Repair of a Bacterial Small β-Barrel Toxin Pore Depends on Channel Width

ABSTRACT Membrane repair emerges as an innate defense protecting target cells against bacterial pore-forming toxins. Here, we report the first paradigm of Ca2+-dependent repair following attack by a small β-pore-forming toxin, namely, plasmid-encoded phobalysin of Photobacterium damselae subsp. damselae. In striking contrast, Vibrio cholerae cytolysin, the closest ortholog of phobalysin, subverted repair. Mutational analysis uncovered a role of channel width in toxicity and repair. Thus, the replacement of serine at phobalysin´s presumed channel narrow point with the bulkier tryptophan, the corresponding residue in Vibrio cholerae cytolysin (W318), modulated Ca2+ influx, lysosomal exocytosis, and membrane repair. And yet, replacing tryptophan (W318) with serine in Vibrio cholerae cytolysin enhanced toxicity. The data reveal divergent strategies evolved by two related small β-pore-forming toxins to manipulate target cells: phobalysin leads to fulminant perturbation of ion concentrations, closely followed by Ca2+ influx-dependent membrane repair. In contrast, V. cholerae cytolysin causes insidious perturbations and escapes control by the cellular wounded membrane repair-like response.

IMPORTANCE Previous studies demonstrated that large transmembrane pores, such as those formed by perforin or bacterial toxins of the cholesterol-dependent cytolysin family, trigger rapid, Ca 2ϩ influx-dependent repair mechanisms. In contrast, recovery from attack by the small ␤-pore-forming Staphylococcus aureus alpha-toxin or aerolysin is slow in comparison and does not depend on extracellular Ca 2ϩ . To further elucidate the scope of Ca 2ϩ influx-dependent repair and understand its limitations, we compared the cellular responses to phobalysin and V. cholerae cytolysin, two related small ␤-pore-forming toxins which create membrane pores of slightly different sizes. The data indicate that the channel width of a small ␤-pore-forming toxin is a critical determinant of both primary toxicity and susceptibility to Ca 2ϩdependent repair. P ore-forming proteins are widely used by bacteria to directly damage cells (1), promote intracellular growth (2,3), or introduce virulence factors into the cytosol (4,5). Intriguingly, nucleated cells are able to restore structural and functional plasma membrane (PM) integrity after damage by bacterial pore-forming toxins (PFTs) (6), permitting, for example, recovery from major Staphylococcus aureus alpha-toxindependent losses of cellular ATP (7). Restoration of PM integrity has been also documented for streptolysin O (SLO) (8), pneumolysin (9), aerolysin, and listeriolysin (LLO) (10); it occurs in various cell types in culture and has been shown in Caenorhabditis elegans to operate in vivo (11). Efficient repair of the PM after wounding or attack by proteins forming large pores, such as SLO, perforin, or complement, is thought to require Ca 2ϩ influx (8,(12)(13)(14)(15). Downstream mechanisms include endocytosis of lesions and replacement of the PM by lysosomal exocytosis (15)(16)(17)(18)(19)(20)(21)(22) and/or blebbing of the PM and ectocytosis (23,24). These pathways might act in a complementary manner (25).
Caveolin has been implicated in endocytosis of SLO pores (22). More recently, a requirement of the endosomal complex required for transport (ESCRT) for membrane repair after damage by laser light has been reported; this pathway could also be involved in the repair of membrane pores (26). Notably, the recuperation of cells from an attack by small ␤-pore-forming S. aureus alpha-toxin or aerolysin is significantly slower than that of cells treated with SLO or LLO, and it proceeds in the absence of extracellular Ca 2ϩ (10,27,28). Furthermore, comparative studies showed that recovery following attack by S. aureus alpha-toxin or aerolysin, but not by SLO or LLO, involves p38 mitogen-activated protein kinase (p38 MAPK), autophagy, and phosphorylation of the ␣-subunit of eukaryotic initiation factor 2 (eIF2␣) (10,28,29). Evidently, the mode and efficacy of PM repair and cellular recovery depend on the type of PFT (reviewed in reference 5). In order to comprehend differential cellular tolerance for various PFTs and PFT-producing bacteria, it will be important to elucidate the scope and limitations of Ca 2ϩ influx-dependent repair. Here, we have investigated cellular responses to phobalysin P (PhlyP) and the orthologous Vibrio cholerae cytolysin (VCC) (30)(31)(32)(33)(34)(35), two related small ␤-PFTs of Photobacterium damselae subsp. damselae and V. cholerae, respectively. V. cholerae is the notorious cause of a profuse, life-threatening diarrhea in humans. P. damselae subsp. damselae is a pathogen of marine animals that may infect wounds and lead to hyperaggressive necrotizing soft tissue infection or sepsis in humans. In addition to other proteins (36,37), VCC and PhlyP are considered to serve as virulence factors of these bacteria (35,38,39). We exploited the similar, yet distinct structures of these toxins to gain insight into the function or failure of Ca 2ϩ influx-dependent repair after attack by small ␤-PFTs.

PhlyP and VCC perturb ion concentrations in epithelial cells with different kinetics.
PhlyP is a small ␤-PFT that is related to VCC (39), but in contrast to VCC, it lacks a C-terminal ␤-prism domain (Fig. 1A). Moreover, homology-based modeling of the PhlyP transmembrane pore using the known structure of VCC (40) as a scaffold predicted a wider narrow point of the channel (Fig. 1B) and fewer charged residues clustering in the channel-forming region of PhlyP (Fig. 1C). Therefore, in spite of their homology-50% identity on the amino acid level-it is conceivable that PhlyP and VCC exert different effects on target cells. This conjecture was confirmed by the finding that only PhlyP made epithelial cells (HaCaT cells) permeable to propidium iodide (PI) (39), prompting us to also compare changes of ion concentrations in epithelial cells. Loss of intracellular K ϩ is a hallmark of PM permeabilization by all PFTs investigated so far. Treatment of HaCaT cells with purified PhlyP caused dose-dependent loss of cellular K ϩ within 2 min; little further decrease was observed thereafter ( Fig. 2A). In contrast, the loss of K ϩ was progressive in samples treated with VCC (Fig. 2B). Given that PhlyP made cells permeable to PI (molecular weight [MW] of 668.4), we surmised that it would also permit influx of Ca 2ϩ ions. Cells permeabilized by PhlyP retained the exquisitely Ca 2ϩ -sensitive probe Fluo-8 AM (MW of~1,000) (see Fig. S1A in the supplemental material), which was exploited to detect whether the toxins caused changes of intracellular calcium ion concentrations [Ca 2ϩ ] i . PhlyP (400 ng/ml) led to a significant increase of fluorescence in Fluo-8 AM-loaded cells within 30 s after exposure (Fig. 2C); half-maximal effects were reached at 100 ng/ml and saturation at~200 ng/ml (data not shown). VCC at 400 ng/ml led to a final increase of fluorescence like that of PhlyP (Fig. 2D), although 100 ng/ml VCC remained ineffective (data not shown). Conspicuously, the VCC-dependent increase in fluorescence commenced significantly later than the PhlyP-dependent increase (~60 s versus~12 s; P ϭ 9.5 ϫ 10 Ϫ8 ). This raised the question of whether the two toxins increased [Ca 2ϩ ] i via different mechanisms. Purinoceptors have been implicated in cellular responses to PFT and in the regulation of Ca 2ϩ influx (41)(42)(43). Therefore, we tested the effect of suramin, an inhibitor of P2 receptors, on PFT-dependent changes of [Ca 2ϩ ] i . Suramin exerted a moderate inhibitory effect on the PhlyP-dependent rise of [Ca 2ϩ ] i ( Fig. 2C) but virtually blocked the VCC-dependent increase (Fig. 2D).
Epithelial cells replenish K ؉ after perforation by PhlyP. To investigate whether PhlyP-treated epithelial cells were able to recover, we measured cellular K ϩ levels immediately after a brief incubation with toxin or after incubation for various times in the absence of toxin. Following incubation of cells with PhlyP (100 ng/ml for 10 min at 37°C), the cellular K ϩ levels were reduced to~10%, but they returned to normal within 1 h after the removal of unbound toxin (Fig. 3A). A similar recovery was observed when cells were treated with 500 ng/ml PhlyP (see Fig. S1B in the supplemental material). In contrast, the loss of cellular K ϩ in response to VCC was sustained (Fig. 3A). Notably, the combination of both toxins behaved like VCC alone, indicating that the rescue process, apparently triggered by PhlyP, cannot save cells simultaneously intoxicated by VCC (Fig. 3B). Although incubation of cells with PhlyP for 8 min sufficed to cause significant influx of PI, membrane integrity was reconstituted after the washing out of PhlyP. Resealing was observed whether cells were treated with purified PhlyP (data not shown) or extracellular products (ECPs) from strain AR119, a P. damselae subsp. damselae strain expressing PhlyP ( Fig. 3C) (39,44).
PhlyP elicits a wounded membrane repair-like response, but VCC does not. Because EGTA prevented the restoration of K ϩ levels ( Fig. 4A), and because the depletion of K ϩ was irreversible in Ca 2ϩ -free medium (see Fig. S1C in the supplemental material), we investigated whether mechanisms proposed to act downstream from Ca 2ϩ influx-dependent repair of large membrane pores (15,18,(22)(23)(24) were also involved here. Therefore, we measured the release of ␤-hexosaminidase, a marker of lysosomal exocytosis (45). Notably, PhlyP causes no leakage of lactate dehydrogenase (39), and EGTA blocked the release of ␤-hexosaminidase (data not shown), demonstrating that ␤-hexosaminidase release faithfully reported lysosomal exocytosis. PhlyP induced release of the enzyme from HaCaT cells (Fig. 4B), but VCC was ineffective ( Fig. 4B and C). In line with a role of lysosomal exocytosis for recovery from PhlyP attack, desipramine (DPA), an inhibitor of acid sphingomyelinase (ASM), which impairs the reversal of SLO-dependent membrane permeabilization (18), reduced the replenishment of cellular K ϩ without altering the initial toxin-dependent loss of this ion (see Fig. S1D). Consistent with this, inhibition of ASM did not aggravate the VCC-dependent loss of K ϩ (see Fig. S1E). PhlyP caused the formation of large, dynamic blebs in HaCaT cells, some of which appeared to detach to form large vesicles (see Movie S1), and blebbistatin stalled the formation of free vesicles (see Movie S2). However, neither alone nor in combination with DPA did blebbistatin alter the replenishment of cellular K ϩ after attack by PhlyP (see Fig. S1D), thus not supporting a major role of blebbing for cellular recovery from PhlyP. And yet, PhlyP led to increased endocytosis of fluorescently labeled bovine serum albumin (BSA), a cargo of caveolar uptake (see Fig. S2A), Small Pore-Forming Toxins May Trigger or Avoid Repair ® and SDS-stable oligomers were coisolated with caveolin and the exosomal marker protein flotillin in supernatants of target cells, suggesting sequential endocytosis and exocytosis of PhlyP (see Fig. S2B). The wounded membrane repair-like response following membrane perforation by SLO has been demonstrated by RNA interference (RNAi) to depend on caveolin (22), and a current model is depicted in Fig. S2D. Here, we exploited mouse embryonal fibroblasts (MEF) lacking caveolin expression (MEFcav Ϫ/Ϫ ) to investigate whether this protein is also important for cellular defense against the small ␤-pore-forming toxin PhlyP. As in HaCaT cells, VCC and PhlyP both caused loss of K ϩ from wild-type MEF (MEFwt) or MEFcav Ϫ/Ϫ (Fig. 4D and E). Also, restoration of the intracellular K ϩ concentration was not observed after treatment with VCC (Fig. 4E), whereas replenishment of K ϩ was efficient in PhlyP-treated wild-type cells. Importantly, replenishment of K ϩ failed in MEFcav Ϫ/Ϫ exposed to PhlyP (Fig. 4D). That PhlyP triggered a wounded membrane repair-like response in MEF was further suggested by the facts that it increased the fraction of cells carrying ceramide ( Fig. 4F and G) and that it led to exposure of a luminal epitope of the lysosomal marker protein LAMP-1 at the cell surface (Fig. 4H) (17); only minor effects were discernible after treatment with VCC. The protective role of caveolin was confirmed when we analyzed PI influx by flow cytometry. Exposure to PhlyP for 5 min led to some influx of PI in wild-type and caveolin-deficient cells. However, wild-type cells soon excluded PI again (15 min), whereas the fraction of MEFcav Ϫ/Ϫ positive for PI had increased (Fig. 4I). By 45 min, two-thirds of MEFcav Ϫ/Ϫ but only about one-fourth of wild-type cells were heavily stained with PI. Recovery from PhlyP is not blocked by the p38 MAPK inhibitor SB203580. In contrast to PhlyP, other small ␤-pore-forming toxins, i.e., aerolysin and S. aureus alpha-toxin, have been previously shown to trigger not Ca 2ϩ influx-dependent repair but slower, p38 MAPK-dependent recovery processes (28). Like VCC, PhlyP activates p38 MAPK (see Fig. S3A in the supplemental material). However, an inhibitor of activated p38 MAPK, SB203580, did not impede the replenishment of cellular K ϩ in PhlyP-treated cells (see Fig. S3B).
Channel narrow points impact fluxes of Ca 2؉ . The differential effects of PhlyP versus VCC on cellular homeostasis, influx of vital dyes, and repair could be due at least in part to differences in their transmembrane channels. Specifically, the bulky side chain of W318 in the VCC channel forms a heptad, reminiscent of the phenylalanine clamp in the anthrax protective antigen pore (4,40), and it could restrict the flux of ions or dyes. In contrast, serine 341, predicted to form the narrow point in PhlyP pores, is expected to be less obstructive. To investigate a potential impact of channel narrow points on toxin function, we generated single-amino-acid exchange mutants of the VCC and PhlyP protoxins pVCC and pPhlyP, in which tryptophan at position 318 (W318) of pVCC and serine at position 341 of pPhlyP were swapped, creating mutants pVCC(W318S) and pPhlyP(S341W) (Fig. 5A). HaCaT cells were loaded with Fluo-8 AM and treated with wild-type or mutant toxins, and fluorescence was recorded; the experiment was performed in the presence or absence of suramin. Intriguingly, pPhlyP(S341W)hereinafter termed pPhlyP S/W-caused a significantly lower suramin-insensitive increase of fluorescence than pPhlyP (Fig. 5B). For comparison of wild-type and mutant VCC (Fig. 5C), the mature toxins were generated with trypsin, because maturation by cellular proteases appeared comparably inefficient; VCC(W318S)-termed VCC W/S below-caused increases of fluorescence similar to those caused by wild-type VCC, but only in the case of VCC W/S was a significant portion of that signal insensitive to suramin (Fig. 5C).
Narrowing the PhlyP channel limits primary damage and repair; widening the VCC channel enhances toxicity. Next, we asked whether the above-described point mutations had an impact on membrane damage or repair. Cells were treated with wild-type or mutant protoxins, stained with PI and Hoechst stain, and examined by fluorescence microscopy. Only minimal influx of PI was noted upon a short incubation with either wild-type or mutant pVCC, but loss of membrane integrity progressed von Hoven et al.

®
inexorably despite washing out of toxin ( Fig. 6A and B; see also Fig. S4 in the supplemental material). Notably, cells deteriorated more rapidly after exposure to pVCC W/S, as suggested by particularly strong influx of PI and nuclear condensation (see Fig. S4 and 6B and C, respectively). The swapping of serine and tryptophan in PhlyP led to more pronounced changes: only in half of the cells was influx of PI observed, and this only to a small extent, upon treatment with pPhlyP S/W, while the majority of cells treated with wild-type pPhlyP were brightly stained. Surprisingly, pPhlyP S/W-treated cells continued to permit low-level ingress of the dye, while membrane integrity was restored in pPhlyP-treated cells. Quantification of SDS-stable oligomers and analysis by electron microscopy did not indicate alterations in the ability to form oligomers (see Fig. S5). Individual PhlyP pores were of equal or slightly higher conductance than mutant pores for all salts tested, but the traces for the PhlyP S/W mutant pores showed increased levels of flickering and frequent breakdown of conductance, which lasted for seconds in some cases (see Fig. S6). We also found that PhlyP pores (mutant or wild type) showed higher conductance than VCC pores (about 300 versus 22 pS in KCl) and  Fig. 1A. Colored letters in cytolysin domain indicate residues exchanged in mutants. (B and C) HaCaT cells were pretreated or not with 150 g/ml suramin, loaded with Fluo-8 AM in the presence or absence of inhibitor, and exposed to pPhlyP or pPhlyP S/W (B) or to VCC or VCC W/S (C) at 400 ng/ml. Fluorescence intensity was recorded at intervals of 5 s for a period of 3 min. Mean values Ϯ SE are shown (n Ն 5). P values were determined with Student's t test.
Small Pore-Forming Toxins May Trigger or Avoid Repair ® were rather cation selective, while VCC was moderately anion selective (46). Differential behavior of pPhlyP and pPhlyP S/W was also evident in MEF cells. First, the influx of PI was more pronounced in response to pPhlyP (see Fig. S7A). Second, pPhlyP caused a far stronger release of ␤-hexosaminidase in HaCaT cells (Fig. 7A) or caveolin-deficient MEF (Fig. 7B) than did pPhlyP S/W. This release was completely blocked by EGTA (data not shown). The binding of pPhlyP to wild-type or knockout MEF was equal (Fig. S7B), but MEFcav Ϫ/Ϫ cells were more sensitive to pPhlyP (Fig. S7A). The difference between PhlyP and pPhlyP S/W was more pronounced in wild-type cells (Fig. S7A). Similarly, DPA sensitized MEFwt for both PhlyP and pPhlyP S/W (see Fig. S7C). Thus, late steps of the wounded membrane repair-like response (ASM-and caveolin-dependent steps; see Fig. S2D) appear to be involved in ongoing defense of MEF against PhlyP or pPhlyP S/W. The different capacities of PhlyP and pPhlyP S/W to trigger the release of ␤-hexosaminidase surfaced when caveolin-dependent tolerance was disabled (Fig. 7B).

DISCUSSION
The present work reveals that different small ␤-pore-forming toxins may either trigger or subvert calcium influx-dependent repair. Furthermore, the data suggest that the channel width of small ␤-pores codetermines the kinetics and degree of primary damage, as well as susceptibility to repair. PhlyP, in contrast to the closely related VCC, caused fulminant breakdown of membrane integrity but permitted resealing by a process which until now has only been implicated in the repair of much larger membrane lesions, for instance, pores formed by cholesterol-dependent cytolysins. The modeling-based hypothesis that the narrow point in the PhlyP channel is wider than that in VCC is supported by conductance measurements.
Repair of PhlyP pores involves Ca 2ϩ influx, lysosomal exocytosis, and caveolin, but MAPK p38 is dispensable, supporting the idea that Ca 2ϩ influx-dependent repair supersedes the requirement for alternative salvage pathways. In striking contrast to PhlyP, VCC subverts Ca 2ϩ influx-dependent repair. Thus, even brief exposure of cells to low concentrations of VCC sufficed to initiate the progressive demise of human epithelial cells. That nanomolar concentrations of VCC are required to increase [Ca 2ϩ ] i , although picomolar concentrations are sufficient to kill cells (34), provides an explanation for VCC's propensity to subvert repair. However, even concentrations of VCC sufficient to increase [Ca 2ϩ ] i did not elicit a membrane repair response. Inappropriate topology, timing, or the degree of VCC-dependent increases of [Ca 2ϩ ] i may be responsible: that VCC-dependent increases of [Ca 2ϩ ] i were blocked by suramin indicated that they are mediated by P2 receptors, G-protein-coupled receptors, or other targets of the drug, which might cause Ca 2ϩ fluxes not to occur in sufficient proximity to VCC pores to allow repair. Second, compared to the PhlyP-dependent Ca 2ϩ influx, the VCCdependent Ca 2ϩ influx was delayed and comparatively slight. Therefore, we believe that the progressive damage by VCC is due to inadequate Ca 2ϩ influx through the small and anion-selective pore (34,40), in the face of otherwise severe perturbations of cellular physiology (e.g., loss of K ϩ ). Sure enough, VCC proved to be an inefficient trigger of lysosomal exocytosis. Notably, the PhlyP-dependent responses were unable to compensate for VCC's inability to trigger repair in mixing experiments. This could happen if VCC inhibits a step of the repair program downstream from lysosomal exocytosis, for instance, caveolar endocytosis. Because the abilities of VCC and PhlyP to trigger or subvert Ca 2ϩ influx-dependent repair appeared to correlate with their different channel narrow points, mutational analysis was a sensible approach. Single residues presumed to form channel narrow points of PhlyP and VCC were swapped to reveal their contributions to functional phenotypes. Changes in the Ca 2ϩ influx resulting from mutations of channel narrow points might impact both the cytotoxic power Small Pore-Forming Toxins May Trigger or Avoid Repair ® and ability to trigger repair responses; it was not predictable which effect would prevail. And yet, the interpretation of the data obtained with these constructs was straightforward: W318 restricts the influx of calcium ions through VCC pores; the obstacle falls away in the W318S mutant. Conversely, the replacement of S341 in pPhlyP with tryptophan reduces the influx of calcium ions. A principle finding made with the mutant protoxins was that the effect of channel width on cellular responses depends on the molecular context. The wider narrow point of the PhlyP channel promotes lysosomal exocytosis and recovery. However, it fails to do so if transplanted to VCC; in fact, it enhances toxicity in that context. The reason could be that the moderately increased influx of Ca 2ϩ through mutant VCC pores is sufficient to enhance toxicity but too low to trigger Ca 2ϩ influx-dependent repair. As a matter of fact, pVCC W/S caused only slightly greater increases of Ca 2ϩ influx than did wild-type pVCC, and the increase in the ␤-hexosaminidase release was statistically insignificant. That pVCC W/S did not elicit a stronger suramin-insensitive increase of [Ca 2ϩ ] i could be due to additional restraints in VCC pores. To sum up, VCC and PhlyP, two related small ␤-PFTs, pose quite different challenges to cell autonomous defense, which may, at least in part, be attributed to different channel widths (Fig. 8): whereas PhlyP acts fast and thus could overrun host responses if present in sufficient quantities, VCC causes insidious damage and subverts membrane repair. The results highlight the function or failure of Ca 2ϩ influx-dependent repair as a defense against small ␤-PFTs; this may help to better understand the pathogenesis of diseases caused by bacteria producing these widespread toxins.

MATERIALS AND METHODS
Toxins. The preparation of PhlyP and VCC was as described previously (39). In brief, PhlyP was purified by preparative isoelectric focusing and ion exchange chromatography from extracellular products of P. damselae subsp. damselae. Recombinant protoxins pVCC, pVCC W/S, pPhlyP, and pPhlyP S/W were expressed in Escherichia coli as N-terminally His 6 -tagged fusion proteins and purified by affinity chromatography; VCC was generated from pVCC with trypsin (33). Single-amino-acid-exchange mutants of pPhlyP and pVCC were generated with the aid of the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies). For primer sequences and technical details, see Text S1 in the supplemental material.
In silico modeling of the PhlyP pore. The sequence of PhlyP was modeled on the X-ray structure of VCC (PDB identifier [ID] 3O44 [40]). Alignment and structural modeling were performed by using MODELLER version 9.13 (47). See Text S1 in the supplemental material for details.
Flame photometry for measurement of K ؉ . The loss and replenishment of cellular K ϩ levels after an initial loss is a valuable proxy of perturbation and reconstitution of membrane integrity after attack by small ␤-PFTs and other PFTs (10, 28). Cellular K ϩ was quantified by flame photometry as described PhlyP, a small ␤-barrel pore-forming toxin which forms comparatively wide pores, triggers rapid Ca 2ϩ influx, lysosomal exocytosis, and repair similarly to the large pore-forming streptolysin O (SLO). In contrast, small ␤-barrel pore-forming toxins like Vibrio cholerae cytolysin (VCC) form narrower channels and subvert this response. CDC, cholesterol-dependent cytolysins.
von Hoven et al. previously (10). In brief, cells were washed three times with ice-cold K ϩ -free choline buffer. Cells were subsequently lysed by incubation for 30 min in choline buffer-0.5% Triton X-100 at room temperature on a shaker. Lysates were analyzed for K ϩ with an M401 flame photometer (Sherwood, United Kingdom) using propane gas. Fluo-8 AM-based Ca 2؉ assay. PFT-induced changes of [Ca 2ϩ ] i in HaCaT cells were monitored by using Fluo-8 AM from Santa Cruz Biotechnology, Inc., in a TriStar LB 941 instrument from Berthold Technologies, as detailed in Text S1 in the supplemental material.
Fluorescence microscopy. Immunofluorescence analysis of ceramide and the lysosomal marker protein LAMP-1 was performed with MEF because the available antibodies yielded unspecific staining in HaCaT cells. The staining protocols for LAMP-1 and ceramide were as described in the supplemental material.
PI influx. The PI influx assay was performed as described previously (39). See Text S1 in the supplemental material for details.
Statistics. The data shown are from Ն3 independent experiments if not otherwise stated. Error bars represent plus-or-minus standard errors of the means. The statistical significance of differences between mean values was assessed with the two-sided Student's t test or with one-way analysis of variance (ANOVA) for multiple comparison; significance was assumed when the P value was Յ0.05.