Insight into structure-function relationship in phenol-soluble modulins using an alanine screen of the phenol-soluble modulin (PSM) α3 peptide

Abstract

Phenol-soluble modulins (PSMs) are a family of peptides with multiple functions in staphylococcal pathogenesis. To gain insight into the structural features affecting PSM functions, we analyzed an alanine substitution library of PSMα3, a strongly cytolytic and proinflammatory PSM of Staphylococcus aureus with a significant contribution to S. aureus virulence. Lysine residues were essential for both receptor-dependent proinflammatory and receptor-independent cytolytic activities. Both phenotypes also required additional structural features, with the C terminus being crucial for receptor activation. Biofilm formation was affected mostly by hydrophobic amino acid positions, suggesting that the capacity to disrupt hydrophobic interactions is responsible for the effect of PSMs on biofilm structure. Antimicrobial activity, absent from natural PSMα3, could be created by the exchange of large hydrophobic side chains, indicating that PSMα3 has evolved to exhibit cytolytic rather than antimicrobial activity. In addition to gaining insight into the structure-function relationship in PSMs, our study identifies nontoxic PSMα3 derivatives for active vaccination strategies and lays the foundation for future efforts aimed to understand the biological role of PSM recognition by innate host defense.—Cheung, G. Y., Kretschmer, D., Queck, S. Y., Joo, H.-S., Wang, R., Duong, A. C., Nguyen, T. H., Bach, T.-H., Porter, A. R., DeLeo, F. R., Peschel, A., Otto, M. Insight into structure-function relationship in phenol-soluble modulins using an alanine screen of the phenol-soluble modulin (PSM) α3 peptide.

Staphylococcus aureus is a notorious human pathogen, which can cause a multitude of diseases, ranging from moderately severe skin infections to potentially fatal diseases such as endocarditis, toxic shock syndrome, or necrotizing pneumonia (1). S. aureus infections are often difficult to treat, owing to the widespread presence of antibiotic-resistant strains and the lack of an effective S. aureus vaccine (2, 3). The severity of an S. aureus infection is at least in part determined by the infecting strain's repertoire of aggressive virulence determinants (4). Phenol-soluble modulins (PSMs) in particular have recently drawn much attention as major determinants of the capacity of S. aureus to evade killing by human immune cells and cause serious infections (5–7).

PSMs are a family of staphylococcal peptides that are secreted by a dedicated transport system (8) and grouped together based on similar physicochemical properties, namely the formation of an amphipathic α helix (6, 9). Due to their amphipathic properties, PSMs are believed to behave similar to biological detergents, resulting in the capacity to disrupt biological membranes (10), as shown for S. aureus δ-toxin (11), and molecular interactions in biofilms (12, 13). These properties also cause partition into the phenol phase during hot phenol extraction, after which the name PSMs was coined (14). Two PSM types can be distinguished: α-type PSMs are relatively short (∼20–25 aa) with the entire peptide forming an amphipathic α helix, while β-type PSMs are about double the size of α-type PSMs and contain the amphipathic α-helical part at their C-terminal half. PSMs are usually core genome encoded, and all strains of a given staphylococcal species, in particular those with high pathogenic potential, produce a species-specific repertoire of PSMs (15). The only exception is PSM-mec, which is encoded on a mobile genetic element (16). Strong expression of PSMs contributes to increased virulence, as shown for community-associated methicillin-resistant S. aureus (6, 17).

Several PSMs of the α-type, in particular PSMα3 of S. aureus, efficiently lyse neutrophils, monocytes, and erythrocytes and have a pronounced effect on disease outcome during experimental S. aureus infection (6, 7, 9). In addition, all PSMs have proinflammatory properties that are mediated primarily via interaction with the formyl peptide receptor 2 (FPR2; ref. 18). These include neutrophil chemotaxis, priming, and stimulation of cytokine release. Of note, interaction with FPR2 is not involved in the cytolytic activities of PSMs (18). This is in accordance with previous studies on δ-toxin, an α-type PSM, and artificial membranes, which indicated that PSMs disrupt membranes by a receptor-independent mechanism (19). Notably, PSMα peptides appear to exert their effect on pathogenesis to a large extent by lysing immune and other cell types after phagocytosis. This has been demonstrated for neutrophils and osteoblasts (7, 8, 20, 21). In human serum, the cytolytic and proinflammatory activities of PSMs are diminished by human serum lipoproteins (22). It is not yet clear whether PSMs are active in the extracellular environment in locations where lipoproteins are not as abundant, for example, during abscess formation or osteomyelitis, where they have a demonstrated crucial effect on disease progression (6, 23).

To learn about the structural determinants underlying the diverse biological activities of PSMs, we here analyzed an alanine substitution peptide library of PSMα3. Our results give important insights into the mechanisms of PSM receptor and membrane interaction, the biofilm-structuring activity of PSMs, and the role of the antimicrobial activities reported for PSMs (24, 25). Furthermore, by identifying peptide derivatives with only proinflammatory or only cytolytic activity, our study provides the basis for future experiments designed to gain a better understanding of the biological role of PSM recognition by FPR2. Finally, the discovery of nontoxic PSMα3 derivatives represents the 1st step to use this major cytolysin of S. aureus for active vaccination efforts similar to those undertaken with S. aureus α-toxin (26).

MATERIALS AND METHODS

Peptide synthesis and dilutions

PSM peptides were synthesized by commercial vendors at >95% purity. All peptides were synthesized with an N-terminal formyl group, unless otherwise noted.

Leukocyte assays

Human neutrophils were isolated from venous blood of healthy volunteers in accordance with protocols approved by the Institutional Review Board for Human Subjects at National Institute of Allergy and Infectious Diseases, U.S. National Institutes of Health, and the University of Tübingen, Germany, as described previously (27, 28). All donors signed the appropriate informed consent forms. HL60 cells stably transfected with human FPR1, FPR2, and FPR3 have been described recently (29, 30). These cell lines were grown in RPMI medium (Biochrom, Berlin, Germany) supplemented with 10% FCS (Sigma-Aldrich, St. Louis, MO, USA), 20 mM HEPES (Biochrom), penicillin (100 U/ml), streptomycin (100 μg/ml; Life Technologies, Frankfurt, Germany), and 1× Glutamax (Life Technologies). Transfected cells were cultivated in the presence of G418 (Biochrom) at a final concentration of 1 mg/ml. Surface expression of CD11b and Ca²⁺ flux measurements were performed as described previously (6). Briefly, for measurement of Ca²⁺ fluxes, leukocytes were loaded with 2 mM Fluo-3-AM (Molecular Probes, Life Technologies) for 20 min at room temperature under agitation, washed, and resuspended at 10⁶ cells/ml. Then, Ca²⁺ fluxes were analyzed with a FACScalibur (Becton Dickinson, Heidelberg, Germany). Lysis of neutrophils by synthetic PSMs was determined essentially as described previously (28, 31). Synthetic PSMs were added to wells of a 96-well tissue culture plate containing 10⁶ neutrophils, and plates were incubated at 37°C for 30 min. Neutrophil lysis was determined by release of lactate dehydrogenase (LDH; Cytotoxicity Detection Kit, Roche Applied Sciences, Penzberg, Germany).

Hemolysis assays

Hemolytic activities of PSM peptides were determined by incubating samples with human erythrocytes (2% v/v in Dulbecco's PBS) for 1 h at 37°C as described previously (6).

Biofilm assays

To determine the effect of PSM peptides on S. aureus biofilm development, the naturally Agr-dysfunctional strain S. aureus SA113 was used as indicator strain in microtiter plate assays that were performed as described previously (12).

Antimicrobial activity assays

To determine antimicrobial activities of PSM derivatives, Micrococcus luteus and Streptococcus pyogenes were used as test strains using agar diffusion assays as described previously (25).

Modeling of the PSMα3 structure

PSMα3 was modeled based on the published structure of S. aureus δ-toxin (protein databank file KAM2.pdb) using SYBYL7.3 by replacing the residues of δ-toxin residues with those of PSMα3. This was followed by a short energy minimization using the MMFF94 force field with MMFF94 atom types and charges. A distance-dependent dielectric was used, simulating solvent.

Circular dichroism (CD) measurements

The structures of synthetic PSM peptides were analyzed by CD spectroscopy on a Jasco spectropolarimeter (model J-720; Jasco, Inc., Easton, MD, USA). Solutions of PSM peptides, each at 1.0 mg/ml, were prepared in 50% trifluoroethanol. Measurements were performed in triplicate, the resulting scans were averaged and smoothed, and the buffer signal was subtracted.

Mouse peritonitis model and flow cytometry

CD-1 Swiss mice (age 6–wk; Charles River Laboratories, L'Arbresle, France) were injected peritoneally with 1 ml of 150 μM PSMα3 or derivatives diluted in sterile PBS. After 5 h of exposure, mice were euthanized by CO₂ inhalation. The peritoneal cavities of all animals were lavaged with 10 ml of cold PBS containing 0.5% BSA (w/v) to collect peritoneal leukocytes. Residual erythrocytes were lysed with ACK lysis buffer (Gibco, Carlsbad, Ca, USA), and live cells were then quantified using trypan blue exclusion.

For flow cytometry, washed cells (1–2×10⁶) were blocked with monoclonal mouse anti-CD16/CD32 antibody (BD Biosciences, San Jose, CA, USA) for 15 min and then stained with monoclonal anti-mouse CD45 (eBioscience, San Diego, CA, USA) for 30 min. Cells were washed twice and then fixed with fixation buffer (eBioscience) for 20 min. Stained cells were analyzed on a LSR2 Fortessa flow cytometer (Becton Dickinson) with FlowJo 9.6.4 software (Tree Star, Ashland, OR, USA). The number of leukocytes from each mouse peritoneum was calculated from the total number of cells as determined by trypan blue staining. All incubation steps were performed at 4°C in the dark.

Statistical analyses

Statistical analyses were performed using Graph Pad Prism 5 (GraphPad, San Diego, CA, USA), using 1-way ANOVA with Dunnett's posttest, comparing to the PSMα3 unchanged peptide control. Correlation analyses were performed using Pearson's test, and 2-tailed P values were computed. Error bars show means ± sd.

RESULTS

To gain insight into the structure-function relationship in PSMs, we used an alanine substitution peptide screen, which represents a common method to link biological activities to specific amino acid positions in a protein or peptide. We chose PSMα3 because it is the most strongly cytolytic PSM and has been shown to be by itself mostly responsible for the effect of PSMs on the lysis of human neutrophils (6). Furthermore, it has a significant contribution to all other phenotypes associated with PSMs, including the lysis of red blood cells, biofilm structuring, and proinflammatory activities (6, 13, 18, 32).

Proinflammatory activities

The known bacterial or endogenous FPR2 ligands are virtually unrelated (18), and it has not been established which structural features define an efficient FPR2 ligand. To gain insight into the structural requirements of PSMs as FPR2 ligands, we first analyzed the influence of amino acid substitutions in PSMα3 on activities that are receptor mediated. To that end, we used a HL60 cell line stably transfected with FPR2 or the related FPR1 or FPR3 receptors, which have only very low affinity for PSMs (18). Amino acid positions that affected Ca²⁺ flux as a readout of receptor activation predominantly included positively charged amino acids (K6A, K12A, and K17A substitutions) and the 2 C-terminal asparagine residues (N21A and N22A). In accordance with previous results, FPR1 or FPR3 caused only residual responses to PSMs as peptide concentrations had to be 40× higher to reach similar levels as with FPR2 (Fig. 1). The peptide positions that were important for FPR2 activation also led to a further reduction of the residual FPR1- or FPR3-mediated responses, suggesting that the 3 receptors share a similar mode of interaction with PSMs.

Figure 1.

Activation of formyl peptide receptors. PSMα3 alanine substitution peptide library was tested for activation of the formyl peptide receptors FPR1, FPR2, and FPR3 using a HL60 cell line stably transfected with one of the receptors. Nontransfected HL60 cells were used in a control experiment. Peptides were added to a final concentration of 25 nM (FPR2) or 1 μM (FPR1, FPR3) to achieve activities in the same measurement range. Control, buffer only. Bars corresponding to peptides showing significant differences compared with PSMα3 are shaded in gray. *P < 0.05; **P < 0.01; ***P < 0.001.

We then analyzed Ca²⁺ flux and surface expression of CD11b on human neutrophils to directly determine the effect of PSMα3 amino acid substitutions on the activation of that main target cell type of PSM activity. Overall, the results were very similar to those achieved with the transfected HL60 cell line, with the 2 C-terminal asparagine residues in particular and at least some of the lysine residues having a strong effect on PSMα3 activity (Fig. 2).

Figure 2.

Priming of human neutrophils. Induction of Ca²⁺ flux (A) and surface expression of CD11b (B) were used to measure priming of human neutrophils with the PSMα3 alanine screen peptide library. Bars corresponding to peptides showing significant differences compared with PSMα3 (black bar) are shaded in gray. Diagonal bar represents nonformylated PSMα3. Peptides were added to a final concentration of 100 nM for Ca²⁺ flux and 160 nM for CD11b expression experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

Taken together, our results identified 2 main groups of amino acids with a strong effect on the proinflammatory, receptor-mediated activities of PSMα3: the C terminus, in particular the C-terminal 2 asparagine residues N21 and N22, and 3 of the 4 lysine residues of the peptide (K6, K12, and K17).

Cytolytic activities

To analyze structural determinants affecting the cytolytic activities of PSMα3, we first measured lysis of human neutrophils. Amino acid positions in which exchange with alanine led to strongly decreased lysis of human neutrophils were the 3 lysine residues at positions 6, 12, and 17 (K6, K12, and K17) and the leucine in position 7 neighboring K6 (L7) (Fig. 3A). In addition, we determined lysis of erythrocytes (hemolysis). In contrast to previous experiments with sheep blood (6), we used human blood to better reflect pathogenesis in humans. L7 and K12 had the strongest effect on hemolysis, with a variety of other positions also influencing hemolysis to a lesser extent (Fig. 3B). Notably, the 2 C-terminal asparagine residues that strongly affected proinflammatory activities had no influence on the cytolysis of human neutrophils or erythrocytes.

Figure 3.

Cytolysis. Lysis of human neutrophils (A) was measured by release of LDH and lysis of human erythrocytes (B) by hemolysis assays. Peptides were added to a final concentration of 2.6 μM for neutrophils and 26 μM for erythrocytes. Bars corresponding to peptides showing significant differences compared with PSMα3 (black bar) are shaded in gray. *P < 0.05; **P < 0.01; ***P < 0.001.

Biofilm formation

We showed recently that PSM peptides contribute significantly to biofilm structuring and detachment in vitro and in vivo (12, 13). Accordingly, when we added PSMα3 to growing biofilm cultures of a PSM-deficient, naturally Agr-dysfunctional S. aureus biofilm-forming clinical isolate, we observed that increasing concentrations of PSMα3 reduce biofilm formation (Fig. 4). PSMα3 promoted biofilm formation at low concentrations, while higher concentrations strongly inhibited biofilm formation. Of note, amino acid positions that showed a strong effect on the biofilm phenotype were clearly different in pattern from those affecting cytolysis and proinflammatory capacity, and almost exclusively comprised amino acids with large hydrophobic side chains such as phenylalanine and leucine (Fig. 4).

Figure 4.

Biofilm formation. Biofilm formation of S. aureus SA113 (agr) on addition of different concentrations of PSMα3 variants was determined using a microtiter plate assay. Peptides were added at the time of inoculation from precultures to the given final concentration. Biofilms were measured after 24 h of static growth by safranin staining. Variant peptides showing a strong, moderate, and weak effect on biofilm formation are shown in red, orange, and black, respectively. Only the K9A variant showed strongly increased detachment capacity already at low concentrations (green). PSMα3 is shown in blue. de-f. PSMα3, N-deformylated PSMα3 (turquoise).

Antimicrobial activity

PSMs and PSM-like peptides have sporadically been described as antimicrobial. For example, synthetic derivatives of S. aureus δ-toxin and a naturally occurring PSMβ-like peptide termed gonococcal growth inhibitor, produced by S. haemolyticus, are inhibitors of microbial growth, sometimes in a species-specific fashion (25, 33). Furthermore, we previously found S. epidermidis δ-toxin, PSMδ, and processed derivatives of S. aureus PSMα1 and PSMα2 to have activity against S. pyogenes (24, 25). The latter results prompted us to determine potential antimicrobial activities of PSMα3 derivatives, despite the lack of antimicrobial activity of natural PSMα3 itself (25). We found that several PSMα3 derivatives had moderate to strong activities against the test strains M. luteus and S. pyogenes (Fig. 5). Strikingly, positions that affected antimicrobial activities were exclusively such that contained large hydrophobic side chains, in particular phenylalanine residues. These results indicate that large hydrophobic side chains in PSMα3 prevent antimicrobial activity. Interestingly, the pattern of amino acid positions that were found to prevent antimicrobial activity was very similar to that affecting biofilm structuring and detachment.

Figure 5.

Antimicrobial activity. Antimicrobial activities of PSMα3 derivatives were measured using agar diffusion tests on M. luteus and S. pyogenes test plates. Peptides were applied onto filter disks (200 μg each) and dried, filters were placed on the plates, and plates were incubated for 24 h before diameters were measured. Inhibition is expressed as diameter of inhibition zone with the filter diameter subtracted.

α-Helicity

We performed CD measurements to determine whether the degree of α-helicity, i.e., the capacity of the peptide to form an α helix, correlates with PSM-mediated phenotypes. Our results indicated that amino acids in the N-terminal region, in particular phenylalanine 3 and lysine 6, are more important for α-helicity than amino acids in the C-terminal part of PSMα3 (Fig. 6). Of note, we did not find a significant correlation between amino acid residues influencing α-helicity with those influencing other phenotypes. This suggests that the formation of an amphipathic α helix by itself is not sufficient for PSM molecules to achieve biological activities.

Figure 6.

Overview of PSMα3 residues important for cytolysis, proinflammatory activities, biofilm detachment, antimicrobial activities, and α-helicity. A) Table of relative effects. All data are expressed as percentage of effect with that of PSMα3 set to 100%. Increasing saturation of red indicates a stronger effect of the relative position on the phenotype (potential to stimulate Ca²⁺ flux, CD11b expression, neutrophil and erythrocyte lysis, leukocyte influx, capacity to prevent biofilm formation at 500 μg/ml, capacity to prevent antimicrobial activity, α-helicity). Darkest red was attributed to the amino acid exchange with the most pronounced effect in every column (phenotype). Occasionally occurring effect in the other direction is expressed in green, with the most pronounced effect in every column receiving the darkest green. α-Helicity was determined by CD measurements and the SELCON3 program. ND, not determined. B) Amino acid positions affecting cytolytic, proinflammatory, biofilm detachment, and antimicrobial activity depicted in a PSMα3 structural model. Left panel: hydrophilic side of PSMα3, where most amino acids are located that influence cytolytic and proinflammatory activities. Right panel: hydrophobic side of PSMα3, where most amino acid positions are located influencing biofilm detachment and antimicrobial activities. Purple is used for positions influencing both cytolytic and proinflammatory, or biofilm and antimicrobial activities. Red and blue are used for amino acid positions influencing primarily only one activity. N-terminal methionine is labeled for orientation.

In vivo chemotactic activities

We used selected PSMα3 derivatives to investigate whether in vivo neutrophil chemotactic activities follow our in vitro observations on proinflammatory activities. To that end, we employed a mouse peritonitis model as used before to analyze the effect of wild-type and psmα deletion mutants of S. aureus on leukocyte influx. The total numbers of leukocytes that had migrated into the peritoneal cavity differed significantly depending on which PSMα3 derivative was injected (Fig. 7). Notably, influx was in general correlated with the in vitro readouts of proinflammatory activities (Ca²⁺ flux and CD11b surface expression) in neutrophils: the K6A, K12A, and N21A derivatives caused strongly reduced influx compared with the L7A and K17A derivatives, which did not promote leukocyte influx that was significantly reduced compared with that caused by PSMα3 (Fig. 7). These findings are in accordance with our previous observations showing that neutrophil chemotaxis is triggered by PSM interaction with the FPR2 receptor (18) and demonstrate that PSMα3 derivatives with reduced proinflammatory activities as determined in vitro cause less leukocyte chemotaxis in vivo. The percentage of dead leukocytes that we observed in these experiments was overall low, likely due to the strong dilution effect in the peritoneal cavity and in accordance with previous findings indicating that sequestration by serum lipoproteins inhibits PSM cytolytic activities (22).

Figure 7.

In vivo neutrophil chemotaxis. Influx of neutrophils into the peritoneal cavities of mice was determined 5 h after injection of PSMα3 or derivatives and analyzed by trypan blue staining and flow cytometry. *P < 0.05; **P < 0.01; ***P < 0.001; N.S., not significant.

DISCUSSION

In the present study, we analyzed the importance of amino acid positions in PSMα3 to gain insight into the structure-function relationship of PSM peptides. Amino acid sequences of PSMs are only barely conserved. However, we believe that general observations, such as those regarding the effects found with substitutions of cationic or hydrophobic amino acids, are likely also applicable to other PSMs, owing to the fact that PSMs share a common 3-dimensional structure and physicochemical features. The most important findings are illustrated in Fig. 6. Most strikingly, cytolytic and proinflammatory capacities were both strongly affected by substitution of the lysine residues in positions 6 and 12, whereas the C-terminal amino acids (N21, N22) were exclusively important for proinflammatory activities. In contrast, leucine 7 and lysine 17 were only crucial for cytolytic activity. These observations are likely applicable to lysis of cell membranes both from the outside and after phagocytosis, as the membrane of, for example, the phagosome in neutrophils, originates from invagination and the mode of action of PSMs is receptor independent (18).

As for PSM proinflammatory activities, a modest, yet significant, effect of the N-terminal formyl group on receptor-mediated activities of some PSM peptides, including PSMα3, which was shown previously in Ca²⁺ flux and chemotaxis experiments (18), was confirmed in our study using CD11b expression as an additional measurement of neutrophil priming. The effect pattern of amino acid substitutions was significantly correlated among the receptor-mediated phenotypes (Ca²⁺ flux in HL60 cells, Ca²⁺ flux in neutrophils, CD11b expression in neutrophils; for all comparisons P≤0.027). Similarly, cytolytic phenotypes (hemolysis and neutrophil lysis) were correlated (P=0.0002). Notably, α-helicity as determined by CD measurements was not correlated with any other phenotype.

Recently, we showed that PSMs determine biofilm maturation and detachment (12, 13). This is presumably mediated by a detergent-like effect of the amphipathic PSM molecules, via disruption of intermolecular attractions of staphylococcal surface molecules (34). Notably, in contrast to cytolytic and proinflammatory activities, biofilm formation was determined predominantly by hydrophobic amino acids the side chains of which are located on the hydrophobic side of the amphipathic PSMα3 molecule. Only 1 substitution (K7A) resulted in clearly less, rather than increased, biofilm formation. Whether this is related to the recently shown positive effect of PSMs on biofilm formation via fibril formation (35) remains to be determined.

While PSMα3 does not exhibit antimicrobial activity, alanine substitution of large hydrophobic side chains of PSMα3 resulted in pronounced antimicrobial activities in a pattern very similar to that affecting biofilm-structuring activity (correlation; P=0.016). These results indicate that large hydrophobic side chains in PSMs inhibit antimicrobial activity, very likely by preventing the disruption of bacterial membrane integrity. It is important to stress that those residues did not affect cytolytic activities toward the human cells tested, emphasizing that the structural features of PSMs lead to a different effect on prokaryotic vs. eukaryotic membranes.

Our findings have several important implications. First, recent efforts to find vaccines for the treatment of S. aureus infections include active vaccination efforts with toxins that play a key role in S. aureus disease (26, 36). To that end, nontoxic variants are indispensable, such as the H35L variant of α-toxin (37). Our study identifies nontoxic variants of PSMα3, an established major contributor to S. aureus pathogenesis (6), to be used in such efforts.

Second, while lysis of immune cells very likely is of key importance for the strong effect of α-type PSMs on staphylococcal virulence, the biological role of PSM proinflammatory capacity and receptor-mediated recognition of PSMs by the human innate immune system is not understood. In particular, it is not known whether this interaction is part of the staphylococcal pathogenesis program or rather a means of host defense to recognize dangerous bacterial invaders. The use of PSMα3 derivatives identified herein that have only cytolytic but strongly reduced proinflammatory activity, or vice versa, will be extremely valuable to delineate the biological role of PSM-receptor interaction. Ultimately, these investigations will require in vivo analysis of constructs in which PSM peptides are expressed from the genome in a PSM null background. Our laboratory is currently trying to construct the necessary mutants, but initial observations indicate that differences of expression, possibly due to differential export of the peptide analogs by the Pmt system (8), pose a significant hurdle to overcome in this endeavor.

Third, our results indicate that PSMs contribute to biofilm structuring by disrupting hydrophobic interactions between bacterial surface molecules, with the hydrophilic side oriented toward the aqueous environment. While this appears as the most likely mechanism of PSM biofilm structuring activity, based on the detergent-like, amphipathic structure of PSMs and the fact that the bacterial surface is hydrophobic compared with the surrounding fluid, our findings for the first time provide experimental evidence supporting this concept.

Fourth, the fact that PSMα3 contains large hydrophobic side chains such as phenylalanine that prevent antimicrobial activity indicates that PSMs such as PSMα3 have evolved to avoid antimicrobial activity while increasing cytolytic capacity. This makes biological sense as it contributes to producer immunity. Nevertheless, other PSMs, such as S. epidermidis PSMδ, may have evolved to additionally exhibit antimicrobial activities. In support of this concept, the antimicrobially active PSMδ and δ-toxin of S. epidermidis both contain only 1 phenylalanine residue, which is in stark contrast to PSMα3, which lacks such activity and contains 5. It is also important in that regard to note that the PSM secretion system provides protection from the antimicrobial activities of self and nonself PSMs (8).

Fifth, our study highlights the key importance of positively charged amino acids for PSM cytolytic activity. This mechanism is reminiscent of the interaction of cationic antimicrobial peptides with bacterial membranes (38). The finding that positively charged amino acids are crucial for PSM cytolytic activity is in accordance with the fact that the noncytolytic PSMβ peptides of S. aureus and S. epidermidis all have a negative net charge.

Finally, our finding that the lysine side chains of PSMα3 also play a crucial role for proinflammatory activities may suggest that binding of the peptide to the membrane is a prerequisite for receptor activation (illustrated in Fig. 8). Such a model, in which agonist membrane attachment is a structural feature that is recognized by FPR2 would be consistent with the structural differences found among PSMs and other FPR2 ligands (39) and the general membrane interaction features of PSMs. The observed very specific dependence of proinflammatory activities on the C-terminal amino acids of PSMα3 suggests a mechanism of receptor interaction in which the C terminus interacts with a ligand-binding pocket of FPR2 (Fig. 8). The formylated N-terminal part may have similar capacity, at least in some PSMs, possibly owing to structural similarity with N-formylated peptides as described ligands of FPR-type receptors. This is in keeping with our previous observation indicating a limited role of the PSM N-terminal formyl-methionine for FPR2 interaction (18). Clearly, it will be necessary in the future to elucidate the costructure of PSMs and FPR2 in a membrane environment to achieve a well-defined picture of the interaction mode and determine whether activation of FPR2 by PSMs requires PSM membrane attachment or insertion.

Figure 8.

Putative mechanisms of PSM-mediated cytolysis and PSM-receptor interaction. Top: initial interaction with the eukaryotic membrane is assumed to be crucial for both subsequent receptor interaction and cytolysis. Bottom left: receptor interaction supposedly occurs via the N or C termini. Bottom right: cytolysis occurs via disintegration of the membrane at higher peptide concentrations. Brown depicts hydrophobic parts; blue depicts hydrophilic parts of the PSM peptide and the phospholipid bilayer of the membrane.