Calcium-dependent ESCRT recruitment and lysosome exocytosis maintain epithelial integrity during Candida albicans invasion

Summary Candida albicans is both a commensal and an opportunistic fungal pathogen. Invading hyphae of C. albicans secrete candidalysin, a pore-forming peptide toxin. To prevent cell death, epithelial cells must protect themselves from direct damage induced by candidalysin and by the mechanical forces exerted by expanding hyphae. We identify two key Ca2+-dependent repair mechanisms employed by epithelial cells to withstand candidalysin-producing hyphae. Using camelid nanobodies, we demonstrate candidalysin secretion directly into the invasion pockets induced by elongating C. albicans hyphae. The toxin induces oscillatory increases in cytosolic [Ca2+], which cause hydrolysis of PtdIns(4,5)P2 and loss of cortical actin. Epithelial cells dispose of damaged membrane regions containing candidalysin by an Alg-2/Alix/ESCRT-III-dependent blebbing process. At later stages, plasmalemmal tears induced mechanically by invading hyphae are repaired by exocytic insertion of lysosomal membranes. These two repair mechanisms maintain epithelial integrity and prevent mucosal damage during both commensal growth and infection by C. albicans.


In brief
The human fungal pathogen Candida albicans invades epithelial cells and secretes the pore-forming toxin candidalysin. Westman et al. demonstrate that epithelial cells implement two membrane repair mechanisms-ESCRT-mediated plasma membrane pinching and lysosome exocytosis-that remove candidalysin and confer resistance to the invading fungus INTRODUCTION Candida albicans colonizes the mucosal surfaces of the gut, oral cavity, or vaginal tracts. The fungus is a commensal constituent of the microbiome of most healthy individuals. A balanced microbiome and an intact immune system are required to maintain the fungus in the commensal state, preventing invasion, epithelial damage, and mucosal infection. Accordingly, antibiotic treatments and immunosuppression are predisposing factors for C. albicans infection (Kumamoto et al., 2020).
Superficial C. albicans mucosal infections are common. However, in severe cases, such as in immunocompromized patients, invasion and translocation from the gut enable the fungus to disseminate via the blood stream to internal organs, causing life-threatening infections (Perlroth et al., 2007;Pfaller and Diekema, 2007;Mayer et al., 2013). A key aspect of this commensal-to-pathogen shift is the fungal morphology. C. albicans exists primarily in yeast form, but it can also adopt filamentous hyphal forms. Hyphae, which can grow several microns per hour, can invade epithelia forming an invasion pocket (Moyes et al., 2015). Importantly, hyphal formation per se is not sufficient to cause epithelial damage (Moyes et al., 2016).
Rather, a combination of hyphal formation and extension, and the expression of hyphal-associated genes is essential to damage epithelial cells (Westman et al., 2019;Mogavero et al., 2021). In particular, the hypha-associated peptide toxin candidalysin encoded by the ECE1 gene is essential to cause damage (Moyes et al., 2016;Allert et al., 2018;Naglik et al., 2019). However, the presence of C. albicans hyphae is not necessarily associated with infection, and moderate levels of invading hyphae are tolerated (Moyes et al., 2010). How epithelial cells preserve their integrity when challenged by growing hyphae and by release of the pore-forming candidalysin remains unknown. The extending hyphae can burrow into the mucosal cells, inflicting severe mechanical stress, while concomitantly membrane permeabilization by candidalysin disrupts the cellular ionic and osmotic balance. It is remarkable that mucosal integrity is maintained when continuously confronted by these challenges.
We sought to understand the manner whereby epithelial cells overcome the stresses imposed by growing C. albicans hyphae. We report that when confronted by growing hyphae, epithelial cells enable a combination of protective mechanisms that repair mechanical damage to the membrane, eliminate candidalysin, and restore ionic homeostasis.

RESULTS
Characterization of the C. albicans-containing invasion pocket When growing on epithelial monolayers, C. albicans hyphae depress the host cell surface, generating an invasion pocket reminiscent of the deformations observed in infected primary tissues (Zakikhany et al., 2007;Dalle et al., 2010). To study the consequences of such deformation, we grew C. albicans for up to 6 h on TR146 cells, an immortalized line of oral epithelial origin. Invasion pockets generated by extending hyphae could be identified by visualizing the entire fungus with calcofluor white (CFW), while staining exposed regions of the cell wall with Alexa Fluor-conjugated concanavalin A (ConA); the lectin is excluded from invasion pockets by the tight adherence of the fungal wall to the invaginated host membrane (Figure 1). Short invasion pockets (<20 mm in length) had no discernible effect on the host cell plasma membrane (PM), which retained its continuity (assessed with the fluorescent plasmalemmal probe PM-RFP) and maintained an intact layer of cortical F-actin (stained with phalloidin) ( Figure 1A). Of note, short pockets were devoid of PtdIns(3)P (detected using PX-GFP) and of LAMP-1 (visualized expressing LAMP-1-GFP), implying that neither early endosomes nor late endosomes/lysosomes had fused with the invaginated PM (Figures 1B and 1C). This contrasts with the invaginations formed in macrophages around similarly sized hyphae. As reported (Maxson et al., 2018), the structures formed by macrophages are akin to frustrated phagosomes, characterized by a thick actin cuff at the neck ( Figure 1D) that enables retention of PtdIns(3)P and late endosomal/lysosomal markers that fuse with the limiting membrane ( Figures 1E and 1F), yet devoid of actin along the rest of the invaginated membrane.
As the fungus continued to grow into the epithelial cell, the appearance of the invasion pocket was markedly altered. Long (>20 mm) pockets showed progressive loss of cortical actin, PM markers showed discontinuities ( Figure 1G), and lysosome insertion became apparent ( Figure 1H). We found no evidence of early endosome fusion, as pockets remained negative for PtdIns(3)P ( Figure 1I). The patchy depolymerization of cortical actin (Figure 1G) was suggestive of significant damage to long invasion pockets.
Damage to the invasion pocket leads to influx of extracellular Ca 2+ , hydrolysis of PtdIns(4,5)P 2 , and loss of cortical actin The mechanism that leads to the apparent PM damage was investigated next. Cortical actin is stabilized in part by ezrin, radixin, and moesin proteins through interactions with PtdIns(4,5)P 2 . We therefore tested whether depletion of PtdIns(4,5)P 2 accompanies the loss of cortical actin. The distribution of the phosphoinositide was visualized by expressing the pleckstrin homology (PH) domain of phospholipase C-d (PH-PLCd). Short invasion pockets retained PtdIns(4,5)P 2 along with cortical actin (Figure 2A). In contrast, a loss of plasmalemmal PtdIns(4,5)P 2 was evident in long pockets that displayed discontinuous cortical actin ( Figure 2A). As shown in Figure 2B, the depletion of PtdIns(4,5)P 2 was not attributable to loss of PtdIns(4)P, the substrate required for its synthesis. Instead, it appeared more likely that accelerated degradation was responsible for PtdIns(4,5)P 2 loss. Activation of phospholipase C by Ca 2+ is often responsible for the hydrolysis of PtdIns(4,5)P 2 . As shown in Figure 2C, elevation of cytosolic Ca 2+ ([Ca 2+ ] c ) by means of ionomycin also effectively depleted plasmalemmal PtdIns(4,5)P 2 . Importantly, the depletion of PtdIns(4,5)P 2 induced by ionomycin was accompanied by loss of cortical F-actin, which depolymerized and redistributed to intracellular structures ( Figure 2C). This raised the possibility that changes in [Ca 2+ ] c were responsible for the alterations caused by deeply penetrating hyphae. Such changes could in principle be inflicted mechanically or, more likely, by the release and accumulation of candidalysin-a pore-forming toxin-in the pocket. To evaluate the latter possibility, we used two a-candidalysin nanobodies conjugated to Alexa Fluor dyes (Mogavero et al., 2021). Of note, the nanobodies were sufficiently small to enter invasion pockets without the need to permeabilize cells after fixation, enabling us to distinguish secreted candidalysin from that still retained inside C. albicans' secretory organelles. In cells infected with wild-type C. albicans the nanobodies could consistently detect candidalysin inside invasion pockets ( Figure 2D). Candidalysin often accumulated at sites of cortical actin breakdown, suggesting that the toxin is responsible for the breakdown ( Figure 2D). In contrast, invasion pockets containing the candi-dalysin-deficient ece1D/D mutant failed to show immunostaining, confirming the specificity of the nanobodies ( Figure 2D). Importantly, actin integrity was largely maintained along the invasion pockets formed by candidalysin-deficient hyphae, implying that the toxin is responsible for the effects caused by the wildtype C. albicans ( Figure 2D).
To directly assess the ability of candidalysin to alter PtdIns(4,5) P 2 and cortical actin, TR146 cells were treated with pure toxin generated synthetically; for comparison, parallel samples were treated with a synthetic version of Ece1-VIIa, another peptide derived from the Ece1 polypeptide that lacks the pore-forming properties ascribed to candidalysin (Moyes et al., 2016). As shown in Figure 2E, addition of candidalysin sufficed to deplete PtdIns(4,5)P 2 and F-actin from the PM, while peptide Ece1-VIIa was without effect. The effect of candidalysin required extracellular Ca 2+ (cf. top right and bottom right panels in Figure 2E). The effects of the peptides on PtdIns(4,5)P 2 are quantified in Figure 2F. Together, these findings suggest that candidalysin depletes plasmalemmal PtdIns(4,5)P 2 -and consequently also cortical actin-by increasing the permeability of the PM to extracellular Ca 2+ .
To validate the notion that PtdIns(4,5)P 2 hydrolysis is dependent on candidalysin, the phosphoinositide was analyzed in TR146 cells infected with the ece1D/D mutant. Unlike the wildtype C. albicans hyphae, which caused extensive depletion of PtdIns(4,5), hyphae formed by the ece1D/D mutant were without effect ( Figure 2G).

Candidalysin-induced membrane rupture causes (Ca 2+ ) c oscillations
The preceding observations suggested that candidalysin increases the permeability of the host membrane to Ca 2+ . That candidalysin promotes Ca 2+ influx was previously reported using Fura-2 (Moyes et al., 2016). We confirmed and extended these observations using the genetically-encoded fluorescent Ca 2+ sensor GCaMP6s . GCaMP6s has a similar affinity for Ca 2+ as Fura-2 but has faster response kinetics and a much greater dynamic range. Representative results are presented in Figure 3A and Video S1. Candidalysin caused a rapid increase in [Ca 2+ ] c , but only when Ca 2+ was present extracellularly ( Figure 3A). The responsiveness of GCaMP6s in cells bathed in Ca 2+ -free medium was validated by adding thapsigargin (TG), an inhibitor of the SERCA pump that causes net Ca 2+ release from the endoplasmic reticulum, with an accompanying elevation of [Ca 2+ ] c ( Figure 3A). Intriguingly, in contrast to the measurements by (Moyes et al., 2016), the increase in [Ca 2+ ] c induced by candidalysin was rarely sustained; instead, multiple short-lived [Ca 2+ ] c transients were generally recorded  Figure 3B). Some of the cells showed up to five [Ca 2+ ] c flashes within the first 15 min after candidalysin treatment (Video S1). Of note, the epithelial cells often managed to reduce the [Ca 2+ ] c back to the basal level between or after oscillations. As shown in Figures 3B and 3C, the effect of candidalysin on GCaMP6s required the presence of extracellular Ca 2+ . Lastly, the control peptide Ece1-VIIa did not induce [Ca 2+ ] c changes, corroborating the specificity of the changes triggered by candidalysin.

Epithelial cells respond to candidalysin by forming PMderived blebs
The transient nature of the candidalysin-induced Ca 2+ permeability changes was suggestive of intervening inactivation or extrusion of the pore-forming toxin. To explore this possibility, we visualized the PM during the course of exposure to candida-lysin using the lipophilic fluorescent dye FM4-64, which is sufficiently small to enter C. albicans-containing invasion pockets that otherwise restrict the entry of larger molecules such as ConA ( Figure 4A). Strikingly, the dye revealed a marked convolution of the host membrane in the distal regions of the pocket, where candidalysin is most often detected, with formation of multiple blebs that were trapped between the hypha and the epithelial cell ( Figure 4A). Membrane blebbing was only prominent in long invasion pockets, which take 4-6 h to form, while short pockets remained intact. Interestingly, the fluorescence intensity of FM4-64 in blebs was significantly higher than in regions of the membrane that were intact, suggesting that the dyewhich does not permeate intact bilayers-reaches both sides of the bleb membrane; we interpret the increased fluorescence as indicative of rupture of the bleb membrane. That candidalysin is responsible for formation of the blebs was suggested by (legend continued on next page) 6 Cell Reports 38, 110187, January 4, 2022 Article ll OPEN ACCESS analysis of comparable long pockets generated by ece1D/D hyphae ( Figure 4A). In these instances, labeling with FM4-64 showed the membrane lining the pocket to be continuous, with no evidence of blebs or discontinuities. Similarly, we found no evidence of damage or bleb formation in pockets formed by ece1D/D + ECE1 D184-279 hyphae, which express all Ece1-derived peptides except candidalysin ( Figure 4A). PM blebs were also formed when synthetic candidalysin was added to TR146 cells ( Figure 4B). In this case, bleb formation was visible as early as 10 min after addition of the toxin and coincided with the [Ca 2+ ] c oscillations (Video S2). Candidalysin induced PM blebbing only when Ca 2+ was present in the medium ( Figure 4B), which suggests that the blebs form in response to the previously documented Ca 2+ influx. Accordingly, no PM damage or bleb formation was visible when the cells were incubated with peptide Ece1-VIIa ( Figure 4B), which similarly failed to induce Ca 2+ influx.
Next, we investigated whether candidalysin-induced PM blebs contained candidalysin by staining the synthetic toxin with the a-candidalysin nanobodies. The labeled CaLH1 nanobody localized to PM blebs induced by synthetic candidalysin, but it did not stain cells incubated with peptide Ece1-VIIa (Figure 4C). Candidalysin-positive blebs could be visualized apically within 10 min of candidalysin exposure ( Figure S1A). At later time points, the blebs pinched off and separated from the cells, removing PM material and candidalysin from the epithelial surface in the process ( Figure S1B). The PM blebs did not retract back into the cell body ( Figure S1C; Video S3). Moreover, bleb staining remained higher than that of the PM, indicating discontinuity between the bleb and the PM. Since FM4-64 diffuses laterally along the membrane, the observed diffusion barrier likely reflects separation of the bleb from the PM.
The CaLF1 nanobody detected secreted candidalysin in the blebs of wild-type C. albicans invasion pockets, but not in those of the ece1D/D strain ( Figure 4D). We also investigated whether invasion pockets generated by other C. albicans clinical isolates developed PM-derived blebs similar to the ones formed by the SC5314 strain used throughout this study. We found that induction of PM blebs by the clinical strains correlated with their ability to secrete candidalysin. C. albicans isolates HUN96 and 101 secreted detectable levels of candidalysin and induced PM blebs ( Figure S1D). In contrast, we failed to detect candidalysin and PM blebs in pockets formed by isolate 529L ( Figure S1D); this strain expresses a variant candidalysin that is secreted at a reduced rate . We attribute the failure to form blebs to the reduced secretion, rather than to structural differences between the candidalysins, because ab initio structure predictions for the candidalysin peptides secreted by SC5314 and 529L showed only minor differences that do not markedly affect their hydrophobicity, charge, or polarity (Figures S1E vs. S1F). Accordingly, the calcium influx, cytokine secretion, and ultimate cell damage effected by treatment with synthetic 529L and SC5314 candidalysins were indistinguishable (J.N., unpublished). We therefore propose that entry of Ca 2+ via candidalysin triggers formation of PM-derived blebs that contain the toxin. These blebs are trapped inside the invasion pocket but can be released into the culture medium when induced by synthetic candidalysin. Shedding of the toxin along with the blebbing membrane terminate the influx of Ca 2+ , at least temporarily, accounting for the transient nature of the oscillations detected by GCaMP6s.
Permeabilization of the PM is followed by recruitment of ALG-2 and ESCRT, leading to shedding of blebs and PM repair Most cells have the ability to repair their PM upon rupture, a process that restores internal homeostasis and prevents cell death. Repair responses are often mediated by the influx of extracellular Ca 2+ . One such repair mechanism is the ALG-2-dependent recruitment of the ESCRT-III complex. ALG-2 has two high-affinity and one low-affinity Ca 2+ -binding sites; upon binding Ca 2+ , ALG-2 undergoes conformational changes that enable it to associate with the membrane (Jia et al., 2001). As it accumulates at sites of membrane rupture, where Ca 2+ influx occurs, ALG-2 recruits ALG-2-interacting protein X (ALIX) and CHMP proteins of the ESCRT-III complex (Jimenez et al., 2014;Scheffer et al., 2014). Assembly of ESCRT-III components constricts and extrudes the damaged membrane. The ensuing scission of the damaged patch results in repair and shedding of the agent that caused the membrane damage. We tested whether this machinery was involved in the blebbing and in the termination of the permeability enhancement caused by candidalysin. To this end, the epithelial cells were transfected with fluorescent chimeras of CHMP2a, CHMP4B, and CHMP6. As shown in Figures 5A and S2, these ESCRT-III components were recruited to invasion pockets containing wild-type C. albicans. Furthermore, CHMP2a accumulated in PM blebs stained by FM4-64 inside long invasion pockets containing wild-type C. albicans (Figure 5B), but not in invasion pockets formed by the ece1D/D hyphae ( Figure 5B). Similarly, CHMP2a was recruited to the base of blebs formed by TR146 cells treated with synthetic candidalysin; no such recruitment was observed when cells were exposed to Ece1-VIIa ( Figure 5C).
To confirm the Ca 2+ -dependence of the recruitment of the ESCRT-III machinery, C. albicans-induced invasion pockets were stained for ALG-2. Like the ESCRT-III components CHMP2a/4B/6, ALG-2 accumulated in invasion pockets formed by wild-type C. albicans hyphae, but not those induced by the ece1D/D mutant ( Figure 5D). ALG-2 accumulation was dependent on the length of the invasion pocket, increasing with the depth of the invagination ( Figure 5E). Additionally, ALG-2 as well as ALIX accumulated under nascent PM blebs induced by synthetic candidalysin, in a Ca 2+ -dependent manner ( Figures  5F and 5G). Our data suggest that the observed blebs inside invasion pockets and at the surface of candidalysin-treated TR146 cells are indicative of a repair response triggered by Ca 2+ influx that leads to the sequential recruitment of ALG-2, ALIX, and ESCRT-III.
Silencing ALG-2 prevents ESCRT-III-mediated membrane repair and curtails epithelial viability We hypothesized that recruitment of the ESCRT-III machinery to sites of invasion was essential to maintain epithelial viability during infection. This premise was tested by assessing the effects of candidalysin on the survival of cells where ALG-2 had been depleted. Using siRNA, ALG-2 was consistently depleted by z70% ( Figures 6A and 6B). Synthetic candidalysin induced markedly fewer PM blebs in ALG-2-silenced cells than in the paired controls ( Figures 6C and 6D). Of note, the [Ca 2+ ] c accumulation induced by candidalysin was greater in ALG-2-silenced cells ( Figure 6E), consistent with the notion that scission of blebs containing the toxin antagonizes its permeabilizing effects, enabling transient restoration of near-normal [Ca 2+ ] c . Indeed, in silenced cells the candidalysin-induced oscillatory transients seen in otherwise untreated cells were replaced by greater and more prolonged increases in [Ca 2+ ] c ( Figure 6F). Upon treatment with candidalysin, the mitochondrial membrane potential (measured as accumulation of rhodamine 123), which is preserved in cells treated with control siRNA, was largely dissipated in cells depleted of ALG-2 (insets, Figure 6C). The latter observation is suggestive of cell morbidity or death in the depleted cells. Two conventional approaches were applied to more specifically assess cell viability: staining with the nominally impermeant dyes DAPI and propidium iodide. As shown in Figure 6C, ALG-2-depleted cells displayed increased staining with DAPI. A systematic study revealed that cell death was significantly greater in cells where ALG-2 had been silenced (Figures 6H-6G). Transient overexpression of CHMP4B did not protect TR146 cells from the cytotoxic effects of candidalysin, suggesting that overexpression of a single component of the ESCRT machinery is insufficient to protect the cells against the toxin (Figures S3A and S3B). These findings suggest that ALG-2/ ESCRT-III-mediated membrane repair is required for the cells to tolerate sublethal concentrations of candidalysin.
We also tested the effects of IL-22, which was reported to protect epithelia during oral candidiasis in vivo (Conti et al., 2009;Bichele et al., 2018;Aggor et al., 2020). Pre-incubation with recombinant IL-22 for 48 h did not protect TR146 cells against candidalysin-induced cell damage ( Figure S3C).
Lysosome exocytosis repairs mechanical damage caused by very long C. albicans hyphae Mechanisms other than ESCRT-III-mediated membrane scission can also affect membrane repair. These include the exocy-tosis of lysosomes, a process that is similarly triggered by elevated [Ca 2+ ] c . Because candidalysin mediates Ca 2+ influx, we considered whether lysosomal fusion is also promoted by the toxin. Addition of synthetic candidalysin to the TR146 cells induced lysosome exocytosis when extracellular Ca 2+ was present. As depicted in Figure 7A, upon treatment with candidalysin an epitope on LAMP-1 normally facing the lumen of lysosomes becomes exposed on the membrane surface and is detectable in non-permeabilized cells, indicating exocytic insertion into the PM. The exocytosis of lysosomes triggered by candidalysin was often observed within the first hour of co-incubation with C. albicans, but it became statistically significant only after 2 h ( Figure 7B). A comparable treatment with the control Ece1-VIIa peptide did not induce lysosome secretion ( Figure 7B).
Unexpectedly, LAMP-1 staining was also observed in a fraction of the cells in the absence of extracellular Ca 2+ . However, close examination by serial confocal sections of cells exposed to candidalysin in Ca 2+ -free medium revealed that the PM was not labeled ( Figure 7A), ruling out the occurrence of exocytosis. Instead, endomembrane structures were immunolabelled with the LAMP-1 antibody. We concluded that, in the absence of Ca 2+ , the monovalent cation fluxes catalyzed by candidalysin alter the ionic balance of endomembrane organelles, including lysosomes, causing their lysis and exposing the luminal aspect of LAMP-1 to the antibody. The ionic imbalance is also likely to account for the loss of mitochondrial potential ( Figure 6C).
Because ESCRT-III-mediated PM repair mitigated the increases in [Ca 2+ ] c prompted by candidalysin, we predicted that the secretion of lysosomes would be enhanced in its absence. This was tested by silencing ALG-2 prior to challenging the cells with the toxin. Consistent with the requirement for elevated [Ca 2+ ], candidalysin-induced lysosome exocytosis was increased moderately when ALG-2 was depleted, although the change did not reach statistical significance (p = 0.16; Figure 7C). We suggest that ESCRT-III-mediated membrane blebbing and lysosome exocytosis serve complementary roles to mitigate candidalysin-induced PM damage.
As mentioned earlier, LAMP-1 was detectable in long invasion pockets. We assumed that localized secretion of candidalysin was responsible for this occurrence. Surprisingly, distinct insertion of LAMP-1 into the lining of the long invasion pockets was also noted when cells were infected with ece1D/D mutant hyphae ( Figures 7D and 7E), suggesting that extreme mechanical strain can promote Ca 2+ entry and lysosome exocytosis even in the absence of candidalysin. Mechanosensitive Ca 2+ channels or transient membrane rupture could enable Ca 2+ influx when membrane invagination is extreme. Accordingly, LAMP-1 was not found in invasion pockets shorter than 20 mm ( Figure 7D), suggesting that lysosome exocytosis is a late, secondary response to PM damage.
Two other features are noteworthy. First, plasmalemmal LAMP-1 was generally detected in regions where F-actin had been lost, likely as a consequence of Ca 2+ entry. Secondly, despite the insertion into the membrane lining the invasion pocket of lysosomes, which are rich in proton-pumping V-AT-Pases, the fluid within the invagination did not become detectably acidic ( Figure S4), in all likelihood because H + was able to diffuse readily out of the invasion pocket.

DISCUSSION
We investigated the manner by which C. albicans invades epithelial cells and the mechanisms whereby the epithelium contends with the stresses imposed by growing hyphae. Our studies indicate that, unlike the frustrated phagosome that forms when macrophages confront C. albicans hyphae, the invagination that the fungus forces when growing on epithelial cells is largely of plasmalemmal origin. The frustrated phagosome is separated from the bulk of the PM by a sturdy actin cuff that forms a diffusional barrier that contains endomembrane proteins and lipids within the invagination. However, in the epithelial pocket, the specialized actin cuff is missing, and in fact, discontinuities in the submembranous cortical actin appear progressively as the hyphae burrow deeper into the cell. The detachment of cortical actin is secondary to the loss of plasmalemmal PtdIns(4,5)P 2 , which is in turn caused by influx of extracellular Ca 2+ through candidalysin. This notion is supported by the finding that the candidalysin-deficient ece1D/D strain had no discernible effect on PtdIns(4,5)P 2 ( Figure 2G) or on cortical actin ( Figure 2D). That only long pockets show loss of membrane integrity suggests that candidalysin-which starts being secreted when the yeast transition to hypha-needs to accumulate over time to reach concentrations that effect permeabilization. Accordingly, the loss of PtdIns(4,5)P 2 and the associated actin depolymerization could be replicated by addition of synthetic candidalysin to otherwise untreated cells, implying that the toxin suffices to produce the alterations observed in infected cells. Another peptide encoded by the ECE1 gene, peptide VIIa, was without effect, validating the specificity of candidalysin ( Figure 2D). Two lines of evidence support the conclusion that Ca 2+ influx through candidalysin triggers the changes in PM structure. First, the loss of PtdIns(4,5)P 2 and F-actin was only observed in the presence of extracellular Ca 2+ , failing to occur when the cation was omitted. Second, [Ca 2+ ] c was found to increase when cells were treated with candidalysin, but not with the Ece1-VIIa peptide, and only when external Ca 2+ was present. Interestingly, the [Ca 2+ ] c increases were generally transient, followed by restoration of the basal levels, which suggests that the pores formed by candidalysin were short-lived. This may reflect inherent instability of the multimeric pore formed by candidalysin, but it is most likely an indication of membrane repair events that extrude or inactivate the toxin. In support of this, oscillatory recruitment of membrane repair machinery has been observed in cells treated with the bacterial pore-forming toxin streptolysin (SLO) (Babiychuk et al., 2009;Keyel et al., 2011). Similarly, we found that [Ca 2+ ] c oscillations in response to candidalysin became more pronounced and sustained when repair mechanisms were overwhelmed or impaired. As with SLO, the host membrane repair mechanisms culminate in the release of blebs from the host PM, which contained candidalysin (Figures 4, S1 and Video S3). Of note, bleb release has also been observed for two short toxin peptides that are structurally similar to candidalysin: melittin ( Cerne et al., 2013) and gomensin (Paredes-Gamero et al., 2012).
As a commensal and opportunistic pathogen, C. albicans continuously interacts with the host. We found that the host implements robust responses to offset the stresses C. albicans creates on the epithelial PM during invasion. By electron microscopy, the invasion pocket was reported recently to have inflated membranes with numerous associated vesicles (Lechat, 2021), reminiscent of the blebs we document. Our cell-based studies are also in accordance with in vivo mouse and zebrafish candidiasis models that link ECE1 expression with C. albicans pathogenicity and invasiveness (Moyes et al., 2010;Swidergall et al., 2019). C. albicans clinical isolates have been studied in vivo and characterized as host-invading or colonizing (Schö nherr et al., 2017;Kirchner et al., 2019). Interestingly, we found that C. albicans isolates that secreted detectable amounts of candidalysin, regardless of invading or colonizing capability in vivo, induced blebs in invasion pockets ( Figure S1D). One isolate tested, 529L, formed hyphae but did not secrete detectable candidalysin nor did it induce membrane blebs. Notably, the 529L isolate encodes a variant of Ece1, which is processed less effi-ciently . Consequently, less mature candidalysin is delivered into the invasion pocket ( Figure S1F), and damage levels caused by invading 529L hyphae remain low.
As the ability to produce hyphae has been conserved in most clinical isolates, hyphal formation likely confers an evolutionary advantage to C. albicans, including during its commensal phase (Witchley et al., 2019). Therefore hyphal formation per se would be insufficient to cause damage. Only an increased burden of invasive hyphae, which deliver a critical level of candidalysin into the invasion pocket (Mogavero et al., 2021), causes epithelial damage followed by necrotic cell death (Ho et al., 2019;Blagojevic et al., 2021). Indeed, our work demonstrates that epithelial repair is an additional major player in this protective mechanism. Given that epithelial cells must tolerate moderate levels of invading hyphae in a commensal setting, only an overwhelming level of candidalysin delivered into the invasion pocket by invading hyphae, which cannot be compensated by epithelial repair, would be associated with epithelial damage, necrosis, and disease.
The key finding of this study is that epithelial cells implement two different types of repair responses to counteract the actions of C. albicans hyphae. These responses not only protect the host from fungal dissemination and detrimental infection, but maintain the homeostatic balance with commensal C. albicans to preserve the ostensibly beneficial relationship. The protective responses consist of PM blebbing that disposes of damaged membranes along with the permeabilizing candidalysin, and exocytosis of endomembranes that patch the ruptured surface. Both of these responses are triggered by elevation of [Ca 2+ ] c , which is also a mediator of the cellular damage. While seemingly paradoxical, these findings imply that the epithelial cells are finely tuned to sense and counteract the very parameter that effects the damage. ALG-2 and ESCRT-III are the first responders. They accumulate in sites of hyphal infection and at the base of the blebs induced by synthetic candidalysin. The toxin is in fact detected in such blebs, confirming that they serve to eliminate it from the cell surface. While ESCRT-III repair has been characterized in cells subjected to bacterial or chemical damage (Jimenez et al., 2014;Scheffer et al., 2014), it had not previously been reported during interaction with fungal cells. When triggered by candidalysin, the ESCRT-III system protects the host cell; this is indicated by the reduced viability of epithelial cells that had been depleted of ALG-2 prior to treatment with the toxin ( Figures  6E and 6F). It is noteworthy that ESCRT-generated vesicles are generally small (100-200 nm), below the limit of resolution of conventional light microscopy, whereas at least a fraction of the vesicles generated by candidalysin were significantly larger. This observation, however, is not unprecedented, since similar vesicles-ranging from 200 nm to 2 mm-were also reported to be extruded by cells treated with SLO, which also triggers an ESCRT-dependent process (Keyel et al., 2011) and in cells exposed to pneumolysin (Wolfmeier et al., 2016).
Unlike the recruitment of the ESCRT-III complex, lysosome exocytosis occurred both in the wild-type and the ece1D/D mutant. We therefore concluded that the invasion pocket can be damaged by means other than candidalysin. C. albicans secretes different hydrolytic enzymes that could conceivably cause membrane damage (Niewerth and Korting, 2001;Schaller et al., 2005;Westman et al., 2019). Alternatively, the damage may be the consequence of mechanical strain exerted by the growing hyphae, which can attach firmly to host receptor proteins. Indeed, mechanically induced rupture has been shown to occur during C. albicans filamentation inside macrophages, leading to phagosomal alkalinization and expansion (Westman et al., 2018(Westman et al., , 2020. That long invasion pockets showed evidence of lysosome fusion suggests that other lysosomal components must have been delivered to the invasion pocket, including lysosomal hydrolases and the vacuolar H + -ATPase. However, we were unable to detect localized acidification within the invasion pocket using acidotropic dyes ( Figure S4), in all likelihood because H + can readily diffuse outward, inasmuch as significantly larger molecules-such as the nanobodies used for candidalysin detection-access the invagination.
In summary, we describe that two types of injuries-one mediated by the pore-forming toxin candidalysin, the other by mechanical stress-result from the growth of C. albicans hyphae on epithelial cells. To minimize the ensuing damage, host cells activate two distinct repair mechanisms that counteract the damage minimizing epithelial damage, inflammation and transepithelial dissemination. In this manner, epithelia can maintain a homeostatic balance that enables the commensal relationship with the most common opportunistic fungal pathogen, C. albicans.

Limitations of the study
That the effects of candidalysin on mitochondrial potential are caused indirectly by changes in cytosolic ion concentrations remains speculative and needs to be validated directly. Similarly, the proposed dependence of actin dissociation on PtdIns(4,5) P 2 hydrolysis must be confirmed experimentally. Lastly, additional studies are needed to validate the impact of the repair mechanisms in vivo.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: Epithelial invasion assay TR146 cells were plated on 18 mm glass coverslips at 1.5 3 10 5 cells mL À1 in DMEM/F12 containing 10% FBS for 16 to 48 h. The day of the experiment, medium was changed to a minimal medium (140 mM NaCl, 3 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 15 mM HEPES, 5 mM glucose, pH 7.4), and C. albicans strains added to monolayers at MOI of 0.25 to 0.5. After addition of C. albicans, plates were centrifuged for 1 min at 1500 rpm, then incubated at 37 C under 5% CO 2 for up to 5 h. Unbound C. albicans cells were washed 3 times in PBS and samples were fixed with 4% paraformaldehyde (PFA). External C. albicans were labeled for 20 min at room temperature using a solution of 5 mg.mL À1 fluorescent conjugated concanavalin A. In some cases, C. albicans were stained with 10 mg.mL À1 calcofluor white. Candidalysin was visualized staining with CalH1 (1:100) or CalF1 (1:500) antibodies, as indicated in the text. To stain actin filaments, cells were permeabilized 5 min with 0.1% Triton X-100 and incubated 30 min with a 1:1000 dilution of fluorescent phalloidin or Acti-stain.

Cytosolic Ca 2+ measurements using pGP-CMV-GCaMP6s
After transfection of TR146 cells with pGP-CMV-GCaMP6s for 16 h, cells were monitored by time-lapse confocal microscopy for Ca 2+ transients in the minimal medium with or without 1 mM Ca 2+ . 30 mM candidalysin or Ece1-VIIa was added at time point 0 s and (Ca 2+ ) recorded every 20 s for 12 min. Thapsigargin was added to a final concentration of 500 nM at the 12 min mark and the GCaMP6s fluorescence intensity was recorded for an additional 3 min. For TR146 cells that were silenced for ALG-2 or control siRNA, fluorescence intensity was measured for each cell every minute for 12 min. Column statistics was performed on all of the GCaMP6s fluorescence intensity values to calculate the standard deviation of each cell.

OPEN ACCESS
Visualization of TR146 membrane blebbing in response to candidalysin TR146 cells were grown on glass coverslips as described above, and incubated with 30 mM candidalysin, 30 mM Ece1-VIIa, or infected with C. albicans yeast. At indicated time points, samples were washed in ice-cold PBS and incubated 10 min at 10 C to arrest endocytosis. Monolayers were stained for 5 min with 20 mM FM4-64 in cold PBS and imaged immediately for FM4-64-positive membrane blebs. In cases were TR146 cells were infected with C. albicans strains, external C. albicans were labeled adding 5 mg.mL À1 of fluorescent ConA during the initial 10 min of incubation at 10 C, and all C. albicans stained with 10 mg.mL À1 calcofluor during the FM4-64 staining step.
ALG-2 immunostaining TR146 cells were grown on glass coverslips as described above, and incubated with 30 mM candidalysin, 30 mM Ece1-VIIa, or infected with C. albicans yeast. At indicated time points, samples were washed in PBS and fixed with 4% PFA for 20 min at room temperature. Samples were washed twice in PBS and the PM was labeled with fluorescent ConA (1:1000 in PBS, 20 min, room temperature). After two washes in PBS, samples were permeabilized with 0.2% Triton X-100 (15 min, room temperature) and incubated with blocking buffer (2% FBS, 2% BSA) for 60 min at room temperature. Samples were immunostained with rabbit anti-ALG-2 antibody (1:100 in blocking buffer, 60 min) and a fluorescent secondary antibody (1:1000 in blocking buffer, 60 min).
siRNA-mediated silencing and immunoblotting of ALG-2 ALG2-targeting stealth RNAi (siRNA) were purchased from ThermoFisher Scientific. Non targeting Stealth siRNAs were purchased from Dharmacon. siRNA delivery was performed using the Neon transfection system (Life Technologies, ThermoFisher Scientific). TR146 cells were resuspended to 5 3 10 6 cells mL À1 , and 100 mL of suspension mixed with 200 pmol control or ALG-2 siRNA. Electroporation was done using two 20 ms pulses of 1300 V. After electroporation, cells were immediately transferred to DMEM/F12 containing 10% heat-inactivated FBS, before seeding on coverslips at concentration of 2.5 3 10 6 cells mL À1 . siRNA-treated cells were used for experiments 48 h after electroporation. In some cases, after 24 h, cells were transiently transfected with mammalian expression vectors as described.
siRNA-mediated knockdown was confirmed at 48 h by immunoblotting. siRNA-treated cells were lysed in Laemmli buffer (Bio-Rad). Lysates were separated by SDS-PAGE, followed by transfer to a polyvinylidene difluoride membrane. The membrane was blocked in TBS containing 1% skim milk and 0.05% Tween 20 for 30 min at room temperature, followed by primary antibody staining for ALG-2 (1:1000 in 1% skim milk and 0.05% Tween 20 for 60 min) or the loading control Vinculin (1:1000 in 1% skim milk and 0.05% Tween 20 for 60 min) for 1 h at room temperature, in blocking buffer. After washing the membrane in TBS containing 0.05% Tween 20, samples were incubated 30 min at room temperature with an HRP-conjugated secondary antibody at 1:5000 dilution. Blots were visualized using BioRad ChemiDoc MP Imaging System and Image Lab software 5.2.1.
Cell damage assays using siRNA-treated TR146 cells TR146 cells were grown as described above and silenced using electroporation as described above. For some experiments, monolayers were first washed with 1X PBS and incubated in minimal medium with 10 mg.mL À1 rhodamine123 for 15 min at 37 C to label mitochondria (Baracca et al., 2003). Following this, monolayers were washed three times with 1X PBS and incubated for 60 min in the presence or absence of 10 mM candidalysin. The cells were then washed in ice-cold PBS and incubated 10 min at 10 C, then stained for 5 min in cold minimal medium containing 5 mg.mL À1 DAPI and 20 mM FM4-64 to visualize cell death and blebbing, respectively, and imaged immediately by confocal microscopy.
Alternatively, TR146 cells silenced with ALG-2 siRNA or scrambled siRNA were incubated with 10 mM candidalysin in minimal medium at 37 C. Following incubation, the nuclei of dead cells were stained with 20 mM propidium iodide, permeabilized with 0.2% Triton X-100, and counter-stained for total nuclei with SYTOX Green. Candidalysin-induced cell death was calculated as (propidium iodide-positive cells/SYTOX Green-positive cells) x 100.
LAMP-1 immunostaining of invasion pockets TR146 cells were grown on glass coverslips and infected with C. albicans as described above. At indicated time points, samples were washed in PBS and fixed with 4% PFA for 15 min at room temperature. Samples were washed extensively in PBS and blocked for 60 min with a blocking buffer containing 2% BSA and 2% FBS, immunostained with anti-LAMP-1 (1:100 in blocking buffer, 60 min) and a secondary fluorescent antibody (1:1000 in block buffer, 60 min).
Exofacial LAMP-1 immunostaining TR146 cells were grown on glass coverslips as described above and incubated with 30 mM candidalysin or the control peptide Ece1-VIIa. At indicated time points, cells were washed in PBS and fixed with 4% PFA. Cells were washed in PBS and incubated for 60 min in blocking buffer containing 2% BSA and 2% FBS. Samples were immunostained with anti-LAMP-1 hybridoma (1:100 in blocking buffer, 60 min) and a fluorescent secondary antibody (1:1000 in blocking buffer, 60 min). The plasma membrane of TR146 cells was labeled with fluorescent ConA (1:1000 in blocking buffer). After immunostaining, samples were permeabilized in 0.1% Triton X-100 and cell nuclei were labeled with 20 mM propidium iodide.