As the occurrence of multidrug resistant strains of pathogens is reaching alarming proportions, there is an increasing urgency to develop novel therapeutic approaches (1). Antimicrobial peptides (AMPs) offer a nature-inspired solution to this problem, in case the human cytotoxicity can be addressed by altering the peptide sequence (2). Most effort on turning AMPs into antibiotics focused on membrane disrupting AMPs (2, 3). These AMPs are 12-50 amino acid long cationic peptides that permeabilize the cytoplasmic membrane by a non-receptor-mediated process, that is, by opening holes in the phospholipid bilayer (4, 5, 6). It is a widely-held view that the activity of these peptides is governed by the overall physicochemical properties, such as charge, helicity, amphiphilicity etc. of the peptide (2, 7, 8). Yet, there is mounting evidence that certain residues and moieties play specific roles in the membrane interaction (9, 10). The chemistry of the C-terminus, in particular its naturally occurring amidation, was identified as one of the key factors in membrane disrupting activity (4). This modification greatly enhances the antibacterial activity compared to the same peptide with a free carboxylic acid at the terminus, e.g. in case of dermaseptin s3 the activity increased ~10-fold (11).
As an explanation for this enhanced activity, it was proposed that C-terminal amidation stabilizes α-helical conformation due to the provision of an additional hydrogen bond (12, 13). Helicity in the presence of lipid membranes correlates well with the activity of cationic membrane disrupting AMPs (14, 15). However, the effect is not limited to α-helical peptides: enhanced activity was also recorded in the case of the β-sheet indolicidin (16). It was also considered that, by eliminating the negative charge of the carboxyl group, C-terminal amidation yields a higher cationic charge compared to identical peptides with a free C-terminus, therefore enhancing peptide binding to negatively charged bacterial membranes (9). Yet it was demonstrated before that charge interaction plays a minor role in the peptide-membrane interaction (10). It was also suggested that C-terminal amidation influences activity by an entirely different, and indirect, pathway by preventing the enzymatic degradation of AMPs (17). This does not explain the activity difference observed in vitro (10).
While there is empirical data in support of each of the hypotheses outlined above, designing negative controls had proven problematic as the amidation changes several aspects of the chemistry of the C-terminus simultaneously. The most important of these is a change in the hydration state, caused by eliminating the negative charge of the carboxylate moiety while introducing a stronger dipole moment and two hydrogen bond donors (18). These changes may also affect the interaction of the terminal region with the polar moieties of the lipid headgroups. Thus, while recognizing that there might be individual differences between AMPs, it can be hypothesized that the amidation exerts its effect via polar interactions.
In a previous work, we have confirmed experimentally and with computer simulations that C-terminal amidation is crucial for the activity of aurein 1.2, a small anuran surface active peptide (GLFDIIKKIAESF-NH2) (10). The short sequence makes this an ideal peptide to study the role of individual moieties in the mechanism of action. Here we report on experiments with a modified peptide where instead of the primary amide of the wild type, a secondary amide is formed at the C-terminus, with a methyl group at the end of the backbone (GLFDIIKKIAESF-NH-CH3). If the hypotheses assigning key roles to charge and/or helicity, i.e. the hydrogen bonding ability of the C terminus are correct, the activity of this peptide would not be different from the wild type peptide. For comparison, results for the non-amidated peptide (GLFDIIKKIAESF-COOH) are also shown.
Materials and Methods
3-mercaptopropionic acid (MPA; HPLC grade, >99%) was purchased from Fluka (Switzerland). Chloroform (99.8%, Analytical grade), methanol (>99.9%, Spectrophotometric grade), and ethanol (HPLC/ Spectrophotometric grade) were purchased from Sigma Aldrich (Castle Hill, NSW, Australia). Hydrogen peroxide (H2O2) solution (30%), and ammonia solution (28%, ACS reagent) were purchased from Sigma-Aldrich (Castle Hill, NSW, Australia). Potassium dihydrogen phosphate (KH2PO4) and potassium hydrogen phosphate (K2HPO4) were purchased from Fluka (Switzerland) at ACS reagent grade; sodium chloride (pro analysis) was purchased from Merck (Darmstadt, Germany). 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), the sodium salt of 1,2-dimyristoyl-sn-glycero-3-phosphoglycerol (DMPG), and cholesterol were purchased from Avanti Polar Lipids (Alabaster, AL, USA). 5(6)-carboxyfluorescein (CF) was purchased from Sigma-Aldrich (USA). Lipids were dissolved in chloroform; for DMPG 3% methanol was added to improve solubility. Desired ratios of lipids: neat DMPC, DMPC/ DMPG (4:1), DMPC/cholesterol (9:1), DMPC/DMPG (3:2), were measured into round bottom glass test tubes. The solvent was evaporated under a gentle stream of N2 and dried overnight. Liposomes were hydrated in the assay buffer (20 mM phosphate buffered saline solution containing 100 mM NaCl at pH 6.9) at 37°C for 30 min, vortexed (~1 min) and briefly sonicated (~30 s) before use.
Quartz Crystal Microbalance
Quartz Crystal Microbalance with Dissipation Monitoring (QCM) measurements were performed with a Q-SENSE E4 system (Q-Sense, Sweden). The sensor crystals used were 5 MHz, AT-cut, polished quartz disks with evaporated gold sensor surface. The changes of resonance frequency (Δf) and energy dissipation (ΔD) upon mass deposition were measured simultaneously at four different eigenmodes (3rd, 5th, 7th and 9th) of the crystal resonance, but only the seventh harmonic of the fundamental frequency of the sensor chip was used for analysis.
Sensor chips were cleaned in the so called base piranha solution: 3:1:1 mixture of 18.2 MΩ deionized water (Ultrapure, Sartorius AG, Germany), hydrogen peroxide and ammonium hydroxide solutions (concentrations as per above) for 20 min at 70 ºC. Consecutively the chips were rinsed with deionized water and dried. Chips were modified with 2% (w/w) MPA in propan-2-ol overnight to form a self-assembled monolayer as a support for the biomimetic membrane. Propan-2-ol wash was used to remove excess MPA. During the QCM experiments, first water was injected into the chambers to hydrate the surface of the chips and then assay buffer was introduced, followed by the lipid (in vesicular form) to deposit a membrane. Peptides were injected into the QCM chamber once the formation of a single bilayer membrane: -13-15 Hz frequency and 2.5-3.0×10-6 dissipation change was confirmed (19, 20). QCM experiments were repeated at least 3 times for each peptide and at each concentration. Control experiments on MPA coated gold confirmed the absence of peptide binding to the MPA self-assembled monolayer. Hence, all measured structural changes correspond to the interaction between the peptides and the membranes.
Plotting the two main signals of QCM: dissipation change against frequency change (f-D curve) for a biomolecular interaction process can be used to provide a mechanistic fingerprint. A detailed description was provided in a previous work (10). QCM fingerprinting was developed and evaluated primarily for the characterization of α-helical AMPs, as interpretation of the trends relies on the geometrical specifics and known membrane disrupting properties of these peptides.
Dye release upon peptide interaction was monitored using 5(6)-carboxyfluorescein (CF) loaded vesicles. Liposomes were prepared as described above using a buffer containing 20 mM CF (i.e. close to the lower limit of self-quenching concentration). Excess dye was removed from the medium by dialysis. Leakage of CF from liposomes was detected as an increase in fluorescence intensity, monitored using Spectramax M5 spectrophotometer (Molecular Devices, Silicon Valley, CA, USA) with excitation wavelength of 480 nm while the emission was monitored in a range of 500-530 nm. Measurements were repeated three times.
Circular Dichroism Spectroscopy
CD spectra from 200 to 250 nm were recorded with a Model 420 CD spectrophotometer (AVIV, USA) at 25 °C. 1-mm path length quartz cell was used for all measurements, containing 10 μM peptide and 100 μM lipid in PBS buffer. Each experiment was repeated three times to confirm the trends. Data are shown as mean residue molar ellipticity (deg cm-1 dmol-1). The percentage of α-helical content was estimated from the molar ellipticity at 222 nm (θ222) using Equation 1 (21).
The simplest test of the membrane disrupting activity is a dye leakage experiment. As shown in Table 1, exposing dye loaded neat DMPC liposomes to wild type aurein1.2-NH2 lead to substantial increase in fluorescence, indicating breach of the membrane integrity as expected for a membrane disrupting peptide. Without C-terminal amidation (aurein1.2-COOH), the peptide exhibited much reduced, but nonzero activity, as seen with QCM before (10). In contrast, the aurein 1.2 sequence with a secondary amide at the C terminus (aurein1.2-NH-CH3) did not cause any dye release.
Dye leakage results with neat DMPC liposomes
|Peptides||Fluorescence intensity increase (%)|
|Aurein 1.2 –NH2 (wild type)||30|
|Aurein 1.2 –COOH||1|
|Aurein 1.2 –NH-CH3||0|
Helicity of the peptide is used as an indirect indicator of membrane binding: most antimicrobial peptides are unstructured in solution but become helical upon binding to the membrane (22). Therefore, conformations of the three aurein 1.2 variants in the presence of different phospholipid mixtures were characterized using CD spectroscopy (Fig. S1). Percentage helicity data are shown in Table 2. All peptides were largely unstructured in the buffer solution. As expected, the helicity of the wild type peptide increased in the presence of phospholipids, due to membrane binding, and so did the helicity of aurein1.2-COOH that was shown before to bind to membranes but do not disrupt them (10). In contrast, the helicity of aurein1.2-NH-CH3 did not change appreciably. This is counterintuitive given the helicity-enhancing role of the amide moiety (23). The most likely explanation is that the peptide cannot enter the headgroup zone of the membrane that is believed to facilitate folding.
Results of CD measurements
QCM experiments were performed with membrane mixtures that mimic mammalian (zwitterionic) and bacterial (charged) membranes. The primary data, (f-t) and (D-t) sensograms are shown in the supplementary material (Figs. S2-S5). For analysis, dissipation change versus frequency change plots (f-D curves) were used. f-D curves show changes in the viscoelastic character of the membrane when it is exposed to the peptides, revealing stages of the molecular mechanism of action (24). Figure 1 shows the viscoelastic fingerprint analysis of the interactions of aurein1.2-NH-CH3 with mammalian model membranes. For zwitterionic PC membranes, there is a short [-f, +D] trend that can be explained with weak membrane association. In case of neat DMPC (Fig. 1 left panels) a second stage (marked with arrow) appears at higher concentrations, revealing a structural change that is likely related to peptide aggregation and/or partial membrane penetration. Addition of 10 % cholesterol eliminated this second stage (Fig. 1 right panels), instead the amount of binding increased somewhat at higher concentrations.
In the case of charged membranes (Fig. 2) the amount of initial binding is higher, ~ -4 Hz for (4:1) and ~ -6 Hz for (3:2), suggesting that charge might play a role in the association; in case of the higher charged lipid content there are signs of some membrane disruption, although it is still not full disintegration as in case of the wild type peptide.
The comparison of ΔD vs Δf plots for aurein1.2 (wild type) and aurein1.2–CH3 exposed to different membranes is shown in Figure 3 (neutral membranes) and Figure 4 (charged membranes). Here the scale is adjusted to accommodate the trendlines of the wt peptide; the traces shown are taken from (10). As it can be seen, only negligible activity was observed for the aurein 1.2-NH-CH3 compared to the wild type peptide, indicating the importance of C-terminal amidation for the peptide activity.
The three different termini are shown in Figure 5 for comparison. As our previous results showed, both aurein1.2-NH2 and aurein1.2-COOH could bind to all membranes, in spite of the different net charge (0 vs. +1) (10). In contrast, the terminal CH3 group reduced the amount of membrane binding substantially, even though also providing +1 charge as in case of the primary amide. This is further evidence that membrane binding is not charge driven. Dye leakage experiments also confirmed the lack of any membrane disrupting activity of aurein1.2-NH-CH3; as the primary and secondary amides at the C terminus yield equivalent net charge for the peptide, the difference in overall activity cannot be related to charge effects either.
The strong dipole and consistently higher hydrogen-bonding ability of the amide, and the resulting increased helicity is also cited as a source of increased activity of C-amidated peptides. There is no significant difference between the hydrogen bonding ability of primary and secondary amides, thus aurein1.2-NH-CH3 should exhibit comparable helicity to the wild type peptide, which it does not do. It is believed that α-helicity of this highly amphipathic peptide is the result of a templating process, between the lipid headgroup and tail zone, where the hydrophobic residues align to the membrane core and the hydrophilic residues are exposed to water (25). From the fact that aurein1.2-NH-CH3 is not becoming α-helical at all in the presence of membranes one can speculate that the peptide is not entering even the headgroup zone of the membrane, the weak binding seen in QCM is likely caused by loose surface association and/or a local change of buffer viscosity due to the presence of the peptides (26).
Thus, the presence of the small methyl group inhibits activity while in all other physicochemical characters the wild type peptide and aurein1.2-NH-CH3 are identical. Hence, the two main reasons cited for increased activity of the C-terminal amidated peptides are ruled out. Inhibition of membrane binding may be caused by a feature of the chemistry of the C-terminus. The main difference between the wild type aurein 1.2 and aurein1.2-NH-CH3 is the hydration state of the terminus: one fewer hydrogen-bond donor and the methyl moiety restricts water access to the amide. Thus, the observed difference in activity is likely related to the hydration of the C terminus of the peptide.
It is challenging to extrapolate the molecular mechanism of action identified in model membranes to the expected activity, or the lack thereof, against bacteria. Changing physicochemical characteristics of an AMP can interfere with its ability to reach the membrane where it can exert its primary, membrane disrupting mode of action. It is known that wild type aurein 1.2 exhibits moderate activity against both Gram positive and Gram negative bacteria with MIC values in the 50-100 μg/ml concentration range (27) thus the membrane disrupting mechanism appears to proceed irrespective of the membrane/cell wall structure. This suggests that the peptide can easily diffuse through to the bacterial membrane surface. The modification discussed here would inhibit interaction with the membrane itself; therefore it is expected that the effect of the modification would be the same in both Gram positive and Gram negative bacteria, that is, lack of activity. Hence it is feasible to assume that the antimicrobial action of aurein1.2 is also dependent on the presence of a primary amide at the C-terminus.
It was shown by using quartz crystal microbalance, dye leakage and circular dichroism experiments that the activity-regulating effect of the C-terminal amidation of aurein1.2 peptide is not related to charge or helical stability of the peptide. A secondary amide cannot serve the same purpose as the primary amide, in particular, the modified peptide exhibited a much-reduced ability to bind to the membrane. Thus, the primary amide at the C-terminus is required to bind to the membrane. We hypothesize that this difference is related to the hydration state of the terminus in the presence of the methyl moiety. We have demonstrated before that in the mechanism of aurein1.2 action phenylalanine residues act as membrane anchors; the results presented here suggest that membrane anchoring is the second step of membrane attachment, whereas the first step involves the primary amide at the C-terminus.
Moore A Devine D Bibby M. Pept Res 1993 7 265-269.