Characterization of peptides
The peptides were synthesized by standard solid-phase peptide synthesis protocols. Chemical synthesis of GS requires an additional step of cyclization in dilute solution, following the cleavage of the linear construct from the resin. Hence, despite having the smallest number of amino acids, GS synthesis is more demanding than the production of linear TL and IDR. Luckily, GS is readily available by bacterial fermentation. As it undergoes the same purification steps (high-performance liquid chromatography, HPLC) as the other two peptides, irrespective of the production route, its biosynthetic production should be preferred. Indeed, when we compared the activities of synthetic and biosynthetically produced GS, we observed no difference between the two. In the following, only biosynthetic GS was used to avoid batch-to-batch uncertainties.
All three peptides are short (less than 13 residues), amphipathic, and hence are active against lipid bilayers. They carry a net cationic charge, thus possessing an electrostatically mediated selectivity towards anionic membranes. They are composed of different types of positively charged amino acids, however, (2 non-canonical ornithines in GS, 1 Agr & 1 Lys in TL, and 4 Arg in IDR) and vary in charge density, besides the obvious differences in secondary structures. Figure 1 shows molecular models of their functionally relevant conformations and summarizes the predicted physicochemical properties.
The decameric GS has a symmetric cyclic structure that is rather compact. The two linear peptides are larger in size, and IDR is the most highly charged, which results in its lowest absolute hydrophobicity and lowest ability to aggregate. A principal difference between the three peptides obviously lies in their conformational propensities. In contrast to GS, which maintains a largely constant structure independent of the environment, TL and IDR are linear peptidylamides and hence possess much higher conformational plasticity39,40,41. Indeed, TL and IDR are unfolded in aqueous solvents, and TL folds into an α-helix only in membrane-mimicking environments, whereas IDR in the presence of membrane models can fold either as an α-helix or as a β-turn.
Compositional differences translate into functional properties, as observed in the physicochemical experiments addressing the peptides´ mechanisms of action (Fig. 2). Determined under reversed-phase chromatography on a standard C18 column, the apparent hydrophobicity in the partially folded state is in the order IDR < TL < GS, corroborating the AGGRESCAN values. The ability to perturb supported zwitterionic lipid monolayers as established models42 for bilayers was measured by rapid cyclic voltammetry (RCV) and correlates best with the charge densities, with IDR being the weakest monolayer perturbant (IDR«TL < GS). The same outstanding electrostatic-driven binding by IDR was observed in anion-binding studies (Fig. 2B). Here, we compared the interactions of the 3 AMPs with various phosphate-containing cellular metabolites (P-metabolites), by measuring 31P-NMR spectra in equimolar P-metabolite/AMP aqueous solution. Judging by the disappearance of the signals (often with visible precipitate formation) and/or modulation of the chemical shifts, we could distinguish two types of interaction in the binary mixtures – (i) no binding, and (ii) 31P-NMR-detected interactions. Most P-metabolites – AMP, ADP, GDP, phosphoenolpyruvate, a short-chain phosphatidylcholine (1,2-dihexanoyl-sn-glycero-3 phosphocholine) and monophosphate – were not influenced by the presence of the peptides, whereas disctinct interactions with ATP, GTP, pyrophosphate, butyryl phosphate and ppGpp suggested a molecular basis for an interference of the peptides with the energy metabolism in bacterial cells. In all cases, IDR was found to be the strongest in binding, followed by TL and GS, which resemble the order in which the charge density decays.
We further compared the model pharmacokinetic parameters of the peptides by measuring their response to biomimetic chromatography conditions. As summarized in Fig. 3, the chromatographic hydrophobicity indices (CHIs) of the peptides, irrespective of the ionizing conditions, always revealed the same order IDR < TL < GS and showed corresponding binding to the C18 stationary phase in reversed-phase HPLC (Fig. 2A). However, the CHI values failed to explain the differential binding to immobilized artificial membranes (IAM). In IAM chromatography, where the peptides are partitioning into phosphatidylcholine monolayers (the immobilized lipid being 1,2-dimyristoyl-sn-glycero-3 phosphocholine), GS showed again the strongest hydrophobic interaction, but TL-binding was significantly weaker than IDR. Interestingly, when the peptides were exposed to other species of phosphatidylcholines, the rankings of the ability to bind lipids were different. We thus note the obviously strong membrane-perturbing abilities of TL shown in the experiments exploiting DOPC (1,2-dioleoyl-sn-glycero 3-phosphocholine) as a membrane model (see LoD (=limit of detection) values and RCV plots (Figs. 2C, 3A), and the apparent inability of all AMPs to bind 1,2-dihexanoyl-sn-glycero-3-phosphocholine in solution (see 31P-NMR results). This discrepancy can be resolved if we consider the lasting prevalence of the hydrophobic binding forces over the initial electrostatics-mediated attraction to be the major determinants of the peptide-membrane interactions. Accordingly, TL and GS both could be suggested as immersing deeper into the apolar core of the outer bilayer leaflet to reveal their membranolytic action. Interestingly, the ability to bind to acidic proteins – water-soluble human serum albumin (HSA, isoelectric point ~4.7) and to α1-acid glycoprotein (AGP, isoelectric point ~3.3) must include hydrophobic contributions from the net cationic peptides, as for both parameters the most highly charged IDR revealed only intermediate binding values. Nontheless, all three peptides were found to be >90% bound when exposed to HSA- or AGP-immobilized chromatography columns. This high nonspecific binding should additionally influence the in vivo pharmacodynamics of all three AMPs, making it tremendously difficult to maintain high bolus blood concentrations. On the other hand, systemic toxicity will also decay correspondingly.
Additional pharmacologically relevant differences in the action of AMPs were evident in the evaluation of eukaryotic toxicity using various assays (Fig. 3B). In vivo, all three peptides were shown to be acutely toxic. Upon extracorporeal application in a zebrafish embryotoxicity model, LD50 values (50% lethality dose) in the 4–10 µM range were consistently found. Interestingly, neither the order of toxicity (e.g. IDR < TL~GS for 3 h exposure LD50, vs. GS < IDR < TL at 5 µM in the haemolysis assay), nor the absolute toxicity levels correlated between the two toxicity assays (IDR and GS taken at LD50 concentrations did not show significant haemolysis). This finding suggests that in our in vivo experiments, haemolysis was not the major lethality factor, but other toxicity mechanisms could and should have contributed. We also noted that the kinetics of in vivo toxicity varied between the peptides. TL was the only peptide where LD50 did not change between 1 hour and 3 hours of exposure, whereas IDR and GS killed more embryos at longer incubation times. This result corroborates the anticipated proteolytic stability of the three AMPs. TL should be the most labile – it contains a canonical trypsin cleavage site (Lys-Phe); IDR should be intermediate as a folded bactenecin analogue43; and GS, due to its cyclic nature and the presence of non-canonical amino acids, is essentially not susceptible to proteolysis.
Characterization of bacterial strains
Prior to systematically addressing the antibacterial properties of the peptides, we characterized the resistance patterns of the available strains. The resistance of clinical isolates to conventional antibiotics was determined in the original laboratories, using the VITEK-2 system. Additionally, we tested all staphylococci against mupirocin6, and the control strain S. aureus DSM 1104 and its SCV towards methicillin, oxacillin (OXA), gentamicin, tetracycline, and streptomycin, using the standard broth microdilution procedure44. We also verified the MICs of demeclocycline (DMC), vancomycin and gentamicin against the control strain E. faecalis DSM 2570, and E. faecalis isolates WW4 and WW6. The results are summarized in Supplementary Information (SI), Table S1. The MDR strains, classified as possessing resistance to at least one agent in three or more antimicrobial categories45, were S. aureus MRSA9, MRSA9 SCV, MRSA538 SCV, all E. faecalis TRE, and E. faecium VRE strains.
Next, we analysed all 16 strains for their biofilm-forming abilities in different nutrient media: a Todd-Hewitt (TH) broth, a Mueller-Hinton (MH) broth, and a minimal medium (MM); the latter had been applied previously for the study of biofilm eradication38. The purpose of this examination was to determine the conditions that promoted growth with the largest biofilm biomass, and to select the most potent biofilm-forming strains for analyzing the antibiofilm activities of the peptides. Our results (Fig. 4) indicate that cultivation in TH broth is uniformly the most appropariate condition that consistently enables the growth of robust biofilms for all bacteria. In the first 24 hours, vigorous biofilm development was observed among S. aureus for DSM 1104 SCV, MRSA538 SCV, and MRSA8 SCV, but not for MRSA9 SCV. The latter isolate grew slowly and, in the conditions of a 96-well microtiter plate, required at least 48 hours to complete biofilm development (data not shown). The strongest biofilm-forming E. faecalis strains were TRE2, WW4, and WW6, among which only TRE2 is an MDR strain (SI Table S1).
We further analysed the stability of the available S. aureus SCVs. Their reversion into the classical large colony variants (LCVs) was defined by streaking of the overnight cultures onto TH agar. As shown in SI Fig. S1, the appearance of LCVs in the cultures of DSM 1104 SCV and MRSA9 SCV clearly suggests the transient nature of both variants. Additionally, DSM 1104 SCV develops distinct agglomerates in the first hours of growth in the liquid culture, followed by passage into fully planktonic growth only after 22 hours of cultivation (SI Fig. S2). This observation clearly highlights a unique biofilm-forming ability of DSM 1104 SCV even in liquid medium. The remaining S. aureus SCVs grew planktonic when in liquid cultures. MRSA8 SCV and MRSA538 SCV both appeared to be stable SCV forms, as they did not revert into LCV (data not shown).
We analysed the peculiarities of the biofilm matrix composition in the transient S. aureus SCVs (DSM 1104 SCV and MRSA9 SCV), their parental variants, and one stable SCV (MRSA8 SCV) by Congo red staining. The interaction with Congo red was studied in two different assays: bacteria were grown as biofilms on hydroxyapatite discs (HAD), and as colonies on brain heart infusion agar plates (SI Fig. S3). A distinct black colour suggested the elevated production of PIA by both phenotypic variants of DSM 1104, but not by any of the tested MRSA strains. These data corroborate previously described biofilm phenotypes of the MSSA strains, which are highly enriched in PIA46. Interestingly, the ability to synthesize PIA does not provide DSM 1104 variants with adaptive tolerance advantages, as planktonic cells remain sensitive to the action of antibiotics (SI Table S1).
Minimum inhibitory and bactericidal concentrations
Next, we analysed the influence of the peptides on all bacterial strains by determining the MIC and minimum bactericidal concentration (MBC) values.
Comparison of the three peptides consistently reveals the best overall antimicrobial activity for GS and the worst for IDR (Table 1). The MIC values of GS against S. aureus were mostly 4 µg/ml, only for enterococci they were seemingly one dilution higher, approximately 8 µg/ml. The same selectivity trend could be suggested for TL. Interestingly, the opposite tendency was apparent for IDR, which was overall more active against enterococci. Nevertheless, except for the activity against VRE2, IDR showed the highest MIC values, indicating its globally lower antimicrobial activity. Although the MIC values of all three peptides were similar for two clinical E. faecium strains, GS clearly had the lowest MBC values.
Time- and concentration-dependent killing effect of peptides
To characterize the killing kinetics, we analysed the exposure of 108 CFU/ml (CFU = colony-forming units) planktonic bacteria to supra-MIC concentrations and monitored the cell number as a function of time. With all three peptides, when exposed to 5 × MIC, S. aureus counts dropped to 101 CFU/ml (LoD) in less than one hour (Fig. 5A). However, the same conditions were less effective against the cells of E. faecalis (SI Fig. S5). Whereas GS at 5 × MIC accomplished killing within 60 min in all cases, TL showed comparable effectiveness only against DSM 2570 and TRE2 but was not able to eliminate WW6. IDR at 5 × MIC (160 µg/ml) was the worst in performance: it killed only E. faecalis TRE2 cells and required the full 60 min. The cells of the other two strains, though reduced in numbers, remained viable (SI Fig. S5). At 10 × MIC (Fig. 5B), GS (80 µg/ml) and TL (160 µg/ml) were again able to kill all E. faecalis cells within 20 min. IDR (320 µg/ml) was similarly effective against DSM 2570, exhibited slower killing of TRE2, and could not complete its action against WW6 planktonic cells within 60 min of incubation.
In all cases, we observed a monophasic killing process. At sufficiently high concentration, rapid and complete killing by all antimicrobial peptides suggests that they are effective against persister cells. The results of the experiments are also consistent with the MIC/MBC evaluations, collectively allowing the peptides to be ranked as GS > TL»IDR in terms of effective bactericidal action against planktonic staphylococci and enterococci.
Minimum biofilm inhibitory concentrations
The MBIC90 (MBIC90 = minimal peptide concentration at which a bacterial strain develops <10% of the biofilm biomass of the untreated control) values for GS, TL and IDR were determined using a two-fold microdilution procedure, exploiting the best biofilm-forming strains identified above (Fig. 4). This parameter quantifies the ability of the peptides to prevent biofilm outgrowth. As summarized in Fig. 6A, GS was again the most effective.
Biofilms on hydroxyapatite discs: regrowth and scanning electron microscopy
To characterize the viability of the biofilm cells, HAD with pregrown biofilms were exposed for 18 hours to supra-MBIC90 concentrations of the AMP, and – after washing – were placed into fresh TH broth. This procedure serves to determine possible planktonic regrowth from the cells that survived the treatment. The mature extended PIA-enriched biofilms of DSM 1104 SCV were tolerant not only to TL and IDR, but also to OXA (Fig. 6B), which was effective against the planktonic cells at 0.25 µg/ml (SI Table S1). At the same time, the complete lack of regrowth of these biofilms after exposure to GS reflects the excellent effectiveness for this particular antibiotic (Fig. 6B). The presence of GS abolishes regrowth for both antibiotic-susceptible E. faecalis WW6 and MDR E. faecalis TRE2 biofilms.
The other two peptides were effective only against the weakest biofilm-former MRSA8 SCV. The surface-attached biofilms were also examined by scanning electron microscopy (SEM) to identify potential morphologic changes after peptide treatment. Without any peptide the SEM of the PIA-producing DSM 1104 SCV biofilms (Fig. 7A) revealed an external layer, which covered the cells and kept them together. This feature did not look like a soft gel, but rather resembled a glazed layer, as is often observed for polysaccharides. The PIA in the biofilm of DSM 1104 SCV obviously protected the cells from OXA, TL and IDR, but failed against GS action. This finding correlates with the reduced binding affinities of GS to proteins and glycoproteins described above (Fig. 3A). In contrast, the adhesion in the PIA-poor biofilms of MRSA8 SCV was observed to be mediated by surface proteins and unidentified fibrils (SI Fig. S6). The absence of PIA could be a prime reason for the high susceptibility of surface-grown MRSA8 SCV biofilms to the action of all three peptides.
In S. aureus, after exposure to AMPs, the extracellular matrix appeared significantly reduced, suggesting a mechanism for biofilm dispersion (Fig. 7). Notably, GS treatment not only caused the highest degree of matrix reduction and the lowest number of remaining cells, but also resulted in profound alterations in the morphology of the remaining cells. Their surfaces appeared wrinkled, many cells were swollen, and some were clearly disrupted. Such alterations were less pronounced upon exposure to TL and IDR.
Intact E. faecalis TRE2 biofilms were observed as layers of cells with a diameter of about 1 µm, seemingly covered by an unknown cell-associated material (Fig. 7B). After GS treatment, this material also disappeared, as could be judged from the perceived cell size reduction to about 0.8–0.9 µm. Single and dividing cells no longer appeared to be attached together, indicating a dispersion of the biofilm, as in the case of S. aureus. Here, again, TL and IDR exhibited poorer dispersion levels and influenced the remaining matrix to a lesser extent. More matrix continued to cover and adjoin the remaining cells, which stayed within distinct slime-enclosed aggregates. Many cells in both S. aureus DSM 1104 SCV and E. faecalis TRE2 biofilms, after co-incubation with TL and IDR, appeared healthy, which correlates with the higher MBIC90 values and positive regrowth results of these peptides. Notably, none of the peptides was able to eradicate the pregrown biofilms completely from the HAD.