Bacterial Periplasmic Oxidoreductases Control the Activity of Oxidized Human Antimicrobial β-Defensin 1

ABSTRACT

The antimicrobial peptide human β-defensin 1 (hBD1) is continuously produced by epithelial cells in many tissues. Compared to other defensins, hBD1 has only minor antibiotic activity in its native state. After reduction of its disulfide bridges, however, it becomes a potent antimicrobial agent against bacteria, while the oxidized native form (hBD1ox) shows specific activity against Gram-negative bacteria. We show that the killing mechanism of hBD1ox depends on aerobic growth conditions and bacterial enzymes. We analyzed the different activities of hBD1 using mutants of Escherichia coli lacking one or more specific proteins of their outer membrane, cytosol, or redox systems. We discovered that DsbA and DsbB are essential for the antimicrobial activity of hBD1ox but not for that of reduced hBD1 (hBD1red). Furthermore, our results strongly suggest that hBD1ox uses outer membrane protein FepA to penetrate the bacterial periplasm space. In contrast, other bacterial proteins in the outer membrane and cytosol did not modify the antimicrobial activity. Using immunogold labeling, we identified the localization of hBD1ox in the periplasmic space and partly in the outer membrane of E. coli. However, in resistant mutants lacking DsbA and DsbB, hBD1ox was detected mainly in the bacterial cytosol. In summary, we discovered that hBD1ox could use FepA to enter the periplasmic space, where its activity depends on presence of DsbA and DsbB. HBD1ox concentrates in the periplasm in Gram-negative bacteria, which finally leads to bleb formation and death of the bacteria. Thus, the bacterial redox system plays an essential role in mechanisms of resistance against host-derived peptides such as hBD1.

INTRODUCTION

The primary barrier against bacterial infection is provided by the innate immune system. Antimicrobial peptides (AMPs) are key effector molecules that protect the body from an overgrowth of pathogenic and commensal bacteria (1). AMPs show antimicrobial activity against many microbes, bacteria, fungi, and some viruses (2–5). One of the most important and intensively studied groups of AMPs are defensins (2), a group of very small cationic peptides characterized by three intramolecular disulfide bridges (2, 6). In contrast to most other defensins, the human β-defensin 1 (hBD1) is constitutively expressed by all human epithelia (7–9). Its antimicrobial activity against microorganisms was underestimated for a long time, and only recently our group discovered that hBD1 exhibits different antimicrobial killing activities depending on the redox properties of the environment. The antimicrobial activity of hBD1 is highly increased when the disulfide bonds of this defensin are reduced (10). Reduced hBD1 (hBD1red) has a distinct antimicrobial profile and function compared to that of the native form. Additionally, as an independent mechanism, the reduced peptide provides broad protection by entrapping bacteria in extracellular net structures, preventing bacterial invasion (11). HBD1 is secreted as an oxidized peptide and can be reduced by present reductases such as thioredoxin (12) or a reducing environment. However, it is likely that in vivo a mixture of reduced and oxidized hBD1 (hBD1ox) exists. HBD1ox, which contains three closed disulfide bonds, shows no antimicrobial activity against tested Gram-positive bacteria (11) but specific activity against the Gram-negative bacterium Escherichia coli (10). Yet, the distinct exact antimicrobial mechanisms of oxidized hBD1 are still unknown.

The envelope of Gram-negative bacteria, e.g., E. coli, is a complex macromolecular structure consisting of the outer membrane, the inner membrane, and the periplasmic space between (13). This periplasmic space, which is missing in Gram-positive bacteria (14), is the site where redox systems fold bacterial proteins by introducing disulfide bonds. While the disulfide bond formation in Gram-positive bacteria is not fully understood, it is known that secreted proteins can also be oxidized and contain multiple cysteine residues (15). Thus, differences between these two bacterial groups include the formation of bacterial cell wall components consisting of different transporters, such as FepA, FhuA, Omp proteins, and especially the periplasmic redox system. In E. coli, the two oxidoreductases DsbA and DsbB are important components of this redox system and catalyze the formation of disulfide bridges into proteins. DsbA, which acts as a thiol oxidase, is maintained oxidized in the periplasm by the inner membrane protein DsbB. Electrons are transferred from DsbB to ubiquinones and finally to terminal oxidases. Under anaerobic conditions, the electrons flow from DsbB to menaquinone (16).

Previous data from our group demonstrated that hBD1red is active against Gram-positive and Gram-negative bacteria by targeting mainly the bacterial membrane and trapping bacteria in a net structure (11). Due to the different activity spectra of hBD1red and hBD1ox under various environmental conditions, we here aimed to investigate whether the redox-active periplasm of Gram-negative bacteria contributes to the observed difference in hBD1 activity.

RESULTS

HBD1ox shows specific toxicity against E. coli.

To determine the antibacterial spectra of hBD1 under different conditions, we first investigated its activity against a variety of Gram-negative bacteria in a radial diffusion assay (RDA). HBD1red showed antimicrobial activity against all tested bacteria (Fig. 1A and B). Remarkably, the oxidized form was active against all tested Gram-negative bacterial strains under aerobic conditions (Fig. 1A), even against pathogenic Salmonella enteritidis or Acinetobacter baumannii. Conversely, Gram-positive bacteria were resistant against hBD1ox (10, 11), which indicates that hBD1ox activity may target a mechanism that is specific to Gram-negative bacteria. Interestingly, under anaerobic conditions, the hBD1ox completely lost its antimicrobial activity (Fig. 1B), suggesting the bacterial target to be redox sensitive. Thus, since hBD1ox, a folded globular polypeptide, is able to specifically kill Gram-negative bacteria only under aerobic conditions, we hypothesized that specific components of the periplasm are involved in hBD1ox activity.

FIG 1.

The antimicrobial activity of hBD1ox is specific for Gram-negative bacteria. HBD1ox and hBD1red were incubated with Gram-negative bacteria in an aerobic (A) or anaerobic (B) environment in the radial diffusion assay. Antimicrobial activity was measured by analyzing the diameter of the inhibition zone. A diameter of 2.5 mm (dotted line) is the diameter of the punched well. Data are presented as mean ± SEM from three independent experiments.

To study the relevance of proteins involved in periplasmic processes that are specific for Gram-negative bacteria (Fig. 2A shows a schematic model), we generated E. coli MC1000 mutant strains that were deficient in different membrane proteins (Fig. 2B and C). The outer membrane proteins of the Omp family are small β-barrel proteins playing a structural role in the cell envelope, which might be relevant for the entry of hBD1ox. LamB is involved in maltose transport, while CirA, FhuA, and FepA are involved in the import of iron and act as receptors for a number of different colicins (17, 18). When testing the antimicrobial sensitivity of E. coli MC1000 mutants deficient in any of these outer membrane proteins, we could not detect any differences in sensitivity toward hBD1ox compared to that of the wild-type (WT) control strain, except for the FepA mutant, which displayed increased resistance against hBD1ox (**, P = 0.0014). In addition, neither of our bacterial mutants that were deficient in cytosolic TrxA, TrxC, Gor, or GshA exhibited altered sensitivity (Fig. 2C). However, when testing mutants that lacked distinct periplasmic proteins involved in the Dsb redox system, a deletion of either DsbA, DsbB, or DsbC led to a significantly decreased sensitivity against hBD1ox (ΔdsbA mutant, ***, P = 0.0002; ΔdsbB mutant, **, P = 0.0054) but not hBD1red (Fig. 2D). In contrast, bacteria lacking DsbG did not change in their sensitivity against any hBD1 forms.

FIG 2.

Antimicrobial activity against bacteria with mutations in the genes of interest. (A) Schematic overview of located membrane and periplasmic proteins in bacteria. (B to D) Two micrograms of hBD1ox and hBD1red was tested against E. coli strains with different protein knockouts in the outer membrane (B), cytosol (C), and periplasmic space (D) by radial diffusion assay. (D) HBD1ox shows a decreased antimicrobial activity against bacteria without DsbA or inner membrane proteins DsbB and TonB in comparison to the WT E. coli MC1000. The diameter of the inhibition zone was measured in the radial diffusion assay to determine the antimicrobial activity. A diameter of 2.5 mm (dotted line) is the diameter of the punched well. Results from experiments with wild-type E. coli MC1000 were pooled (n = 18) and used as the control for all the times when E. coli MC1000 mutants (n = 3) are assessed. All data are presented as mean ± SEM.

Another potentially interesting periplasmic protein is TonB, which is located in the inner membrane. TonB provides energy for the three high-affinity ferric iron uptake systems, CirA, FhuA, and FepA. Additionally, TonB facilitates cell penetration of some bacterial antimicrobial peptides, such as colicins (19, 20). Accordingly, the TonB deletion mutant became insensitive to hBD1ox, highlighting its central role in energizing essential proteins and potential AMP uptake. Based on these observations, we assume that the uptake of hBD1ox depends on the outer membrane receptor FepA, powered by TonB, and that functionality of the Dsb redox system is required for antimicrobial activity of hBD1ox.

Bacteria without the DsbA/DsbB complex are resistant against hBD1ox.

In the periplasm, the bacterial redox system DsbA/DsbB introduces disulfide bridges into proteins (21). To further test whether the absence of a functional disulfide bond formation pathway impacts hBD1ox activity, we generated different double mutants of E. coli MC1000 and analyzed the antimicrobial activity of hBD1ox (Fig. 3A). Indeed, bacterial mutants lacking both DsbA and DsbB became fully resistant against hBD1ox (****, P < 0.0001) (Fig. 3A, red box). In contrast, deletion of DsbC and DsbG, which are further down in the Dsb pathway, did not affect antimicrobial sensitivity against hBD1ox. FepA and FhuA both have disulfide bridges that are required for folding. A double mutant lacking both FepA and FhuA displayed decreased sensitivity toward hBD1ox (Fig. 3A); however, it was not lower than that of the single mutant lacking FepA (Fig. 2B). This indicates that FepA, but likely not FhuA, is important for hBD1ox antimicrobial activity. It is thus possible that in the absence of a functional disulfide bond formation pathway, TonB-dependent FepA does not fold properly, and therefore the cells become resistant to hBD1ox. To confirm our results from the radial diffusion assay, we tested bacterial killing in an assay for membrane depolarization. While reduced hBD1 led to effective depolarization of the bacterial membrane in the E. coli ΔdsbA ΔdsbB mutant, hBD1ox did not (Fig. 3B), thus supporting the bacterial resistance from the diffusion assay. Consistent with published data with WT E. coli MC1000, only hBD1red had a rapid bactericidal effect (11).

FIG 3.

Bacterial oxidoreductases DsbA and DsbB are essential for the activity of hBD1ox. (A) Two micrograms of hBD1ox and hBD1red was tested against E. coli with protein double knockouts in the redox system and outer membrane. Bacteria without both oxidoreductases obtain resistance against hBD1ox (**, P = 0.0044; ****, P < 0.0001). (B) Membrane depolarization assay of E. coli MC1000 ΔdsbA ΔdsbB incubated with reduced and oxidized hBD1 for 1 h. HBD3ox (50 μg/ml) was used as positive control (+), and the negative control (−) was incubated without any peptide. HBD1ox is not able to significantly affect the bacterial membrane (**, P = 0.0067). (C) Four micrograms of hBD1ox was used in the plasmid induction experiment with either E. coli MC1000 ΔdsbA ΔdsbB, E. coli MC1000 ΔdsbA ΔdsbB harboring plasmids pQE60::dsbB and pBAD33::dsbA to generate the wild-type phenotype, or, as a control, E. coli MC1000 ΔdsbA ΔdsbB with empty plasmids pQE60 and pBAD33. Plasmid expression was during the 2.5 h of incubation, with addition of 0.4% l-arabinose and 2 mM IPTG every 30 min. HBD1ox shows antimicrobial activity against the mutant containing plasmids pQE60::dsbB and pBAD33::dsbA (**, P = 0.0036; ****, P < 0.0001). The diameter of the inhibition zone was measured in a radial diffusion assay to determine the antimicrobial activity. Results of representative radial diffusion assays are shown. A diameter of 2.5 mm (dotted line) is the diameter of the punched well. Data are presented as mean ± SEM from at least three independent biological replicates. Student's t test was used for comparison of normally distributed data.

To further confirm the essential role of DsbA and DsbB in hBD1ox-mediated bacterial killing, we introduced both proteins by genetic complementation of both proteins (see Fig. S1 in the supplemental material). Indeed, in the complemented ΔdsbA ΔdsbB (pQE60::dsbB, pBAD33::dsbA) double mutant, the sensitivity toward hBD1ox was fully restored (Fig. 3C). These data confirm that DsbA and DsbB are indispensable for the antimicrobial activity of hBD1ox.

It is known that the E. coli strain lacking DsbA exhibits defective formation of flagellar components (16, 22–24), resulting in reduced motility. To exclude side effects due to impaired motility, we tested E. coli Nissle 1917 strains with diminished swarming capabilities by analyzing mutants lacking different flagellum components (see Fig. S2A in the supplemental material) (24, 25). Furthermore, we tested the influence of lipopolysaccharide (LPS), a major surface molecule of Gram-negative bacteria, containing a unique phospholipid domain known as lipid A (Fig. S2B and C). However, neither bacterial motility nor lipid A structure seemed essential for the activity of hBD1ox, as neither bacterial mutants displayed differences in sensitivity (Fig. S2A to C). In conclusion, this emphasizes the specific importance of DsbA/DsbB in hBD1ox-mediated killing and suggests no mechanistic connection between motility and other redox and membrane components.

HBD1ox leads to membrane vesicles and is located predominantly in the periplasm.

After we demonstrated that a functional DsbA/DsbB system is crucial for the antimicrobial activity of hBD1ox, we investigated morphological changes in the wild-type and ΔdsbA ΔdsbB mutant bacteria after hBD1ox treatment. Bacteria were incubated with hBD1ox for 1 h and subsequently fixed in Karnovsky's reagent and visualized by scanning electron microscopy (Fig. 4A). It is known that treatment with antimicrobial peptides often results in formation of membrane vesicles (blebs) in bacteria (11, 25, 26), a sign of a bacterial stress response (25). While no blebs were observed in untreated wild-type bacteria, treatment with hBD1ox resulted in bleb formation on the bacterial surfaces (Fig. 4A). In contrast to the wild type, the DsbA/B double mutant had a strikingly lower number of blebs after hBD1ox treatment, indicating a decreased stress reaction (Fig. 4A and B), which was in concordance with the increases in resistance observed in the RDA. This observation visually shows the antimicrobial effect of hBD1ox and further indicates that DsbA/DsbB play a central role in the antimicrobial process of hBD1ox.

FIG 4.

HBD1ox induces bleb formation in E. coli. (A) WT E. coli MC1000 and E. coli MC1000 ΔdsbA ΔdsbB were treated with hBD1ox for 2 h. The samples were fixed in Karnovsky's reagent, and the morphology was analyzed with scanning electron microscopy. Electron microscopy pictures from one representative experiment are shown. Scale bar, 2 μm. (B) Visual analysis of electron microscopy pictures. Outer membrane vesicles (blebs) were counted and determined by four different experts. Data are presented as mean ± SEM from the analyzed bacteria (n = 10). The statistical significance was evaluated by using Student's t test (**, P = 0.0031).

Based on the differences in bleb formation, we studied the localization of hBD1ox in bacterial compartments in the wild-type and the ΔdsbA ΔdsbB mutant to better understand potential bacterial targets. For this experiment, we performed immunoelectron microscopy (IEM) with WT E. coli or the ΔdsbA ΔdsbB mutant. Bacteria were incubated with hBD1ox for 2 h, and the fixed samples were treated with primary antibodies against hBD1ox and secondary antibody conjugated to gold particles. HBD1ox was detectable in all bacteria (Fig. 5, black points marked by arrows). In the hBD1ox-treated wild-type strain, the outer membrane showed signs of detachment from the periplasmic space and the inner membrane, indicating bacterial disintegration. In contrast, the resistant E. coli ΔdsbA ΔdsbB strain showed no obvious structural damage of the bacterial cell. In addition to the structural difference, oxidized hBD1 was detected primarily at the outer membrane or in the periplasm in wild-type E. coli (78.5% of detected peptide) but at only a minor amount (21.4%) in the cytosol (Fig. 5A [left] and B). In the double mutant strain, however, the localization of hBD1ox differed clearly from the that in the wild type (Fig. 5A [right] and B), as hBD1ox was detected at a higher frequency in the cytosol (45.5%). In summary, these findings indicate that the DsbA/DsbB system inhibits the translocation of hBD1 into the cytosol. It can thus be speculated that the bacterial periplasm is the site of antimicrobial action of the human host peptide, since hBD1ox remains largely at this compartment in wild-type bacteria with a functional DsbA/DsbB system. Taken together, these data show that DsbA and DsbB are indispensable for the activity of hBD1ox, which presumably performs its antimicrobial activity in the periplasm.

FIG 5.

Localization of hBD1ox in bacterial compartments. WT E. coli MC1000 and E. coli MC1000 ΔdsbA ΔdsbB were treated with hBD1ox for 2 h. The samples were fixed and incubated with antibodies against the hBD1ox. The secondary antibodies were conjugated to 6-nm gold particles (black points) and visualized by electron microscopy. (A) Electron microscopy pictures from one representative experiment. Scale bar, 0.2 μm. (B) Visual analysis of electron microscopy pictures. Gold-labeled hBD1ox was counted in bacterial periplasm or cytosol per bacterium. Gold particles per bacterium (percent) are represented (WT, n = 39; ΔdsbA ΔdsbB mutant, n = 30) as determined by four independent individuals.

DISCUSSION

Here we show that functional bacterial disulfide bond machinery is required for the antimicrobial activity of the oxidized form of hBD1. Remarkably, analysis of the gold-labeled peptide suggests for the first time that the periplasmic bacterial space is the site of antimicrobial action, while translocation of hBD1ox into the cytoplasm does not affect bacterial viability.

Although antimicrobial peptides are small, they show a huge variety in antimicrobial spectra and mechanisms. For example, the insect defensin A kills bacteria by reducing the cytoplasmic potassium concentration. Another AMP, buforin II, penetrates and accumulates in the cytoplasm, where it interacts with nucleic acids (26, 27). Different lines of investigations show that a variety of known antimicrobial peptides, including hBD1, Paneth cell-derived HD6, and the S100 protein psoriasin, exert distinct antimicrobial activity depending on the environmental conditions after breaking up disulfide bonds (10, 12, 28–31). In addition, as just shown recently, only the reduced form of hBD1 can entrap bacteria, including hBD1red-resistant strains such as Klebsiella pneumoniae strains, in a net-like structure. These nets, which are also formed by the Paneth cell-derived HD6, are functionally capable of preventing bacterial migration independent of bacterial killing (11, 32). More-detailed studies also showed that only the reduced form of hBD1 targets the bacterial membrane by reducing the membrane integrity or potential, while the oxidized form (hBD1ox) does not (11). In summary, these investigations support the idea that one host defense peptide can have distinct anti-infective mechanisms depending on the environmental conditions. Despite the recently discovered activation of hBD1 against a variety of microbes after reduction, we were intrigued by the specific antimicrobial activity of oxidized hBD1 against several E. coli strains. Thus, we aimed here to better understand killing mechanisms, especially against Gram-negative strains such as E. coli strains. The antimicrobial activity of the native oxidized form of hBD1 targets specific Gram-negative bacteria only in aerobic environments (Fig. 1A). Other bacterial proteins in the flagellum system or changes in the bacterial lipid A composition (33) did not affect the antimicrobial function of HBD1ox (see Fig. S2A and B in the supplemental material). One major characteristic of Gram-negative bacteria is the periplasmic space including a redox system (34). Of note, this bacterial redox system is active mainly under aerobic conditions (16, 22), which could explain the dependence of hBD1ox activity on the environmental status. We hypothesized here that the periplasmic redox system of Gram-negative bacteria might be necessary for the activity of the oxidized form of ubiquitous hBD1. Furthermore, an inner membrane protein, TonB, preferentially interacts with and energizes outer membrane transport proteins, e.g., FhuA, FepA, and CirA. These proteins are multifunctional and transport ferrichrome and the antibiotics albomycin and rifamycin across the outer membrane or serve as receptors for colicins (35). It is known that these iron uptake systems are dependent on the TonB function. However, it should be noted that TonB is repressed in E. coli under anaerobic conditions (19).

After testing several mutants, we conclude that FepA, TonB, and especially DsbA and DsbB are essential proteins for the activity of hBD1ox. We observed a dramatically decreased antimicrobial activity against bacteria without DsbA or DsbB and full resistance in the double-knockout mutant (Fig. 3A). Since ΔdsbB strains are more sensitive than ΔdsbA strains, DsbA function might be more prominent than that of DsbB. This is in concordance with other observations showing that the role of DsbA is irreplaceable by other proteins (36). Furthermore, DsbA is an essential catalyst promoting the correct folding of secreted or surface-presented factors and seems to be important for intracellular survival (37). In addition to the double-knockout mutant, Gram-positive bacteria which are naturally missing this redox system also are resistant against hBD1ox (34). To confirm our findings, we expressed DsbA and DsbB plasmids in the double-knockout mutant to restore the wild-type phenotype by genetic complementation. After excluding any differences in the composition of membranes between the wild type and the double mutant (Fig. 3B), we could show a strong significant antimicrobial activity after plasmid expression. It is conceivable that hBD1ox can enter the bacterial periplasm through FepA, which is an outer membrane receptor for colicin (19). A decreased sensitivity in bacteria without FepA and TonB could be caused by a difficulty in hBD1ox uptake in the periplasm by inner membrane protein active transporters (Fig. 2B to D). Our data demonstrate an important role of DsbA/B as key players in this antimicrobial mechanism; however, we cannot exclude that other, still-unknown proteins might be involved, e.g., FepA. We hypothesize that this molecular mechanism of TonB-dependent outer membrane active transport is additionally involved but not fully understood.

Different mechanisms suggest different locations within the bacteria. To localize the bacterial target we used specific techniques. As antimicrobial peptides are very small molecules and their activity depends strongly on their conformation (10, 31, 32), we avoided tagging any compounds to the small peptide which might interfere with the antimicrobial activity. We therefore performed immunogold labeling, in which hBD1 is detected by specific antibodies after incubation with the bacteria and final fixation. With this technique, we were able to detect hBD1 in different bacterial compartments. Also, the immunogold images suggest that the peptide is able to diffuse through the outer membrane into the periplasm. High concentrations of hBD1ox in this compartment can correlate with antimicrobial activity, since the resistant mutant of hBD1 shows a large amount of gold-labeled hBD1ox in the cytosol. Thus, these data suggest that a deficiency of DsbA/DsbB enables the peptide to translocate into the cytosol by an unknown mechanism. Under normal conditions in the presence of the oxidoreductases, hBD1ox seems to be restricted to the periplasm. This peptide accumulation at this compartment could potentially be necessary for antimicrobial action. Despite the high specificity of the immunogold labeling without modifying the natural target peptide, experimental limitations include the qualitative nature of this technique. Electron microscopy imaging without specific labeling provides additional information about the mode of action of hBD1ox. Bacteria treated with active peptide bacteria were characterized by the presence of bacterial membrane vesicles, called blebs, at the bacterial surfaces (Fig. 4A, left). This morphology as observed here indicates decreased viability, as blebs are correlated with membrane damage and cell death (25, 26).

Chileveru et al. have recently described that the intestinal antimicrobial peptide HD5 can induce morphological changes in E. coli. These bacteria displayed cell elongation, clumping, and additional membrane vesicles, which are typically formed at the site of cell division or cell poles (25). As expected, the DsbA/DsbB double mutant, as shown here, displays significantly decreased bleb formation after hBD1ox treatment (Fig. 4B, right). Thus, the depletion of DsbA/B might lead to a stable membrane conformation and a lower stress reaction.

In summary, this study uncovers crucial components of antimicrobial mechanisms of host peptides and potential bacterial resistance. Our study shows that DsbA and DsbB are essential factors for the activity of the oxidized form of hBD1 in Gram-negative bacteria. Based on the data as shown here, we hypothesize that HBD1ox diffuses in the bacterial periplasm via TonB-dependent outer membrane active transporters, e.g., FepA, and exerts its antimicrobial killing in the bacterial periplasm by a still-unknown mechanism. We expect that the bacteria are damaged when the concentration of hBD1ox in the bacterial periplasm is high enough (a schematic model is shown in Fig. 6).

FIG 6.

Schematic model describing a potential mechanism of hBD1ox activity. (A) Proposed model for cell death in case of an intact DsbA/DsbB complex. HBD1ox can diffuse in the periplasm by using FepA or FhuA, where it can interact directly or indirectly with DsbA/DsbB. Subsequently, the accumulation of hBD1ox in the periplasm induces bleb formation and finally bacterial cell lysis by an unknown mechanism. (B) Hypothetical model for hBD1ox resistance in bacteria without a traditional DsbA/DsbB complex. Without DsbA/DsbB, hBD1ox diffuses into the cytosol and is finally degraded.

So far, not enough is known about the full antimicrobial killing mechanisms of host antimicrobial peptides, and further research regarding this aspect might be important for developing innovative therapeutic strategies against bacteria. However, this study highlights the complex variety of dynamic environmentally controlled antimicrobial killing mechanisms even within the same molecule. This complexity likely contributes to the success and stability of this highly conserved ancient host defense system in the constant fight against microbes.

MATERIALS AND METHODS

Bacterial strains.

E. coli (O26:H-) DSM 8695 and Acinetobacter baumannii DSM 30007 were obtained from Deutsche Sammlung von Mikroorganismen und Zellkultur GmbH (Braunschweig). Salmonella enteritidis was provided by the Department for Laboratory Medicine at Robert-Bosch-Hospital Stuttgart. All E. coli MC1000 strains were kindly provided by the lab of Jean-Francois Collet (Université Catholique de Louvain, Brussels, Belgium). E. coli Nissle strains were obtained from Tobias Oelschläger (Institute for Molecular Infection Biology, University of Würzburg, Würzburg, Germany). E. coli JM83 strains were provided by Julia Frick (Institute of Medical Microbiology and Hygiene, Tübingen, Germany).

Bacteria were grown in tryptic soy broth (TSB) or on LB agar plates overnight at 37°C. Anaerobic bacteria were cultivated using AnaeroGen pouches (Oxoid, UK) to generate an anaerobic atmosphere. For the E. coli mutants, it was necessary to add antibiotics to the medium at the following final concentrations: kanamycin (Kan) and ampicillin (Amp), 50 μg/ml; chloramphenicol (Cm), 25 μg/ml.

Peptides.

Reduced hBD1 (hBD1red) was chemically synthesized by Emc Microcollections (Tuebingen, Germany). Oxidized hBD1 (hBD1ox) was ordered from Peptide Institute, Japan.

RDA.

The antimicrobial activity of peptides was tested with an established method, a radial diffusion assay (RDA). The assay was modified from that described in reference 38 and performed as described earlier (10). Bacteria were grown (anaerobic bacteria with AnaeroGen [Oxoid, UK]) for 16 h in liquid TSB medium. Fresh log-phase bacteria were washed with 10 mM sodium phosphate buffer, and 4 × 10⁶ CFU/ml bacteria were used for the killing assay. Incubation of bacteria with peptides was performed in 10 ml of 10 mM sodium phosphate buffer (pH 7.4) containing 3 mg of powdered or liquid TSB medium and 1% (wt/vol) low EEO-agarose (AppliChem) under anaerobic (anaerobic bacteria with AnaeroGen [Oxoid]) or aerobic conditions with 2 or 4 μg of synthetic, oxidized hBD1 and 2 or 4 μg of synthetic, reduced hBD1 for 3 h at 37°C. After incubation the plates were overlaid with 10 ml sterile liquid agar containing 6% (wt/vol) TSB powder, 1% agar, and 10 mM sodium phosphate buffer (overlay gel). After 16 h at 37°C (anaerobic bacteria were cultivated using AnaeroGen pouches), the diameter of the inhibition zone was measured.

Flow cytometry assay.

To determine whether the periplasmic proteins are responsible for the permeabilization of the bacterial membrane, we used approximately 1.5 × 10⁶ CFU log-phase E. coli MC1000 ΔdsbA ΔdsbB in a final volume of 95 μl TSB (1:6 diluted in H₂O). We incubated these bacteria with peptide concentrations of 50, 75, and 95 μg/ml in a final volume of 10 μl for 1 h at 37°C. After adding 1 μg/ml of the membrane potential-sensitive dye DiBAC₄(3) [bis-(1,3-dibutylbarbituric acid)trimethine oxonol] (Thermo Fisher Scientific, USA), the suspension was incubated for 10 min at room temperature. After centrifugation (5 min at 4°C and 7,000 rpm), the supernatant was removed and bacteria were diluted in 300 μl phosphate-buffered saline (PBS). Using a FACSCalibur flow cytometer (BD, Sparks, MD, USA), the percentage of depolarized fluorescent bacteria was determined (33).

Plasmid induction experiments in the RDA.

For plasmid induction experiments, we used E. coli MC1000 ΔdsbA ΔdsbB, which harbors plasmids pQE60::dsbB and pBAD33::dsbA. Induction experiments were performed in TSB medium, and E. coli MC1000 ΔdsbA ΔdsbB (pQE60::dsbB, pBAD33::dsbA) was grown at 37°C under antibiotic selection without isopropyl-β-d-thiogalactopyranoside (IPTG) and l-arabinose. All induction experiments were also performed in TSB medium at 37°C under antibiotic selection (chloramphenicol [Cm], 25 μg/ml; ampicillin [Amp], 200 μg/ml). On the next day, we inoculated fresh TSB medium at an optical density at 600 nm [OD₆₀₀] of 0.01. During the incubation time, every 30 min we added 0.4% l-arabinose and 2 mM IPTG in 10 ml TSB medium. Radial diffusion assay was started when bacteria reached an optical density at 600 nm of 0.2 to 0.4. Finally, we added the antibiotics, 0.8% l-arabinose, and 4 mM IPTG in the overlay gel.

Scanning electron microscopy of bacteria.

Scanning electron microscopy was based on protocols described by Schroeder et al. (31). Briefly, around 1.2 × 10⁹ CFU/ml of WT E. coli MC1000 and MC1000 ΔdsbA ΔdsbB was incubated with 200 μg/ml hBD1ox for 2 h at 37°C. Bacteria were centrifuged, and the pellet was fixed in Karnovsky's reagent. Bacteria were washed with PBS and additionally fixed with 1% OsO₄ in H₂O. Samples were dehydrated to 100% ethanol, critical-point dried from CO₂, and analyzed by scanning electron microscopy at the Max Planck Institute for Developmental Biology (Tuebingen, Germany). Analyses of the bacterial surface were done visually by four independent experts. Blebs were counted by using 10 completely displayed bacteria.

Immunogold labeling and transmission electron microscopy.

Immunoelectron microscopy (IEM) was performed as previously described (39). We used around 1.2 × 10⁹ CFU/ml of WT E. coli MC1000 and MC1000 ΔdsbA ΔdsbB. Bacteria were incubated with 200 μg/ml peptide for 2 h at 37°C. Treated and untreated E. coli cells were centrifuged, and the pellet was fixed with 3.0% paraformaldehyde and 0.01% glutaraldehyde, followed by an additional centrifugation step, and the resulting pellet was embedded in 4% agarose at 37°C and then cooled to room temperature. Small parts of agarose blocks were embedded in Lowicryl K4M (Polysciences, Germany). The blocks were cut with an ultramicrotome (Ultracut; Reichert, Vienna, Austria). Ultrathin sections (30 nm) were mounted on Formvar-coated nickel grids and incubated with rabbit-anti-hBD1ox (Cell Concepts, Germany). Finally, the samples were incubated with goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories, Germany), conjugated with 6-nm gold. Samples were examined using a Zeiss Libra 120 transmission electron microscope (Zeiss, Oberkochen, Germany) operating at 120 kV. Gold-labeled hBD1ox (black points) was counted in bacterial periplasm or cytosol per bacterium and correlated with the total number of bacteria in percent by three different experts (WT, n = 39; ΔdsbA ΔdsbB mutant, n = 30).

Statistical analysis.

Differences between E. coli MC1000 wild-type and E. coli MC1000 mutant strains were analyzed using t tests. All experiments were replicated at least three times. All results are displayed as mean values ± standard errors of the means (SEM). P values of <0.05 were considered statistically significant. Data were analyzed using GraphPad Prism 5.