Membrane potential is vital for rapid permeabilization of plasma membranes and lipid bilayers by the antimicrobial peptide lactoferricin B

Lactoferricin B (LfcinB) is a cationic antimicrobial peptide, and its capacity to damage the bacterial plasma membrane is suggested to be a main factor in LfcinB's antimicrobial activity. However, the specific processes and mechanisms in LfcinB-induced membrane damage are unclear. In this report, using confocal laser-scanning microscopy, we examined the interaction of LfcinB with single Escherichia coli cells and spheroplasts containing the water-soluble fluorescent probe calcein in the cytoplasm. LfcinB induced rapid calcein leakage from single E. coli cells and from single spheroplasts, indicating that LfcinB interacts directly with the plasma membrane and induces its rapid permeabilization. The proton ionophore carbonyl cyanide m-chlorophenylhydrazone suppressed this leakage. Next, we used the single giant unilamellar vesicle (GUV) method to examine LfcinB's interaction with GUVs comprising polar lipid extracts of E. coli containing a water-soluble fluorescent probe, Alexa Fluor 647 hydrazide (AF647). We observed that LfcinB stochastically induces local rupture in single GUVs, causing rapid AF647 leakage; however, higher LfcinB concentrations were required for AF647 leakage from GUVs than from E. coli cells and spheroplasts. To identify the reason for this difference, we examined the effect of membrane potential on LfcinB-induced pore formation, finding that the rate of LfcinB-induced local rupture in GUVs increases greatly with increasing negative membrane potential. These results indicate that membrane potential plays an important role in LfcinB-induced local rupture of lipid bilayers and rapid permeabilization of E. coli plasma membranes. On the basis of these results, we discuss the mode of action of LfcinB's antimicrobial activity.

Lactoferricin B (LfcinB) 3 is produced by the hydrolysis of transferrin-like bovine lactoferrin in milk (e.g. pepsin digestion in the human stomach) (1,2), which has antimicrobial activities against Gram-negative and -positive bacteria, yeast, and filamentous fungi (2)(3)(4), and hence, LfcinB is one of the antimicrobial peptides (AMPs). LfcinB is a 25-residue cationic peptide (FKCRRWQWRMKKLGAPSITCVRRAF) with 5 Arg residues and three Lys residues (3). The three-dimensional LfcinB structure exhibits an amphipathic character because one surface of the peptide is hydrophobic, whereas its opposite surface is hydrophilic (5). This amphipathic character of the three-dimensional structure of peptides is common to most AMPs (6 -10). Shorter versions of LfcinB also have antimicrobial activity (4,(11)(12)(13)(14)(15). A shorter version of LfcinB composed entirely of D amino acid residues had the same antibacterial activity as that composed of natural L amino acid residues (16,17), indicating that proteins are not involved in the antibacterial activity of this peptide.
In our previous paper (18), we investigated the interaction of LfcinB with single giant unilamellar vesicles (GUVs) composed of negatively-charged dioleoylphosphatidylglycerol (DOPG) and electrically neutral dioleoylphosphatidylcholine (DOPC) mixtures containing a water-soluble fluorescent probe, calcein, in their lumen using the single GUV method (18). It is wellknown that using the single GUV method we can separate a process of AMP-induced damage of a membrane such as pore formation from a process of leakage (membrane permeation) of fluorescent probes through the damage (19). At the beginning of the interaction of LfcinB with a single GUV, the calcein concentration in the GUV lumen remained constant, but suddenly calcein started to leak from the GUV and the leakage completed within 5 s after starting the leakage, with a concomitant decrease in GUV size. We performed the same experiments using many "single GUVs," and found that the leakage started at different times in all GUVs but the time required for complete leakage after starting was less than 5 s in all GUVs. To express the LfcinB-induced membrane damage inducing this rapid leakage, we used the term "local rupture" instead of "large pore." Moreover, in the interaction of LfcinB with Escherichia coli cell suspension in the presence of a membrane-impermeant fluorescent probe, SYTOX Green, it was observed that the fluorescence intensity of SYTOX Green of the suspension increased gradually with time, suggesting LfcinB-induced influx of SYTOX Green from the outside of E. coli into the cytoplasm (18). These results suggest that LfcinB-induced plasma membrane damage is a main factor in its antimicrobial activity. However, experimental data demonstrating the direct interaction of LfcinB with single bacterial cells is needed.
In the present study, we examined the interaction of LfcinB with single E. coli cells to reveal the mechanism of its antimicrobial activity. For this purpose, we investigated the interaction of LfcinB with live single E. coli cells containing calcein in the cytoplasm using confocal laser scanning microscopy (CLSM). We found that LfcinB induced rapid leakage of calcein from single E. coli cells, indicating that LfcinB induced rapid permeabilization due to damage or pore formation in the plasma membrane. Gram-negative bacteria such as E. coli are composed of an outer membrane, a peptidoglycan layer, and a plasma membrane (inner membrane); hence, there may be several targets of LfcinB. To examine whether direct interaction of LfcinB with the plasma membrane plays a role in the calcein leakage, we investigated the interaction of LfcinB with single spheroplasts derived from E. coli cells using CLSM. Spheroplasts do not have an outer membrane and peptidoglycan, and hence only the plasma membrane and cytoplasm remain (20 -23). We prepared spheroplasts from filamentous E. coli cells and loaded calcein in the cytoplasm. We found that LfcinB induced rapid leakage of calcein from single spheroplasts. To elucidate the elementary processes and the mechanism of LfcinB-induced damage of the plasma membrane, we examined the interactions of LfcinB with GUVs of E. coli polar lipid extract (E. coli-lipid) using the single GUV method (19, 24 -26). LfcinB induced stochastic local rupture in E. coli-lipid-GUVs, but higher concentrations of LfcinB were required to induce a significant rate of local rupture compared with that observed in E. coli cells and spheroplasts. We hypothesized that this difference is due to the membrane potential in the cells and spheroplasts, because the membrane potential of E. coli cells is Ϫ100 ϳ Ϫ130 mV at pH 6.0ϳ7.0 (27) and it is expected that the negative membrane potential could affect the interaction of highly positively charged LfcinB with the plasma membrane due to strong electrostatic interactions. To verify this hypothesis, we examined the effect of membrane potential on LfcinBinduced local rupture or pore formation in single GUVs and also the effect of a proton ionophore, carbonyl cyanide m-chlorophenylhydrazone on the LfcinB-induced leakage of calcein from single E. coli cells and spheroplasts. On the basis of these results, we discuss the mode of action of LfcinB's antimicrobial activity.

Interaction of LfcinB with single E. coli cells
To elucidate the mechanism of action of LfcinB in live bacterial cells, we examined the interaction of LfcinB with single live E. coli cells in EZ rich medium using CLSM. EZ medium is a synthetic cell culture medium, allowing us to control for unexpected interactions of peptides with unknown components contained in most media (28). We loaded calcein to the cytoplasm of E. coli cells, using the interaction between calcein-AM and E. coli cells. It is well-known that hydrophobic calcein-AM can pass through the plasma membrane to enter the cytoplasm of bacterial cells, where it is hydrolyzed by esterases to produce calcein (29,30). We applied the same method to load calcein in E. coli spheroplasts, and successfully loaded spheroplasts with calcein (see the next section), indicating the validity of the above interpretation.
First, we investigated the interaction of 3.0 M LfcinB with single E. coli cells in EZ rich medium in a microchamber at 25°C. Under this condition before the interaction of LfcinB, E. coli cells grew and the average increase in cell length for 10 min was 11 Ϯ 3%. A LfcinB solution was continuously provided to the vicinity of an E. coli cell through a micropipette, so that the LfcinB concentration near the cell became constant at a steady state, and was almost the same as that in the micropipette (31). Two kinds of cells were present, nonseptating and septating cells. Fig. 1A shows a typical result for septating cells. During the interaction of 3.0 M LfcinB, the fluorescence intensity (FI) of the cell due to calcein was constant during the initial interaction (up to 26 s), then rapidly decreased (Fig. 1A, and green line in C and D), indicating that LfcinB induced rapid permeabilization. This is due to LfcinB-induced damage or pore formation in the plasma membrane through which calcein leaked rapidly. On the other hand, the length of the cell increased before the addition of LfcinB (6% increase for 10 min), but subsequently the length gradually decreased (7% decrease for 6 min after the start of the interaction) (red squares in Fig. 1, C and D). The same experiments were performed using 10 cells (i.e. 7 nonseptating and 3 septating cells), and the same result (i.e. first a rapid leakage of calcein occurred, then a gradual decrease in cell length) was observed in 7 cells (i.e. 5 nonseptating and 2 septating cells). Thus, the fraction of LfcinB-induced leaked cells among all examined cells after 6 min reaction (P leak (6 min)), which is a measure of the rate of pore formation or local rupture induced by peptides (24,32), was 0.70 for 3.0 M LfcinB. We conducted five independent experiments (n ϭ 5) for the 3.0 M LfcinB concentration, and obtained the mean Ϯ S.D. of the fraction of leaked cell. P leak for 3.0 M LfcinB was 0.71 Ϯ 0.05. Fig. 1E shows the time course of FI of several cells during the interaction of 3.0 M LfcinB. Each curve corresponds to the time course of FI for each cell. This result indicates a wide distribution of rates of LfcinB-induced leakage of calcein from a single cell. The rate of leakage in septating cells was larger than that in nonseptating cells. To compare the rate more quantitatively, we estimated the rate of LfcinB-induced leakage of calcein from a single cell using halftime (i.e. t1 ⁄ 2 ), which is defined as the time required for leakage of 50% of total calcein from a single cell after the start of the leak-Membrane potential is vital for AMP-induced permeabilization age. We plotted the distribution of t1 ⁄ 2 using a histogram (Fig.  S1). This indicates that the t1 ⁄ 2 values of all septating cells (n ϭ 28) were less than 50 s, but the t1 ⁄ 2 values of nonseptating cells (n ϭ 44) was widely distributed from 6 to 259 s. This quantitative analysis supports the conclusion that the rates of leakage from septating cells were larger than those from most nonseptating cells and there is a wide distribution of the rates of leakage from nonseptating cells.
Before the interaction of LfcinB the cell length of E. coli cells increased, but after the start of the interaction of LfcinB the cell length started to decrease, indicating that the LfcinB-induced leakage from cells stopped cell growth. These results indicate that first LfcinB induced permeabilization in plasma membrane of E. coli cells due to damage or pore formation in the plasma membrane, through which calcein rapidly leaked, and then cell growth stopped, resulting in a decrease in cell length. A similar decrease in the cell length of E. coli cells was reported for the AMP, LL-37 (33).
Next, we investigated the interaction of a lower concentration (0.5 M) of LfcinB with single E. coli cells in the medium at 25°C. Fig. 1B shows a typical result for nonseptating cells. During the interaction of 0.5 M LfcinB, the FI of a cell due to

Membrane potential is vital for AMP-induced permeabilization
calcein was constant initially (up to 110 s), and then gradually decreased (Fig. 1B, and green line in F), indicating that LfcinB induced rapid permeabilization due to damage or pore formation in the plasma membrane through which calcein leaked. On the other hand, cell length increased before the interaction of LfcinB (10% increase for 10 min), but subsequently cell length gradually decreased (9% decrease for 6 min) (red squares in Fig.  1F). Fig. 1G shows the LfcinB concentration dependence of P leak (6 min). At 0.1 M, no leakage was observed (i.e. P leak ϭ 0), whereas at and above 0.5 M, the P leak increased with an increase in LfcinB concentration. This result indicates that the rate of damage or pore formation in the plasma membrane of E. coli cells increased in a LfcinB concentration-dependent manner. On the other hand, the rate of LfcinB-induced leakage of calcein through damages or pores also greatly depended on LfcinB concentration. Fig. 1H shows the time course of FI of several cells due to calcein among all examined cells (n ϭ 17 for 2 independent experiments) during the interaction of 5.0 M LfcinB, indicating that LfcinB induced leakage of calcein from a single cell at a great rate and the values of this rate exhibited a sharp distribution. The mean Ϯ S.E. of leakage half-time (t1 ⁄ 2 ) were obtained for each LfcinB concentration (the number of cells, n, was 15-17, except for 3.0 M (n ϭ 72) and 0.5 M (n ϭ 5)) ( Fig. 1G). For example, t1 ⁄ 2 for 2 and 5 M LfcinB was 100 Ϯ 20 and 10 Ϯ 1 s, respectively. Leakage half-time decreased with an increase in LfcinB concentration (Fig. 1G), indicating that the rate of leakage from the pores increased in an LfcinB concentration-dependent manner.
It is reported that the minimum inhibitory concentration (MIC) of LfcinB against E. coli (JM-109) is 3 Ϯ 1 M (18). For measurement of MIC, we use a suspension of many E. coli cells. As the interaction of LfcinB with many cells proceeds, the effective LfcinB concentration outside the cells decreases. For example, if we use 3 M LfcinB initially, the effective concentration outside the cells is smaller than 3 M. Moreover, the MIC values are determined after ϳ20 h culture. In contrast, in the experiments of interaction of LfcinB with the single E. coli cells described above, we continuously add LfcinB solution, and hence the LfcinB concentration remains constant from the beginning (e.g. 3 M). Therefore, LfcinB concentration near E. coli cells in a single cell experiment is larger than that in the MIC measurement using a suspension of cells. This is the reason why P leak (6 min) was large for 3.0 M LfcinB. To compare quantitatively the result of single cell experiments shown in Fig.  1 and that of MIC measurement, we need a systematic investigation in the near future.
To elucidate the role of membrane potential in the interaction of LfcinB with single live E. coli cells, we investigated the effect of a proton ionophore, CCCP, on the interaction of LfcinB with the cells. It is well-known that CCCP decreases the transmembrane difference of the electrochemical potential of proton (or the proton-motive force) in the plasma membrane of bacterial cells, such as E. coli, by increasing proton permeability in the membrane, dissipating the membrane potential, ⌬, as well as the transmembrane pH difference, ⌬pH (34). If the pH of the external medium is neutral, ⌬pH is small (27). The pH of EZ medium used in the above experiments is pH 7.0, and hence, we can consider that CCCP mainly dissipated ⌬ under our experimental conditions.
In the presence of 100 M CCCP in the EZ medium (final concentration), we investigated the interaction of 5.0 M LfcinB with single E. coli cells in the medium at 25°C. The FI of the cell due to calcein was constant during the interaction (up to 6 min) (Fig. 2, A and B), indicating LfcinB did not induce any damage in the plasma membrane through which calcein leaked. The same experiments were performed using 12 cells (i.e. 10 nonseptating and 2 septating cells), and the same result was observed. We conducted two independent experiments using 12-15 E. coli cells (n ϭ 2), and obtained the same results. These results indicate that membrane potential in the plasma membrane plays an important role in LfcinB-induced permeabilization due to damage or pore formation in the plasma membrane of E. coli cells.

Interaction of LfcinB with single E. coli spheroplasts
To examine the direct interaction of LfcinB with the plasma membrane of E. coli cells, we investigated the interaction of LfcinB with single spheroplasts derived from E. coli cells using CLSM. We loaded calcein in the lumen of spheroplasts by interacting calcein-AM with spheroplasts using the same method used for E. coli cells, as described in the previous section.
First, we investigated the interaction of 2.0 M LfcinB with single spheroplasts in 10 mM Tris-HCl (pH 7.8) containing 1.5 mM KCl, 48.5 mM NaCl, and 0.73 M sucrose at 25°C (the concentrations of K ϩ and Na ϩ are almost the same as those in the EZ rich medium). A LfcinB solution was continuously provided to the vicinity of a spheroplast through a micropipette. During the interaction of 2.0 M LfcinB, the FI of the spheroplast lumen due to calcein was constant up to 66 s, and then rapidly decreased to zero within a few seconds (Fig. 3, A and B), indicating that LfcinB induced rapid permeabilization due to damage or pore formation in the plasma membrane of the sphero-

Membrane potential is vital for AMP-induced permeabilization
plast through which calcein leaked. When we performed the same experiments using other spheroplasts, similar rapid leakage of calcein occurred stochastically within 6 min after a gradual decrease in FI (Fig. 3C). This initial gradual decrease in FI was also observed in spheroplasts in the absence of LfcinB, and hence we can consider reasonably that it is due to the photobleaching of calcein (Fig. 3C). For 2.0 M LfcinB, 13 spheroplasts exhibited leakage by 6 min among all examined ones (n ϭ 16) (i.e. P leak (6 min) ϭ 0.81).
Next, we investigated the interaction of higher concentrations of LfcinB with single spheroplasts at 25°C. During the interaction of 5.0 M LfcinB, the FI of spheroplasts due to calcein was constant, and then rapidly decreased (Fig. 3D), indicating that LfcinB induced rapid permeabilization due to damage or pore formation in the plasma membrane of the spheroplasts. Fig.  3E shows the LfcinB concentration dependence of P leak (6 min). We conducted two independent experiments for each LfcinB concentration, and obtained the mean Ϯ S.D. of the fraction of leaked spheroplast. At 0.1 M, no leakage was observed (i.e. P leak ϭ 0), whereas at 2.0 M, P leak ϭ 0.76 Ϯ 0.07, and at 5.0 M, P leak ϭ 1.0; thus, P leak increased in an LfcinB concentration-dependent manner. This result indicates that the rate of damage or pore formation in the plasma membrane of spheroplasts increased with an increase in LfcinB concentration.
These results indicate that the direct interaction of LfcinB with the plasma membrane of spheroplasts induces rapid permeabilization due to damage or pore formation in the membrane, through which the internal contents rapidly leak. The dependence of the fraction of leaked spheroplast for 6 min on the LfcinB concentration was similar to that of E. coli cells, indicating that one of the targets of LfcinB in E. coli cells is the plasma membrane. In contrast, the time course of the LfcinBinduced leakage from spheroplasts was different from that of E. coli cells. The starting time and the rate of leakage in E. coli cells was slower and had a wider distribution compared with that in spheroplasts. This difference can be explained as follows. In E. coli cells, LfcinB molecules must pass through the outer membrane and peptidoglycan layer in the periplasm to

Membrane potential is vital for AMP-induced permeabilization
reach the plasma membrane, which decreases and varies the rate of increase in LfcinB concentration in the plasma membrane. Moreover, for the leakage of calcein molecules to the outside of the cells, they must pass through the peptidoglycan layer and the outer membrane, which decreases and varies the rate of leakage of calcein. Therefore, the difference of the time course of the leakage is due to the barriers (i.e. the peptidoglycan layer and the outer membrane) of the cells.
Several experimental results indicate that spheroplasts from E. coli cells have membrane potential (35). To elucidate the role of membrane potential in the interaction of LfcinB with single spheroplasts, we examined the effect of a CCCP on the interaction of LfcinB with the spheroplasts. In the presence of 100 M CCCP in 10 mM Tris-HCl (pH 7.8) containing 1.5 mM KCl, 48.5 mM NaCl, and 0.73 M sucrose, we investigated the interaction of 5.0 M LfcinB with single spheroplasts at 25°C. A gradual decrease in FI occurred during the interaction (up to 6 min) (Fig. 4, A and B) due to the photobleaching of calcein, indicating LfcinB did not induce any damage in the plasma membrane through which calcein leaked. The same experiments were performed using 15 spheroplasts, and the same result was observed (i.e. P leak ϭ 0). We conducted two independent experiments using 15 spheroplasts (n ϭ 2), and obtained the same results. This result indicates that the membrane potential in the spheroplasts plays an important role in the LfcinB-induced rapid permeabilization due to damage or pore formation in their membranes.

Interaction of LfcinB with single E. coli-lipid-GUVs
To elucidate the interaction of LfcinB with the E. coli plasma membrane, we prepared GUVs of lipid bilayers with a lipid composition similar to the E. coli plasma membrane and examined the interaction of LfcinB with these GUVs using the single GUV method. For this purpose, we used an E. coli polar lipid extract (E. coli-lipid), whose lipid composition is phosphatidylethanolamine/phosphatidylglycerol/cardiolipin (67/23/10 (weight % ratio)). We first investigated the interaction of LfcinB with single E. coli-lipid-GUVs containing a water-soluble fluorescent probe, AF647. For the GUV experiments, buffer C (10 mM PIPES, pH 7.0, 1 mM EGTA, 50 mM NaCl) was used. Fig. 5A shows a typical experimental result of the interaction of 40 M LfcinB with single GUVs. The FI of the GUV lumen due to AF647 remained essentially constant during the addition of LfcinB solution for the first 124 s, then the FI rapidly decreased (Fig. 5, A(2) and B). At 125 s, the FI fell to effectively zero; a differential interference contrast (DIC) image of the same GUV (Fig. 5A(3)) showed that the spherical GUV structure remained, albeit with an apparent decrease in diameter. This is essentially the same phenomenon as that observed in the interaction of LfcinB with DOPG/DOPC (1/1)-GUVs (18). As discussed in our previous report (18), the rapid decrease in FI occurred as a result of the rapid leakage of the fluorescent probe through the LfcinB-induced local rupture of the GUV. Thus, the time at which the FI began to rapidly decrease corresponded red to the time at which local rupture occurred in the GUV. When the same experiments were carried out using 12 single GUVs, we observed the stochastic occurrence of a similar rapid leakage of AF647 from each GUV (Fig. 5C).
As demonstrated in our previous studies on peptide-induced damages in GUV membranes (18,25), the rate constant of LfcinB-induced pore formation or local rupture in the membrane can be obtained by analyzing the time course of the fraction of intact GUVs (i.e. those from which the fluorescent probe did not leak) among the examined GUVs, P intact (t), over time t. Fig. 5D shows that the value of P intact of E. coli-lipid-GUVs decreased with time during the interaction of 40 M LfcinB. If we assume a two-state transition from the intact state to the locally ruptured state of single GUVs, we can obtain the theoretical equation of the time course of P intact (t) as follows (18,25), where k P is the rate constant of the LfcinB-induced local rupture and t eq is a fitting parameter that denotes the time required for the binding equilibrium of LfcinB from the aqueous solution to the GUV membrane. The experimental data of the time course of P intact showed a good fit with the single exponential decay function defined by Equation 1. To determine the mean value for k P , two independent experiments (n ϭ 2) using 12-15 single GUVs to obtain the time course of P intact were carried out, and these curves also showed a good fit by Equation 1. The mean Ϯ S.D. of k P for 40 M LfcinB was (8 Ϯ 3) ϫ 10 Ϫ3 s Ϫ1 (n ϭ 2). Fig. 5E (OE) indicated that the k P increased with an increase in LfcinB concentration. Fig. 5E (red f) also revealed that the fraction of leaked GUV among all examined GUVs (P leak [ϭ1 Ϫ P intact ]) after 10 min (i.e. P leak (10 min)) increased with an increase in LfcinB concentration. At less than 5 M LfcinB, no leakage of AF647 was observed, indicating that local rupture did not occur in E. coli-lipid-GUVs at low LfcinB concentrations.

Measurement of the membrane potential of GUVs using a fluorescent probe
As described in the previous section, LfcinB induced rapid, stochastic leakage of AF647 from single E. coli-lipid-GUVs;

Membrane potential is vital for AMP-induced permeabilization
however, higher concentrations of LfcinB were required to induce a significant rate of local rupture in GUVs compared with E. coli cells and spheroplasts. On the other hand, the results shown in Figs. 2 and 4 indicate that the membrane potential across the plasma membrane of E. coli cells and spheroplasts increases the rate of LfcinB-induced pore formation or local rupture of the plasma membrane. Based on these results, we inferred that membrane potential may increase the rate of LfcinB-induced local rupture in E. colilipid-GUVs. To test this hypothesis, we examined the effect of membrane potential on LfcinB-induced leakage from single E. coli-lipid-GUVs.
To apply a membrane potential across a GUV membrane, we included a small amount of gramicidin A in the GUV membrane and generated a difference in K ϩ concentration between the lumen and the outside of the GUVs by diluting GUVs prepared in buffer K (containing 50 mM KCl) with buffer T (con-taining 50 mM tetraethylammonium chloride (TEAC)) at various ratios. It is well-known that gramicidin A is a small ion channel for monovalent cations (36,37), and tetraethylammonium ions (TEA ϩ ) cannot permeate through the gramicidin A channel nor block the channel (38). Therefore, only the ratio of the K ϩ concentration in the outside of a GUV, [K ϩ ] out , to that in the GUV lumen, [K ϩ ] in , determines the membrane potential, ⌬, according to the Nernst equation (i.e. Equation 2 under "Experimental procedures"). For example, when [K ϩ ] in ϭ 50 mM, ⌬ ϭ Ϫ59 mV for [K ϩ ] out ϭ 5.0 mM, and ⌬ ϭ Ϫ102 mV for [K ϩ ] out ϭ 0.94 mM at 25°C.
To detect a change in the membrane potential produced by various K ϩ gradients across GUV membranes, we examined the interaction of 1 nM of the membrane potential-sensitive dye 3,3Ј-dihexyloxacarbocyanine iodide (DiOC 6 (3)) with E. colilipid-GUVs at various membrane potentials using CLSM. Fig.  6A shows representative CLSM images of one of the E. coli-

Membrane potential is vital for AMP-induced permeabilization
lipid-GUVs under various membrane potentials. The FI of the GUV membrane (i.e. the rim intensity of the GUV) due to DiOC 6 (3) increased with an increase in negative membrane potential (or the absolute value of membrane potential) (Fig. 6,  A and B).
It is thought that DiOC 6 (3) can enter the cytoplasm of cells under negative potentials by translocating across the plasma membranes and hence the DiOC 6 (3) concentration in the cytoplasm increases (39 -42). In the case of GUVs, with an increase in negative membrane potential, the DiOC 6 (3) concentration in the GUV lumen becomes larger than that in the outside of the GUV. The binding equilibrium of DiOC 6 (3) between the GUV lumen and inner monolayer of the GUV holds, and thus, the DiOC 6 (3) concentration in the inner leaflet of the GUV increases with an increase in negative membrane potential. It can be reasonably inferred that the rim intensity due to DiOC 6 (3) is proportional to the DiOC 6 (3) concentration in the GUV membrane at a low concentration, although at higher DiOC 6 (3) concentrations the rim intensity would decrease due to quenching of DiOC 6 (3) (39). Therefore, Fig. 6 indicates qualitatively that the DiOC 6 (3) concentration in the GUV membrane increased with an increase in membrane potential (calculated by the Nernst equation), i.e. with an increase in the K ϩ concentration gradient. We can reasonably infer that the binding constant of DiOC 6 (3) from aqueous solution to lipid monolayers is large because DiOC 6 (3) has two hydrocarbon chains per molecule, and thus, most DiOC 6 (3) dyes exist in the membrane, not in aqueous solution of the GUV lumen. Therefore, the sensitivity of detection of membrane potential in the measurement of the rim intensity is greater than that in the measurement of FI of GUV lumen. The result shown in Fig. 6 also indicates that the GUVs prepared using our method showed negative membrane potentials and their absolute value increased with a decrease in K ϩ concentration in the outside of a GUV, [K ϩ ] out . Hence, we can use these GUVs to investigate the effect of membrane potential on the LfcinB-induced local rupture of GUVs.

Effect of membrane potential on the interaction of LfcinB with single E. coli-lipid-GUVs
We examined the effect of membrane potential (⌬) on the interaction of 5.0 M LfcinB with single E. coli-lipid-GUVs (Fig.  7). In the absence of membrane potential, 5.0 M LfcinB induced leakage of AF647 from single E. coli-lipid-GUVs with a low probability, i.e. the fraction of leaked GUV at 10 min was 0.09 (Fig. 5E). First, we investigated the effect of ⌬ ϭ Ϫ102 mV. After applying ⌬ ϭ Ϫ102 mV to the GUVs, the LfcinB solution was continuously provided to the vicinity of the GUV through a micropipette under confocal microscope. The FI of the GUV lumen due to AF647 remained essentially constant during the addition of LfcinB solution for the first 50 s, then the FI rapidly decreased (Fig. 7, A and B). At 51 s, the FI fell to effectively zero; a DIC image of the same GUV ( Fig. 7A(3)) showed that the spherical GUV structure remained, albeit with an apparent decrease in diameter. The time at which the FI began to rapidly decrease corresponded to the time at which local rupture occurred in the GUV. When the same experiments were carried out using 15 single GUVs, we observed the stochastic occurrence of a similar rapid leakage of AF647 from each GUV (Fig. 7C). Fig. 7D shows that the P intact of E. colilipid-GUVs decreased with time in the presence of 5.0 M LfcinB, which was fit with the single exponential decay function defined by Equation 1. To determine the mean value for k P , two independent experiments (n ϭ 2) using 12-15 single GUVs to obtain the time course of P intact were carried out and these data were also fit well with Equation 1. The mean value of k P for 5.0 M LfcinB at ⌬ ϭ Ϫ102 mV was (4.9 Ϯ 0.2) ϫ 10 Ϫ3 s Ϫ1 (n ϭ 2). We also examined the effect of various membrane potentials. For example, Fig. S2 shows the results at ⌬ ϭ Ϫ86 mV. Fig. 7E indicated that the k P increased with an increase in negative membrane potential (or the absolute value of membrane potential). Fig. 7F indicated that the fraction of leaked GUV at 10 min also increased with membrane potential. At ⌬ ϭ Ϫ102 mV, the fraction of leaked GUV at 10 min increased in an LfcinB concentration-dependent manner (Fig. 7G).

Discussion
The results in this study show that in the interaction of LfcinB with single E. coli cells and single E. coli spheroplasts, LfcinB induced rapid, complete leakage of calcein from the cytosol, indicating that LfcinB induced rapid permeabilization due to local damage or pore formation in the plasma membrane, through which calcein molecules rapidly leaked out. Therefore, we can conclude that the antimicrobial activity of LfcinB is mainly due to permeabilization in plasma membrane of E. coli cells induced by the direct interaction of LfcinB with the plasma membrane. In the interaction of LfcinB with single E. coli-lipid GUVs, LfcinB induced rapid, stochastic leakage of AF647 due to local rupture of the GUVs; however, higher LfcinB concentrations were required to induce a significant rate

Membrane potential is vital for AMP-induced permeabilization
of permeabilization compared with that observed in E. coli cells and spheroplasts. This result indicates other factors are necessary to induce LfcinB-induced rapid local rupture in E. coli-lipid-GUVs. We considered membrane potential as one of these factors. We found that the rate constant of LfcinB-induced local rupture in GUVs increased greatly with an increase

Membrane potential is vital for AMP-induced permeabilization
in negative membrane potential, and at ⌬ ϭ Ϫ102 mV, the rate of local rupture at the same LfcinB concentrations became similar in the GUVs and E. coli cells. In the case of the interaction of LfcinB on single spheroplasts and E. coli cells, the CCCP-induced dissipation of membrane potential greatly decreased the rate of permeabilization in the plasma membrane. To the best of our knowledge, this is the first report that indicates the significant role of membrane potential in the permeabilization activity of AMPs in GUVs of lipid bilayers and spheroplasts, whereas some researchers have suggested the role of membrane potential in the interaction of AMPs with bacterial cells (43,44).
Some previous results have suggested the effects of membrane potential on AMP activities. Yeaman and colleagues (43) found that the activity of platelet microbicidal protein-2 (PMP-2) to induce permeabilization against Staphylococcus aureus strain 6850 with ⌬ ϭ Ϫ150 mV was greater than that against another S. aureus strain (JB-1) with ⌬ ϭ Ϫ100 mV, indicating that ⌬ plays an important role in the antimicrobial activity of PMP-2 against S. aureus. Hancock and colleagues (44) found that in the interaction of AMPs with planar bilayers, high transmembrane voltage (usually Ϫ180 mV) was indispensable to induce conductance, and the increase in KCl concentration in the outside of E. coli cells (i.e. decrease in ⌬) did not alter the antimicrobial activity of several ␣-helical cationic peptides and a ␤-structured peptide (Gram 474) (judging from MIC values) but decreased the antimicrobial activity of gramicidin S, Gram 4112, and indolicidin (44), suggesting that ⌬ plays an important role in the antimicrobial activity of some AMPs.
Recently, it is well-recognized that membrane potential greatly affects the localization of many proteins in bacterial cells (34). Therefore, the results of the effects of the change in membrane potential on activities of AMPs in bacterial cells can be interpreted by various ways. For example, a change in localization of a specific protein in the cells may affect the activity of AMPs. However, the results in this study show that membrane potential greatly affects the activity of LfcinB in E. coli spheroplasts, indicating that the effect of membrane potential on the interaction of LfcinB with the plasma membrane (neither periplasm nor outer membrane) of bacterial cells plays an important role in the LfcinB-induced leakage. Moreover, the results in this study show that the negative membrane potential greatly increases the activity of LfcinB in E. coli-lipid-GUVs, indicating that the effect of membrane potential on the interaction of LfcinB with lipid bilayer regions of the plasma membrane of bacterial cells plays a vital role in the LfcinB-induced leakage. Therefore, to the best of our knowledge, this is the first report to indicate that membrane potential plays a vital role in the interaction of AMPs with lipid bilayer regions of the plasma membrane of bacterial cells to induce their bactericidal activity.
In this study, we observed the interaction of LfcinB with single cells, and hence, we detected variations of the LfcinB-induced permeabilization of the plasma membrane of each cell. For low concentrations of LfcinB (e.g. 3.0 M), the rates of LfcinB-induced leakage from septating cells were larger than those from most nonseptating cells and the rate of LfcinB-induced leakage of calcein from single nonseptating E. coli cells exhibited a wide distribution (Fig. 1E and Fig. S1). Generally, the rate of leakage depends on degree of the damage, such as pore formation induced by peptides/proteins in lipid membranes and plasma membrane, such as the size and the number of pores or local rupture (45,46). Therefore, the rate of permeabilization in plasma membrane depends on AMP concentration in the membrane and the degree of interaction of AMPs with the membrane. We can reasonably consider that each E. coli cell exists in a different step of the bacterial cell cycle (e.g. replication, Z-ring formation, end of replication, septation, division) (47,48). The degree of damage in the plasma membrane induced by LfcinB may greatly depend on the step of cell cycle of E. coli cells. In the cell cycle, only the septating cells are clearly recognized from their optical microscopic images. The result that the rates of LfcinB-induced leakage from septating cells were larger than those from most nonseptating cells is evident, but currently its mechanism is not clear. Further investigation is required. In contrast, for the higher concentration of LfcinB (5.0 M), the leakage started stochastically (i.e. the onset times of the leakage were random) but the rate of LfcinB-induced leakage of calcein was similar in all examined cells. This result suggests that the degree of the damage induced by LfcinB does not depend on the step of the cell cycle at a higher concentration of LfcinB.
At present, the mechanism of the effect of membrane potential on LfcinB-induced local rupture or pore formation in lipid bilayers of GUVs is not clear. We can expect that the negative membrane potential induces a large electric field inside the lipid bilayer, which could alter the interaction modes of LfcinB with the membrane, such as degree of insertion of peptides into the membrane interface, which may change the physical property of the membranes and their stretching. Further studies on the effect of membrane potential on the mechanisms of LfcinBinduced pore formation are indispensable.
In this study, we demonstrated that the electrochemical potential of proton or membrane potential plays a vital role in the activity of LfcinB (i.e. LfcinB-induced permeabilization) in bacterial cells and lipid vesicles by affecting the interaction of LfcinB with the membranes. So far almost all studies on the interactions of AMPs with lipid bilayers and the structures of AMPs in lipid bilayers have been performed in the absence of membrane potential. Therefore, it is indispensable to reconsider and reinterpret all the results obtained in these studies. Furthermore, we have to reveal the mechanisms of activities of AMPs by including the effects of membrane potential. Currently it is recognized that there are several modes of actions of AMPs (54). Thus, it is necessary to examine the effects of membrane potential on each mode of action of AMPs. To design new peptides that have an antimicrobial activity, it is indispensable to consider the effect of membrane potential on the interaction of the peptides with lipid bilayers.

Conclusion
The results of the effect of LfcinB on single E. coli cells and single spheroplasts indicate that the direct interaction of LfcinB with plasma membranes induced rapid permeabilization. We can reasonably consider that this permeabilization induced growth cessation of the cells, based on the results of changes in the cell length during LfcinB interaction with single cells. At

Membrane potential is vital for AMP-induced permeabilization
present it is not clear that this rapid permeabilization occurs due to large pore formation or local rupture in the plasma membrane. The presence of CCCP (hence the decrease in the membrane potential) suppressed this leakage. On the other hand, LfcinB induced local rupture in single E. coli-lipid-GUVs, through which AF647 leaked rapidly from the GUV lumen. The rate of this LfcinB-induced local rupture in single GUVs increased with an increase in negative membrane potential of the GUVs. These results indicate that the membrane potential plays an important role in LfcinB-induced local rupture of lipid bilayers and in the interaction of LfcinB with lipid bilayer regions of the plasma membrane of bacterial cells to induce rapid permeabilization, resulting in their bactericidal activity. The results in this study are indispensable in revealing the mechanism of action of LfcinB and also other AMPs against bacterial cells.

Peptide synthesis and identification
LfcinB with an amide-blocked C terminus was synthesized by the FastMoc method using a 433A peptide synthesizer (PE Applied Biosystems, Foster City, CA) (18). The methods of cleavage, peptide purification, and measurement of mass have been described previously (18). LfcinB concentration in buffer was determined by absorbance using the molar extinction coefficient of Trp at 278 nm (i.e. 5,500 M Ϫ1 cm Ϫ1 ).

CLSM investigation of the interaction of LfcinB with single E. coli cells containing calcein
An E. coli (JM-109) suspension was subcultured on nutrient agar plates at 37°C overnight to obtain single colonies, and a single bacterial colony was then grown in Nutrient Broth medium for 6 to 7 h (exponential phase cells) (18). The bacterial suspension was diluted to obtain a final bacterial concentration of 1 ϫ 10 6 CFU/ml, which was pelleted by centrifugation (620 ϫ g, 10 min); the pellet was washed and resuspended in EZ rich medium (containing 50 mM NaCl and 1.3 mM K 2 HPO 4 as main ion sources (28)). Calcein loading of E. coli was performed according to the method of Dubey and Ben-Yehuda (29). In brief, a 50-l aliquot of calcein-AM (1.0 g/l in DMSO solution) was added into 1.0 ml of the bacteria suspension, which was shaken using a rotary shaker for 2 h at room temperature under darkness. Then, the suspension was centrifuged at 620 ϫ g for 10 min, and the pellet was resuspended in fresh EZ rich medium. This procedure was repeated twice. The E. coli suspension was transferred into a handmade chamber formed on a glass slide by putting a U-shaped silicone-rubber spacer between the glass slide and a coverslip (18), and E. coli cells had settled onto the coverslip surface after 15 min and were adsorbed thereafter. To fix live E. coli cells to the glass surface of coverslips, we generated poly-L-lysine coatings on coverslips using the standard method.
Interactions of LfcinB with single E. coli cells containing calcein were observed under a confocal laser scanning microscope (FV-1000, Olympus, Tokyo, Japan) at 25 Ϯ 1°C with a stage thermocontrol system (Thermoplate, Tokai Hit, Shizuoka, Japan), in reference to a previously reported method (30). For CLSM measurements, fluorescence images of calcein (473 nm laser) and DIC images were obtained using a ϫ60 objective (UPLSAPO060X0, Olympus) (NA ϭ 1.35) (25,30). During interaction of the peptide with single E. coli cells, various concentrations of LfcinB in the medium were added continuously to the vicinity of the single E. coli cells through a 20-m diameter glass micropipette positioned by a micromanipulator (18,31). The distance between the single cells and the tip of the micropipette was 50 m, and the applied pressure, ⌬P (ϭ P in Ϫ P out , where P in and P out are the pressure of the inside and outside of a micropipette, respectively) was 30 pascals. ⌬P was measured using a differential pressure transducer (DP15, Validyne, Northridge, CA), pressure amplifier (PA501, Validyne), and a digital multimeter. Glass micropipettes were prepared by pulling 1.0-mm glass capillaries composed of borosilicate glass (G-1, Narishige, Tokyo, Japan) using a puller (PC-10, Narishige). Details of these methods have been described previously (31,49).
For the experiments of the interaction of LfcinB with single E. coli cells in the presence of CCCP in the EZ medium, we mixed CCCP solution in EZ media containing DMSO with the suspension of E. coli cells containing calcein (final concentrations were 100 M CCCP, 0.5% DMSO, and 1 ϫ 10 6 CFU/ml E. coli cells), and then transferred this mixture into a handmade chamber. After more than 10 min incubation, we investigated the interaction of LfcinB with single E. coli cells.

CLSM investigation of the interaction of LfcinB with single spheroplasts containing calcein
Spheroplasts of E. coli cells were prepared using a similar method to Martinac et al. (21). Details of the method are described in the supporting data. In brief, first, single-cell filaments were prepared by growing E. coli cells in the presence of cephalexin, which prevents cell division (50 -52). For hydrolysis of the peptidoglycan layer, the single-cell filaments were reacted with lysozyme, DNase I, and EDTA. MgCl 2 was added to stop the reaction. Finally, MgCl 2 was replaced with KCl and NaCl because Mg 2ϩ affected the interaction of LfcinB with spheroplasts. The final buffer of the purified spheroplasts was 10 mM Tris-HCl (pH 7.8) containing 1.5 mM KCl, 48.5 mM NaCl, and 0.73 M sucrose. We used these spheroplasts within 3 h after the purification, without storing in the freezer.
For calcein loading to spheroplasts, we used a similar method of loading calcein in E. coli cells (29). A 50-l aliquot of calcein-AM (1.0 g/l in DMSO solution) was added to 1.0 ml of Membrane potential is vital for AMP-induced permeabilization a spheroplast suspension (final DMSO concentration was 1%), which was incubated for 30 min at room temperature under darkness. Then, the suspension was transferred into a microchamber and used in the interaction of LfcinB under a confocal laser scanning microscope. The interaction of LfcinB with single spheroplasts containing calcein was investigated at 25 Ϯ 1°C using a method similar to that used for single E. coli cells, as described in the above section.
For the experiments of the interaction of LfcinB with single spheroplasts in the presence of CCCP, we mixed a CCCP solution in the same buffer containing DMSO with the suspension of spheroplasts containing calcein (final concentrations are 100 M CCCP and 0.5% DMSO), and then transferred this mixture into a handmade chamber. After more than 10 min incubation, we investigated the interaction of LfcinB with single spheroplasts.

GUV preparation
GUVs of E. coli polar lipid extract (E. coli-lipid-GUVs) were prepared by incubating buffer C (10 mM PIPES, pH 7.0, 1 mM EGTA, 50 mM NaCl) containing 0.10 M sucrose and 6.0 M AF647 with dry lipid films by the natural swelling method at 37°C for 2-3 h (18). The membrane filtering method was used to remove untrapped fluorescent probes (53).
Purified GUV suspension (300 l: 0.1 M sucrose in buffer C as the internal solution; 0.1 M glucose in buffer C as the external solution) was transferred into a handmade microchamber (18). To prevent strong interaction between the glass surface and GUVs, the inside of the microchamber was coated with 0.10% (w/v) BSA in buffer C containing 0.10 M glucose (18).

CLSM investigation of the interaction of LfcinB with single E. coli-lipid-GUVs
GUVs were observed using a confocal laser scanning microscope (FV-1000, Olympus) at 25 Ϯ 1°C using a stage thermocontrol system (Thermoplate, Tokai Hit). For CLSM measurements, fluorescence images of AF647 (excited by a laser at ϭ 635 nm) and DIC images were obtained using the ϫ60 objective (19,30,39). To investigate the effect of LfcinB on single GUVs, various concentrations of LfcinB in buffer C containing 0.10 M glucose were added continuously to the vicinity of the GUV through a 20-m diameter glass micropipette positioned using a micromanipulator. The distance between the single cells and the tip of the micropipette was 50 m, and the applied pressure, ⌬P, was 30 pascals (31). Consequently, the LfcinB concentration near the GUV approached steady state and was almost the same as the concentration inside the micropipette (31).

Application of membrane potential to single GUVs
To induce a membrane potential in GUVs, we created a K ϩ concentration gradient across the GUV membrane. To induce permeability of K ϩ in the GUV membrane, gramicidin A molecules were incorporated in the GUV membrane. For this purpose, we prepared GUVs of E. coli-lipid/gramicidin A (molar ratio: 100/0.01) from a dry lipid film in buffer K (10 mM PIPES, pH 7.0, 50 mM KCl, and 1 mM EGTA) containing 0.10 M sucrose and 6.0 M AF647 by the natural swelling method, and purified them by the membrane filtering method. Then, we diluted the purified GUV suspension with buffer T (10 mM PIPES, pH 7.0, 50 mM TEAC, and 1 mM EGTA) containing 0.1 M glucose at various ratios. The total concentration of KCl and TEAC in the buffer outside the GUVs was 50 mM, which had the same osmotic pressure as the buffer inside the GUVs. The membrane potential difference across a GUV membrane, ⌬, or the membrane potential inside a GUV, in , was measured using the following Nernst equation, where F is the Faraday constant, R is the gas constant, T is the absolute temperature, [K ϩ ] out and [K ϩ ] in are the K ϩ concentrations outside and inside the GUV, respectively. The numerical value (25.7) in the last equation in Equation 2 is for 25°C. For example, when we diluted the GUV suspension with buffer T at a GUV suspension to buffer T ratio of 1:9, ⌬ ϭ Ϫ59 mV at 25°C. To investigate the interaction of LfcinB with single GUVs with various membrane potentials, we used the same method described in the above section. We used the same buffer mixture to prepare LfcinB solution as for the outside of the GUVs. We waited 10 min after application of the K ϩ gradient to attain an equilibrium, and the GUVs were observed using CLSM. We also measured the membrane potential using a fluorescent probe, DiOC 6 (3), which is sensitive to the membrane potential (39 -42). First, we mixed the membrane potential probe DiOC 6 (3) in methanol solution with the buffer, which was the same as for the outside of the GUVs, and then mixed this DiOC 6 (3) solution in the buffer with a suspension of GUVs with a membrane potential that had been produced due to a K ϩ concentration gradient across the GUV membrane. The final concentrations were 1 nM DiOC 6 (3) and 0.006% methanol. Then, the GUV suspension was transferred into a handmade microchamber. To prevent strong interaction between the glass surface and the GUVs, the inside of the microchamber was coated with 0.10% (w/v) BSA in the same buffer as used for the experiments. We waited 10 min after the application of the K ϩ gradient to attain an equilibrium, and the GUVs were observed using a confocal laser scanning microscope (FV-1000) at 25 Ϯ 1°C using the stage thermocontrol system (Thermoplate). Fluorescence images of DiOC 6 (3) (excited by a laser at ϭ 473 nm) and DIC images were obtained using the ϫ60 objective. The fluorescence intensities of the images obtained by CLSM were measured using the software supplied with the CLSM (Fluoview, version 4.1, Olympus). The fluorescence intensity of the rim of a GUV due to the binding of DiOC 6 (3) dye was measured as the peak intensity in a line crossing the GUV (19).
We investigated the interaction of LfcinB with single E. colilipid-GUVs with various membrane potentials using the same method described in the above section (CLSM investigation of the interaction of LfcinB with single E. coli-lipid-GUVs). Various concentrations of LfcinB in the same buffer as for the outside of the GUVs (i.e. mixture of buffer T and buffer K with various ratios) containing 0.10 M glucose were added continuously to the vicinity of the GUV through a 20-m diameter glass micropipette positioned using a micromanipulator.