair of damage caused by CDCs is activated by MEK (Ray et al., 2022). To determine whether MEK signaling contributes to survival following aerolysin challenge, we blocked MEK activation with the MEK inhibitor U0126 (Fig. 4). U0126 blocked MEK activation (Fig. 4B,D), confirming inhibitor activity. We then challenged MEK-inhibited cells with aerolysin or SLO. MEK inhibition did not increase cellular sensitivity to aerolysin (Fig. 4; Fig. S3). As previously observed (Ray et al., 2022), MEK inhibition increased CDC-dependent cell death (Fig. 4C,E; Fig. S3B,C). Based on these data, we conclude that MEK-dependent repair is not a major contributor to cell survival from aerolysin.

If MEK is not a major contributor to survival from aerolysin, another could MAP kinase mediate protection. Both p38 and JNK kinase families are implicated in long-term viability, protecting cells (Huffman et al., 2004) and recovering ion homeostasis after pro-aerolysin challenge (Gonzalez et al., 2011). We tested if they contribute to membrane repair. Neither p38 nor JNK inhibition altered repair responses to aerolysin or SLO in HeLa cells (Fig. 4E; Fig. S3C). However, JNK inhibition increased the sensitivity of HeLa cells to ILY (Fig. 4E; Fig. S3C). Thus, we conclude that MAP kinases are not a major repair mechanism against aerolysin.

We next tested one proposed mechanism of membrane repair – caveolar endocytosis. Although we have shown that caveolar endocytosis restores homeostasis instead of mediating repair (Keyel et al., 2011; Romero et al., 2017) following CDC challenge, caveolar endocytosis might mediate aerolysin resistance. To test the role of caveolar endocytosis in aerolysin resistance, we used cells where cavin-1 (also known as PTRF) was inactivated by CRISPR (Liu and Pilch, 2016). Cavin-1 deficiency prevents caveolae formation (Hill et al., 2008; Liu and Pilch, 2008), and cavin-1-deficient murine embryonic fibroblasts (MEFs) may have repair defects (Corrotte et al., 2020). We used 3T3-L1 cells instead of MEFs because 3T3-L1 cells can be differentiated into adipocytes, which are richer in caveolae (Scherer et al., 1994; Thorn et al., 2003). If caveolar endocytosis is important for repair, any phenotype would be expected to be exacerbated in adipocytes. We differentiated two distinct cavin-1-deficient 3T3-L1 cell lines (K5 and K3) and one CRISPR-control 3T3-L1 cell line (KN; wild type). To control for any differences due to duration of differentiation time, we compared earlier (day 9) and later (day 14) adipocytes. Consistent with previous results (Liu and Pilch, 2016), and similar to what we found in HeLa cells, cavin-1 expression was restricted to 3T3-L1 wild-type cells (Fig. 5A). To determine the role of caveolar endocytosis in repair in cells with low levels of caveolae, we challenged undifferentiated wild-type, K3 or K5 3T3-L1 cells with aerolysin or SLO. Cavin-1 deficiency did not reduce the LC₅₀ for either toxin (Fig. 5B; Fig. S4A). These data suggest caveolar endocytosis is not involved in cellular protection or membrane repair against aerolysin or SLO.

We next tested cellular resistance under conditions with higher caveolar levels. The number of caveolae increase during adipocyte differentiation (Thorn et al., 2003). We challenged 3T3-L1 cells differentiated into adipocytes for 9 or 14 days with aerolysin or SLO, using HeLa cells as a control. All 3T3-L1 adipocytes were resistant to both SLO and aerolysin, regardless of time point (Fig. 5C,D). One cavin-1^−/− line, K5, had increased toxin resistance (Fig. 5C) at day 9. However, this did not hold at day 14. The toxins were active because HeLa cells challenged at the same time with the same toxin aliquot remained sensitive (Fig. 5C,D). We were unable to calculate the LC₅₀ for the adipocytes because the sensitivity was above baseline cell death, but not dose-dependent in the toxin range used. This resistance could be due to an inability of toxins to bind to differentiated adipocytes. To test whether adipocytes no longer bound to toxins, we challenged HeLa cells or 3T3-L1 adipocytes with fluorescently conjugated pro-aero^K244C or SLO ML (SLO ML Cy5). Both HeLa cells and 3T3-L1 adipocytes bound toxins in a dose-dependent manner (Fig. S4B–E). We interpret these data as indicating that adipocytes are resistant to pore-forming toxins like SLO and aerolysin independently of caveolae. We conclude that caveolar endocytosis does not protect cells from aerolysin or SLO.

We next examined the ability of annexins to protect cells from aerolysin. To determine the role of annexins in repair against aerolysin, we knocked down three different annexins using RNAi (siRNAs) – annexin A1 (A1), annexin A2 (A2), and annexin A6 (A6). We selected these annexins because A2 and A6 are recruited early to plasma membrane damage, whereas A1 is recruited later, based their Ca²⁺ sensitivity (Roostalu and Strähle, 2012; Wolfmeier et al., 2015), and we previously characterized their activity during SLO challenge (Ray et al., 2022). The knockdown efficiency was 65–95% (Fig. 6A,B). We then challenged these cells with aerolysin and found no significant difference in the sensitivity (Fig. 6C; Fig. S5A). However, knockdown of annexins increased cell sensitivity to SLO, and showed a trend for increasing sensitivity to ILY (Fig. 6C; Fig. S5B,C). These data suggest that annexins are unlikely to contribute to aerolysin resistance to the same degree they contribute to CDC resistance.

We followed up the annexin RNAi by examining annexins during toxin challenge using high resolution live cell imaging. We measured cytosolic depletion and membrane recruitment of A6-YFP after challenging HeLa cells with a sublytic dose of aerolysin or SLO for 45 min (Fig. 6D,E; Movie 3). After 45 min, we added 1% Triton X-100 as a positive control for membrane translocation and TO-PRO3 uptake. Active toxins, but not aerolysin^Y221G, induced A6 translocation from the cytosol to the plasma membrane (Fig. 6D,E; Fig. S5D,E, Movie 3). Trypsin activation of aerolysin accelerated A6 cytosolic depletion and membrane translocation (Fig. 6D,E; Fig. S5D,E). A6 cytosolic depletion took 10 min for aerolysin, or 15 min for pro-aerolysin compared to ∼6 min for SLO (Fig. 6D). The faster cytosolic loss of A6 for SLO was paralleled by a faster membrane recruitment of A6 (∼5 min) (Fig. 6E; Fig. S5D,E). We found a similar trend of membrane enrichment of A6 for aerolysin (∼10 min) or pro-aerolysin (∼15 min) (Fig. 6E). Given that annexin depletion does not measure viability or permeability, we analyzed TO-PRO3 uptake. We found that all toxins induced A6–YFP translocation without TO-PRO3 accumulation (Fig. S5F, Movie 3).

We next analyzed microvesicle shedding by toxin challenged cells. As in previous studies (Babiychuk et al., 2009; Ray et al., 2022, 2019), we used A6 as a marker for microvesicle shedding. Pro-aerolysin and aerolysin induced a ∼2-fold lower shedding of A6-enriched microvesicles compared to SLO (Fig. 6F). Trypsin treatment of aerolysin did not change the rate of microvesicle shedding (Fig. 6F). Aerolysin-triggered shedding was increased compared to that induced by non-hemolytic oligomeric aerolysin^Y221G, which was minimal (Fig. 6F). These data are consistent with the inability of aerolysin to activate MEK-dependent repair (Fig. 4). We interpret these data to suggest that aerolysin might activate the same MEK-independent shedding observed with CDCs (Ray et al., 2022). Thus, aerolysin does not trigger microvesicle shedding to the same extent as CDCs.

We next examined microvesicle shedding by ultracentrifugation and western blotting. We isolated microvesicles shed by non-transfected HeLa cells following toxin challenge by ultracentrifugation. To control for spontaneously released vesicles, we included an equivalent amount of trypsin used to activate aerolysin. We detected aerolysin using an affinity-purified custom rabbit polyclonal antibody raised against aerolysin^Y221G. We validated the antibody specificity against both aerolysin and aerolysin^Y221G by western blotting, ELISA, and immunofluorescence (Figs S5G, S6). We found that aerolysin and SLO were both shed on microvesicles (Fig. 6G; Fig. S5G). Aerolysin^Y221G localized at low levels on microvesicles (Fig. 6G; Fig. S5G). Notably, SLO was shed to a greater degree than aerolysin, consistent with our A6 shedding results (Fig. 6F). Along with active toxins, we found that the membrane repair proteins A1 and Alix were shed in microvesicles (Fig. 6G), consistent with the previous results (Jimenez et al., 2014; Romero et al., 2017). To control for cellular debris from lysed cells in our shed microvesicles, we probed for nuclear protein Lamin A/C. We found that Lamin A/C localized to cell pellet only (Fig. 6G), indicating that the microvesicle fraction did not contain detectable cellular debris. Trypsin alone did not induce shedding of any repair proteins (Fig. 6G). These findings confirm that cells shed microvesicles following aerolysin challenge, but at a reduced extent compared to what is seen after challenge with CDCs. Based on these data, we conclude that microvesicle shedding is not the primary mechanism used to resist aerolysin pores.

Finally, we tested the role of patch repair in resisting aerolysin. We used the well-characterized muscle patch repair protein dysferlin because it participates in patch repair (Cooper and McNeil, 2015) and is also amenable to both loss-of-function and gain-of-function studies. We tested loss of function using C2C12 myoblasts stably transfected with control or dysferlin shRNA. We validated dysferlin knockdown by real-time PCR. We found that dysferlin shRNA-expressing C2C12 cells had 80% knockdown of dysferlin at the mRNA level compared to control cells (Fig. S7A). These myoblasts had a significant increase in sensitivity to aerolysin and SLO compared to control C2C12 cells (Fig. 7A; Fig. S7B). These data suggest the loss of the patch repair protein dysferlin from myoblasts curtails repair against aerolysin.

Owing to redundancies in patch repair, we also used a gain-of-function approach. We used exogenous expression of dysferlin in HeLa cells to measure increases in patch repair. We transfected HeLa cells with GFP or GFP–dysferlin, and challenged cells with toxins. GFP transfection did not significantly increase cellular resistance to aerolysin, SLO or ILY (Fig. 7B,C; Fig. S7C,D). Dysferlin increased SLO resistance beyond GFP alone, but was not effective in protecting cell integrity against another CDC, ILY (Fig. 7B). Dysferlin increased cellular resistance to aerolysin to a greater extent than resistance to SLO (Fig. 7B,C; Fig. S7C,D). To confirm the Ca²⁺ dependence of dysferlin-mediated protection, we challenged transfected cells with or without extracellular Ca²⁺. Dysferlin-mediated repair was dependent on Ca²⁺ (Fig. 7C). Thus, dysferlin expression protects cells against aerolysin.

Although dysferlin is associated with patch repair, it might not mediate resistance via patch repair. To confirm that dysferlin mediated patch repair, we measured GFP and GFP–dysferlin dynamics during toxin challenge by high resolution imaging. GFP was cytosolic, and active toxins depleted GFP from the cell over time (Movie 4). GFP–dysferlin localized to the plasma membrane and vesicles independently of extracellular Ca²⁺ (Fig. 7D,E; Movie 5). When challenged with a sublytic aerolysin dose in the presence of Ca²⁺, we found that dysferlin-enriched vesicles moved to the plasma membrane from cytosol and protected the integrity of plasma membrane by patching and blebbing (Fig. 7D; Movie 5). The non-hemolytic oligomeric aerolysin^Y221G did not trigger dysferlin-mediated patching (Fig. 7D; Movie 5). To study the Ca²⁺ dependency of dysferlin-positive vesicle fusion with the plasma membrane, we removed extracellular Ca²⁺ with 2 mM EGTA and challenged cells with sublytic toxin doses. We found less depletion of dysferlin-positive vesicles from cytosol (Fig. 7E; Movie 6) and cells were more resistant to aerolysin, confirming our previous results (Figs 2B, 4A,C). When challenged with sublytic SLO or ILY in presence of EGTA, cells showed larger bleb formation and less dysferlin-positive vesicle transport to the plasma membrane (Fig. 7E; Movie 6). However, at later stages dysferlin-positive vesicles moved from the cytosol to plasma membrane, suggesting that intracellular Ca²⁺ could activate patch repair as a last resort. Based on these data, we conclude that patch repair is the primary mechanism used to resist aerolysin pores.

Taken together, our data support a model where different toxins are resisted by different hierarchies of repair. CDCs are primarily resisted by MEK-dependent microvesicle shedding (Ray et al., 2022), and partly resisted by annexins, which could mediate clogging of the pores (Fig. 8). Patch repair is activated as a last resort (Fig. 8). In contrast, aerolysin does not trigger MEK-dependent microvesicle shedding. We propose that patch repair is the primary mechanism of repair for aerolysin, with support from annexins (Fig. 8). We attribute this difference to the activation of a Ca²⁺-dependent death pathway. An alternative explanation is that repair is toxin-independent, but instead depends on the extent of membrane damage. This explanation is unlikely in our opinion because we observed consistent differences in cell resistance across a range of toxin doses normalized to hemolytic units. Furthermore, these activities were compared in the same cell line that supports binding of both toxins. Neither toxin saturates binding in this cell line (Fig. S4) (Romero et al., 2017). Thus, we conclude pore-forming toxins trigger distinct repair mechanisms.

In this study, we demonstrate that patch repair is the primary repair response used by cells to protect themselves against the small pore-forming toxin aerolysin. Aerolysin is more toxic to nucleated cells compared to CDCs. The increased susceptibility to aerolysin depends on Ca²⁺ toxicity, and the limited activation of protective repair pathways. Caveolar endocytosis did not protect cells against any toxin we tested. However, intracellular Ca²⁺ activated delayed annexin recruitment, limited shedding and patch repair. Of these repair mechanisms, patch repair was the primary repair response, which contrasts with repair activated by CDCs. This suggests that toxins trigger distinct membrane repair pathways downstream of Ca²⁺ flux. Overall, our results provide clarity into repair responses activated by aerolysin.

We found that aerolysin was more potent against nucleated mammalian cells compared to CDCs. In order to compare toxins with distinct sizes, subunits in the complete pore and relative killing capacity, we normalized cytotoxicity using hemolytic activity. Given that erythrocytes are not thought to have any specific membrane repair mechanisms, they provide a baseline for toxin activity. When normalized to equal hemolytic activity, we found that cells are highly sensitive to aerolysin compared to CDCs. In contrast to SLO, but similar to PFO, we found that sensitivity increased over time (Ray et al., 2019). Our results are consistent with previous results suggesting that aerolysin is a potent toxin (Gonzalez et al., 2011; Larpin et al., 2021). Thus, these findings suggest that aerolysin triggers minimal repair compared to CDCs.

Although limited repair is one possible interpretation of the results, another possibility is that aerolysin triggers a death pathway that overcomes any repair pathways. We investigated this possibility by controlling ion flux in cells challenged with aerolysin or CDCs. In contrast to prior results suggesting that K⁺ depletion was important during aerolysin challenge (Cabezas et al., 2017), we found no impact on cell survival at 30 min when blocking K⁺ depletion. However, we measured acute responses in HeLa cells, whereas other studies used a longer time period (1–6 h) in HT29 (Gonzalez et al., 2011) or BHK (Cabezas et al., 2017) cells. Although prior studies did not report hemolytic activity of their toxin preparations, we used a higher mass of toxin, potentially due to differences in purification. In our system, we found that Ca²⁺ influx was essential for aerolysin cytotoxicity.

Compared to CDCs, aerolysin triggered a delayed and sustained Ca²⁺ flux characterized by few oscillations, as measured by volatility. We interpret the sustained influx and limited volatility as the activation of cellular Ca²⁺-conducting protein(s) and/or lipids. In contrast, we interpret the CDC-driven volatility as the cell attempting to control intracellular Ca²⁺ concentrations, possibly due to membrane resealing, or possibly attempting to reverse activation of cellular Ca²⁺-conducting protein(s) and/or lipids. These results are consistent with previous findings (Krause et al., 1998) supporting aerolysin-dependent Ca²⁺ flux. However, our results diverge from a previous study (Larpin et al., 2021) that used less toxin and did not detect Ca²⁺ flux in THP.1 or U937 cells. One possibility is that THP.1 and U937 cells are unable to activate the cellular Ca²⁺-conducting protein(s) and/or lipids. The activation of a Ca²⁺ death pathway could also obscure activation of repair mechanisms against aerolysin. Overall, activation of a cell type-specific Ca²⁺ death pathway that obscures any repair could reconcile conflicting interpretations (Krause et al., 1998; Larpin et al., 2021) of the impact of Ca²⁺ influx on aerolysin. However, it remains to be determined how Ca²⁺ mediates toxicity in aerolysin-challenged cells.

We determined which of the repair mechanisms triggered by CDCs were activated in response to aerolysin. Caveolar endocytosis has been proposed as a repair mechanism (Corrotte et al., 2020), but we found caveolar endocytosis did not promote repair against either aerolysin or CDCs. However, we tested pre-adipocyte and adipocyte cell lines instead of MEFs. Cavin-1 could be required for the membrane homeostasis of cholesterol or other key repair lipids/proteins in MEFs. Notably, intact toxin pores perforating the membrane have been observed on shed blebs (Keyel et al., 2011) but not in the membrane of internal organelles. Our findings here are consistent with our prior finding (Romero et al., 2017) that caveolar endocytosis does not contribute to repair but might restore cellular homeostasis by removing inactive toxin and toxin laden blebs after repair is complete.

Aerolysin, unlike CDCs, was not resisted by MEK-dependent repair. Given that MEK dependent repair accounted for 70% of the repair against CDCs (Ray et al., 2022), failure of this mechanism to restrict damage could account for the increased cytotoxicity of aerolysin. We also tested p38 and JNK activation because they were implicated in restoring ion homeostasis and long-term viability following aerolysin challenge (Gonzalez et al., 2011; Huffman et al., 2004). In contrast to previous results (Gonzalez et al., 2011; Huffman et al., 2004), we did not find changes in repair responses after blocking p38 or JNK MAPK pathways. This difference could be due to the differences in toxin preparation, duration of toxin challenge or the difference in time frame we used (30–60 min), compared to previous studies (6+ hours/days). Overall, this suggests that p38 and JNK do not contribute to proximal membrane repair events.

We next compared the ability of three annexins (A1, A2 and A6) to promote repair following CDC or aerolysin challenge. Deletion of A1, A2 or A6 did not increase cellular sensitivity to aerolysin, but did sensitize cells to SLO and ILY. This suggests that annexins either serve redundant functions in aerolysin resistance, or A1, A2 and A6 are minimally involved. We propose they are minimally involved because they also showed delayed translocation. The delayed translocation of annexins is consistent with the slower increase in Ca²⁺ that we observed. Delayed, steady Ca²⁺ influx could account for the failure of annexins to contribute to protecting cells from aerolysin. It remains possible that other annexins resist aerolysin. It remains unknown how previously described interactions of annexins with other repair proteins, including dysferlin, MG53 and ESCRT III (Bittel et al., 2020; Demonbreun et al., 2016; Lennon et al., 2003; Roostalu and Strähle, 2012; Sonder et al., 2019), impact repair.

Although annexins are associated with clogging, they also mark cellular shedding from the membrane. Shedding of A6⁺ microvesicles was reduced by approximately half for cells challenged with aerolysin compared to SLO. Given that we observed a 2-fold reduction in shedding when MEK was inhibited (Ray et al., 2022), we interpret these results as indicating that aerolysin activates the MEK-independent shedding also observed with CDCs. Ultracentrifugation confirmed that shedding occurs, despite being reduced. Whether the MEK-independent repair is lipid mediated (Keyel et al., 2011) or ESCRT mediated (Jimenez et al., 2014), remains to be determined. However, a low contribution of repair by MEK and annexins is consistent with the enhanced cytotoxicity of aerolysin.

In contrast to other repair pathways, we found that enhancing patch repair made cells resistant to aerolysin. Reduction of dysferlin in myoblasts reduced their ability to resist aerolysin and SLO. Overexpression of the muscle patch repair protein dysferlin protected cells from aerolysin, to a lesser extent from SLO, and not from ILY. Interestingly, both ILY and aerolysin target GPI-anchored proteins, although ILY is restricted to human CD59 (Abrami et al., 1998b; Giddings et al., 2004). We did not rule out a direct effect of dysferlin on toxic Ca²⁺ levels. Dysferlin could reduce Ca²⁺ toxicity in aerolysin-challenged cells either by sequestering excess cytoplasmic Ca²⁺ with its seven Ca²⁺-binding C2 domains, as reported in t-tubules (Kerr et al., 2013), or by blocking the activation of the Ca²⁺-dependent death effector protein(s) and/or lipids. However, two lines of evidence suggest that patch repair is the mechanism of action in our system. First, survival did not correlate with the extent of dysferlin overexpression, which would be expected if dysferlin acted as a Ca²⁺ sink. For example, very bright dysferlin-positive cells were not more resistant than dim dysferlin-positive cells. Second, we observed depletion of dysferlin-positive vesicles as they fused with the plasma membrane during toxin insult. Thus, we conclude that patch repair is the primary resistance mechanism against aerolysin.

One new horizon opened by our work is determining how dysferlin-enhanced patch repair prevents aerolysin-induced death. Although vertex fusion (McNeil and Kirchhausen, 2005) to displace aerolysin pores from the plasma membrane is one possibility, another possibility is that endolysosomal proteins and/or lipids delivered to the membrane clog or remove the aerolysin pore. Alternatively, dysferlin and/or endolysosomal components could interfere with the Ca²⁺-dependent death pathway via an unknown mechanism.

Our work has limitations. Although we showed Ca²⁺ is toxic to cells after aerolysin challenge and removing Ca²⁺ with EGTA protected cells, the signaling intermediate between Ca²⁺ influx and the death effector remains unknown. Finally, we focused on three key repair responses, but did not examine ESCRT-mediated repair, sphingomyelin/ceramide, other annexins or other repair proteins. These factors could also contribute to repair against aerolysin. Similarly, although this work builds a foundation on which to investigate interactions between repair pathways, we did not pursue potential overlap or redundancy of pathways.

Overall, the key contribution of this study is the comparison of multiple repair responses, and determining which is used by cells against the small PFT aerolysin. Although MEK-independent microvesicle shedding was observed, neither MEK nor annexins appear to drive major cellular resistance. Both of these responses were reduced compared to responses against CDCs. Given that CDCs are less potent than aerolysin against nucleated cells, we propose aerolysin evades the key Ca²⁺-dependent repair mechanisms of shedding and annexin recruitment. However, patch repair is activated as a last resort when other repair mechanisms fail to reseal the damage.

All reagents were from Thermo Fisher Scientific (Waltham, MA, USA) unless otherwise noted. The MEK inhibitor U0126 was from Cell Signaling Technology (Danvers, MA, USA) (Cat# 9903S) or Tocris (Minneapolis, MN, USA) (Cat# 1144). The p38 MAPK inhibitor SB203580 (Cat # AG-CR1-0030-M005) was from AdipoGen Life Sciences (San Diego, CA, USA). The MAPK9/JNK1/2 inhibitor SP600125 (Cat # S1460) was from Selleckchem (Houston, TX, USA). siRNAs for annexin A1 and A2 were designed and ordered from Cytiva (Marlborough, MA, USA). Negative control siRNA (Cat # 462001) and annexin A6 siRNA were from Ambion (Cat# AM16708). BAPTA-AM (Cat# 15551) was from Cayman Chemical (Ann Arbor, MI, USA). Propidium iodide (PI) was from Sigma-Aldrich (St. Louis, MO, USA) (Cat# P4170-100MG) or Biotium (Fremont, CA, USA) (Cat# 40016). DAPI (Cat# D9542), dexamethasone (Cat# D2915-100MG) and insulin (Cat# I6634-50MG) were from Sigma-Aldrich. 3-Isobuty-1-methylxanthine (IBMX) (Cat# 195262) was from MP Biomedicals (Solon, OH, USA). MTT reagent (Cat# 2809-1G) was from BioVision (Milpitas, CA, USA).

The anti-annexin A1 monoclonal antibody (mAb) (Clone: CPTC-ANXA1-3-s) was deposited to the Developmental Studies Hybridoma Bank (DSHB) by Clinical Proteomics Technologies for Cancer (DSHB Hybridoma Product CPTC-ANXA1-3) and anti-Lamin A/C (MANLAC-4A7-s) was deposited to the DSHB by GE Morris (DSHB Hybridoma Product MANLAC1(4A7)). Both were obtained from the DSHB created by the NICHD of the NIH and maintained at the University of Iowa, Department of Biology (Iowa City, IA, USA). Anti-β-actin AC-15 mAb (Cat# GTX26276) was from GeneTex (Irvine, CA, USA). Anti-SLO antibody 6D11 mAb (catalog: NBP1-05126) was from Novus Biologicals (Littleton, CO, USA). Anti-Alix 3A9 mAb was from BioLegend (San Diego, CA, USA). Anti-PTRF rabbit polyclonal Ab (Cat# ab48824) was from Abcam (Cambridge, MA, USA). Anti-MEK (Cat# 9122L), anti-phospho-MEK[Ser²¹⁷/Ser²²¹] (Cat# 9121S), anti-ERK p44/42 (Cat# 9102S), anti-phospho-ERK[Thr²⁰²/Tyr⁴⁰⁴] (Cat# 9101S), anti-p38 (Cat# 9212S), and anti-phospho-p38 (Cat# 94511S) rabbit polyclonal antibodies were from Cell Signaling Technologies (Danvers, MA, USA). Anti-mouse-IgG (Cat# 711-035-151) and anti-rabbit-IgG (Cat# 711-035-152) HRP-conjugated antibodies were from Jackson Immunoresearch (West Grove, PA, USA).

Non-hemolytic oligomeric aerolysin Y221G (Aero^Y221G) was purified and concentrated to 5 mg/ml at 95% purity. Two New Zealand White rabbits were immunized and bled by Pacific Immunology (Ramona, Ca, USA), as overseen by their Institutional Animal Care and Use Committee. After bleeding to collect pre-immune serum, rabbits were immunized once in complete Freund's adjuvant, a second time on day 21 in incomplete Freund's adjuvant, a third time on day 42 in incomplete Freund's adjuvant, and a fourth time on day 70 in incomplete Freund's adjuvant. Antisera were collected on days 49, 63, 77, 91, with final exsanguination bleeds on days 98 and 101. The antisera were affinity purified against pro-aerolysin. For affinity purification, pro-aerolysin was dialyzed with coupling buffer overnight at 4°C. Then, pro-aerolysin was coupled to cyanogen bromide-activated Sepharose beads for 2 h at room temperature (RT) and washed with coupling buffer. Remaining sites on the Sepharose were quenched using 0.1 M Tris-HCl (pH 8) for 2 h at RT. Coupled beads were then washed alternatively with acid wash (0.1 M sodium acetate, 0.5 M sodium chloride, pH 4) and alkali wash (0.1 M Tris-HCl, 0.5 M sodium chloride, pH 8) buffers four times. Anti-aerolysin antiserum was then incubated with beads for 2 h at RT and the mixture loaded on an empty column. The flowthrough was collected and saved as depleted serum. Beads were washed with PBS followed by 150 mM NaCl, and the absorbance at 280 nm (A₂₈₀) was measured until it reached <0.01. Once the A₂₈₀ was <0.01, high-affinity antibodies were eluted with 0.1 M glycine elution buffer (0.1 M glycine, pH 2.5 with HCl) and collected into 1 ml fractions with 50 µl 1 M Tris-HCl (pH 8) until A₂₈₀ was low. The column was washed with 0.1 M Tris-HCl. Next, antibodies were again eluted into 1 M Tris using 0.1 M triethylamine (TEA) elution buffer (0.1 M TEA, pH 11.5). For each fraction, the antibody concentration was calculated using 1.4 A₂₈₀=1 mg/ml antibody. Relevant elutions were pooled together and final antibody concentration was calculated. Antibody validation is summarized in Fig. S6.

The pET22b plasmids encoding His-tagged aerolysin and aerolysin^Y221G were kind gifts from Gisou van der Goot (École Polytechnique Fédérale de Lausanne, Canton of Vaud, Switzerland) (Tsitrin et al., 2002). The wild-type aerolysin contained Q254E, R260A, R449A and E450Q mutations and the C-terminal KSASA sequence was replaced with NVSLSVTPAANQLE HHHHHH compared to sequences in GenBank (M16495.1). The K244C mutation was introduced into wild-type aerolysin by Quikchange mutagenesis. Cys-less, His-tagged, codon-optimized SLO C530A and ‘monomer-locked’ SLO G398V/G399V (SLO ML) were previously described (Ray et al., 2022). Cys-less His-tagged PFO in pET22 (Shepard et al., 1998) and Cys-less His-tagged ILY in pTrcHisA (Giddings et al., 2004) were generous gifts from Rodney Tweten (University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA). Human Annexin A6 fused to YFP was a generous gift from Annette Draeger (University of Bern, Bern, Switzerland) (Babiychuk et al., 2009). GFP–dysferlin was cloned into pcDNA4.

All experimental mice were housed and maintained according to Texas Tech University Institutional Animal Care and Use Committee (TTU IACUC) standards, adhering to the Guide for the Care and Use of Laboratory Animals (8th edition, NRC 2011). TTU IACUC approved mouse use. Casp1/11^−/− mice on the C57BL/6 background were purchased from the Jackson Laboratory (Bar Harbor, ME, USA) (stock #016621). Mice of both sexes aged 6–15 weeks were used to prepare BMDMs. Sample size was determined as the minimum number of mice needed to provide enough bone marrow for experiments. Consequently, no randomization or blinding was needed. Mice were euthanized by asphyxiation through the controlled flow of pure CO₂ followed by cervical dislocation.

All cell lines were maintained at 37°C, 5% CO₂ and tested for mycoplasma contamination. HeLa cells (ATCC, Manassas, VA, USA; CCL-2) were cultured in DMEM (Corning, Corning, NY, USA) supplemented with 10% Equafetal serum blend (Atlas Biologicals, Fort Collins, CO, USA) and 1× L-glutamine (D10 medium). 3T3 and 3T3-L1 cells (ATCC, CL-173) were cultured in D10 medium supplemented with 1 mM sodium pyruvate (Corning) and 1× non-essential amino acids (GE Healthcare, Pittsburgh, PA, USA). 3T3-L1 cell lines with cavin-1 inactivated by CRISPR (K5 and K3) and the CRISPR control 3T3-L1 cell line (KN; wild type) were generous gifts from Libin Liu (Boston University School of Medicine, Boston, MA, USA) (Liu and Pilch, 2016). 3T3-L1 cells were differentiated to adipocytes for 3 days in Adipocyte Differentiation medium (D10 medium supplemented with 1 µM dexamethasone, 0.5 mM 3-isobutyl-1-methylxanthine and 10 µg/ml insulin) and maintained for the next 10 days in Adipocyte Maintenance medium (D10 medium supplemented with 10 µg/ml insulin) by changing the medium every other day. We used adipocytes after 9 or 14 days of differentiation for assays. Casp1/11^−/− BMDMs were isolated from bone marrow and cultured as previously described (Ray et al., 2022; Romero et al., 2017). BMDMs were differentiated for 7–21 days in DMEM supplemented with 30% L929 cell supernatants, 20% fetal calf serum (VWR Seradigm, Radnor, PA, USA), 1 mM sodium pyruvate and 1× L-glutamine. C2C12 myoblasts stably expressing control (ATCC CRL-3419) or Dysf shRNA (ATCC CRL-3418) were cultured in DMEM supplemented with 10% Equafetal serum blend and 2 µg/ml puromycin.

Toxins used in assays were induced in Escherichia coli BL21 GOLD cells and purified as previously described (Keyel et al., 2012a; Ray et al., 2022, 2019; Romero et al., 2017). Briefly, toxins were induced with 0.2% arabinose (SLO, SLO ML) or 0.2 mM IPTG (PFO, ILY, or pro-aerolysin and pro-aerolysin^Y221G) for 3 h at room temperature, frozen, lysed and then purified using Nickel-NTA resin. For antibody production, or to determine the impact of purity on toxin activity, pro-aerolysin or pro-aerolysin^Y221G were further purified by FPLC Q anion exchange chromatography (Fig. S8A). The protein concentration was determined by Bradford Assay (Table S1). Purity was assessed by SDS-PAGE. The hemolytic activity of each toxin was determined as previously described (Keyel et al., 2012a; Ray et al., 2022, 2019; Romero et al., 2017) using human red blood cells (Zen-Bio, Research Triangle Park, NC, USA) (Table 1). One hemolytic unit (HU) is defined as the quantity of toxin required to lyse 50% of a 2% human red blood cell solution in 30 min at 37°C in 2 mM CaCl₂, 10 mM HEPES, pH 7.4 and 0.3% BSA in PBS (Keyel et al., 2012a; Romero et al., 2017). In effect, 1 HU is the LC₅₀ of the toxin against human red blood cells. To determine the optimal amount of trypsin needed to activate aerolysin, we incubated pro-aerolysin with decreasing amounts of trypsin (Fig. S8B). The minimal dose of trypsin needed to activate aerolysin was chosen. Notably no oligomerization of aerolysin was observed during trypsin activation (Fig. S8B). We used HU/ml to normalize toxin activities in each experiment and to achieve consistent cytotoxicity across toxin preparations. To measure batch-to-batch variation, the LC₅₀ were plotted by toxin type for toxin challenge with wild type cells (Fig. S8C,D). The highest toxin concentration that killed <20% of target cells as assessed by live-cell imaging (e.g. TOPRO uptake) was defined as the sublytic dose (Ray et al., 2019). For live-cell imaging experiments in HeLa cells, the sublytic dose used was 250 HU/ml for SLO, 250 HU/ml for ILY and 62 HU/ml for aerolysin. SLO ML was fluorescently labeled as previously described (Romero et al., 2017) whereas pro-aerolysin^K244C was fluorescently labeled using maleimide chemistry. Briefly, purified Cys-less SLO ML was gel filtered into 100 mM sodium bicarbonate (pH 8.5) using a Zeba gel filtration column according to the manufacturer's instructions. Sufficient monoreactive Cy5 dye (Cytiva) to label 0.25 mg protein was reconstituted in 100 mM sodium bicarbonate, added to the SLO ML, and incubated for 1 h at room temperature. Any unconjugated dye was removed by gel filtration of conjugated SLO ML into PBS. For the conjugation of Alexa Fluor 647 (AF647) to pro-aerolysin^K244C, the toxin was eluted without DTT, then reduced with 6.32 mM tris(2-carboxyethyl) phosphine hydrochloride for 30 min on ice. Alexa Fluor 647 maleimide (Sigma) dye was reconstituted as 1 µmol in DMSO and added to pro-aerolysin as a 10–20-fold molar mass excess and incubated overnight at 4°C. Unconjugated dye was removed by gel filtration of conjugated pro-aerolysin^K244C into Tris salt buffer (50 mM Tris-HCl, 300 mM NaCl, pH 8). 5 mM DTT was added to the toxin, which was snap frozen on dry ice.

Activation of pro-aerolysin was performed as previously described (Iacovache et al., 2011), with the following changes. Pro-aerolysin was activated by adding 1 µl of 0.02% trypsin in nuclease free water to 10 µl pro-aerolysin and incubating at RT for 10 min. The trypsin dilution was determined empirically based on the extent of pro-aerolysin cleavage.

HeLa cells were plated at 2×10⁵ cells per 35 mm glass-bottom dish or per well of a six-well plate and transfected with 750 ng peGFP-N1 or GFP–Dysferlin using Lipofectamine 2000 in Opti-MEM 2 days before imaging or cytotoxicity assays. D10 medium was replaced on the day after transfection. Transfection efficiencies for each construct ranged between 65 and 80%. For RNAi, HeLa cells were plated at 2×10⁵ cells in 6-well plates and transfected the following day with 10 nM siRNA using Lipofectamine 2000 in Opti-MEM for 48–72 h as previously described (Ray et al., 2022).

The MTT assay was performed and analyzed as previously described (Ray et al., 2019). Briefly, 2×10⁴ HeLa cells per well were plated in a 96-well tissue culture plate. On the next day, the medium was aspirated. Cells were challenged with aerolysin (31-2000 HU/ml) for 1, 3, 6, 12 or 24 h at 37°C, then centrifuged at 1200 g for 5 min at 4°C to remove the medium. Phenol red-free DMEM with 1.1 mM 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was added, and cells were incubated at 37°C for 4 h. SDS-HCl was used to dissolve formazan at 37°C overnight and absorbance was measured at 570 nm using a plate reader (Bio-Tek, Winooski, VT, USA). We calculated the percentage of viable cells as: % Viable=(Sample−Background)/(Control−Background)×100. Then, the specific lysis was calculated as 100−%Viable.

Cytotoxicity was assessed by flow cytometry as previously described (Haram et al., 2022; Keyel et al., 2012a; Ray et al., 2022, 2019; Romero et al., 2017). Briefly, 10⁵ cells were challenged in suspension with various concentrations of toxins for 5, 15, 30 or 60 min at 37°C in RPMI supplemented with 2 mM CaCl₂ (RC) and 20 μg/ml PI. Cells were analyzed on a 4-laser Attune Nxt flow cytometer. Assay variations included changes to the medium [replacing 2 mM CaCl₂ with 2 mM EGTA, adding 150 mM KCl or using Tyrode's buffer (140 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 10 mM glucose, 10 mM HEPES, pH 7.4) instead of RPMI] or serum-starvation of cells in DMEM for 30 min at 37°C and pretreatment with DMSO or inhibitors in serum-free medium for 30 min at 37°C. For analysis of cell lysis, we gated out the debris and then quantified the percentage of cells with high dye fluorescence (2–3 log shift) (dye high), low dye fluorescence (∼1 log shift) (dye low), or background dye fluorescence (dye negative). Populations of cells marked ‘dye negative’ and ‘dye low’ remain metabolically active, indicating that only the ‘dye high’ population are dead cells (Keyel et al., 2011). We calculated specific lysis as: % Specific Lysis=(% Dye High^Experimental−% Dye High^Control)/(100−% Dye High^Control)×100. Similar calculations were used for transiently permeabilized cells by using the ‘dye low’ population instead of the ‘dye high’ populations. The toxin dose needed to kill 50% of cells was defined as the lethal concentration 50% (LC₅₀) and was determined by logistic modeling using Excel (Microsoft, Redmond, WA, USA) as previously described (Haram et al., 2022). The sublytic dose was defined as the highest toxin concentration that gave <20% specific lysis.

HeLa cells were plated at 2×10⁵ cells per 35 mm glass-bottom dish two days before imaging. Cells were labeled with 4 μM Fluo-4 AM (Cat# F14201) for 30 min at 37°C. After a brief wash, Fluo-4 AM-labeled cells were challenged with the following sublytic toxins: 250 HU/ml SLO, 62.5 HU/ml aerolysin or a mass equivalent of Aero^Y221G, and imaged at 37°C in RPMI, 25 mM HEPES pH 7.4, 2 mM CaCl₂, and 2 μg/ml TO-PRO3 for 45 min using a Fluoview 3000 confocal microscope (Olympus, Tokyo, Japan) equipped with a 60×, 1.42 NA oil immersion objective and recorded using the resonant scanner at 1–3 s per frame. Fluo-4 AM was excited using the 488 nm laser, whereas TO-PRO3 was excited with the 640 nm laser. After 45 min, the assay was stopped by adding an equal volume of 2% Triton X-100 to give a final concentration of 1%. Triton X-100 addition served as a control for TO-PRO uptake and dye loss. The extent of Ca²⁺ flux in cells was assessed by measuring intensity from the middle of the cell over time by plotting Z-profiles in FIJI (ImageJ, NIH, Bethesda, MD, USA). TO-PRO3 uptake was determined in cells by plotting the Z-profile of a nuclear subsample over time. We calculated fluorescence intensity over time for individual cells in each toxin group and then normalized each group to their brightest point to measure the temporal changes. We calculated the number of dead cells in each time point per experiment by converting monochromatic blue (TO-PRO-3) channel into masks and analyzing particles in FIJI. Experimental cells which expressed at least two-thirds of the maximal fluorescence intensity of TO-PRO3 (post Triton X-100) were considered dead. We categorized cells based on their survival: ‘died in 5 min’, ‘died between 5 and 45 min’, and ‘survived >45 min’.

To measure Ca²⁺ oscillations, we used realized volatility. Realized volatility measures the variance over a defined time period. We chose 10 s periods using Nyquist's theorem. We calculated the difference in the log fluorescence intensity at time t and time t+10, where t is in seconds. Then, we summed the square of these differences, and took the square root of this sum to get the final realized volatility.

HeLa cells were plated at 2×10⁵ cells per 35 mm glass-bottom dish and transfected with 750 ng A6-YFP using Lipofectamine 2000 at 2 days prior to imaging. The transfection efficiency was ∼65–80%. Transfected cells were challenged with a sublytic toxin dose for SLO (250 HU/ml), aerolysin (62.5 HU/ml) or mass equivalent of aerolysin^Y221G and imaged at 37°C in RPMI, 25 mM HEPES pH 7.4, and 2 mM CaCl₂ with 2 μg/ml TO-PRO3 for 45 min using a Yokogawa CSU-X spinning disc confocal microscope (Intelligent Imaging Innovation, Denver, CO, USA). A6-YFP was excited using a 488 nm laser, whereas TO-PRO3 was excited with a 640 nm laser. Fluorescence was collected using a 60×, 1.49 NA oil immersion objective and recorded using an Evolve 512 EMCCD camera (Photometrics, Tucson, AZ, USA) at 1–3 s/frame. After 45 min, an equal volume of 2% Triton X-100 was added to give a final concentration of 1%. The total number of A6–YFP⁺ microvesicles released was counted manually using every second frame and was expressed as number of microvesicles shed/number of cells/minute. The percentage of cells showing A6–YFP translocation was determined by comparing the A6–YFP intensity at 15 min to the intensity 1 min after toxin addition. If these values were below 80% of the initial value, cells were considered to show A6–YFP translocation. In cells showing A6–YFP translocation, the extent of A6–YFP depletion from cells was assessed by measuring A6–YFP intensity from the middle of the cell over time by plotting Z-profiles in ImageJ (NIH, Bethesda, MD, USA). TO-PRO3 uptake was determined in A6–YFP translocated cells by plotting the Z-profile of a nuclear subsample over time. For both A6–YFP and TO-PRO3, the data were normalized to the integrated intensity of the brightest point. The t_1/2 for annexin cytosolic depletion, membrane accumulation or TO-PRO3 uptake was calculated by measuring the half-maximal intensity between the starting intensity and the mean intensity of the last 4 min prior to Triton X-100 addition. Membrane accumulation of A6 over time was calculated by determining the changes in the fluorescence intensity along the edges of the cells of translocated cells by plotting the Z-profile in FIJI. The intensity was then normalized on a per-cell basis and individually analyzed cells were plotted. From the multi-plane images, several cells (>20) were analyzed from at least three independent experiments and graphed using Microsoft Excel and quantified values for individual cells were represented with Prism 8.1 (GraphPad, San Diego, CA, USA).

Images were exported as .tif format files and split into individual monochromatic red, green and blue channels in FIJI. For display (but not analysis), bleach correction was performed as previously described (Ray et al., 2019) on the ‘green’ channel cells by histogram matching using FIJI followed by a median pass filter. The monochromatic images were then merged to form RGB .tif files, time-stamped, annotated, and exported as AVIs for the supplementary movies.

Immunofluorescence was performed as previously described (Keyel et al., 2011). Briefly, HeLa cells (10⁵) were plated on coverslips, transfected with 750 ng of annexin A6–YFP and challenged with aerolysin or mass equivalent of aerolysin^Y221G diluted in RPMI supplemented with 2 mM CaCl₂ (RC) for 0, 15, 30, or 60 min at 37°C. Cells were washed in PBS, fixed in 2% paraformaldehyde, washed, permeabilized in 10% goat serum with 0.2% saponin, probed with rabbit polyclonal anti-aerolysin (1:500) for 1 h, washed, probed with goat Alexa-Fluor-647-conjugated anti-rabbit-IgG (1:1000) for 1 h, washed, DAPI stained, washed and mounted on slides in gelvatol. Cells were imaged on a Fluoview 3000 confocal microscope (Olympus, Tokyo, Japan) equipped with a 60×, 1.42 NA oil immersion objective. Images were processed using ImageJ (NIH, Bethesda, MD, USA).

Microvesicles were isolated as previously described (Keyel et al., 2011; Romero et al., 2017). Briefly, 5×10⁶–10×10⁶ cells were harvested, resuspended in RC at 2.5×10⁶–5×10⁶ cells/ml, challenged with a sublytic dose of toxin (hemolytic toxins), equivalent mass (inactive toxins) or 0.02% trypsin (control), and incubated for 15 min at 37°C. Challenged cells were then pelleted at 2000 g for 10 min, solubilized at 95°C in SDS-sample buffer for 5 min, and sonicated. Supernatants were spun at 100,000 g in a Beckman Coulter Optima XPN-80 ultracentrifuge using a Beckman SW 41 Ti rotor for 40 min at 4°C. The microvesicle pellet was solubilized at 95°C in 4× SDS-PAGE sample buffer. We previously demonstrated (Keyel et al., 2011, 2012b) that vesicles triggered by CDCs best fit the definition of microvesicles. These microvesicles are nontoxic to cells and do not fuse with the plasma membrane of cells that internalize them (Keyel et al., 2012b).

Samples were resolved on 10% polyacrylamide gels at 170 V/90 min and transferred to nitrocellulose in an ice bath with transfer buffer (15.6 mM Tris and 120 mM glycine) at 110 V for 90 min. Blots were blocked using 5% skim milk in 10 mM Tris-HCl, 150 mM NaCl and 0.1% Tween 20 at pH 7.5 (TBST), and portions of the blot were incubated with one of the following primary antibodies for 2 h in 1% skim milk in TBST at RT: 6D11 anti-SLO mAb (1:1000), CPTC-A1–3 anti-annexin A1 (1:250) mAb, anti-Annexin A2 (1:1000) mAb, anti-Alix (1:250) mAb, MANLAC-4A7 anti-Lamin A/C (1:250) mAb, AC-15 anti-β-actin (1:5000) or with rabbit polyclonal antibodies: anti-aerolysin, anti-annexin A6, anti-MEK, anti-phospho-MEK [Ser²¹⁷/Ser²²¹], anti-ERK p44/42, anti-phospho-ERK[Thr²⁰²/Tyr⁴⁰⁴], anti-p38, or anti-phospho-p38, each at 1:1000 dilution. Next, blots were incubated with HRP-conjugated anti-mouse- or anti-rabbit-IgG antibodies (1:10,000) for 1 h in 1% skim milk in TBST at RT and developed with enhanced chemiluminescence (ECL): either 0.01% H₂O₂ (Walmart, Fayetteville, AR, USA), 0.2 mM p-Coumaric acid (Sigma), 1.25 mM Luminol (Sigma) in 0.1 M Tris-HCl pH 8.4, ECL Plus Western Blotting Substrate (Cat# 32134), ECL Prime Western Blotting Reagent (Cat# RPN2232, Cytiva), or Pierce ECL western blotting substrate (Lot # UH288963A, Thermo Fisher). Full western blots are included in the supplemental data (Figs S9, S10).

ELISA plates were coated with 10 ng of pro-aerolysin, pro-aerolysin^Y221G, aerolysin or aerolysin^Y221G, and incubated overnight at 4°C. Plates were then washed three times in PBS with 0.05% Tween 20 (PBST), blocked for 1 h with 1% BSA in PBST, and then serially diluted pre-immune serum, the affinity-purified anti-aerolysin antibody or the depleted serum post-affinity purification were added for 2 h at 37°C. Plates were washed three times in PBST, incubated with HRP-conjugated goat anti-rabbit-IgG (1:20,000) and developed using 0.2 mg/ml TMB (Sigma), 0.015% H₂O₂ in 100 mM sodium acetate, pH 5.5. The reaction was stopped with 0.5 M H₂SO₄. A₄₅₀ was measured on a plate reader (Bio-tek, Winooski, VT) and plotted against antibody dilution.

Total RNA was extracted from C2C12 control and Dysf shRNA cells using TRI-Reagent. cDNA was generated using Superscript II. Dysferlin and β-actin levels were measured on a CFX96 Deepwell (Bio-Rad, Hercules, CA) using SybrGreen and compared using ΔΔCt as previously described (Bhattacharjee and Keyel, 2018; Keyel et al., 2014). The mean Ct for control cells in each assay was considered 1-fold change so that variation in control cell dysferlin expression could also be measured. Primer sequences for dysferlin are 5′-CTTCCTGGGTGTCTTTGTCTAC-3′ and 5′-AAGCCAAGAGTTTATCCAGAGG-3′ and for β-actin are 5′-GAAATCGTGCGTGACATCAAAG and 5′-TGTAGTTTCATGGATGCCACAG-3′.

Prism 8.1 (GraphPad, San Diego, CA, USA) or Excel were used for statistical analysis. Data are represented as mean±s.e.m. as indicated. The LC₅₀ for toxins was calculated by logistic modeling in Excel as previously described (Haram et al., 2022). Sample size was determined based on prior work (Haram et al., 2022; Ray et al., 2022). Exclusion criteria for data was failure of positive or negative controls. Statistical significance for normally distributed data was determined by regular or repeated measures one-way or two-way ANOVA as specified; P<0.05 was considered to be statistically significant. For non-normally distributed data, the Kruskal–Wallis test was used. Graphs were generated in Excel, Prism and Photoshop (Adobe, San Jose, CA, USA).

The authors would like to thank members of the Keyel lab for critical review of the manuscript. We thank colleagues for the generous gifts of reagents. We thank the College of Arts & Sciences Microscopy for use of facilities.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.261018.reviewer-comments.pdf