The Bacillus subtilis cannibalism toxin SDP collapses the proton motive force and induces autolysis

Summary

Bacillus subtilis SDP is a peptide toxin that kills cells outside the biofilm to support continued growth. We show that purified SDP acts like endogenously produced SDP; it delays sporulation, and the SdpI immunity protein confers SDP resistance. SDP kills a variety of Gram-positive bacteria in the phylum Firmicutes, as well as E. coli with a compromised outer membrane, suggesting it participates in defense of the B. subtilis biofilm against Gram-positive bacteria as well as cannibalism. Fluorescence microscopy reveals that the effect of SDP on cells differs from that of nisin, nigericin, valinomycin and vancomycin-KCl, but resembles that of CCCP, DNP and azide. Indeed, SDP rapidly collapses the PMF as measured by fluorometry and flow cytometry, which triggers the slower process of autolysis. This secondary consequence of SDP treatment is not required for cell death since the autolysin-defective lytC, lytD, lytE, lytF strain fails to be lysed but is nevertheless killed by SDP. Collapsing the PMF is an ideal mechanism for a toxin involved in cannibalism and biofilm defense, since this would incapacitate neighboring cells by inhibiting motility and secretion of proteins and toxins. It would also induce autolysis in many Gram-positive species, thereby releasing nutrients that promote biofilm growth.

Introduction

Bacterial species must compete with a multitude of other species in the environment for both space and nutrients. Due to the diversity and density of microbes present in the environment, competition is constant as each species attempts to find and maintain a niche. In the environment, many bacteria exist primarily in biofilms, communities of cells held together by exopolysaccharides, proteins and other substances (Hibbing et al., 2010, Abee et al., 2011). Biofilms enable bacteria to claim territories by excluding competitors, by adhering to surfaces and they confer increased resistance to a variety of environmental stresses, including antibacterial compounds. Differentiation of cells within the biofilm can allow a great variety of niches to be simultaneously occupied, by one species or by a mixed population (Hibbing et al., 2010, Haussler, 2010, Lopez et al., 2010). Thus, biofilms provide a ubiquitous means for bacteria to compete and cooperate with other species, and to form spatially organized and functionally differentiated communities to maximally exploit the environment.

A commonly implemented competition mechanism is the production of secreted metabolites that have diverse effects on intra- and interspecies interactions. This includes modulating development, promoting biofilm dispersal, killing competitors, or inhibiting toxin production in a competitor (Straight & Kolter, 2009, Hibbing et al., 2010, Yang et al., 2009, Gonzalez et al., 2011). These secreted metabolites can also stimulate growth by facilitating acquisition of iron and other metals or by facilitating extracellular electron transfer to insoluble electron acceptors for respiratory growth (D’Onofrio et al., 2010, Bird et al., 2011, Skaar, 2010). Bacillus subtilis secretes many secondary metabolites, including subtilosin, surfactin, plipistatin and bacillaene (Shelburne et al., 2007, Babasaki et al., 1985, Stein, 2005, Asaduzzaman & Sonomoto, 2009), some of which play dual functions as antimicrobial compounds and signaling molecules. For example, the lipopeptides surfactin and plipistatin have antimicrobial properties (Carrillo et al., 2003, Vanittanakom et al., 1986, Gonzalez et al., 2011) and together repress the production of toxin molecules by Staphylococcus aureus (Gonzalez et al., 2011). Surfactin also mediates various behaviors, such as B. subtilis swarming motility (Kearns & Losick, 2003) and biofilm formation (Lopez et al., 2009a), and the inhibition of aerial hyphae formation in Streptomyces coelicolor (Straight et al., 2006, Yang et al., 2009). Similarly, the polyketide bacillaene both inhibits bacterial translation (Patel et al., 1995) and inhibits production of prodiginine and several other molecules in S. coelicolor (Yang et al., 2009). Thus, secreted metabolites are critical for mediating the outcome of interspecies interactions due to both their antibacterial effects and their signaling roles.

The B. subtilis biofilm also secretes antimicrobial products that mediate cannibalism by killing non-biofilm forming siblings (Gonzalez-Pastor et al., 2003). These cannibal cells lyse genetically identical siblings and susceptible neighbors of other species (Nandy et al., 2007, Liu et al., 2010) to release nutrients that allow continued growth (Gonzalez-Pastor et al., 2003, Lopez et al., 2009b). This delays the initiation of sporulation, a starvation-induced developmental process that produces dormant spores. Sporulation is a lengthy committed pathway (Dworkin & Losick, 2005, Parker et al., 1996), so if nutrients become available, sporulating cells are at a competitive disadvantage as they cannot resume growth until sporulation is complete. Therefore, using cannibalistic toxins to lyse susceptible neighbors would provide an advantageous mechanism to delay sporulation until absolutely necessary (Gonzalez-Pastor et al., 2003). Cannibalism might also serve to eliminate cheater cells not directly contributing to biofilm development (Smukalla et al., 2008, Mitri et al., 2011) and to eliminate differentiated cell types that are no longer beneficial to the population, thereby allowing the biofilm to use the nutrients from these cells to produce new beneficial cell types. Thus, secreted metabolites may be used to sculpt a differentiated bacterial biofilm by removing and replacing unnecessary cell types, thereby allowing the biofilm to adapt even during starvation.

Critical to understanding the role of secreted metabolites in microbial populations is elucidating the mechanism by which these molecules affect bacterial cells. Secreted metabolites that inhibit bacterial growth do so by inhibiting a variety of essential cellular pathways, from transcription to cell wall biosynthesis. Furthermore, at the sublethal concentrations that one might expect to predominate early in an interaction, different compounds have distinct effects on gene expression and behavior (Romero et al., 2011, Rogers et al., 2007, Nalca et al., 2006). Teasing apart the role of these molecules in microbial population dynamics will therefore require understanding the effects they have on cells at both lethal and sublethal concentrations.

We here describe studies on the mechanism of action of the SDP cannibalistic toxin, which we previously demonstrated to cause lysis of B. subtilis, S. aureus and S. epidermidis (Liu et al., 2010). The mature SDP toxin is a 42 amino acid peptide with a single disulfide bond (Liu et al., 2010) that is produced by processing and secretion of the product of the sdpC gene. We here demonstrate that purified SDP acts in a manner consistent with that of endogenously produced SDP, as it delays sporulation (Gonzalez-Pastor et al., 2003) and resistance to purified SDP is mediated by SdpI (Ellermeier et al., 2006). We further demonstrate that SDP rapidly collapses the proton motive force (PMF) of B. subtilis, as well as the lptD4213 E. coli mutant that has a compromised outer membrane. This subsequently induces the dramatic autolysin mediated lysis that is a secondary consequence of SDP treatment in autolysis susceptible Gram-positive species, such as B. subtilis, S. aureus and S. epidermidis (Tipper, 1969, Sieradzki et al., 1998, Jolliffe et al., 1981). Collapsing the PMF is an ideal mechanism of action for a cannibalistic and defensive toxin, as it would rapidly eliminate the ability of neighboring species and non-biofilm producing B. subtilis cells (which do not produce SdpI) to respond by moving away, while autolysis would release nutrients that can be readily used to promote biofilm growth.

Results

Purified and endogenously produced SDP have the same biological effects

Prior to investigating the mechanism by which purified SDP kills cells, we first verified that purified SDP affected Bacillus subtilis cells by a mechanism relevant to in vivo studies that used endogenously produced or inducible SDP (Gonzalez-Pastor et al., 2003, Ellermeier et al., 2006, Butcher & Helmann, 2006). SDP was originally identified as delaying the onset of sporulation (Gonzalez-Pastor et al., 2003), so we first tested if treatment with purified SDP delayed sporulation. To do so, we used the undomesticated strain 3610 because PY79 did not sporulate in the small scale culture conditions used to document the cell biological effects of purified SDP (Liu et al., 2010). Briefly, in this method, cells are grown in LB to mid-log phase, then concentrated tenfold in the same medium and 15 μl aliquots transferred to microfuge tubes for further incubation with or without SDP. Under these conditions, 3610 showed significant levels of sporulation by 5 hours (Fig. 1A-B), when asymmetrically positioned sporulation septa at various stages of engulfment were visible in 36% of DMSO-treated cells. SDP treatment reduced the frequency of sporulation to 1%. Thus, SDP treatment dramatically reduces the frequency of sporulation, in keeping with previous publications (Gonzalez-Pastor et al., 2003, Lopez et al., 2009b).

Figure 1. Purified SDP acts in a manner similar to endogenously produced SDP.

(A-B) Purified SDP delays sporulation. Strain 3610 was grown for 5 hr after treatment with DMSO (A) or 20 μg ml^-1 SDP (B), which is 2-5X above the MIC (Supplemental Data). Membranes were stained with FM 4-64 (red) before fluorescence microscopy. Arrows indicate cells with sporulation septa at various stages of engulfment. Scale bar represents 1 μm. (C-F) SdpI is sufficient for protection against purified SDP. P_hyspacsdpIΔsdpABCIRΔsigW (TPM758) cells were treated for 3 hr with 0.5% DMSO (C,E) or 20 μg ml^-1 SDP (D,F) either without (C,D) or with (E,F) induction of sdpI with 1 mM IPTG. Cells were stained with FM 4-64 (red), and two DNA stains, DAPI (blue) and SYTOX Green (green). SYTOX Green is membrane impermeable and only stains permeabilized cells. (G) Effect of SdpI production on viability of P_hyspacsdpI ΔsdpABCIR ΔsigW (TPM758) cells treated with SDP (n=2). Viability is shown as the ratio of colony forming units (CFU) at the indicated time and treatment to the CFU at t₀ for the DMSO control (CFU^C).

Next, we tested if previously identified SDP-resistance mechanisms also provided resistance against purified SDP. The primary SDP resistance mechanism is SdpI, a membrane protein that is induced in response to SDP and conveys resistance to high levels of SDP (Ellermeier et al., 2006). A secondary resistance mechanism, which functions in the absence of SdpI, is mediated through the extracytoplasmic function (ECF) sigma factor σ^W ((Butcher & Helmann, 2006) and Fig. S1), and this requires two potential SDP resistance systems, yknW-Z, an ABC transporter, and yhfL, an SdpI paralogue (Butcher & Helmann, 2006). Thus, if purified SDP kills cells by the same mechanism as endogenously produced SDP, overexpression of SdpI should be sufficient for high level SDP resistance even in the absence of the backup systems induced by σ^W. We used fluorescence microscopy to test this prediction. To do so, we employed a ΔsdpABCIR ΔsigW strain lacking the normal resistance mechanisms that also had sdpI under the control of the IPTG inducible promoter P_hyspac (TPM758). In the absence of IPTG, and hence SdpI, SDP treatment caused significant lysis by three hours (Fig. 1D), and viable cell counts showed that viability dropped nearly 5 logs in one hour (Fig. 1G, red). However, when sdpI was induced with IPTG, we failed to observe lysis (Fig. 1F) or any loss of viability (Fig. 1G, purple). These results demonstrate that SdpI is sufficient for protection against purified SDP, as expected based on previous results with endogenously produced SDP (Ellermeier et al., 2006, Butcher & Helmann, 2006).

SDP affects both Gram-positive and Gram-negative bacteria

We next explored the ability of SDP to kill additional Gram-positive bacteria, to extend previous studies showing that it kills Listeria monocytogenes (Palmer et al., 2009) Staphylococcus aureus and S. epidermidis (Liu et al., 2010). We chose five species from the phylum Firmicutes and one from the phylum Actinobacteria. Of the species tested, only the Actinobacterium Micrococcus luteus was resistant to SDP (Fig. 2A, red) at the concentration tested. All of the Firmicutes species tested were sensitive to SDP, with B. megaterium showing the most sensitivity, as it was no longer viable at the initial timepoint (Fig. 2A, purple). S. epidermidis was the least sensitive to SDP, but still showed a 2.5 log drop in viability by 5 hours (Fig. 2A, green). This experiment demonstrates that SDP is active against a wide variety of Gram-positive bacteria within the phylum Firmicutes.

Figure 2. Effects of SDP on Gram-positive species and E. coli.

(A) Effect of 20 μg ml^-1 SDP on viability of a variety of Gram-positive species (n=2), calculated as in Fig. 1G. We were unable determine if M. luteus was sensitive to higher concentrations of SDP due to our limited supply. (B) Fluorescence micrograph of E. coli wt (MC4100) treated with 0.5% DMSO or 20 μg ml^-1 SDP for 5 hrs and stained with FM 4-64 (red) and DAPI (blue). Scale bar represents 1 μm. (C) Fluorescence micrograph of lptD4213 (NR698) cells treated with 0.5% DMSO or 20 μg ml^-1 SDP for 5 hr. SDP treatment produces cells that are hyper-permeabilized to DAPI and that fail to enter stationary phase and therefore remain elongated. (D) Effect of SDP on E. coli wt (MC4100) and lptD4213 (NR698, lptD4213) viability, calculated as in Fig. 1G. Error bars show the standard error of ≥3 experiments.

Our previous studies failed to show any effect of SDP on the Gram-negative species tested (Liu et al., 2010), suggesting that either the outer membrane of Gram-negative bacteria provides a barrier to entry of SDP or that SDP has a Gram-positive specific target. To test these hypotheses, we took advantage of an E. coli strain that has a defect in the lptD gene that encodes a protein required for outer membrane biogenesis. This mutation permeabilizes the outer membrane and renders E. coli sensitive to drugs such as vancomycin and nisin that normally target only Gram-positive bacteria (Braun & Silhavy, 2002, Wu et al., 2006, Sampson et al., 1989). Fluorescence microscopy and viable cell counts demonstrated that wild type E. coli (MC4100) is unaffected by SDP treatment, and that by 3 hours of treatment the cells had the small cell morphology characteristic of stationary phase cells (Fig. 2B). However, upon SDP treatment, the lptD (NR698) cells stopped dividing, became slightly wider and longer than untreated cells and were hyper-permeabilized to DAPI (Fig. 2C). Cell viability also dropped almost 4 orders of magnitude (Fig. 2D, purple). Therefore, SDP recognizes a target that is broadly distributed in bacteria, but is unable to penetrate the outer membrane of Gram-negative species.

Comparison of the cytological effects of SDP and other membrane active compounds

SDP induces dramatic lysis of B. subtilis cells that is characterized by the extrusion of membrane vesicles and tubules after 1-2 hours of treatment (Liu et al., 2010). We used fluorescence microscopy to investigate this characteristic lysis phenotype by comparing the cell biological effects of SDP to antibiotics with known mechanisms of action. We focused on antibacterial compounds that target the cell wall, the cytoplasmic membrane and cellular energy production (vancomycin, nisin, CCCP, DNP, azide; Fig. 3A-H). Three compounds, the energy poisons CCCP, DNP and sodium azide, showed SDP-like phenotypes and a similar rate and extent of viability loss. These compounds killed cells within 1-2 hours (Fig. 3K), although SDP treated cells showed increasing numbers of viable cells over time. This is likely because at the onset of treatment, a few cells in the culture express SdpI, and this SDP resistant population survives and resumes growth. Fluorescence microscopy revealed that after 2 hours of treatment with SDP, CCCP, DNP and sodium azide, some cells showed large gaps in the cytoplasmic membrane, extracellular membrane vesicles of variable size, permeability to SYTOX Green, increased permeability to DAPI and diffuse chromosomal DNA (Fig. 3B-E).

Figure 3. Effects of antibacterial compounds on B. subtilis cell architecture.

(A-J) Fluorescence micrographs of growing cells of PY79 treated with (A) 0.5% DMSO (B) 20 μg ml^-1 SDP (C) 10 mM sodium azide (D) 100 μM CCCP (E) 2 mM DNP (G) 250 μg ml^-1 vancomycin (H) 10 μg ml^-1 nisin (I) 62.5 μM nigericin (J) 125 μM valinomycin with 200 mM KCl (F) incubated statically, for 2 hr (A-F, I-J), 1 hr (H) or 3 hr (G). Cells are stained with FM 4-64 (red), DAPI (blue), and SYTOX Green (green), as in Fig. 1. Arrowheads indicate small membrane vesicles and debris. Arrows indicate large vesicles. Scale bar represents 1 μm. (K) Effects of treatments on PY79 viability, calculated as in Fig. 2. Error bars show the standard error of ≥3 experiments. (L) Cartoon demonstrating the mechanism of PMF component collapse for CCCP, DNP, nigericin, and valinomycin + KCl (M) Cartoon showing representations of the cytological profiles (A-E, G-H, J).

The channel forming toxin nisin also caused cell lysis, as evidenced by cells with gaps in the cytoplasmic membrane and increased permeability. However, unlike the other compounds nisin uniformly hyper-permeabilized the cells to DAPI and SYTOX Green (Fig. 3H), likely because of the large, non-specific channels it forms in the cytoplasmic membrane ((Lubelski et al., 2008); these changes in permeability are discussed and quantified below). The membranes of nisin-treated cells showed highly uneven staining and no external vesicles were observed (Fig. 3H). The cell wall-active drug vancomycin caused the cells to accumulate extra membrane material in a few locations along the lateral cell wall (Fig. 3G), perhaps because vancomycin titrates peptidoglycan precursors and slows peptidoglycan synthesis without slowing membrane biogenesis (Molenkamp & Veerkamp, 1976). Thus, compounds with different mechanisms of action have different effects on bacterial cell architecture. SDP produces a cytological profile similar to CCCP, DNP and azide suggesting that it depletes cellular energy stores (the PMF and/or ATP).

SDP rapidly collapses the PMF

Many antimicrobial peptides collapse the PMF by forming channels in the cytoplasmic membrane (Wilmes et al., 2011, Nan et al., 2011b, Peters et al., 2010). We therefore hypothesized that this was more likely to be the primary mechanism of action for SDP. Prior to determining if SDP collapses the PMF, we tested how rapidly it affected cell viability and growth. Short term viable cell counts showed that SDP treated cells showed a several log drop in viability by 10 minutes (Fig. 4D), although lysis was not observed until >60 minutes. Timelapse fluorescence microscopy demonstrated that cells treated with SDP for 20 minutes failed to divide, with lysis starting ~60 minutes after treatment (Fig. S2). Thus, SDP is fast acting and its effects are irreversible within 10 minutes (Fig. 4D).

Figure 4. SDP rapidly depletes the PMF and slowly induces autolysis.

(A-C) Flow cytometry assay of DiBAC₄(5) stained E. coli lptD4213 (NR698) cells. (A-B) Cells were treated for 10 min (light) or 20 min (dark) with DMSO (green), 4 μg ml^-1 SDP (red) or 500 μM CCCP (blue) and subjected to flow cytometry. (C) Cells were treated for 20 min or 60 min, as indicated, with DMSO (dark green), Water (green), 200 mM Azide (crimson), 100 mM Azide (pink), 125 μM Nigericin (aqua), or 20 mM DNP (orange) and subjected to flow cytometry. (D) Treatment of PY79 with 20 μg ml^-1 SDP reduces B. subtilis PY79 viability (calculated as in Fig. 2) within 10 minutes; the limited quantities of SDP available precluded doing additional shorter timepoints. Error bars represent the standard error of ≥3 experiments (E) Fluorometry assay of DiSC₃(5) stained B. subtilis PY79 cells treated with DMSO (purple), 10 μg ml^-1 vancomycin (red), 1 μg ml^-1 SDP (green) or 11 μg ml^-1 nisin (blue).

We used two complimentary assays to determine if SDP collapses the PMF, fluorometry and flow cytometry. First, to measure the effect of SDP on B. subtilis, we used a fluorometry assay that allows the effects of compounds on the PMF to be followed over time, starting as soon as it is possible to load the sample after treatment, about 30 seconds with the small sample volume used here (which precluded adding SDP during the experiment). This assay measures the average effects on all the cells in the population. This assay uses DiSC₃(5), a dye that enters polarized cells, is quenched, and is released when the cells depolarize, leading to increased fluorescence (Strahl & Hamoen, 2010, Sims et al., 1974). SDP caused depolarization in B. subtilis strain PY79 at concentrations as low as 1 μg ml^-1 with a timescale similar to that of nisin (Fig. 4E, green and blue). In contrast, 10 μg ml^-1 vancomycin had no impact on the PMF. Thus, SDP rapidly collapses the PMF in B. subtilis.

Second, we used flow cytometry to quantify the fraction of cells affected by various treatments. This assay cannot be performed with B. subtilis because it grows in chains that preclude analysis by flow cytometry without either gating the cell size to analyze only individual cells, which comprise <5% of the population, or by fixing and physically separating the cells (Aguilar et al., 2010), which makes it impossible to measure the PMF. We therefore used the E. coli lptD strain (NR698;(Ruiz et al., 2005)), which grows as single cells. Cells were treated with SDP and a variety of control compounds and then resuspended in 1X PBS with 5 μg ml^-1 DiBAC₄(5), a potential sensitive dye that only enters depolarized cells and then binds to membranes and proteins, resulting in increased fluorescence (Jepras et al., 1997, Jepras et al., 1995). Cells treated with SDP at concentrations as low as 4 μg ml^-1 show increased fluorescence within 10 minutes of treatment. (Fig. 4A, red; see SDP titration experiment in Fig. S3A). This shift is comparable to that seen with CCCP (Fig. 4B, blue), DNP (Fig. 4C, orange), nigericin (Fig. 4C, blue) and nisin treatment (Fig. S3B). In contrast, azide only depolarized the membranes in a fraction of the cells at the highest concentration after an hour (Fig. 4C, crimson). Thus, SDP rapidly collapses the PMF of both B. subtilis and E. coli, suggesting that this is its primary mechanism of action.

SDP treatment most closely resembles treatments that collapse both ΔΨ (membrane potential) and ΔpH

The data above indicate that SDP collapses the PMF, but they do not discriminate between decreases in the individual components of the PMF, ΔΨ (membrane potential) and ΔpH. We therefore expanded our list of reference compounds to include those that specifically collapse the ΔΨ or the ΔpH (unlike CCCP, DNP and azide, which collapse both components). To discriminate between these two activities, we tested the effects of nigericin and valinomycin on B. subtilis cells. Nigericin inserts into the membrane and facilitates the equal exchange of extracellular H⁺ for intracellular K⁺ ions, dissipating ΔpH while maintaining ΔΨ since this is a charge neutral exchange. Valinomycin moves K⁺ ions from high to low concentrations, so in the presence of high extracellular concentrations of K⁺ it collapses ΔΨ with no effect on ΔpH, as there is no change in the proton gradient. Somewhat surprisingly, the cell biological effects of these two compounds were easily distinguishable from each other and from SDP, CCCP, DNP and azide. Nigericin induces cell lysis and accumulation of membrane at septa, like those that collapse both components of the PMF, however, the DNA in lysed cells is more compacted, extracellular vesicles are rarely seen, and there are subtle accumulations of membrane pools along the side of the cell (Fig. 3I, with entire fields shown in Supplemental Figure 4). Valinomycin-KCl treatment produces elongated cells with smooth membranes, decondensed chromosomes, and cell lysis produces many extracellular vesicles but small membrane gaps are not observed. Thus, somewhat surprisingly, compounds that specifically deplete the ΔΨ or ΔpH appear different from one another, and different from compounds that collapse both components of the PMF. Our data suggest that SDP might collapse both the ΔΨ and the ΔpH, similar to CCCP and DNP.

Quantitative discrimination between mechanisms of action using cell biology

Our cell biological studies demonstrated that compounds with different mechanisms of action had different effects on membrane and DNA architecture and on cell permeability, even after short times of treatment (20 minutes). Specifically, all of the compounds that target components of the PMF produce cells with decondensed chromosomes and reduced DAPI staining (Fig. 5B,E,F,H,L) compared to untreated cells (Fig. 5A,D,G,K). Quantification demonstrated that these cells have lower DAPI fluorescence intensity per cell than controls (Fig. 5N), suggesting that DAPI might require the PMF for uptake (similar to kanamycin (Taber et al., 1987)). In contrast, nisin treated cells showed increased chromosome condensation and increased DAPI fluorescence intensity per cell compared to the control (Fig. 5A,C). This suggests that intact cells take up less DAPI that nisin-treated cells. Nisin-treated cells were also rapidly permeabilized to SYTOX Green (Fig. 5C) and showed a 130X increase in SYTOX fluorescence intensity (Fig. 5M). This increased cell permeability is likely due to the ability of nisin to form non-specific channels in the cytoplasmic membrane (Lubelski et al., 2008). Vancomycin (at >10X the MIC) has little effect on DAPI or SYTOX Green fluorescence, but FM 4-64 staining was uneven, suggesting that the cells contained membrane invaginations. Thus, at short times of treatment, SYTOX Green permeability and chromosome condensation can discriminate between compounds that affect the PMF, which decondense the chromosomes, and those which produce large membrane channels, which rapidly permeabilized the cells and cause chromosome condensation.

Figure 5. The short-term consequences of various compounds on cell structure.

PY79 cultures were treated with the indicated compounds for 20 min at the concentrations specified in Fig. 3 (except CCCP is 500 μM and DNP is 10 mM), stained with FM 4-64 (red), DAPI (blue), and SYTOX Green (green) and imaged. (A-L) Overlays are shown following adjustment of DAPI images to visualize DNA structure (left) and of the DAPI and SYTOX Green images to reveal changes in fluorescence intensity (right) as described in the methods. Scale bar represents 1 μm. (M) Graph showing the ratio of average SYTOX Green fluorescence intensity/pixel for cells treated with various compounds for 20 min to the appropriate controls. Only nisin significantly permeabilizes the cells. (N) Graph of the ratio of average DAPI fluorescence intensity/pixel for cells treated with various compounds for 20 min to the appropriate controls. SDP, CCCP, DNP, nigericin, and valinomycin treated cells show significantly decreased fluorescence intensity, whereas nisin shows increased fluorescence intensity and vancomycin shows no change.

Autolysis is a secondary consequence of SDP treatment

During our cytological profiling experiments, we noted that cells treated with SDP lyse in a manner identical to that observed when untreated B. subtilis cells are moved from aerated to static culture (Fig. 3F). This suggested that autolysis, which is also induced by CCCP and azide (Jolliffe et al., 1981), might be responsible for the dramatic lysis that is a late consequence of SDP treatment. Autolysis is caused by cellular enzymes called autolysins that digest peptidoglycan during processes such as cell separation and elongation, and can cause cell lysis when deregulated (Smith et al., 2000, Jolliffe et al., 1981). To test the hypothesis that SDP induces autolysis, we used a strain missing the four major B. subtilis autolysins LytC, LytD, LytE, LytF (ALB1111) that is resistant to autolysis induced by a lack of aeration and by certain membrane active compounds, including CCCP and azide (Margot et al., 1999). When treated with SDP, the autolysin deficient strain failed to lyse, showing neither increased SYTOX permeability nor external membrane vesicles, even after five hours (Fig. 6B), although it accumulated internal membrane vesicles. However, viable cell counts showed that the mutant cells lost viability as rapidly as the wild type strain after SDP treatment (Fig. 6D, green). This indicates that autolysis is a secondary consequence of SDP treatment and not the primary cause of cell death.

Figure 6. Effect of autolysins on cell lysis and death.

(A-B) Fluorescence micrographs of autolysin deficient cells ΔlytABC ΔlytD ΔlytE ΔlytF (ALB1111) treated for 5 hrs with (A) DMSO, (B) 20 μg ml^-1 SDP, or (C) statically incubated. Cells are stained with FM4-64 (red), DAPI (blue), and SYTOX Green (green). Scale bar represents 1 μm. No change in SYTOX Green permeability is observed. (D) Cell viability in SDP-treated PY79 and ΔlytABC ΔlytD ΔlytE ΔlytF (ALB1111) cells. (The PY79 curves are the same as those shown in Fig. 3). Both strains lose viability upon SDP treatment. Error bars represent the standard error of ≥3 experiments.

Discussion

Our studies of the mechanism of action of SDP were greatly facilitated by cell biological studies of the effects of membrane active antibacterial compounds on the structure of bacterial cells. These studies indicated that SDP is likely collapses the PMF (as confirmed by more specific assays) and that this subsequently induces autolysis. Cytological profiling is capable of discriminating between two compounds that bind the lipid II precursor for peptidoglycan biogenesis (vancomycin and nisin), as only one (nisin) makes large channels in the membrane that permeabilize the cells to SYTOX Green (Fig. 5C and Fig. S4). This assay also discriminates between compounds that collapse both components of the PMF (CCCP, DNP), those that collapse or dissipate a single component of the PMF (ΔpH or ΔΨ; nigericin, valinomycin) and those that make non-specific channels in the cytoplasmic membrane (nisin). Our studies suggest that SDP rapidly collapses both components of the PMF and that this is responsible for its immediate toxicity to bacterial cells. Thus, cytological profiling can rapidly provide insight into the possible mechanisms of action of newly identified antibacterial compounds that affect the cell membrane, and thereby limit the number of more specific assays required to fully document the mechanism of action.

We favor the hypothesis that SDP collapses both components of the PMF by translocating protons across the membrane, perhaps forming a channel that allows protons but not larger molecules to diffuse freely across the membrane (Figure 7). Interestingly, many cationic antimicrobial peptides produced by the human immune system and bacteria kill cells by inserting into the membrane and assembling channels that collapse the PMF (Peters et al., 2010, Nan et al., 2011b). Thus, SDP, which is also a cationic peptide, shares a mechanism of action with many other peptides that play an evolutionarily ancient role in protecting eukaryotes and bacteria against invading bacteria.

Figure 7. Comparison of the cytological profiles and mechanisms of action and for SDP and nisin.

Key: grey - cell wall; red – cytoplasmic membrane; light green – DNA structure; dark green – cell permeability; + - protons; purple – autolysins; blue – SDP; brown - nisin. (A) SDP (blue) depolarizes the membrane, and we propose that it inserts into the membrane and forms a proton channel, causing PMF collapse and chromosome decondensation. Autolysin activity is deregulated and delocalized peptidoglycan degradation causes holes in the cell wall. After hours, this induces autolysis, with extrusion of membrane tubules and vesicles caused by high internal osmotic pressure (red), and increased SYTOX Green permeability. (B) Nisin (brown) inserts into the membrane and forms a large non-specific channel that rapidly permeabilizes the cells, collapsing the PMF, osmotically equilibrating the cells, and condensing the chromosomes. Autolysin activity is deregulated and delocalized peptidoglycan degradation causes holes in the cell wall. After hours, autolysis occurs, but in the absence of a high internal osmotic pressure, the cell implodes to form internal membranes rather than exploding to form external membrane vesicles and tubules.

A secondary consequence of PMF collapse in many Gram-positive bacteria, including B. subtilis and S. aureus, is autolysis, which we have demonstrated produces the dramatic lysis associated with SDP treatment. Autolysis involves enzymes that cleave peptidoglycan during cell elongation and cell separation, which are deregulated or hyperactivated by PMF collapse, leading to cell lysis. The mechanism by which loss of the PMF induces autolysis is not completely clear, but our observation that different cytological profiles are produced by collapse of either ΔpH or ΔΨ or both suggests that collapse of each component of the PMF might activate enzymes with distinct substrate specificities or localization. Previous results suggest that ΔpH is coupled to autolysin activity via the acidic wall polymer teichoic acid, which maintains a pH gradient across the cell wall. During respiration, teichoic acid is proposed to retain protons close to the cytoplasmic membrane, which are replaced by K+ or Na+ from the medium at the outer surface of the wall. This inhibits autolysin activity near the membrane of respiring cells, thereby restricting autolysin to the outer surface of the cells (Rice & Bayles, 2008, Neuhaus & Baddiley, 2003). Such autolysins would therefore be deregulated upon collapse of the ΔpH but not by collapse of ΔΨ. There are two attractive models for sensing ΔΨ. First, some bacteriophages regulate host cell lysis by PMF collapse, which triggers the release and activation of membrane-anchored endolytic enzymes (the SAR-endolysins; (Park et al., 2007, Rice & Bayles, 2008)). The PMF governs the insertion of charged transmembrane proteins into the membrane (Celebi et al., 2008), likely via the ΔΨ. Second, ΔΨ is required for localization of a variety of cell division proteins (Strahl & Hamoen, 2010), suggesting that depolarization might release the autolytic enzymes from their usual site of activity at the septum or the outer layers of the wall. Further studies are required to identify the autolytic enzymes that are induced by collapse of the ΔpH or ΔΨ, and to understand the mechanism by which collapse of each component of the PMF leads to autolysis.

Regardless of the mechanism by which the ΔpH and ΔΨ govern autolysin activity, delocalized peptidoglycan degradation ultimately produces holes in the cell wall (Rice & Bayles, 2008, Vollmer et al., 2008). In otherwise intact cells, the osmotic pressure would then force the membrane through these holes, leading to the extrusion of membrane vesicles and to permeabilization of the cell to larger molecules such as SYTOX Green. We hypothesize that the number and distribution of the holes in the cell wall is dictated by the specific autolysins activated by each treatment, and that this in turn determines if cells lyse with the extrusion of relatively few large vesicles or with the production of many smaller vesicles. In contrast, cells that been permeabilized with molecules that assemble large non-specific pores, such as nisin, would lack an internal osmotic pressure as the pores would allow the osmotic equilibration of the cell with the medium. Thus, by the time the autolysins produce holes in the cell wall, cell lysis would simply entail an internal collapse of the cell membrane rather than the extrusion of membrane vesicles.

Potential role of SDP in mediating inter- and intra-species interactions

SDP acts as a cannibalistic toxin that kills a subset of B. subtilis cells within the biofilm to allow continued growth thereby delaying the onset of sporulation (Ellermeier et al., 2006, Liu et al., 2010). We here demonstrate that SDP is active against a variety of bacterial species in the Firmicutes phylum, including species closely related to B. subtilis such as B. amyloliquifaciens and also more distantly related species such as Lactobacillus acidophilus. This expands the list of Gram-positive species sensitive to SDP beyond those previously published (Palmer et al., 2009, Liu et al., 2010) and suggests that SDP might play a more general role in defending the B. subtilis colony against invading species. We speculate that in this broader context, depleting cellular energy stores might be the ideal mechanism of action for a defensive extracellular metabolite. First, sublethal concentrations of compounds that collapse the PMF rapidly inhibit flagellar motility in a variety of species (Nan et al., 2011a, Ridgway, 1977, Goulbourne & Greenberg, 1980). This could prevent motility driven invasion of B. subtilis colonies as well as swarming away from the B. subtilis colony. Second, collapsing the PMF disrupts biofilm formation in Pseudomonas aeruginosa and Shewanella oneidensis (Ikonomidis et al., 2008, Saville et al., 2011), suggesting that SDP could inhibit biofilm formation by neighboring species, and thus reduce competition for a niche. Third, the PMF is necessary for protein secretion and the export of many toxic secreted metabolites (Geller, 1991, Driessen, 1992), so SDP might reduce the ability of neighboring species to produce or respond to toxic molecules. Finally, and perhaps most importantly, a variety of Gram-positive species undergo autolysis when the PMF is compromised (Reith & Mayer, 2011, Jolliffe et al., 1981), but not when pathways such as transcription or translation are inhibited by antibiotics (Falk et al., 2010, Jolliffe et al., 1981). Thus, inducing autolysis would be advantageous for a cannibalistic or defensive toxin as it would lead to the release of nutrients from target cells without the need to secrete enzymes to lyse neighboring bacteria. The SDP toxin is therefore ideally suited for its roles in eliminating unnecessary cell types from the colony and in defending it from invasion by competing species.

Experimental Procedures

Strains and culture conditions

The strains used in this study are listed in Table I. All cultures were grown in LB medium at 37°C, except cultures for fluorometry. All cells treated with compounds were treated in concentrated microcultures (described below), except cultures for fluorometry, which were grown in LB supplemented with 50 mM Hepes pH 7.5, 300 mM KCl and 0.1% glucose.

Table I.

Strains used in this study

Strains	Genotype	Background	Reference
B. subtilis
3610	Prototroph, undomesticated parent of 168		(Branda et al., 2001)
PY79	Prototrophic derivative of B. subtilis 168		(Youngman et al., 1984)
EH273	sdpABC∷kan	PY79	(Ellermeier et al., 2006)
TPM758	ΔsdpABCIR∷tet, ΔsigW∷kan amyE∷PhyspacsdpI+Ωspec	PY79	Ellermeier lab
KP3034	lacA∷spec	PY79	BGSC1A785→PY79
SCB751	amyE∷mifM-yidC26-lacZΩcat	PY79	(Chiba et al., 2009)
ALB1039	ΔsigW∷mls	PY79	This study
ALB1088	ΔsdpABCIR∷tet	PY79	This study
ALB1142	ΔsdpABCIR∷tet, ΔsigW∷mls	PY79	This study
ALB1111	lytABC∷neo, lytD∷ tet, lytF∷spec, lytE∷cm	PY79	This study
HB0020	ΔsigW∷mls trpC2 ΔSPβ	168	(Cao et al., 2001)
L16601	Prototrophic derivative of B. subtilis 168		(Margot & Karamata, 1996)
L16648	lytABC∷neo, lytF∷spec, lytE∷cm, lytD∷tet	L16601	(Margot et al., 1999)
CDE433	ΔsdpABCIR∷tet, ΔsigW∷kan	PY79	(Ellermeier & Losick, 2006)
E. coli
MC4100	Prototroph		(Casadaban, 1976)
NR698	lptD4213	MC4100	(Ruiz et al., 2005)
Other Species
ATCC9341	Micrococcus luteus		Lab collection
ATCC12872	Bacillus megaterium		Lab collection
FZB42	Bacillus amyloliquifaciens		Lab collection
ATCC35646	Bacillus thuringiensis		Lab collection
ATCC4356	Lactobacillus acidophilus		Lab collection
ATCC842	Paenibacillus polymyxa		Lab collection
ATCC35984	Staphylococcus epidermidis		Lab collection

Solutions of SDP and other toxins were prepared using the following concentrations and solvents: 400 μg ml^-1 SDP (10% DMSO, purified as described in (Liu et al., 2010), 100 mM sodium azide (H₂O, Sigma), 2 mM CCCP (20% DMSO, Sigma) or 10 mM CCCP (100% DMSO, Sigma), 5 mg ml^-1 valinomycin (100% DMSO, Sigma), 2.5 mg ml^-1 nigericin (100% DMSO, Calbiochem), 400 mM dinitrophenol (100% DMSO, Sigma) 5 mg ml^-1 vancomycin (H₂O, Sigma), and 100 μg ml^-1 nisin (10% DMSO, Sigma). 1mm IPTG was used to induce SdpI in TPM758 cells.

Concentrations of antibiotics in solid culture were as follows: MLS (1 μg ml^-1 erythromycin, 25 μg ml^-1 lincomycin), 10 μg ml^-1 tetracycline, 100 μg ml^-1 spectinomycin. Except when otherwise noted, SDP was used at 20 μg ml^-1, which we estimate is between 2 and 5x the MIC (Table SI).

Fluorescence microscopy

Cells were cultured for fluorescence microscopy as described in (Liu et al., 2010). The effects of compounds on individual B. subtilis cells were investigated in concentrated 15 μl microcultures prepared in the following manner. Cultures were grown in LB media to an OD₆₀₀ of 0.3, centrifuged, resuspended in 1/10 the volume and 14.25 or 13.5 μl (depending on the volume of compound to be added) of concentrated cells were added to 1.7 ml microcentrifuge tubes. At t = 0, 0.75 μl or 1.5 μl of the indicated compound was added to aliquots of cells. Compounds were used at the following final concentrations: 0.5% DMSO, 2.5% DMSO, 5.0% DMSO, 20 μg ml^-1 SDP, 10 mM azide, 100 μM CCCP (Fig. 3), 500 μM CCCP (Fig. 5), 2 mM DNP (Fig. 3), 10 mM DNP (Fig. 5), 62.5 μM nigericin, 125 μM valinomycin, 250 μg ml^-1 vancomycin, 10 μg ml^-1 nisin. Cultures to be treated with valinomycin were grown in LB supplemented with 200 mM KCl (Fig. 3), or 300 mM KCl (Fig. 5). Tubes were capped and incubated at 37°C in a roller. Samples were collected for imaging every hour. 3 μl of cells were added to 0.75 μl of a stain mix containing 30 μg ml^-1 FM 4-64, 2.5 μM SYTOX Green and 1.2 μg ml^-1 DAPI prepared in 1X T-base. E. coli cultures were treated identically to B. subtilis cultures, but the treatment was extended to 5 hours. The pH of control cultures was monitored after incubation, and found to be unchanged from the starting medium (pH 6.5-7.0).

Cells were immobilized on an agarose pad (1/10 LB, 0.375 μg ml^-1 FM 4-64, 0.025 μg ml^-1 DAPI) and imaged on an Applied Precision Spectris Microscope described in (Liu et al., 2006). Images were deconvolved using softWoRx v3.3.6 (Applied Precision) and the medial focal planes shown. The DAPI and SYTOX Green images in Fig. 1-3, 6, S1 and S4 were normalized within each figure based on intensity and exposure length to reflect intensities relative to the treatment with the highest fluorescence intensity in each figure.

Sporulation conditions

To quantify the impact of SDP on sporulation, concentrated microcultures in LB were prepared as described above, and incubated with rolling at 37°C. Samples were harvested after 5 hours, and stained with FM 4-64 and prepared for microscopy as described above. The number of cells with asymmetric septa in a 3610 culture was counted after treatment with 0.5% DMSO or 20 μg ml^-1 SDP for 5 hours. At least 100 cells were counted per treatment per experiment. Numbers reflect the average of three experiments.

Viable cell counts

Viable cell counts were obtained through dilution and plating of cells from the same cultures as those subjected to microscopy. Tenfold serial dilutions were made at the indicated time in 1X T-base and spotted onto LB plates. The first dilution reduces the concentration of SDP and other antimicrobial compounds below their minimal inhibitory concentrations. We therefore infer that cell death is due to compounds that bound the cell before dilution. Colonies were counted after growth and colony forming units (CFU) per ml calculated. Shown is the ratio of CFU at time × (CFU(t_x)) to CFU of the control at t₀ (CFU^C(t₀)).

Flow cytometry

Measurement of the PMF by flow cytometry used NR698 (E. coli lptD4213) grown as for microscopy. After SDP treatment at 37°C for 10 or 20 min (as indicated in the figure), 3-15 μl of culture was resuspended in 1X PBS with 5 μg ml^-1 DiBAC₄(5) and 100,000 events counted using a BD LSR II flow cytometer. Fluorescence data was collected using the 561 nm laser. Area under the curve (dsRed-A) was used as the fluorescence intensity measurement.

Fluorometry

Measurement of the PMF by fluorometry was conducted similarly to (Strahl & Hamoen, 2010). B. subtilis strain PY79 cells were grown in LB supplemented with 50 mM Hepes pH 7.5, 300 mM KCl and 0.1% glucose to an OD₆₀₀ of 0.3-0.35, spun down, then resuspended in ½ volume 50 mM Hepes pH 7.5, 300 mM KCl and 0.1% glucose. DiSC₃(5) (1 μM final), cells, and DMSO (0.5% final), vancomycin (10 μg ml^-1 final), nisin (11 μg ml^-1 final), or SDP (1 μg ml^-1 final) were mixed together in a final volume of 50 μl and transferred into a capillary cuvette. The fluorescence was monitored with a Fluoromax-4 spectrofluorometer at excitation 544-10 nm and emission 660-10nm. The representative curves shown were normalized so that each started at the same level of fluorescence.

Quantification of DAPI and SYTOX Green staining intensity

Cells used for DAPI and SYTOX Green quantification were treated with the indicated compound for 20 min at 37°C prior to immobilization on an agarose pad for fluorescence microscopy. The DAPI and SYTOX staining intensities are sensitive to both cell permeabilization and to chromosome architecture. Thus, more permeabilized cells show brighter staining, as do more condensed chromosomes, which is why we show two images. First, to show chromosome architecture in the panels titled “DNA structure”, the DAPI images were adjusted to allow the optimal visualization of the chromosomes. Second, to show staining intensity relative to untreated controls in the panels titles “DAPI and Sytox intensity”, the DAPI and SYTOX Green images were adjusted to normalize the brightness relative to that of the brightest sample (nisin), based on exposure length and intensity. This more accurately displayed relative fluorescence intensities. Thus, the manner in which we quantified the data in Figure 5 primarily reflects uptake, since we quantified total DAPI staining intensity per cell rather than per area occupied by the chromosome.

Average DAPI and SYTOX Green fluorescence intensity/pixel for each cell was calculated from non-deconvolved images and the background values subtracted. ≥100 cells were measured per treatment. Briefly, a polygon was drawn in Image J using the membrane as a guide. To calculate background fluorescence, polygons were drawn in areas without cells. Values for DAPI and SYTOX Green fluorescence intensity/pixel were adjusted to reflect a 0.3 s exposure time. Finally, the average fluorescence intensity for each sample was divided by the average fluorescence intensity of the appropriate solvent control.

Timelapse microscopy

Timelapse fluorescence microscopy (Becker & Pogliano, 2007) was conducted on concentrated microcultures treated with 0.5% DMSO or 20 μg ml^-1 SDP at 37°C for 20 minutes. After treatment, 3.0 μl of cells were added to 9.0 μl of a stain mix in 1X T-base, containing 0.67 μg ml^-1 FM 4-64 and 0.67 μM SYTOX Green, applied to agarose pads (1/5 LB, 0.3 μg ml^-1 FM 4-64, 0.5 μM SYTOX Green) and grown at 30°C for ~20 min prior to microscopy. Pictures were taken every 10 minutes for 2 hr.

Statistical Analysis

Viable cell counts after drug treatment (Fig. 2C, 3K, 4DE): n≥3 separate experiments, the average value is plotted, and error bars represent standard error. Spot assay viable cell quantification (Fig. S1): Bars represent the average of n≥3 separate experiments. Error bars represent the standard deviation.