Selective Membrane Disruption Mechanism of an Antibacterial γ-AApeptide Defined by EPR Spectroscopy

Abstract

γ-AApeptides are a new class of antibacterial peptidomimetics that are not prone to antibiotic resistance and are highly resistant to protease degradation. It is not clear how γ-AApeptides interact with bacterial membranes and alter lipid assembly, but such information is essential to understanding their antimicrobial activities and guiding future design of more potent and specific antimicrobial agents. Using electron paramagnetic resonance techniques, we characterized the membrane interaction and destabilizing mechanism of a lipo-cyclic-γ-AApeptide (AA1), which has broad-spectrum antibacterial activities. The analyses revealed that AA1 binding increases the membrane permeability of POPC/POPG liposomes, which mimic negatively charged bacterial membranes. AA1 binding also inhibits membrane fluidity and reduces solvent accessibility around the lipid headgroup region. Moreover, AA1 interacts strongly with POPC/POPG liposomes, inducing significant lipid lateral-ordering and membrane thinning. In contrast, minimal membrane property changes were observed upon AA1 binding for liposomes mimicking mammalian cell membranes, which consist of neutral lipids and cholesterol. Our findings suggest that AA1 interacts and disrupts bacterial membranes through a carpet-like mechanism. The results showed that the intrinsic features of γ-AApeptides are important for their ability to disrupt bacterial membranes selectively, the implications of which extend to developing new antibacterial biomaterials.

Introduction

Antibiotic resistance is one of the major threats to public health; therefore, it is essential to develop more effective antibacterial reagents with minimum microbial resistance. Antimicrobial peptides (AMPs), both natural and synthetic, have shown antimicrobial activities without inducing resistance in bacteria (1). This has generated interest in the utility of AMPs to curtail bacterial infection (2). One major difference between conventional antibiotics and AMPs lies in their distinct mechanisms of antimicrobial activities. Antibiotics exert their antibacterial abilities by interfering with the pathways in protein synthesis, cell wall synthesis, or DNA/RNA replication (3). These pathways are prone to be avoided by genetic mutations of bacteria and thereby induce antibiotic resistance due to their defined targets (4). In contrast, the antimicrobial activities of AMPs are attained by the nonspecific interactions of AMPs with bacterial membranes, which lead to membrane disruption. In particular, AMPs are capable of selectively interacting with bacterial membranes versus mammalian cell membranes. The former contains ∼25% negatively charged lipids, whereas the latter is composed of mostly neutral lipids (zwitterionic) with high cholesterol content in the outer leaflets (5, 6, 7). For bacterial membranes, the interaction of cationic AMPs with anionic lipids contributes to peptide-induced membrane permeability changes and cell death, as shown in many AMPs, including magainins (8, 9, 10, 11). In contrast, for mammalian membranes, the presence of a large amount of cholesterol inhibits the membrane disruption activities of AMPs (5, 12, 13, 14).

Several mechanisms have been proposed for the antibacterial activities of AMPs, including carpet, toroidal-pore, barrel-stave, lipid-clustering, and interfacial activity (15, 16, 17, 18). In the carpet model, AMPs accumulate on the membrane surface; the lipid bilayers are destabilized after a critical concentration has been reached (19, 20). In this model, membrane disruption may include the formation of transient pores or a global collapse of the membrane through detergent-like behaviors of some AMPs (19, 20). The barrel-stave mechanism is initiated by the penetration of AMPs into lipid bilayers, leading to hydrophilic pores formed by peptide–peptide interactions (21). In the toroidal-pore model, on the other hand, pores are formed by both AMPs and lipid molecules (22). In the lipid-clustering model, AMPs binding to the membrane surface cause charged lipid clustering and the phase separation of lipids (23). The interfacial-activity model proposes that AMP activity is initiated through the binding of peptides to the membrane interfacial region. This leads to disruption of the segregation of the interfacial and hydrophobic regions of the membrane, which results in membrane leakage accompanied by translocation of lipid molecules, peptides, and solutes (24). These models are able to explain the AMP activities, but the detailed molecular mechanisms and the changes in lipid organization on AMP binding are still not completely understood.

Despite exhibiting broad spectrum activities against Gram-positive and Gram-negative bacteria, the applications of natural AMPs remain limited due to their susceptibility to proteolytic degradation (25). Synthetic peptidomimetics become an attractive alternative by mimicking the antibacterial activities of AMPs with no protease vulnerability (26). Recently, we developed a new family (to our knowledge) of synthetic peptidomimetics: γ-AApeptides (Fig. 1) (27). They are termed “γ-AApeptides” as they are oligomers of n-acylated-n-aminoethyl amino acids. γ-AApeptides can project the same number of side chains as α-peptides with the same lengths. In addition, γ-AApeptides are known to resist proteolytic degradation, and are amendable for derivatization with diverse functional groups. They have shown promise in biological applications, one of which is the development of antimicrobial agents that mimic the mechanisms and functions of AMPs (28, 29, 30). Among a few types of antimicrobial γ-AApeptides, lipo-cyclic-γ-AApeptides display intriguing antimicrobial activity (29). In the structures of lipo-cyclic-γ-AApeptides, lipid tails were introduced to ring structures to increase the lipophilicity of the molecules so as to enhance their interactions with bacterial membranes (Fig. 1). The lead compound 1 (AA1) shows potent and broad-spectrum activity against multidrug-resistant Gram-positive and Gram-negative bacteria (29). In addition, it can mimic AMPs to modulate immune response. Fluorescence studies have suggested that lipo-cyclic-γ-AApeptides might be able to selectively depolarize and disrupt both Gram-positive and Gram-negative bacterial membranes, eventually causing bacterial cell death (29). However, the molecular bases for the selectivity, changes in lipid organization, and the disruption mechanism are not known.

Figure 1.

Structure of lipo-cyclic-γ-AApeptide 1 (A) and comparison of an α-peptide (B) and a γ-AApeptide (C). A γ-AApeptide is comparable to an α-peptide in unit length and half of the side chains of a γ-AApeptide are linked to the amide groups. The lipo-cyclic-γ-AApeptide 1 contains a cyclic γ-AApeptide and a lipid tail.

Electron paramagnetic resonance (EPR) has been shown to be a robust tool for characterizing membrane properties, including dynamics and lipid ordering (31, 32, 33, 34, 35, 36). Moreover, EPR has been used to determine the membrane interactions of CM15 (37), Cecropins (18, 38), and Melittin (39). In this work, we applied and further developed, to our knowledge, novel EPR techniques to characterize biological membranes and protein–lipid interactions. These techniques include: (1) membrane permeability analysis to assess membrane disruption by proteins, (2) lipid lateral ordering induced by protein binding defined using EPR at 94 GHz, and (3) a new (to our knowledge) application of EPR power saturation methods to determine membrane thinning. To obtain an EPR signal, EPR spin probes are attached to different positions of phospholipid molecules (Fig. S1 in the Supporting Material). We expect these spin probes at varied positions will reveal a full picture of membrane properties across the bilayer.

In this study, we employed EPR techniques to elucidate: (1) the molecular basis of the membrane-disruptive activities of AA1 on bacterial membranes; and (2) AA1-induced membrane property changes, including membrane permeability, fluidity, solvent accessibility, and lateral order. The negatively charged bacterial membranes were mimicked by POPC/POPG (4:1 wt/wt) liposomes (40, 41, 42). Because the mammalian cell membranes contain mostly neutral lipids with high cholesterol content ranging from 20 to 50 mol % (43, 44), the corresponding membrane mimic was used with POPC and 30–50 mol % cholesterol. We found that AA1 selectively interacts and disrupts the bacterial-mimic membranes over the mammalian-mimic membranes, possibly through carpet-like interactions. These results will give insight into the mechanism of the action of lipo-cyclic-γ-AApeptides, and therefore provide a basis for the design of more potent and selective antimicrobial agents.

Materials and Methods

Materials

Phospholipids POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine), POPG (1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-rac-(1-glycerol)]), T-PC (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho -tempocholine), N-TP (N-tempoyl palmitamide), 5-PC (1-palmitoyl-2-stearoyl-(5-doxyl)-sn-glycero-3-phosphocholine), 7-PC (1-palmitoyl-2-stearoyl-(7-doxyl)-sn-glycero-3-phosphocho-line), 10-PC (1-palmitoyl-2-stearoyl-(10-doxyl)-sn-glycero-3-phosphocholine), 12-PC (1-palmitoyl-2-stearoyl-(12-doxyl)-sn-glycero-3-phosphocholine), and CHOL (cholesterol) were purchased from Avanti Polar Lipids (Alabaster, AL). 5-SASL (2-(3-carboxypropyl)-4,4-dimethyl-2-tridecyl-3-oxazolidinyloxy), 5-MeSL (2-(4-methoxy-4-oxobutyl)-4,4-dimethyl-2-tridecyl-3-oxazolidinyloxy), 4-PT (4-phosphonooxy-TEMPO), and VC (vitamin C) were purchased from Sigma-Aldrich (St. Louis, MO).

Synthesis of AA1

The peptide was synthesized in the solid phase following the reported protocol (27). Briefly, synthesis and assembling blocks of lipo-γ-AApeptides followed a standard protocol of solid-phase synthesis using Fmoc-chemistry on a Rink amide resin (Sigma-Aldrich). The N-terminus of AA1 was lipidated by reactions with lauric acid, palmitic acid, or oleic acid using DIC (diisopropylcarbodiimide)/DhBtOH (oxohydroxybenzotriazole) as activation agents. After the desired sequences were assembled, the peptides were cleaved from the solid support using 50:45:5 TFA/CH₂Cl₂/triisopropylsilane overnight. The solvent was evaporated and the residues were analyzed and purified on an HPLC instrument (Waters, Milford, MA). The desired fractions (>95% purity) were collected and lyophilized. The molecular weights of the peptides were determined using an AutoFlex MALDI-TOF mass spectrometer (Bruker, Billerica, MA).

Liposome preparation

Large unilamellar vesicles (LUVs) were prepared as described in Hope et al. (45) and Szoka et al. (46). Lipids in chloroform were mixed in a glass tube and dried as thin films under a stream of nitrogen gas. To obtain EPR signals, 1–2 mol % of spin-labeled lipid analogs (Fig. S1) were added to the lipid mixture. To remove residual organic solvent, the lipid films were further dried using a vacuum pump for ∼16 h. The lipids were resuspended in an HK buffer (20 mM HEPES, 150 mM KCl, pH 7) by vortexing for 1–2 min and then subjected to 10–15 freeze-and-thaw cycles. This was followed by extruding the lipid suspension 10–15 times through a mini-extruder with a 100 nm polycarbonate membrane (Avanti Polar Lipids).

EPR spectroscopy

Membrane fluidity and permeability measurements

The measurements were carried out on a model No. E680 X-/W-band continuous-wave and pulsed EPR spectrometer (Bruker) with a high-sensitivity resonator (model No. 4119HS; Bruker) at the National High Magnetic Field Laboratory (NHMFL). Spectra were collected at 2 mW microwave power with a field modulation frequency of 100 kHz and a modulation amplitude of 1–2 Gauss (G) at room temperature. Samples were loaded in 0.6 mm inner diameter glass capillary tubes. For membrane fluidity measurements, LUVs with 1 mol % of spin-labeled lipid analogs (Fig. S1) were prepared in the HK buffer as described earlier with a 10 mM lipid concentration. The AA1 peptides were added to LUVs at a lipid/peptide ratio (L/P) of 10–80:1.

A modified 4-PT/VC quenching assay was performed to determine membrane permeability (47, 48). Briefly, a stock of water-soluble 4-PT spin labels was prepared in the HK buffer. 10 mM LUVs containing 4-PT were prepared using the extrusion method described earlier. 4-PT-loaded LUVs were dialyzed against the HK buffer for 48 h to remove the untrapped 4-PT and were subsequently equilibrated with 1 mM VC for 30 min. The AA1 peptides were added to the liposome samples at desired L/Ps. EPR signal reduction as a result of membrane penetration and 4-PT/VC mixing was recorded for 60 min after adding the peptides. Control experiments in the absence of AA1 showed no significant 4-PT signal reduction for the liposomes used in this study during this timescale (Fig. S2).

Accessibility measurements

To determine solvent accessibility, EPR power saturation experiments were performed on a model No. E680 spectrometer (Bruker) at 9.5 GHz using a loop gap resonator (Molecular Specialties, Milwaukee, WI). LUV samples were loaded in gas-permeable TPX capillary tubes (Molecular Specialties) and purged using either a stream of air or N₂ gas. EPR spectra were collected at microwave powers ranging from 0.4 to 100 mW with a modulation field of 2 G and a modulating frequency of 100 kHz. The accessibility measurements were performed either in the presence of a nonpolar reagent O₂ or an uncharged polar reagent NiEDDA (nickel(II) ethylenediaminediacetate, 50 mM). NiEDDA accessibility measurements were carried out in a degassed environment, i.e., purged with N₂ gas. The accessibility of both reagents was quantified using an accessibility parameter (Π) (49, 50) using DPPH (Bruker) as a standard. The depth parameter, Φ (50), was calculated from the ratio of the accessibility value Π(O₂) to Π(NiEDDA).

Lipid lateral ordering measurements

EPR spectra at 94 GHz or higher are sensitive to lipid lateral ordering (36). High-frequency high-field EPR improves spectral resolution through increased g-factor sensitivity, enabling the determination of the motionally averaged g_xx-g_yy anisotropy, which reflects lipid lateral ordering. A brief description of the calculation of the lateral order parameter is as follows: Eqs. 1 and 2 define the axial and azimuthal anisotropy of the g-tensor, respectively. The motionally averaged g-anisotropy is given by Eqs. 3 and 4, where ${\theta }_A$ is the tilting angle of the z axis of a spin label with regard to the membrane normal (Fig. S3). Angle $\phi$ (with maximum amplitude ± ${\phi }_0$) represents the rotation of the nitroxide x-z plane with regard to the Z (magnetic field axis)-z plane (Fig. S3). The nitroxide mean x axis is tilted at angle $\overline{\phi }$, which is the mean value of φ. The angular brackets represent angular averages:

$$\Delta g=g_{zz}–\frac{1}{2}\left(g_{xx}+g_{yy}\right),$$ (1)

$$\delta g=\frac{1}{2}\left(g_{xx}–g_{yy}\right),$$ (2)

$$\langle \Delta g\rangle =\langle P_2\left(\text{cos}{\theta }_A\right)\rangle \Delta g,$$ (3)

$$\langle \delta g\rangle =\frac{1}{6}\left[2+3\langle \text{cos}{\theta }_A\rangle +\langle P_2\left(\text{cos}{\theta }_A\right)\rangle \right]\langle \text{cos}2\left(\phi –\overline{\phi }\right)\rangle \delta g.$$ (4)

The lateral ordering is represented by $\text{cos}2\left(\phi –\overline{\phi }\right)$ and the conventional order parameter is defined by $\langle P_2\left(\text{cos}{\theta }_A\right)\rangle$, where $P_2\left(\text{cos}{\theta }_A\right)$ = (1/2) (3cos²θ_A – 1).To calculate the lateral ordering, the spin-label motion is constrained in a cone with a half-angle β, in which case $\langle P_2\left(\text{cos}{\theta }_A\right)\rangle =\left(1/2\right)\text{cos}\beta \left(1+\text{cos}\beta \right)$ and $\langle \text{cos}{\theta }_A\rangle =\left(1/2\right)\left(1+\text{cos}\beta \right)$ (51). The value β can be calculated using Eq. 3 and $\langle \text{cos}2\left(\phi –\overline{\phi }\right)\rangle$ can be obtained from Eq. 4. The motionally averaged 〈g_xx〉 and 〈g_zz〉 values are determined from the peak maxima in the low-field and high-field regions of the EPR spectra, respectively. The value 〈g_xx〉 can also be resolved using the derivative of the EPR spectrum, i.e., the second derivative of the absorption spectrum. The value 〈g_yy〉 is obtained from the maximum slope in the intermediate region of the EPR spectrum or the peak maximum in the integrated EPR spectrum (absorption-like). The g-tensor components g_xx, g_yy, and g_zz were determined from the rigid limit spectra at 150 K. The values are 2.0090, 2.0062, and 2.0023, respectively, for POPC/POPG liposomes with 5-SASL.

To examine the lateral ordering of lipids, EPR spectra were recorded on a W-band 94 GHz continuous-wave and pulsed EPR spectrometer (52), with a quasi-optical transmission design, at the NHMFL. The EPR measurements were performed using a nonresonant sample holder containing a thin-cylindrical layer of sample of 20-mm diameter, 0.17-mm thickness, and 53-μL volume (53). The sample holder is operating in induction mode, i.e., samples are irradiated with linearly polarized microwaves and signals are detected as the orthogonal mode of the circularly polarized radiation that is emitted from the sample. The sample holder was fabricated in-house by the NHMFL machine shop. Spectra were collected with 300-G sweep width, 4-G modulation amplitude, and 600–800-Hz modulation frequency at room temperature. LUVs containing 2 mol % 5-SASL were prepared as described before. LUVs were mixed with the AA1 peptides at an L/P at 10:1 before measurements.

Results

AA1 binding increases membrane permeability

To study the permeability changes of lipid vesicles upon binding AA1, we carried out an EPR-based quenching assay. Briefly, liposomes with encapsulated 4-PT were used to give an EPR signal. Membrane permeability was detected using the signal reduction upon adding VC and AA1. Permeability change % is defined as the percentage of EPR signal (4-PT) changes relative to the liposomes without AA1. One-hundred-percent permeability changes indicate complete 4-PT and VC mixing as a result of membrane penetration. Fig. 2 A shows the peptide-induced membrane permeability changes of liposomes with the following lipid compositions: POPC/POPG (4:1 wt/wt), POPC, and POPC with varied cholesterol content. POPC/POPG liposomes show the highest permeability changes of 67 ± 2% in 60 min, after peptide binding. In contrast, reduced membrane permeability changes were found for the neutral POPC liposomes. Moreover, the presence of cholesterol further diminished the peptide-induced permeability changes of the lipid membranes. When cholesterol concentrations were increased from 10 to 50 mol %, the permeability changes at 60 min dropped from 42 ± 2% to 12 ± 2%. In addition, with higher peptide/lipid ratios, increased permeability changes were observed (Fig. 2 B). To be noted, the permeability data can be fitted with the inverted biexponential decay curves represented by solid lines in Fig. 2. Each curve includes a fast component with a decay constant of 4–5 min and a slow component with a constant ranging from one to several hours. Taken together, these results illustrated the selective membrane-disruptive activity of AA1 on negatively charged bacterial-mimic liposomes, whereas the mammalian-mimic liposomes with large amounts of cholesterol are resistant to its perturbation.

Figure 2.

Membrane permeability changes induced by AA1 binding. (A) Comparison of the permeability changes of liposomes with different lipid compositions at an L/P of 10. The following lipids were compared: POPC/POPG, POPC, and POPC with 10, 30, and 50% CHOL. (B) Permeability changes of POPC/POPG liposomes with L/Ps ranging from 10 to 80. Note: A 100% permeability change indicates complete 4-PT and VC mixing and 4-PT signal reduction due to membrane penetration.

It has been proposed that AMPs are able to fluctuate or transfer between cell membranes (54). To ascertain whether this behavior exists for AA1, we compared the permeability changes of peptide-loaded liposomes both with and without adding unloaded bare liposomes. As illustrated in Fig. S4, decreased permeability changes were observed after adding the unloaded bare liposomes to the vesicles loaded with 4-PT and AA1. This confirms that AA1 moves from the loaded liposome surfaces to the unloaded lipid surfaces and subsequently reduces the permeability changes of the loaded liposomes. Moreover, these results are consistent with the biexponential rates shown in Fig. 2. Specifically, the permeability rates may be defined by a fast membrane leakage rate and a slower rate caused by the fluctuation of peptides between liposomes (54).

AA1 binding reduces membrane fluidity

The lipid fluidity or mobility in a bilayer can be perturbed by the binding of peptides to the membrane. Mobility changes in lipid bilayers on peptide binding have been observed for AMPs in previous biophysical studies (55, 56). EPR spectral line shapes represent molecular motion, which is suitable for probing lipid fluidity (31, 57). Greater spectral peak-to-peak splitting and/or broader spectral line shapes indicate slower molecular motion. Here, to determine the mobility changes across the bilayer on AA1 binding, EPR spin probes were attached to different positions of the lipid molecules (Fig. S1). The AA1-induced mobility changes on the membrane surface were reflected in the EPR spectra of T-PC. The mobility of lipid headgroups was accessed using N-TP. In addition, mobility changes in the acyl-chain region were determined using 5-SASL, 5-PC, 7-PC, 10-PC, and 12-PC. We expect these spin probes at varied positions will reveal a full picture of membrane fluidity and peptide-induced changes across the bilayer. Fig. 3 shows the motional parameters of 5-SASL-labeled lipid membranes with different compositions. Lipid mobility from the EPR spectra of 5-SASL was quantified in terms of a motional parameter 2T_||, i.e., the peak-to-peak splitting. For POPC/POPG liposomes, the peak-to-peak splitting of the 5-SASL spectrum is increased on the addition of AA1, and this can be interpreted as reduced mobility of lipid chains. Overall, the bacterial-membrane mimic, POPC/POPG liposomes, shows the greatest change in mobility upon binding of AA1 followed by neutral POPC lipids. Minimal changes were observed for liposomes containing 30% cholesterol.

Figure 3.

Membrane fluidity changes in the presence of AA1. Room temperature EPR spectra and mobility changes upon AA1 binding of liposomes labeled with 5-SASL for (A) POPC/POPG, (B) POPC, and (C) POPC/30% CHOL liposomes at an L/P of 10. EPR spectra of bare liposomes are overlaid with the spectra in the presence of AA1. The 2T_|| changes (Δ2T_||) on AA1 binding are shown on the upper-right side of the spectra. To see this figure in color, go online.

Like 5-SASL, slower lipid motion was also observed for T-PC, N-TP, 5-PC, 7-PC, 10-PC, and 12-PC for POPC/POPG liposomes upon AA1 binding (Figs. S5, S6, S7, and S8). To account for parameter sensitivity for different EPR line shapes, three different EPR parameters were used to analyze mobility changes. The line widths of the central peaks in the EPR spectra (ΔH₀) were compared for all spin labels (Fig. S9 A). A₀/A_–1 ratios were compared for T-PC, N-TP, and 12-PC, where A₀ is the intensity of the central peak and A_–1 is the intensity of the high-field peak of a spectrum (Fig. S9 B (58)). Parameter 2T_|| was used to quantify the mobility of 5-PC, 7-PC, and 10-PC (Fig. S9 C). EPR spectra were collected with L/Ps ranging from 10 to 80 for liposomes containing different spin-labeled lipids. The largest mobility changes were observed at an L/P of 10 (Figs. S5, S6, and S7). Overall, the results show that AA1 induces more mobility changes on N-TP and 5-PC than on 7-PC, 10-PC, and 12-PC (Fig. S9). Because N-TP represents the lipid headgroup region and the spin probes of 5-SASL and 5-PC are adjacent to lipid headgroups, the data suggest larger mobility changes around the headgroup region than in the acyl-chain region deep inside the membrane.

In an effort to compare AA1 interactions with neutral and negatively charged lipid molecules, EPR spectra of liposomes containing 5-MeSL and 5-PC were compared with that of 5-SASL (Fig. S10). These three spin-labeled lipids were chosen because at neutral pH, 5-SASL is negatively charged, whereas there is no charge for 5-MeSL and 5-PC. Interestingly, for POPC/POPG, 5-MeSL- and 5-PC-labeled lipid bilayers show smaller mobility changes upon AA1 binding when compared with 5-SASL (Fig. S10). These data suggest electrostatic interactions between the cationic AA1 and negatively charged lipids. Similar differences between 5-SASL and 5-PC were also observed on the binding of melittin peptides to lipid bilayers (39). On the other hand, for neutral and cholesterol-containing lipids, 5-PC-labeled liposomes showed no changes in EPR spectra similar to 5-SASL (Figs. S6 and S7). Taken together, these data argue that AA1 interacts more strongly with the anionic lipid molecules on the membrane.

AA1 increases lipid lateral ordering

Next, EPR W-band (94 GHz) spectra were used to identify lipid lateral ordering induced by AA1 binding. Lateral order reflects the lateral packing or the lateral pressure of the lipids in a bilayer that can be perturbed by peptide and AMP binding (59, 60, 61). The changes in lateral order can be measured using EPR spectra in W-band or higher frequencies with increased g-factor resolution (51, 62). Fig. 4 shows W-band EPR spectra of POPC and POPC/POPG liposomes with 5-SASL in the absence and presence of AA1. For POPC, no significant changes were observed in the presence of AA1 (Fig. 4 B). In contrast, the EPR spectra of POPC/POPG with AA1 show visible separation of 〈g_xx〉 and 〈g_yy〉 components, as illustrated in Fig. 4 A. POPC/30% CHOL liposomes were not included in the analysis because cholesterol was found to induce significant lateral ordering due to a lateral condensing effect (51). The appearance of g-factor anisotropy (motionally averaged) in the EPR spectra indicates that AA1 increases the lateral ordering of POPC/POPG lipid bilayers. This g anisotropy can be quantified in terms of a lateral-order parameter, which is defined on the basis of the equations described in the Materials and Methods (51). Similar to the conventional axial-order parameter, the lateral-order values vary from zero (disordered) to one (ordered). For POPC/POPG liposomes in the presence of AA1, the 〈g_xx〉, 〈g_yy〉, and 〈g_zz〉 values determined from the EPR spectra are 2.0075, 2.0062, and 2.0029, respectively. The calculated lateral order is 0.52, which indicates significant lateral ordering. No apparent lateral ordering was observed in the absence of AA1. Interestingly, the conventional axial-order parameter only increases from 0.63 to 0.70 upon AA1 binding. This suggests that AA1 binding mainly inhibits lipid lateral motion and induces lipid lateral pressure.

Figure 4.

Lipid lateral ordering. 94 GHz EPR spectra of POPC/POPG (A) and POPC (B) liposomes with 5-SASL in the absence and presence of AA1. Lateral ordering was detected for (A) with arrows showing the 〈g_xx〉, 〈g_yy〉, and 〈g_zz〉 components. The L/P is 10. No sign of lateral ordering was observed for (B). (C) Membrane structure showing transverse and lateral order and the corresponding g-factor anisotropy, indicated by g_xx, g_yy, and g_zz. The averaged principle axes of a nitroxide spin label (5-SASL) aligned with regard to a bilayer are shown. Changes in lipid lateral order are reflected in the x and y components of the g tensors of the spin label. To see this figure in color, go online.

Accessibility changes upon AA1 binding

Lastly, AA1-induced accessibility changes were determined by EPR power saturation techniques (50). Because of their amphiphilic nature, an intrinsic aspect of lipid membranes is a polarity gradient established across the bilayer with a largely nonpolar environment inside the bilayer and a polar environment on the membrane surface. As a result, the lipid acyl chains are more accessible to nonpolar reagents than to polar reagents. This polarity gradient and accessibility profile can be altered by the binding of proteins depending on the extent and nature of protein–lipid interactions (37). This accessibility profile can be measured by EPR spectroscopy using a microwave power saturation technique (50). Briefly, with increasing microwave power, the EPR signal becomes saturated, which is described in terms of a power saturation parameter P_1/2. The presence of polar or nonpolar relaxing reagents in the vicinity of spin probes changes the P_1/2 values due to the collision between the relaxing reagents and spin probes. Based on the P_1/2 values, the accessibility of spin probes to relaxing reagents can be estimated in terms of an accessibility parameter, Π.

To study the effect of AA1 on the accessibility profile of lipids, Π-parameters in the presence of O₂ (nonpolar) and NiEDDA (polar) reagents were determined. The results of 5-SASL-labeled POPC/POPG liposomes are shown in Fig. 5. The data revealed that 5-SASL becomes less accessible to both O₂ and NiEDDA in the presence of AA1. This indicates that peptide binding reduces the collision rate between the spin probes and the relaxing reagents, most likely by extensive surface interactions. On the other hand, minimal accessibility changes of 5-SASL were found for POPC and POPC/30% CHOL liposomes upon AA1 binding, shown in Fig. 5 C. Specifically, at an L/P of 10, POPC/POPG liposomes show ∼30% accessibility changes upon peptide binding, while POPC and cholesterol-containing membranes show <5% changes. Therefore, AA1 affects greater accessibility changes of 5-SASL for bacterial-mimic membranes versus mammalian-mimic membranes. Considering that 5-SASL is negatively charged at neutral pH, an uncharged spin-labeled lipid analog (5-PC) was also used to determine the peptide-induced accessibility changes of POPC/POPG liposomes. Similar to 5-SASL, the accessibility of both O₂ and NiEDDA to 5-PC were reduced for POPC/POPG liposomes in the presence of the peptides (Fig. S11). Moreover, reduced accessibility was also observed for POPC/POPG liposomes containing N-TP (Fig. S11). Taken together, similar to the fluidity data, the above results suggest that AA1 binding reduces solvent accessibility around the headgroup region of POPC/POPG liposomes.

Figure 5.

Accessibility changes upon the binding of AA1. Changes in O₂ (A) and NiEDDA (B) accessibility of POPC/POPG liposomes with 5-SASL on the addition of AA1 are shown. (C) Comparison of the percentage of the O₂ accessibility changes of POPC/POPG, POPC, and POPC/30% CHOL liposomes upon AA1 binding.

In addition, to investigate the effect of AA1 binding on the hydrophobic core of lipid bilayers, we performed accessibility measurements of POPC/POPG liposomes using 7-PC, 10-PC, and 12-PC. AA1 induces similar changes in O₂ accessibility to these spin-labeled lipids as in 5-PC (i.e., decreased Π-values upon peptide binding). However, AA1 binding increases the Π-values of NiEDDA for these three spin-labeled lipids (Fig. S11). The Π-parameters were used to calculate the membrane depth parameter, Φ (Fig. 6). As expected, in the absence of peptides, Φ-values are increased from 7-PC to 12-PC due to the polar gradient across the bilayer. Remarkably, the Φ-values appear to remain constant for 7-PC, 10-PC, and 12-PC after peptide binding (Fig. 6). This indicates membrane thinning and acyl-chain overlay, as elaborated below in the Discussion.

Figure 6.

Depth parameter changes of 7-PC, 10-PC, and 12-PC upon AA1 binding to POPC/POPG liposomes when compared to bare liposomes.

Discussion

AA1 selectively interacts and disrupts anionic bacterial membranes

Membrane permeability

The antibacterial mechanisms of AMPs have been linked to their capabilities to increase membrane permeability, thereby causing the leakage of cellular content and eventually cell death. Here, AA1 was found selectively to induce large permeability changes in anionic POPC/POPG liposomes mimicking bacterial membranes (Fig. 2), which is consistent with the peptide’s efficacious antimicrobial activities against diverse bacterial strains. Notably, the data show incomplete leakage phenomena (<100%) with a higher percentage of permeability changes when decreasing L/Ps from 80:1 to 10:1 (Fig. 2). These ratios seem to be high, but it is not surprising. Similar L/Ps have been observed in previous studies on AMPs. For example, membrane permeability changes induced by LF11, magainin 2, and LL37 were observed at L/Ps ranging from 10:1 to 50:1 (61, 63, 64). Moreover, geometrical studies have demonstrated that the effective AMP concentration in vivo could be two orders-of-magnitude higher than the bacterial lipid concentration (reviewed in Melo and Castanho (65)). To be noted, the AA1-induced permeability changes include a fast phase (several minutes) followed by a much slower phase (a few hours). Incomplete leakage along with a fast rate may be due to transient pore formation or temporary failure of the bilayer structure upon peptide binding (66). The slow phase might be a result of peptide fluctuation between liposomes, as shown in Fig. S4. Interestingly, the permeability changes were reduced for POPC liposomes in the presence of cholesterol. These results argue that lipo-cyclic-γ-AApeptides have the ability to preferentially disrupt bacterial membrane versus mammalian membrane.

Membrane fluidity and lateral motion

Upon AA1 binding, the EPR spectra show decreased motion of lipids across the bilayer, with the greatest mobility changes being around the headgroup region (Fig. 3 and Figs. S5, S6, S7, S8, and S9). This rigidifying effect on lipid bilayers has been observed in the studies of other AMPs such as CM15 (37), and cecropins B1 and B3 (38). EPR spectra at 94 GHz suggest that AA1 induces significant lateral ordering of lipid molecules in POPC/POPG membranes, suggesting reduced lipid motion in the lateral direction perpendicular to the membrane normal. The reduced lateral motion may be an indicator of tighter lipid lateral packing and lateral expansion of lipid bilayers (Fig. 7). AMPs such as Gramicidin S and PGLa have been suggested to cause lateral pressure and lateral expansion of membranes by embedding in the lipids (59). This lateral expansion effect is accompanied by the increasing lateral area of one side of a bilayer, which has been related to the thinning of lipid membranes (59). Taken together, AA1 binds strongly to anionic membranes, decreases membrane fluidity, reduces lipid lateral motion, and possibly induces lipid lateral expansion.

Figure 7.

Membrane interaction and disruption mechanism of AA1. (A) AA1 binds to the membrane through electrostatic and hydrophobic interactions. (B) Insertion of the bulky hydrophobic groups of AA1 into the membrane results in lateral expansion of the upper leaflet of a lipid bilayer. (C) Lipid lateral expansion leads to membrane thinning. Moreover, peptide insertion causes transient membrane disruption or local bilayer disorder, subsequently leading to permeability changes. To see this figure in color, go online.

Solvent accessibility and membrane thinning

EPR power saturation data revealed that AA1 binding significantly decreases the accessibility of POPC/POPG membranes to both polar (NiEDDA) and nonpolar (O₂) reagents for N-TP, 5-SASL, and 5-PC (Figs. 5 and S11). The decreased accessibility is a result of the reduced concentrations of relaxing reagents (O₂ and NiEDDA) in the vicinity of spin probes, which lead to a decreased collision rate between the relaxing reagents and the spin-labeled lipid chains. Most likely, the extensive interactions of AA1 on the surface and headgroups of membranes block the diffusion of the relaxation reagents to the lipids, thereby reducing solvent accessibility to headgroups (N-TP) and atoms adjacent to headgroups (5-SASL and 5-PC). An interesting observation was the different trend of accessibility changes in the hydrophobic core of the POPC/POPG liposomes as determined by 7-PC, 10-PC, and 12-PC. The Π-values of O₂ are decreased while the values of NiEDDA are increased for these positions upon AA1 binding, which indicates increased polarity of these positions. Moreover, in the presence of AA1, the polarity gradient between 7-PC, 10-PC, and 12-PC is diminished and the depth parameters (Φ) of these three spin labels become comparable (Fig. 6). A plausible explanation for these observations is membrane thinning after AA1 peptides are inserted into the lipid membranes (Fig. 7). Intercalation of the peptides into the membrane leads to an increase in the lateral area of one leaflet of a bilayer, providing more space for the lipids of the other leaflet to penetrate into the acyl chains, subsequently causing the thinning of the membrane (Fig. 7). Because of membrane thinning and more acyl-chain overlays, the depth parameters of 7-PC, 10-PC, and 12-PC are averaged and become similar to each other. The thinning of membranes is also accompanied by decreased lipid motion because of the higher density of lipid packing in the acyl-chain region (61). This agrees well with the decreased lipid mobility of 7-, 10-, and 12-PC upon AA1 binding (Fig. S8).

Preventive effect of cholesterol on mammalian cells

Our results argue that the large quantity of cholesterol in mammalian cell membranes might contribute to their resistance to the membrane perturbation caused by AA1. This is supported by the reduced membrane permeability, mobility, and accessibility changes in the presence of cholesterol. These data also agree with the results that AA1 is able to target bacterial cells selectively with low hemolytic activity. The presence of cholesterol stabilizes the membrane structure by increasing membrane order and lateral organization. Cholesterol might inhibit the interaction of AA1 with the membrane, prevent correct docking of the peptides to the membrane, and subsequently affect the membrane-disruptive functions of the peptides. For instance, the role of cholesterol in protecting erythrocytes from antimicrobial activity has also been found in studies on magainin 2 and Melittin (67, 68). In addition, AMPs have been found binding preferably to disordered domains lacking cholesterol, as shown in the studies of MSI peptides and δ-lysin (66, 69, 70). In summary, our results show that cholesterol has a preventive effect for mammalian cell membranes and plays an important role in the selective membrane-disruptive activities of lipo-cyclic-γ-AApeptides.

Mechanism of bacterial membrane disruption by AA1

Based on the EPR data in this study, the membrane-disruption mechanism of AA1 can be explained using the carpet model (15). The carpet model of AA1 is supported by the following EPR observations: (1) Lipid fluidity and ordering—as shown in Figs. 3 and 4, reduced lipid fluidity and increased lipid ordering were observed upon peptide binding, which indicates that AA1 is able to rigidify the lipid molecules. Moreover, this immobilization effect is greater around headgroup regions than in the acyl-chain region (7-PC, 10-PC, and 12-PC), suggesting that the peptides are not deeply inserted into the bilayer; this is consistent with the carpet model. (2) Solvent accessibility—extensive surface interaction and peptide accumulation can also lead to decreased accessibility of the lipid molecules to both polar and nonpolar reagents in the solvent, as illustrated in Figs. 5 and S11. (3) Membrane thinning—in the carpet model, the high density of peptides binding on the membrane surface may lead to thinning of the membrane, as shown for CM15 (37) and magainin 2 (71). This is confirmed by the EPR depth data (Fig. 6), which indicates reduced thickness of the POPC/POPG membranes. (4) Permeability—the permeability data show an incomplete and fast leakage event upon AA1 binding. This indicates a transient disruption of the bilayer, which has been suggested for the carpet model (66). On the other hand, a long-lasting transmembrane pore (such as a barrel-stave or toroidal-pore) is unlikely to be formed because the short hydrophobic groups of AA1 are not long enough to span the bilayer. To be noted, a complete disruption of the membrane or a detergent effect may occur in the carpet mechanism (66). However, this is not consistent with the immobilization of lipid chains across the bilayer (Fig. 3 and Figs. S5, S6, S7, S8, and S9) and the decreased solvent accessibility to 5-PC and N-TP upon AA1 binding (Fig. S11). In the case of the interfacial-activity model, peptide binding would not lead to rigidification of the lipid headgroups and steric hindrance of O₂ and NiEDDA as shown upon AA1 binding. Therefore, the mobility and accessibility changes induced by AA1 are not consistent with the interfacial-activity model. In addition, the lipid-clustering model is not supported by the EPR data. No significant spin-spin interaction or lipid clustering was observed for POPC/POPG liposomes upon AA1 binding, as detected by 5-PC spectra at 200 K (not shown). Taken together, the membrane disruption mechanism of AA1 can be explained using the carpet model.

Specifically, the binding of AA1 is initiated by the electrostatic interactions between the cationic side chains of the peptides and the anionic groups in lipid vesicles, which is evident in the stronger interaction of AA1 with POPC/POPG liposomes when compared with POPC liposomes (Fig. 3). This is also confirmed by the greater mobility changes of the 5-SASL containing liposomes than its uncharged analog 5-MeSL (Fig. S10). This step is followed by the insertion of the long carbon tail and the hydrophobic side chains on the cyclic ring of the peptide into the lipid bilayer. Moreover, the strong electrostatic and hydrophobic interactions decrease the membrane fluidity and increase the ordering of the lipid molecules, resulting in reduced accessibility of lipid headgroups to both polar and nonpolar molecules in the solvent. In addition, the insertion of the bulky hydrophobic groups of the peptide into the membrane may lead to increased lateral pressure, thereby inducing lateral expansion and membrane thinning (Fig. 7). Furthermore, peptide penetration also creates transient membrane disruption or local disorder, subsequently causing leakage of cellular contents and cell death.

In summary, AA1 features that are important for its antimicrobial activities have been revealed, and such information offers insight for designing new antibiotic agents. Our structural and biophysical data suggest that the hydrophobic side chains of AA1 function as the major membrane-disruptive groups. Bulky hydrophobic groups might be more effective to induce lateral expansion and membrane thinning, such as the adamantyl group, the naphthyl group, etc. In addition, a longer tail might be able to insert the aromatic rings deeper into the lipid acyl chains and more efficiently promote lateral expansion and membrane thinning. As such, a C18 alkyl tail may be employed in a future design. Moreover, the data show that the cationic groups of AA1 account for its selective bacterial-membrane targeting capabilities. Therefore, it is important to maintain the amphipathic building blocks with balanced polar and nonpolar groups, i.e., one cationic and one hydrophobic side chain. Furthermore, the data demonstrate that lipo-cyclic peptides can be effective AMPs. With its more rigid ring structure, a cyclic peptide contains a more stable amphiphilic structure than does a linear peptide. The size of the ring might affect a peptide’s capability to induce membrane property changes and this can be further fine-tuned. For instance, a bicyclic ring may be used to increase the area interacting with bacterial membranes, while its rigidity can be retained due to bicyclization. Our results demonstrate that understanding the detailed AMP mechanism at a molecular level will facilitate new antibacterial biomaterial designs.

Conclusions

With an aim to understand the molecular basis of the antimicrobial activity of lipo-cyclic-γ-AApeptides, we investigated the permeability and membrane property changes induced by our lead peptide—AA1. The peptide selectively permeates and structurally modifies negatively charged bacterial-mimic membranes. In contrast, cholesterol-containing neutral membranes mimicking mammalian cells were minimally affected by the molecule. Based on combined analyses of membrane permeability, dynamics, membrane thinning, and accessibility, we proposed a carpet-like mechanism for the antimicrobial activities of AA1. The results provide implications for the development of effective AMPs with robust antibacterial activities against antibiotic-resistant microbes.

Author Contributions

P.K. performed research, analyzed data, and wrote the article; Y.L. synthesized material; J.C. designed research and wrote the article; and L.S. designed research, analyzed data, and wrote the article.

Editor: David Cafiso.