Role of lipid-bound peptidoglycan precursors in the formation of pores by nisin, epidermin and other lantibiotics


Hans-Georg Sahl E-mail; Tel. (228) 287 5704; Fax (228) 287 4808.


It is generally assumed that type A lantibiotics primarily kill bacteria by permeabilization of the cytoplasmic membrane. As previous studies had demonstrated that nisin interacts with the membrane-bound peptidoglycan precursors lipid I and lipid II, we presumed that this interaction could play a role in the pore formation process of lantibiotics. Using a thin-layer chromatography system, we found that only nisin and epidermin, but not Pep5, can form a complex with [14C]-lipid II. Lipid II was then purified from Micrococcus luteus and incorporated into carboxyfluorescein-loaded liposomes made of phosphatidylcholine and cholesterol (1:1). Liposomes supplemented with 0.05 or 0.1 mol% of lipid II did not release any marker when treated with Pep5 or epilancin K7 (peptide concentrations of up to 5 mol% were tested). In contrast, as little as 0.01 mol% of epidermin and 0.1 mol% of nisin were sufficient to induce rapid marker release; phosphatidylglycerol-containing liposomes were even more susceptible. Controls with moenomycin-, undecaprenol- or dodecaprenolphosphate-doped liposomes demonstrated the specificity of the lantibiotics for lipid II. These results were correlated with intact cells in an in vivo model. M. luteus and Staphylococcus simulans were depleted of lipid II by preincubation with the lipopeptide ramoplanin and then tested for pore formation. When applied in concentrations below the minimal inhibitory concentration (MIC) and up to 5–10 times the MIC, the pore formation by nisin and epidermin was blocked; at higher concentrations of the lantibiotics the protective effect of ramoplanin disappeared. These results demonstrate that, in vitro and in vivo, lipid II serves as a docking molecule for nisin and epidermin, but not for Pep5 and epilancin K7, and thereby facilitates the formation of pores in the cytoplasmic membrane.


Lantibiotics are gene-encoded peptides that contain the rare thioether amino acids lanthionine and/or 3-methyllanthionine and that possess antibacterial activity against Gram-positive bacteria. On the basis of differences in their structures and modes of action, two types of peptides have been distinguished, type A and type B lantibiotics. A typical type A lantibiotic is a flexible elongated peptide with a net positive charge, whereas type B lantibiotics are rigid globular molecules that carry either no net charge or a net negative charge (Jung, 1991). The prototype type A lantibiotic is nisin, which is produced by Lactococcus lactis ssp. lactis and is widely used as a food preservative. The nisin-like peptide subtilin, which was isolated from Bacillus subtilis ATCC 6633 and the lantibiotics epidermin, Pep5 and epilancin K7, all produced by Staphylococcus epidermidis, are other members of this group (Sahl et al., 1995; Sahl and Bierbaum, 1998).

Type A lantibiotics exert their antibacterial action by formation of transient pores in the energized cytoplasmic membrane of a sensitive bacterium. This process leads to dissipation of the membrane potential and promotes a rapid efflux of small metabolites, for example amino acids and nucleotides, which in turn abruptly stops all cellular biosynthetic processes (Sahl and Brandis, 1982; 1983; Ruhr and Sahl, 1985; Sahl, 1985). The mechanism of pore formation has been investigated in a number of recent studies that used various physiological and artificial membrane systems (Benz et al., 1991; Driessen et al., 1995; Demel et al., 1996; Giffard et al., 1996; Moll et al., 1996; van den Hooven et al., 1996a,b). NMR data indicate that, upon association with membrane micelles, nisin molecules adopt a rod-like amphipathic conformation such that the hydrophobic amino acid side-chains are immersed in the hydrophobic core of the membrane, whereas the hydrophilic groups interact with the phospholipid head groups (van den Hooven et al., 1996a,b). As lantibiotics are small peptides that can span the membrane only once, it is assumed that several molecules have to associate with the membrane in order to form a pore; however, data that describe the oligomerization process are not yet available. Early experiments demonstrated that a membrane potential is essential for pore formation (Ruhr and Sahl, 1985; Sahl, 1985) in that it forces the peptides into a conducting state, most probably perpendicular to the membrane surface. According to the current wedge model, the peptides insert into the membrane without losing contact with the membrane surface and thereby distort the lipid bilayer to form a short-lived hole (Moll et al., 1996). Although this model is able to explain many of the data obtained with model membrane systems, some in vivo effects still remain enigmatic; for example, the considerable differences in specific activity of a particular lantibiotic against various strains of related or even the same bacterial species.

Type B lantibiotics exert their action by binding to specific membrane lipids. The complex of cinnamycin and phosphatidylethanolamine has been recently characterized in detail (Hosoda et al., 1996), and mersacidin was shown to inhibit cell wall biosynthesis by forming a complex with lipid II (undecaprenyl-pyrophosphoryl-MurNAc-(pentapeptide)-GlcNAc), the ultimate monomeric, membrane-bound peptidoglycan precursor (Brötz et al., 1998). Interestingly, in the early lantibiotic literature, in vitro inhibition of cell wall biosynthesis by the type A lantibiotics nisin and subtilin had been described (Linnet and Strominger, 1973) and was later shown to be due to the formation of a complex between nisin and lipid I or lipid II (Reisinger et al., 1980). It was concluded that the inhibition of cell wall biosynthesis would eventually lead to cell death. However, pore formation causes complete cessation of all biosynthetic processes within minutes and therefore overrides killing by inhibition of cell wall biosynthesis, which is a comparatively slow process. Moreover, inhibition of cell wall biosynthesis in vitro required 10 μM of nisin or higher concentrations, whereas many bacterial strains have nanomolar minimal inhibitory concentrations (MICs) for this lantibiotic (see Discussion). Nevertheless, the interesting observation that preincubation of cells with the N-terminal fragment of nisin, nisin 1-12, inhibits the antibiotic action of nisin (Chan et al., 1996) indicates that nisin may interact specifically with a particular component of the cytoplasmic membrane. As an interaction between nisin and lipid II has already been described, these results prompted us to study lantibiotics further with respect to their affinity for lipid II and to investigate the effect of the presence of lipid II on the ability of nisin and epidermin to form pores.


Epidermin inhibits peptidoglycan biosynthesis in vitro

As reported previously for nisin (Reisinger et al., 1980), epidermin displayed a strong inhibitory effect on in vitro peptidoglycan biosynthesis. In a cell-free system based on a crude membrane preparation of Gaffkya homari, epidermin completely prevented the formation of polymeric peptidoglycan from the soluble cytoplasmic precursors UDP-N-acetylmuramic acid-[14C]-pentapeptide (UDP-MurNAc) and UDP-N-acetylglucosamine (UDP-GlcNAc) at a concentration of 25 μg ml−1 (Fig. 1).

Figure 1.

. Effects of epidermin and nisin on in vitro peptidoglycan synthesis in G. homari ATCC 10400 from the precursors UDP-MurNAc-[14C]-pentapeptide and UDP-GlcNAc. The amounts of polymeric peptidoglycan (•) and lipid-bound peptidoglycan intermediates (▴) formed in the presence of epidermin (—) or nisin (—) are indicated as a percentage of a control sample in the absence of inhibitors. The 100% value corresponds to 14.2 pmol of peptidoglycan and to 0.24 pmol of lipid intermediate per μg of membrane protein respectively.

In the presence of either epidermin or nisin the decrease in peptidoglycan synthesis was accompanied by a strong accumulation of lipid-coupled peptidoglycan intermediates in the membranes. The peptidoglycan biosynthesis cycle comprises two lipid-bound intermediates. The first, lipid I, is formed by the transfer of the phospho-MurNAc-pentapeptide group from UDP-MurNAc-pentapeptide to the membrane-bound lipid carrier undecaprenyl-pyrophosphate (see also Fig. 2). The second lipid intermediate, lipid II, results from coupling of a GlcNAc residue to the MurNAc group of lipid I. In the in vitro G. homari system it was not possible to identify which of the two lipid intermediates accumulated in the presence of epidermin; owing to the use of UDP-MurNAc-[14C]-pentapeptide as a label, the marker had already been incorporated into lipid I. Thus, we determined the activity of epidermin in a second peptidoglycan synthesis assay, which is based on a membrane fraction of Bacillus megaterium and which allows a distinction to be made between the lipid intermediates by the use of different labelled substrate combinations (Table 1). Use of labelled GlcNAc instead of labelled MurNAc (system 2) demonstrated that the amount of lipid II formed in the presence of epidermin and nisin is reduced to 44%. In system 3, UDP-GlcNAc was omitted from the test mixture and UDP-MurNAc-[3H/14C]-pentapeptide was the only substrate present; under these conditions cell wall biosynthesis is halted after formation of lipid I. In the presence of both lantibiotics, a massive accumulation of lipid I (Table 1) was observed in system 3, which demonstrates that the biosynthesis of this compound proceeded unhindered. In conclusion, the accumulation of lipid I and the inhibition of formation of lipid II indicated that, in the in vitro system, both lantibiotics interfered with the conversion of lipid I to lipid II. The lipoglycopeptide ramoplanin, which was tested for comparison, displayed the same effect (Table 1; see also Somner and Reynolds, 1990); in contrast, the lantibiotic mersacidin, which blocks peptidoglycan synthesis at the subsequent transglycosylation reaction (Brötz et al., 1997), did not impede the formation of lipid II.

Figure 2.

. Model of peptidoglycan biosynthesis and central role of the precursors lipid I and lipid II. Positioning of the bactoprenol moiety is schematic; its actual location in a bacterial membrane is not known. M, N-acetylmuramic acid; G, N-acetylglucosamine; P, phosphate; shaded circles represent the pentapeptide chain. Antibiotics mentioned in this study (mersacidin, vancomycin, ramoplanin, bacitracin) and their sites of action are depicted (–▸).

Table 1. . Effects of epidermin and nisin on in vitro peptidoglycan synthesis in B. megaterium. a. All lantibiotics were used at a concentration of 100 μg ml−1.b. System 1, 2 or 3 indicates the use of different labelled substrate combinations for peptidoglycan biosynthesis, which were (i) UDP-MurNAc-[3H/14C]-pentapeptide (labelled in the diaminopimelic acid residue) and UDP-GlcNAc; (ii) UDP-MurNAc-pentapeptide and UDP-[14C]-GlcNAc and (iii) UDP-MurNAc-[3H/14C]-pentapeptide respectively.c. The amounts of peptidoglycan and lipid intermediate formed in vitro are listed in percentage of an untreated control.Thumbnail image of

Epidermin forms a complex with both membrane-bound peptidoglycan intermediates

Reisinger et al. (1980) have shown that nisin inhibits peptidoglycan biosynthesis by binding to lipid I and lipid II. The complex of nisin with lipid II is not detected in the above in vitro peptidoglycan biosynthesis system because isolated membranes are used for biosynthesis of peptidoglycan, and both lipid precursors are easily accessible to the lantibiotic. Therefore, the interaction of nisin with lipid I prevents the synthesis of lipid II and hence formation of a complex between nisin and lipid II. However, when added to living cells, lantibiotics first interact with the outside layer of the cytoplasmic membrane and may only face the inner layer in the course of pore formation; thus lipid II may be the more important interaction partner for the lantibiotics as, in vivo, it is present on the outside, whereas lipid I is only accessible from the inside (see also Fig. 2). Therefore, we applied the same thin-layer chromatography system that had been used for detection of the complex of nisin and the membrane-bound peptidoglycan precursors to epidermin in order to test whether there is an interaction between epidermin and lipid II. Instead of crude membrane fractions we used purified [14C]-lipid II. Figure 3 directly demonstrates the interaction of the lantibiotics and of ramoplanin with the second lipid intermediate. When trapped in a complex with for example nisin, lipid II was retained at the site of application and was clearly resolved from non-complexed lipid II that migrated at an Rf-value of ≈0.7. Epidermin and ramoplanin also immobilized lipid II, thereby demonstrating that their effect on peptidoglycan synthesis could be equally based on the interaction with lipid intermediates. The amount of [14C]-lipid II immobilized by epidermin was smaller than that trapped by nisin, which may be because the assay conditions were optimized for the latter (Reisinger et al., 1980). The fact that it was at all possible in this system to document complex formation of lipid II with both lantibiotics as well as with ramoplanin suggests that, at least qualitatively, similar interactions may be involved. In contrast, the complex between the glycopeptide vancomycin and lipid II was not stable under these conditions (Fig. 3), although it has been well established to occur by other methods (Sheldrick et al., 1978). This complex is stable in aqueous solution and appears to dissociate in the organic solvent system applied here. Also Pep5, another pore-forming type A lantibiotic, did not influence the migration behaviour of lipid II in this system (Fig. 3). Taken together, both the cell-free peptidoglycan synthesis assays (Fig. 1, Table 1[link]) and the thin-layer chromatography system (Fig. 3) demonstrate that the lantibiotics nisin and epidermin can form a complex with both lipid I and lipid II, which differ from each other only by a GlcNAc residue.

Figure 3.

. Complex formation of epidermin and nisin with [14C]-lipid II. Purified [14C]-lipid II was incubated with either vancomycin (lane 1), ramoplanin (lane 2), nisin (lane 3), epidermin (lane 4), Pep 5 (lane 5) or without any supplement (lane 6). The mixture was separated using thin-layer chromatography; the autoradiograph is shown. The site of application (position of complexed lipid II) and of free lipid II are marked; purified lipid II was partially unstable and minor bands occurred after prolonged storage. Incubation buffer and solvent system were as optimized by Reisinger et al. (1980) for the studies on nisin.

Pore formation by nisin and epidermin in ramoplanin-pretreated cells

We have previously shown that incubation of Micrococcus luteus or Staphylococcus simulans with ramoplanin drastically reduces the number of mersacidin binding sites, i.e. the number of lipid II molecules on the outer layer of the cytoplasmic membrane (Brötz et al., 1998). This effect can be primarily attributed to the interference of the lipopeptide antibiotic with the formation of lipid II by interaction with lipid I (Somner and Reynolds, 1990), although the above results clearly show that it also binds to lipid II. Because of these properties we used ramoplanin to reduce the accessibility to lipid II and to test whether this could affect the ability of nisin and epidermin to induce leakage of radiolabelled glutamic acid from intact growing cells. Similar amino acid efflux experiments had been instrumental in proving that in vivo pore formation is the primary mode of action of a number of lantibiotics (Sahl and Brandis, 1983; Ruhr and Sahl, 1985; Schüller et al., 1989). When M. luteus, which has a nisin MIC of 12 nM, was treated with 25 nM nisin, the radiolabel was rapidly released from the cells upon addition of the lantibiotic. In contrast, pretreatment with ramoplanin prevented nisin-induced leakage of glutamate (Fig. 4). Similar results were obtained with S. simulans 22 which has a nisin MIC of 0.75 μM.

Figure 4.

. Inhibition of nisin-induced [3H]-glutamate efflux from Micrococcus luteus cells by preincubation with ramoplanin. Cells grown and treated as described in Experimental procedures were either supplemented with 0.2 μg ml−1 ramoplanin (⊠) or run as a control (×). Cells were filtered and assayed for content of [3H]-glutamate; the MIC of M. luteus for nisin is 12 nM.

A closer inspection of the phenomenon revealed that the ramoplanin effect is only detectable in concentrations close to the MIC value. In general, efflux of cytoplasmic material can already be demonstrated with lantibiotic concentrations below the MIC (Sahl and Brandis, 1983), e.g. with 0.1 μM nisin in Fig. 5. At such low concentrations, ranging from well below to up to several times the MIC, the membrane poration could be prevented by ramoplanin; when the concentrations of the lantibiotic were further increased, the protective effect of ramoplanin disappeared. These results suggest that at some stage of the pore formation process, lipid II is involved, and that this effect is relevant in vivo as it occurs at lantibiotic concentrations that correlate with the MIC. When we replaced ramoplanin by mersacidin, an acceleration rather than an inhibition of pore formation was observed, indicating that through the inhibition of transglycosylation and concomitant accumulation of lipid II in the membrane by this lantibiotic further binding sites for nisin and epidermin may be provided.

Figure 5.

. Influence of pretreatment of Staphylococcus simulans 22 with ramoplanin (2 μg ml−1) on the efflux of [3H]-glutamate at various concentrations of nisin; the MIC of nisin for S. simulans 22 is 0.75 μM. ⊠, cells pretreated with ramoplanin; * controls; further details see Fig. 4.

Pore formation in non-energized liposomes

Several lines of evidence suggest that pore formation by type A lantibiotics is driven by the proton-motive force (for reviews see Benz et al., 1991; Sahl, 1991; Sahl et al., 1995). For example, when membrane vesicles are energized through a valinomycin-induced potassium diffusion potential (e.g. Breukink et al., 1997) or through the activity of a reconstituted proton pump (Gao et al., 1991), they become sensitive to nisin and release markers. In contrast, no leakage of incorporated markers was observed from non-energized liposomes made of asolectin or defined mixtures of phospholipids, although the lantibiotics interacted with the vesicle membranes and influenced their physical properties (e.g. Kordel et al., 1989). Binding of the amphiphilic cationic peptides was found to be greatly enhanced when an excess of anionic phospholipids were incorporated into the membranes (e.g. Breukink et al., 1997). However, in none of these studies was a specific interaction of the lantibiotics with lipid II taken into consideration.

We prepared carboxyfluorescein-loaded liposomes from phosphatidylcholine (PC) and cholesterol and incorporated 0.05 or 0.1 mol% of purified lipid II. In these experiments we did not include negatively charged phospholipids for liposome formation to minimize unspecific adsorption of the cationic peptides due to mere ionic interaction. Control liposomes without lipid II were stable and did not release the fluorescent dye when increasing amounts of nisin, epidermin, Pep5 or epilancin K7 were added (in Fig. 6, a representative nisin control is shown). All lantibiotics were tested in concentrations of at least up to 5 mol% with regard to total lipid; in a previous study with pure dioleoylphosphatidylcholine liposomes (Kordel et al., 1989) it was observed that Pep5 had no effect, even in concentrations of up to 30 mol%, whereas nisin induced partial leakage at 5 mol%. In contrast, 0.05–0.1 mol% nisin was sufficient to induce significant dye efflux from liposomes containing as little as 0.05 mol% or 0.1 mol% membrane-associated lipid II (Fig. 6). In general, epidermin was even more effective and induced leakage at somewhat lower concentrations (0.02 mol%).

Figure 6.

. Carboxyfluorescein release from phosphatidylcholine/ cholesterol liposomes. A. Control liposomes without lipid II; a nisin control, representative for all tested lantibiotics is shown; the peptides were added stepwise (indicated by gaps in the traces) and cumulative concentrations (in mol% of total phosphatidylcholine and cholesterol) are given. B. Liposomes supplemented with 0.1 mol% lipid II; dye release, as indicated by fluorescence increase, started with 0.02 mol% epidermin and 0.1 mol% nisin, respectively, and was not detectable with up to 5 mol% of Pep5 or epilancin K7.

Various control experiments were performed to exclude the possibility that lipid II-containing liposomes are intrinsically less stable and to test whether the effect is specific for nisin and epidermin or whether it occurs with pore-forming lantibiotics in general. Whereas nisin and epidermin share a homologous N-terminal double-ring system, the bridging pattern of Pep5 differs over the entire molecule; moreover, it contains a total of eight positive charges, whereas nisin and epidermin have only three. The effect of Pep5 on the liposomes was comparatively small; ≈ 20 times higher concentrations than with epidermin were needed to induce slow the onset of dye efflux, which was, however, not observed with the control liposomes devoid of lipid II. In this respect, it could be relevant that incorporation of lipid II results in an increase in surface charge, as two negative charges are introduced per molecule of lipid II incorporated in the membrane; this could attract significantly more Pep5 molecules via ionic forces. Epilancin K7 shares general structural features with Pep5 in that it is highly charged and has only three rings; on the other hand, its C-terminal double-ring system is closely related to the nisin ring system and even identical to the subtilin rings (for review on structures see, e.g. Sahl et al., 1995; Sahl and Bierbaum, 1998). However, epilancin K7 showed as little dye release activity as Pep5 (Fig. 6), indicating that the C-terminal part of nisin may not be so relevant for the specific permeabilization of the lipid II-containing liposomes.

In further controls lipid II was replaced by moenomycin, a potent inhibitor of bacterial transglycosylases. Its antibiotic effect is based on its structural relatedness to lipid II, which results in effective competition for the active site of the cell wall-synthesizing enzymes (for a review see van Heijenoort et al., 1988). However, the moenomycin-doped liposomes remained as stable as the controls when nisin or epidermin was added (Fig. 7), demonstrating the specificity of the interaction. Also, incorporation of the commercially available undecaprenol or dodecaprenolphosphate did not render the liposomes susceptible to the lantibiotics. We also tried to antagonize nisin- and epidermin-induced leakage by blocking parts of the lipid II molecule. The binding of mersacidin to lipid II is presumed to mainly include the disaccharide moiety of lipid II (Brötz et al., 1998), whereas vancomycin renders the peptide side-chain inaccessible by binding to the terminal D-Ala–D-Ala. However, neither antibiotic was able to prevent dye release (Fig. 7), a finding which is in agreement with the observation that, in contrast to ramoplanin, mersacidin could not prevent pore formation by nisin in intact cells (see above). Blocking pore formation with ramoplanin, in analogy to the in vivo experiments, was not meaningful as the lipopeptide itself destabilized the liposomes, particularly in the presence of lipid II, further confirming the observation that it binds to both lipid I and lipid II.

Figure 7.

. Carboxyfluorescein release from control liposomes. A. Effect of nisin, epidermin and epilancin K7 on liposomes supplemented with 0.1 mol% moenomycin instead of lipid II. B and C. Liposomes containing 0.1 mol% lipid II were pretreated with vancomycin (left) and mersacidin (right) before the addition of epidermin (B) and nisin (C). D. Nisin activity on liposomes with negative surface charge (composed of 25 mol% PG, 25 mol% PC and 50 mol% cholesterol) supplemented with 0.1 mol% lipid II (top) or without lipid II (bottom); further details see Fig. 6.

As previous work (Breukink et al., 1997) had demonstrated the importance of PG for mediating nisin activity on membranes, we tested liposomes made of 25 mol% phosphatidylglycerol (PG), 25 mol% PC and 50 mol% cholesterol, thus presenting 50% negatively charged phospholipid head groups on the surface. Owing to enhanced binding of nisin, we expected such liposomes to be even more susceptible. Indeed, when containing 0.1 mol% lipid II, dye release was already detectable after the addition of 0.01 mol% nisin (Fig. 7D) compared with 0.1 mol% with the net neutral PC liposomes (Fig. 6B). Controls without lipid II were not rendered leaky with up to 5 mol% nisin tested (Fig. 7D).


The type A lantibiotics, like so many other amphiphilic antibiotic peptides that have been described as unmodified bacteriocins (Jack et al., 1995) or as plant and animal defence peptides (Boman, 1995), exert their activity by disruption of the barrier function of microbial cytoplasmic membranes. It is generally assumed that the peptides associate in oligodynamic aggregates that can form transient pores when oriented across the membrane. Several models have been proposed to describe such pores. Initially, an alamethicin-like barrel-stave model was suggested for type A lantibiotics (Sahl et al., 1987). Later, this was modified to a wedge model, primarily based on spectroscopical investigations in the presence of micelles (Driessen et al., 1995); a similar ‘wormhole’ model was proposed for the frog skin peptide magainin (Ludtke et al., 1996). None of these models takes into account specific interactions with an integral membrane component.

On the other hand, it has been suggested that, at least some, bacteriocins of Gram-positive bacteria (e.g. lactococcin A; van Belkum et al., 1991) may need receptors; these ideas were mainly inspired by the concept of the domain-based activity of colicins and by comparable narrow activity spectra displayed by some of the peptide bacteriocins of Gram-positive bacteria. However, attempts to identify receptors failed so far, possibly because proteins were mainly considered as candidates. Chan et al. (1996) observed a finite number of binding sites for nisin and specific antagonization of nisin activity by the inactive N-terminal nisin fragment nisin 1-12. As nisin acts on protein-free liposomes, these authors concluded that phospholipids, probably phospatidylglycerol, should represent the binding site. However, anionic phospholipids such as phosphatidylglycerol were found to stimulate pore-formation activity best when incorporated in high concentrations (50–60 mol%, Breukink et al., 1997).

The results presented here are well suited to explain some of the apparent discrepancies and also to unite into one mode-of-action model both early observations on nisin acting as a peptidoglycan biosynthesis inhibitor (Linnett and Strominger, 1973; Reisinger et al., 1980) and later findings that type A lantibiotics kill bacteria by membrane poration (Ruhr and Sahl, 1985; Sahl et al., 1987; Benz et al., 1991). Our studies strongly suggest that nisin and epidermin use lipid II as a docking molecule for specific binding to bacterial membranes. In addition, it seems possible that the energy necessary for moving the peptides into a conducting state (‘insertion into the membrane’) is substantially decreased after binding to the lipid II molecule, as is indicated by the fact that liposomes are susceptible without previous energization. We are currently studying this issue on black lipid planar bilayers and first results strongly support such an interpretation.

Is the role of lipid II in the pore formation process adequately described by the term ‘receptor’? Previous research using a number of artificial membrane systems has convincingly demonstrated that nisin and epidermin form pores in the absence of lipid II, i.e. it is not essential, whereas a receptor is usually indispensable for the activity of its ligand. Thus, a descriptive term such as ‘docking molecule’ seems more appropriate. A pore formation model that includes high-affinity interactions of the lantibiotics with an integral membrane component such as lipid II may help to explain better the considerable differences in susceptibility of Gram-positive bacteria. In comparison with the bulk phospholipids, the number of lipid II molecules per cell is much lower; estimates are in the range of 104 to 5 × 105 for different genera, and the molecule probably does not have unrestricted lateral mobility over the entire membrane. Rather, as a central carrier molecule, it should be associated with the enzymes that perform cell wall biosynthesis. It seems reasonable to assume that the membrane-associated steps of cell wall synthesis, with the lipid-bound precursors in various synthetic stages as substrates, are arranged in a highly organized macromolecular complex. Differences in the ultrastructure of such complexes may restrict the accessibility of the precursor for the lantibiotics. Thus, in Micrococcus, which has a comparatively high number of undecaprenol and undecaprenol-coupled molecules (5 × 105; Storm and Strominger, 1974), the precursor may also be easily accessible, resulting in high sensitivity and MIC values for epidermin and nisin in the nanomolar range; in this context it seems interesting to note that this genus is generally highly susceptible to antibiotics acting on the precursor in one or another stage, such as ramoplanin, mersacidin, vancomycin or bacitracin. In contrast, when the number of precursor molecules and the accessibility are low, pore formation may primarily take place at sites different from the cell wall synthesis complex, following a mechanism as proposed by the wedge model. This, probably in combination with other factors such as reduced phosphatidylglycerol content (Verheul et al., 1997), could then result in low susceptibility as observed with Listeria or Enterococcus.

Comparison of structural features of nisin, epidermin, Pep5 and epilancin K7 may give some hints about which parts of the lantibiotics and of lipid II, respectively, may interact. Apparently, the binding to lipid II is specific for nisin and epidermin, which share a homologous N-terminal ring pattern; Pep5 and epilancin K7, the latter shares a similar set up of the C-terminal rings with nisin, have low affinity for the precursor. This agrees well with the results of Chan et al. (1996), who identified the fragment nisin1-12 as a strong antagonizer of nisin binding to cells. With respect to lipid II, it appears that the interaction with epidermin and nisin may involve both the lipid moiety and the disaccharide unit or the peptide side-chain. This can be concluded from the observation that mersacidin or vancomycin, which bind to the latter parts of the precursor, cannot prevent pore formation by the lantibiotics; also, the affinity of nisin and epidermin seems to be the same for both lipid I and lipid II, which are distinguished by a mono-(MurNAc) and a disaccharide (MurNAc-GlcNAc) moiety respectively. On the other hand, the lipid moiety seems not sufficient to promote pore formation as indicated by the results with undecaprenol or dodecaprenolphosphate.

The observation that Pep5 seems not to use lipid II for pore formation is consistent with its comparatively low activity against M. luteus (MIC of 120 nM compared with 12 nM for nisin and 0.6 nM for epidermin, unpublished data for our laboratory indicator strain ATCC 4698). On the other hand, Pep5 kills S. simulans 22 and S. carnosus TM300 at 0.6 nM, whereas nisin and epidermin have MICs of 750 nM and 550 nM respectively (unpublished). Possibly, Pep5 uses a yet unidentified docking molecule that is more prominent or better accessible in these two staphylococcal species. In general, antibiotic activities in the nanomolar range are unusual among antibiotic peptides supposed to act by membrane destabilization and may only be possible when high-affinity interactions with integral components of bacterial membranes are involved. Designing antibiotic peptides with respect to optimum membrane disruption (e.g. on the basis of eukaryotic defence peptides such as magainins or cecropins) has so far not yielded peptides with MICs lower than in the order of 100 nM to 1 μM (e.g. Tossi et al., 1997), and increasing cytotoxicity of such peptides then becomes a serious problem.

Bactoprenol and the lipid-bound peptidoglycan precursors are prominent targets for a number of antibiotic substances, e.g. the glycopeptides vancomycin and teicoplanin, ramoplanin, tunicamycin, bacitracin and colicin M as well as the type B lantibiotics mersacidin and actagardine. In all these cases, the interaction of the antibiotic with the precursors eventually blocks cell wall synthesis. In contrast, the interaction of the type A lantibiotics nisin and epidermin with lipid II promotes pore formation, which kills cells rapidly so that the inhibition of the peptidoglycan biosynthesis, as postulated on the basis of in vitro experiments, does not come into effect. Thus, it appears to be a remarkable feature that binding of lantibiotics from two different subgroups to the same target molecule does have completely different consequences for the bacterial cell and results in entirely different modes of action. Future research must aim at further characterisation of the interaction of nisin and epidermin with lipid II to understand better the molecular mechanisms of this phenomenon.

Experimental procedures

In vitro peptidoglycan biosynthesis

Isolated membrane preparations were used for the synthesis of polymeric peptidoglycan from soluble peptidoglycan precursors. Conditions for the B. megaterium KM systems have been reported previously (Brötz et al., 1997; for differential labelling see also legend to Table 1). In vitro peptidoglycan synthesis in G. homari ATCC 10400 was performed essentially as described by Hammes and Neuhaus (1974); the modified assay mixture contained 50 mM Tris-HCl (pH 8), 42 mM magnesium acetate (pH 7), 8 mM NH4Cl, 1.7 mM dithioerythritol, 3.3 mM ATP, 1.14 UDP-GlcNAc, 0.04 mM UDP-MurNAc-[14C]-pentapeptide (7.9 mCi mmol−1, labelled in the lysine residue) and membranes corresponding to 44 μg of protein in a total volume of 30 μl. The assay mix was incubated for 45 min at 25°C.

Preparation of lipid II and complex formation with lantibiotics

[14C]-Lipid II (2.5 mCi mmol−1, labelled in the GlcNAc moiety) was synthesized in vitro by a membrane preparation of M. luteus ATCC 4698, extracted and purified as described previously (Brötz et al., 1998). Unlabelled lipid II was prepared using the same procedure but replacing the radiolabelled compounds by the non-labelled counterparts. To facilitate detection of unlabelled lipid II, we made use of its affinity for mersacidin; aliquots of the respective chromatographic fractions were added to mersacidin activity tests (105 cells of M. luteus ATCC 4698 in 100 μl of Mueller–Hinton broth, containing 2 μg ml−1 mersacidin, which corresponds to 10 times the MIC); growth of M. luteus under these conditions indicated antagonization of mersacidin by the lipid-bound cell wall precursor. Purified lipid II was stored in chloroform–methanol (1:1, vol) at −20°C. To study complex formation, labelled lipid intermediate (0.18 nmol) was dried in a desiccator and incubated for 1.5 h at 25°C with 2 nmol of either epidermin, nisin, Pep5, ramoplanin or vancomycin in 10 μl of 69 mM Tris-HCl (pH 8.8), 58 mM MgCl2, 11.6 mM NH4Cl, and 5.8 mM sodium dodecylsulphate. Samples were separated using thin-layer chromatography in n-butanol–acetic acid–pyridinium–water (15:3:12:10, v/v/v/v) as reported by Reisinger et al. (1980) and subjected to autoradiography for 3 months at −70°C.

Efflux experiments with intact cells

The principle set up of such experiments was described by Sahl and Brandis (1983) with some modifications introduced by Ruhr and Sahl (1985). In this study we used M. luteus ATCC 4698 and S. simulans 22 as test strains and adapted the protocol as follows. Cells were grown in tryptone soya broth (Oxoid), which was prepared at a lower concentration (10 g l−1) than recommended to reduce the ionic strength. After overnight growth (A600: 2–3) cells were diluted into fresh medium to an A600 of 0.5, further incubated to an A600 of ≈1 and then tested for uptake and efflux of [3H]-glutamic acid. Cells first received 100 μg ml−1 chloramphenicol to prevent incorporation of the amino acid; after 30 min, ramoplanin or mersacidin at concentrations of 10 times the respective MIC (0.2 μg ml−1 ramoplanin and 1 μg ml−1 mersacidin for M. luteus, or 2 μg ml−1 and 110 μg ml−1 for S. simulans) were added to manipulate the availability of lipid II. Immediately afterwards, the radiolabel was applied (1 μCi ml−1 culture; 30 nM glutamate) and samples (500 μl) taken to monitor the uptake. After ≈15 min, nisin and epidermin were added at varying concentrations to test for pore formation. In some experiments, we added ramoplanin just 1 min before the addition of the lantibiotics; this did not have any obvious effect on the efflux of glutamate, nor was its uptake altered in the presence of the lipopeptide.

Liposome experiments

We prepared liposomes composed of PC and cholesterol as used by Tomita et al. (1993) for studying staphylococcal α-toxin. Equimolar amounts of PC with variable chain length (average molecular weight 702 g mol−1) and cholesterol were dissolved in chloroform to a final concentration of 10 μmol each per ml, the solvent evaporated in a desiccator and the sample dried with N2. In some cases 25 mol% phosphatidylglycerol (PG) and 25 mol% PC were used instead of 50 mol% PC. The dried lipids were kept in the desiccator for another 30 min and then taken up in Tris-buffered saline (TBS, 0.85% NaCl, w/v in 10 mM Tris, pH 7.2) containing 0.1 M carboxyfluorescein (predisolved in a small volume of 1 M NaOH and corrected for pH with TBS). The mixture was heated to 55°C for 5 min and vigorously vortexed for 1 min; this cycle was repeated four times. The liposomes were then pelleted and washed with TBS at least five times, until the supernatant was colourless. For incorporation of lipid II or moenomycin, both were dissolved in chloroform–methanol (1:1) and added to the chloroform-dissolved PC–cholesterol mixture in final concentrations of 0.05 or 0.1 mol% with respect to total PC. Release of carboxyfluorescein was monitored as described previously (Kordel et al., 1989).

Bacterial strains and chemicals

M. luteus ATCC 4698 and S. simulans 22 and culture conditions have been described previously (Ruhr and Sahl, 1985). Chemicals for lipid II synthesis and sources for antibiotics are listed by Brötz et al. (1998). Cholesterol, PC, PG, dodecaprenolphosphate and undecaprenol were obtained from Sigma.


This work was supported by grants of the Deutsche Forschungsgemeinschaft (Sa 292/8-1) and through the BONFOR programme of the Medizinische Einrichtungen, University of Bonn. Parts of the work were also supported by the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie (grant 01KI 9705/8). We thank Hoechst AG and Merrel Dow/Lepetit for providing mersacidin and ramoplanin respectively.