Role of lipopolysaccharide in antimicrobial and cell penetrating peptide membrane interactions probed by deuterium NMR of whole cells

Highlights

• •LPS reduces membrane damage by the AMP MSI-78.
• •LPS disruption by EDTA has a small but statistically significant effect on acyl chain disorder.
• •Peptide-induced acyl chain disruption is similar for the AMP MSI-78 and the CPP TP2.
• •LPS slightly increases membrane perturbation by the CPP TP2.

Abstract

Understanding how non-lipid components of bacteria affect antimicrobial peptide (AMP)-induced membrane disruption is important for a comprehensive understanding of AMP mechanisms and informing AMP-based drug development. This study investigates how lipopolysaccharide (LPS) affects membrane disruption by the AMP MSI-78 and compares the results to the effect of TP2, a cell-penetrating peptide that crosses membrane bilayers without permeabilizing them. We destabilize the LPS layer of Escherichia coli (E. coli) cells via chelation of the stabilizing divalent cations. ²H NMR spectra of E. coli demonstrate that EDTA concentrations of 2.5 mM and 9.0 mM alone have very minor effects on lipid acyl chain order. Interestingly, we find that E. coli pre-treated with 9.0 mM EDTA before treatment with MSI-78 are more sensitive to AMP-induced acyl chain disruption, indicating that intact LPS reduces MSI-78-induced membrane disruption in E. coli. Surprisingly, we also found that at the level of ²H_NMR, the peptide-induced acyl chain disruption is similar for MSI-78 and TP2, although MSI-78 permeabilizes the bilayer and TP2 does not. Furthermore, LPS disruption appears to protect the bacteria from TP2, although it sensitizes them to MSI-78.

Graphical abstract

1. Introduction

Antimicrobial peptides (AMPs) have long been studied as a possible novel class of antibiotics that could be used against multidrug-resistant pathogens [1], [2], [3], [4], [5], [6]. AMPs are innate immune effectors found in a wide range of animals and plants. They commonly have broad-spectrum activity against many strains of Gram-positive and Gram-negative bacteria, fungi and viruses [1,7]. AMPs often exert their antimicrobial activity by disrupting the bacterial cytoplasmic membrane to form pores that allow molecules to pass from one side of the bilayer to the other [8], [9], [10]. This membrane permeabilization mechanism does not involve any specific protein target, and it is thus hard for pathogens to develop bacterial resistance. Hence, AMPs can also be effective against antibiotic-resistant strains [11].

Several models have been proposed to explain AMPs' lipid bilayer permeabilizing mechanisms. For example, AMPs may act by forming membrane-spanning pores like toroidal or barrel-stave pores [12], [13], [14]. In addition to the pore models, some AMPs' activity has also been described by non-pore models. For example, the carpet model is one commonly cited membrane destabilization model by AMPs [15,16]. Regardless of the mechanism, permeabilizing AMPs compromise lipid bilayer integrity, leading to bacterial growth inhibition and death [13,15,17].

At present, it is well established that several AMPs have non-lipid targets. AMPs with intracellular targets cross the cytoplasmic membrane and accumulate inside cells, inhibiting critical cellular processes. For example, at their minimal inhibitory concentration (MIC), pleurocidin and dermaseptin inhibit nucleic acid and protein synthesis without damaging the cytoplasmic membrane [14]. Many non-lytic mechanisms involving intracellular targets have been discovered, including inhibiting protein or nucleic acid synthesis and disrupting protein activity, including enzymatic activity [14,18,19].

Despite this substantial progress in understanding AMP mechanisms, important elements of AMP action are not well understood. These include how AMPs traverse the non-lipid components of the cell envelope to reach the bilayer, either to permeabilize it or as the first step in reaching an intracellular target. This question has seen some attention but relatively little compared to AMP-lipid interactions. Most of the work on AMP-induced permeabilization has been performed with model lipid bilayers, which lack the peptidoglycan (PGN) and lipopolysaccharide (LPS) usually present in bacterial cell envelopes. LPS is difficult to study in a model system in large part because it is challenging to make model membranes asymmetric. One way researchers have sought to assess the importance of such non-lipid interactions is to compare the AMP to lipid molar (AMP:L) ratios that cause effects in model lipid bilayers, i.e. without the non-lipid cell envelope components, to the AMP:L ratio needed to harm cells in actual bacteria where the non-lipid cell envelope components are present.

A variety of efforts have estimated or measured the AMP:L ratio needed to see growth inhibition in cells. Wimley [20] estimated that the AMP:L ratio required to see permeabilization in lipid vesicle experiments was about 1:200, while growth inhibition in cells required a cell-bound peptide AMP:L of around 10–100:1. Using an alternate approach, Castanho and co-workers used partition constants [21] to compare model lipid studies with whole-cell studies. Examining two AMPs, melittin and omiganan, they found that the cell-bound AMP:L ratio was 2.3–9.2 times higher than the threshold needed to see membrane permeabilizing effects on lipid vesicles. Since then, the Stella group [22] has developed a helpful experimental approach using a special minimal medium where the bacteria are metabolically active but do not multiply. They studied bactericidal activity against Escherichia coli (E. coli) as well as AMP-cell association and showed that, at the minimum bactericidal concentration (MBC), 10⁷ fluorescently labelled AMP molecules (dansylated PMAP23) are bound to each cell. This corresponds to an AMP:L ratio of ∼1:3 to 5:1. A few other studies have also pointed out that bacterial killing can require 10⁶ to 10⁸ molecules per cell [23,24] which corresponds to AMP: L ratios of ∼1:3 to 5:1, depending on the peptide and bacterial cell type.

Altogether, these findings force us to start questioning why there is a difference in the AMP:L ratio required to show effects in the in vitro and in vivo studies. There are several possible reasons for these differences. For example, some peptides may bind to cellular components other than the bacterial membrane, such as PGN, LPS, teichoic acids, and membrane proteins, significantly impair the cell wall structure and stability of the cell membrane [18,[25], [26], [27]]. The cell envelope of Gram-negative bacteria like E. coli is a complex structure composed of two lipid membrane layers. The outer leaflet of the outer membrane contains LPS. It thus has a sizable carbohydrate component which may capture peptides, sequestering them away from the lipid bilayer and protecting the cell [28]. It has also been suggested, however, that negatively charged LPS may attract more peptides to the cell envelope, resulting in more peptide accumulation on the lipid bilayer and thus potentially more cell damage [27]. The potential importance of the LPS layer has been underlined by findings that AMP Maculatin 1.1 and Melittin are much more efficient at inhibiting growth of E. coli genetically modified to alter their LPS compared to normal E. coli [29].

It is thus necessary to build a bridge that connects the in vitro work on AMPs interacting with model lipid bilayers to the findings that at least some AMPs interact with non-lipid cell envelope components. Several biophysical approaches, for example, atomic force microscopy [30] and Fourier Transform Infrared (FTIR) spectroscopy [31], differential scanning calorimetry (DSC) [18], as well as confocal and electron microscopy [32,33], have been used to help bridge the gap between studying AMP mechanisms in model lipid bilayers with the mechanisms at play in real, intact cells. One tactic is to extend the techniques used to perform experiments in model lipid systems to whole bacteria. For example, ³¹P NMR has been used with whole cells to investigate AMP-induced changes in bacteria's nucleic acid and lipid headgroups [34]. ²H NMR is a technique commonly used with model lipid systems to study perturbations, including bilayer disruptions, that alter lipid acyl chain orientational order. Since AMP-induced permeabilization of lipid bilayer affects the lipid acyl chains, ²H NMR can be used to assess changes in lipid chain orientational order resulting from membrane disruption by AMPs [35], [36], [37], [38], [39]. The Davis group obtained the first ²H NMR spectra of membrane deuterated bacteria in 1979 [40]. Our group and others have since applied whole-cell ²H NMR to bacteria treated with AMPs and refined protocols to incorporate deuterium-labelled acyl chains into the bacterial membrane [41], [42], [43], [44], [45]. These approaches have been used to investigate changes in the ²H NMR spectra of whole cell bacteria induced by AMPs such as MSI-78, BP100 [40], [41], caerin 1.1, aurein 1.2 [44], and CAME (cecropin A (1-8) melittin (1-10)) [45,46]. These peptides may not have the same mechanism of action, but MSI-78 at least has been shown to have an intracellular target [18] in addition to its membrane-permeabilizing mechanism, and many other AMPs have been shown to possess intracellular targets [47,48].

In the present study, we investigate how LPS affects AMP-induced membrane disruption by the AMP MSI-78 and the cell-penetrating peptide (CPP) TP2 in E. coli. MSI-78 is a synthetic analogue of magainin 2, an AMP found in the skin of African frog [49]. It is a potent antibiotic with a MIC value of ≤64 μg/ml [50], a net cationic charge of +9, and a grand average hydropathicity index of −0.159 [51]. TP2 is a cell-penetrating peptide (CPP) and contains the motif, LRLLR, called the spontaneous membrane translocation peptide (SMTP). TP2 is substantially more hydrophobic than MSI-78, with a grand average hydropathicity index of 0.423 and a +3 charge at neutral pH [52]. More generally, CPPs are peptides that can cross the bilayer spontaneously and enter cells by translocating directly across the membrane [52], [53], [54]. In contrast to AMPs, CPPs at moderate concentrations generally interact with lipid bilayers without membrane permeabilization, at least as measured by the ability of the fluorescent dye to cross the membrane [52,53]. In this ²H NMR study of intact cells, the effect of MSI-78 was compared to that of the CPP TP2.

The focus of this study was to assess the influence of LPS on the ability of the AMP and CPP to cause disordering of the lipid acyl chains in the membranes of the bacteria. This was done by comparing peptide-induced membrane disruption with and without a pre-treatment with a chelating agent to destabilize the LPS layer of the cells before the peptide treatment was applied.

2. Material and methods

2.1. Materials

Deuterated palmitic acid (PA-d₃₁) was acquired from CDN Isotopes (Pointe-Claire, QC, Canada). n-dodecyl phosphocholine (DPC) was purchased from Avanti Polar Lipids (Alabaster, AL, USA). Oleic acid (OA), ethylenediaminetetraacetic acid (EDTA), propidium iodide, yeast extract, tryptone and sodium chloride were purchased from Thermo Fisher Scientific BioReagents™ (Markham, ON, Canada). MSI-78’s sequence is GIGKFLKKAKKFGKAFVKILKK-NH₂ and TP2’s is PLIYLRLLRGQWC-NH₂.

2.2. Peptide preparation

MSI-78, and TP2 peptides, C-terminal amidated, were obtained from GenScript and HPLC-purified to purity >75 %. MSI-78 and TP2 were desalted by buffer exchange using Spectrum™Spectra/Por™ dialysis membrane tubing (100–500 and 1000 Da MWCO). The dry peptide was dissolved with a small volume of dialysate (5 % acetic acid and 95 % water) and added to the membrane tubing. Peptides were dialyzed against the dialysate for 24 h at 4 °C. Next, peptides were dialyzed against 100 % purified water for another 24 h. Dialyzed peptides were lyophilized, weighed, and then stored at −20 °C.

2.3. Preparation of membrane deuterated E. coli

JM109 (ATCC-68862) strain E. coli bacteria were grown with a PA-d₃₁-DPC mixture incorporated in the growth medium as described before [45] and outlined below.

Overnight cultures of JM109 were prepared by inoculating 20 mL of Luria Broth (LB) media (10 g/L tryptone, 5 g/L yeast extract and 5 g/L NaCl) with a 1 mL aliquot of frozen glycerol cell stock, incubating at 30 °C with a shaking speed of 150 rpm. The LB media contain 205 µM calcium and 296 µM magnesium (based on Sigma-Aldrich yeast extract product information). Large-scale cultures (200 mL) were grown at 37 °C and 175 rpm starting with 2 mL of overnight culture in 200 mL of fresh LB media containing 0.25 mM deuterated palmitic acid (PA-d₃₁) complexed with 1 mM n-dodecyl phosphocholine (DPC) and 0.25 mM OA/DPC.

To prepare the fatty acid (PA-d₃₁ or OA)/DPC complexes, a 5 ml solution containing 0.25 mM of PA-d₃₁ or unlabelled OA and 1 mM DPC (final concentrations) was prepared with distilled water. Next, using a 0.2-micron syringe filter unit, each solution was filtered and transferred to a new 50 mL Falcon tube. Subsequently, the mixtures were heated in a boiling water bath for 3 min and then submerged in liquid nitrogen until each mixture froze solid. Next, both tubes were warmed in a water bath at room temperature until the ice melted. Finally, the mixture containing PA-d₃₁ complexed with DPC micelles and OA was added immediately to the growth media as the large-scale (200 ml) culture was started.

Cells were harvested in the mid-log phase (after ∼3.5 h of growth), at an absorbance at 600 nm (A600) of 0.6–0.8, and then pelleted by centrifugation at 5670 × g for 20 min at 4°C. The pellet was transferred into a 3.2 mm MAS NMR rotor, and NMR experiments were carried out as described in Section 2.5 below. Samples with EDTA and/or AMP treatment required additional pre-NMR steps that are described in Section 2.4.

2.4. AMP treatment

AMP amounts were added as a % of the dry weight of AMP to the dry weight of bacteria, which facilitates comparing molar AMP:L ratios with other studies [41,45]. Because the amount of cell pellet differs from preparation to preparation, a relationship between A₆₀₀, just before centrifugation to harvest the cells, and the bacteria dry weight was required and obtained as follows.

To derive this relationship, cells were grown under conditions explained in Section 2.3. Cultures were harvested at different points of the log phase and pelleted by centrifugation at 5670 × g for 20 min. The final pellet was weighed and then placed in a vacuum chamber for 48 h. After 48 h, each pellet was re-weighed. A linear fit of ln(dry weight) versus the absorption at 600 nm (A₆₀₀) gives the relationship equation that can be used to calculate the bacterial dry weight and the appropriate amount of AMP to add to each bacterial sample. For JM109 E. coli the relationship obtained was ln (dry weight) = (1.28 ± 0.28) A₆₀₀ + (5.18 ± 0.28), which yields the dry weight in mg per litre of cells grown. For example, at A₆₀₀ optical density 0.62, 23 mg of the peptide was used to obtain 30 % dry weight of AMP to dry weight of bacteria for 0.2 \L of bacterial growth.

While cells were being harvested by centrifugation as described above, the appropriate amount of peptide was calculated then weighed and suspended in 30 mL of LB medium. The media + AMP mixture was then used to resuspend the bacterial pellet, which was then incubated at room temperature for 20 min while undergoing mild shaking on a benchtop shaker. After 20 min, the sample was centrifuged at 5670 × g for 20 min at 4 °C. Finally, the resulting pellet was transferred into the 3.2 mm NMR rotor as detailed in Section 2.6. Bradford assays have been performed on the supernatant after AMP treatment to show that most of the AMP binds the cells and with little found in the supernatant (data not shown).

2.5. EDTA treatment

EDTA chelates the divalent cations that help stabilize the LPS layer of E. coli and thus destabilizes the LPS [55]. For this study we needed to establish an EDTA treatment protocol that would destabilize the LPS gently, without lysing the cells. To accomplish this, we treated E. coli for 45 min with a range of EDTA concentrations from 1.5 to 9.0 mM and observed them under a microscope by Gram staining and by flow cytometry. The E. coli exhibited only minor changes up to 8.0 mM EDTA but showed changes in the shape and Gram-staining colour at 9.0 mM EDTA. Thus, for the rest of the studies 2.5 mM was chosen for the lower concentration of EDTA and 9.0 mM for the upper concentration of EDTA.

For the EDTA treated samples in Figs. 3-5, bacteria were grown and harvested at A₆₀₀ values between 0.6 and 0.8 and then pelleted by centrifugation at 5670 × g for 20 min at 4 °C. The resulting pellet was resuspended in 50 mL of LB media with 2.5 mM or 9.0 mM EDTA at room temperature for 45 min with gentle shaking on a benchtop shaker. For the combined treatment with both EDTA and AMP, cells were first treated with 2.5 mM or 9.0 mM EDTA and then pelleted by centrifugation at 5670 × g for 20 min at 4 °C followed by AMP treatment as described above for 20 min. Finally, the sample was centrifuged at 5670 × g for 20 min.

Fig. 3.

EDTA effects on lipid chain order, cell permeability and cell wall staining. A: ²H NMR spectra of deuterium-enriched E. coli that are untreated (black, SK-53); treated with 2.5 mM EDTA (blue, SK-38) and treated with 9 mM EDTA (red, SK-39) Dashed lines at ±12.5 kHz are included to facilitate the comparison of the spectra. Each spectrum is obtained from 110,000 scans recorded over 12 h at 37 °C in a 600 MHz NMR spectrometer and normalized by area. The top panel shows the overlaid spectra. Additional spectra obtained from samples with EDTA added are shown in Figs. S1 and S2. B: Cell count versus PI fluorescence intensity for E. coli cells treated with 2.5 mM and 9.0 mM EDTA as probed by flow cytometry with PI. The relatively small count numbers for high PI fluorescence intensity show that PI cannot penetrate 2.5 mM and 9.0 mM EDTA treated cells. C: Microscope images of cells treated with 9.0 mM EDTA and 2.5 mM EDTA after Gram staining. These show that 9.0 mM EDTA (but not 2.5 mM EDTA) treatment drastically changes bacterial shape and susceptibility to Gram staining. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 5.

CPP peptide increases the amplitude of lipid chain motions but does not induce membrane permeabilization. A: ²H NMR spectra of deuterium-enriched E. coli that are untreated (black, SK-53); after treatment with 30 % TP2 (pink, SK-71); and after treatment with 30 % MSI-78 (purple, SK-55) B:²H NMR spectra of deuterium-enriched E. coli that are untreated (black, SK-53); after treatment with 30 % TP2 (pink, SK-71) and after treatment with 9 mM EDTA followed by 30 % TP2 (orange, SK-72). Dashed lines at ±12.5 kHz are shown to facilitate the comparison of the spectra. Each spectrum is obtained from 110,000 scans recorded over 12 h at 37 °C in a 600 MHz NMR spectrometer and normalized by area. Only selected spectra are shown; additional spectra are shown in Fig. S6. C: Flow cytometry cell count vs. PI fluorescence intensity for E. coli cells treated with 30 % TP2 and 9.0 mM EDTA+TP2. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

2.6. Packing the rotor for NMR

The NMR rotor was cleaned, washed, and allowed to dry before use. The weight of the bacterial paste was determined by weighing the rotor before and after the bacterial paste (∼40 mg) was added. The bacterial pellet was transferred to the 3.2 mm rotor with the help of 200 μL pipette tips. First, the pipette tip was trimmed slightly to enlarge the opening. Then the pipette tip was inserted into the rotor, and both were placed in a 1.5 ml Eppendorf tube. The bacterial paste was transferred into the pipette tip with a spatula, and the assembly was centrifuged for 30 s at 3217 × g to transfer the bacteria into the rotor.

2.7. ²H NMR

²H NMR experiments were performed at 37 °C with a solid-state Bruker Avance II 600 MHz spectrometer (Milton, Ontario, Canada), operating at a frequency of 92.15 MHz for ²H. All experiments were run with a triple resonance (HCD(N)) magic-angle spinning probe using 3.2 mm diameter rotors. Although a MAS rotor and MAS NMR probe were used, all the data presented in this paper were acquired under static conditions (with no spinning).

Static spectra were obtained using the solid-echo pulse sequence [56], with a 90° pulse width of 5 μs and a delay of 60 μs. Spectra were derived from free-induction decays obtained by averaging 20,000 transients acquired with a dwell time of 5 μs (spectral width of 200 kHz) and a repetition delay of 500 ms. The total experiment time for each spectrum was 18 h. Data were accumulated in 5 h blocks so the spectra could be checked for any changes with time in the spectrometer.

To obtain spectra from free-induction decays, (for processing,) the data were left-shifted by 3 points to start the Fourier transform at the top of the echo [56]. Phasing, baseline correction and 40 Hz line broadening were also applied. Spectra were phased using both 0^th and 1st order phase corrections to achieve a symmetric spectrum and symmetry was checked by comparing the moments calculated from the left and right hand-side of the spectra as shown below. For the spectra displayed in the figures, the baseline was corrected by the automatic polynomial fit in Topspin software; for moment calculations a linear baseline was subtracted as described below.

Most of the static spectra showed two peaks near the center, one from the lipids and the other presumably from HDO or some other metabolic product. When two peaks were apparent, the larger peak closest to negative frequency was identified as 0 Hz. This ensured that the prominent edges of the spectra were symmetric at ±12.5 kHz.

To quantify the spectral shape and changes to it caused by the AMP, spectral moments (M_n) were calculated by using the equation

$$M_n=\frac{{\int }_0^{\infty }{\omega }^{n}f\left(\omega \right)d\omega }{{\int }_0^{\infty }f\left(\omega \right)d\left(\omega \right)}$$ (1)

where ω is the angular frequency relative to the Larmor angular frequency, $\omega$₀, and f(ω) is the spectral intensity. The first moment (with $n=1$) is proportional to the average quadrupolar splitting and the average of the deuteron orientational order parameter. The second moment is proportional to the mean square quadrupolar splitting and the mean square order parameter.

The first and second moments, $M_1$and $M_2$ were used to calculate Δ₂ [57] the relative mean square width of the distribution of order parameters. Δ₂ can be defined in terms of M₁ and M₂ as

$${\Delta }_2=\frac{M_2}{1.35\times M_1^2}-1$$ (2)

Spectral moments $M_1$ and $M_2$ from static spectra were calculated using an algorithm previously used in this group [41,42] but implemented through a new MATLAB script as follows. The average intensity value between 32.48 and 38.06 kHz was subtracted from each spectrum for the baseline correction. To remove the effect of the water and metabolite peaks at the center of the spectrum (as described above), the intensity in the ±0.61 kHz range was set uniformly to the average intensity in the 0.61 to 0.88 kHz range. The intensity of the spectrum between 0 and 24.8 kHz was then used to calculate the moments. The moment calculation was then repeated with the negative half of the spectra and the moments obtained from the two halves, normally identical or nearly so, averaged to give the overall moment.

Since each sample preparation and NMR experiment was repeated several times, it became meaningful to carry out a statistical analysis. The moments and Δ₂ values for each composition were averaged and expressed as average ± standard deviation by using Excel data analysis software. Different treatment groups were compared using a one-way ANOVA statistical test.

2.8. Flow Cytometry

Flow cytometry was performed to observe the AMP-induced cell permeability under the same sample preparation conditions used for the NMR work. Control samples were prepared by treating cells with 70 % isopropanol to completely permeabilize the cells and by treating them with buffer alone to serve as healthy cell control. As for NMR, 200 mL of the cell suspension were harvested at OD 0.6 and pelleted by centrifugation at 5670 × g for 20 min at 4 °C. The cell pellet was resuspended in 50 mL of LB media treated with EDTA at 2.5 mM or 9.0 mM concentration for 45 min at room temperature and centrifuged at 5670 × g for 20 min. For the combined EDTA + AMP treatment, the EDTA treatment was performed first, followed by AMP treatment as described in Section 2.4. Regardless of treatment, all cell samples were harvested by centrifugation at 5670 ×g for 20 min at 4 °C. All samples were then diluted in phosphate buffer saline (PBS) (0.137 M NaCl, 0.05 M NaH₂PO_4, pH 7.4) to 2.5 × 10⁵ CFU/ml (colony-forming unit)/mL of bacterial suspension, followed by washing by adding PBS buffer and centrifuging at 5670 ×g, after which the supernatant was discarded. Cells were resuspended in 100 μL of PBS buffer with propidium iodide (PI) as follows.

To measure membrane permeability, 5–10 μL of propidium iodide (final concentration 10 µM) staining solution (PI Staining solution: 10 μg/ml PI in PBS) was added to the bacterial samples. Tubes were gently mixed and incubated at room temperature for 5–10 min in the dark. After the PI incubation, flow cytometry was started within 15 min. Data were acquired for 30 s using a Beckman Coulter CytoFLEX flow cytometer using the ECD channel (610/20) with the 488 nm laser. Data were analyzed using CytExpert version 2.3 software (Beckman Coulter) by drawing a gate to include the significant cell events in the sample. Optical data (forward scatter and side scatter plots) are provided in supplemental Figs. S7 and S8. Each flow cytometry experiment was repeated at least 2 times, starting from new cells.

2.9. Gram staining

Gram staining was used as the method to evidence the destabilizing of the LPS layer of E. coli cell with EDTA treatment. A 1 mL aliquot of the bacterial sample was set aside in a 1.5 mL micro centrifuge tube before the final centrifugation harvest in Section 2.3. These bacterial cells were spread onto a microscope slide using a sterile cooled loop. Next, the cells were heat fixed onto the microscope slide by passing them three times through the flame of the Bunsen burner. The slide was allowed to air dry. Next, crystal violet was gently added to the slide, assuring full coverage of smear, and left to stand for 1 min. The crystal violet was gently rinsed off with distilled water. Next, iodine solution was gently added to the slide and let stand for 1 min. The iodine solution was gently rinsed off with distilled water. The fixed smear was decolorized using 95 % ethanol for about 5 to 10 s or until the ethanol ran almost clear. The ethanol was gently rinsed off with distilled water. Safranin was gently added to the slide, assuring full coverage of smear, and let stand for 45 s. Finally, the safranin was gently rinsed off with distilled water. The slide was placed on paper towel and allowed to air dry. Once the slide was completely dry, bacterial samples were observed under a conventional light microscope using a 100× immersion oil lens which was equipped with in built-in camera.

3. Results

3.1. ²H NMR spectra of E. coli enriched with deuterated acyl chains

These studies employ bacteria that have deuterons in place of hydrogen in the acyl chains of a number of different phospholipid species in their membranes [42,58] in both the inner and outer membranes. ²H NMR spectra are sensitive to the dynamics of the deuterated lipid acyl chains in the sample, particularly the degree of disorder of each segment along the chain. Obtaining consistent spectra from intact membrane-deuterated bacteria can be challenging. Thus, it is important to demonstrate the reproducibility of independently prepared samples before going on to probe how different treatments affect the bacterial spectra. Fig. 1 shows the ²H NMR spectra of three different preparations of untreated membrane-deuterated E. coli from independent cell preparations. As seen in the overlay at the top of Fig. 1, the spectra from the different samples are highly reproducible. All spectra in Fig. 1 display spectral edges at ∼±12.5 kHz; these edges correspond to the quadrupole splitting for acyl chain carbon-deuterium bonds near the headgroup [42,59], which have relatively constrained motions compared to the segments closer to the center of the bilayer.

Fig. 1.

²H NMR spectra of membrane-deuterated E. coli are reproducible. Shown are spectra from 3 independent preparations of bacteria (purple, SK-53; red, SK-37; green, SK-31). Dashed lines at ±12.5 kHz are included to facilitate the comparison of the spectra. Each spectrum is obtained from 110,000 scans recorded over 12 h at 37°C in a 600 MHz NMR spectrometer and normalized by area. The high degree of overlap in the stacked spectra on top shows that the spectra are reproducible. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

The spectral shapes were quantified by calculating their moments, $M_1$and $M_2,$ along with their ∆₂ values. These are shown with averages and standard deviations to indicate the variability between different preparations (top section of Table 1 and Fig. 2). As detailed in the Methods, $M_1$and $M_2$ relate to the order of the lipid acyl chains while ∆₂ describes the shape of the spectrum (as expanded on in Section 3.3 below).The sample labelled as SK-53 had $M_1$and $M_2$ values closest to the average and was thus selected to show in Fig. 3-5 below as a comparison spectrum.

Table 1.

$M_1$, $M_2$, and ∆₂ values for all the spectra with averages and standard deviations. P-values for ANOVA comparison of selected treatment groups are also included. SK-XX is the sample numbers and referenced spectra not shown in the main text are shown in supplementary data. As detailed in Methods, $M_1$ and $M_2$, are measures of lipid disorder and ∆₂ quantifies the shape of the spectra.

Experiment	$M_1$× 10-4 (s-1)	$M_2$× 10-9 (s-2)	Δ2
No treatment SK-31	3.66	2.04	0.12
No treatment SK-37	4.40	3.02	0.15
No treatment SK-53	4.23	2.77	0.15
Average ± Standard deviation	4.09 ± 0.38	2.61 ± 0.50	0.14 ± 0.02
2.5 mM EDTA SK-38	4.47	3.02	0.11
2.5 mM EDTA SK-40	4.32	2.84	0.12
Average ± Standard deviation	4.39 ± 0.11	2.61 ± 0.13	0.11 ± 0.01
9.0 mM EDTA SK-39	4.04	2.64	0.19
9.0 mM EDTA SK-41	4.61	3.33	0.15
9.0 mM EDTA SK-66	4.4	3.22	0.19
Average ± Standard deviation	4.35 ± 0.29	2.73 ± 0.36	0.18 ± 0.02
p-value (no treatment vs. 9.0 mM EDTA)	0.40	0.28	0.09
30 % MSI-78 SK-47	3.60	2.26	0.30
30 % MSI-78 SK-57	3.29	2.05	0.36
30 % MSI-78 SK-64	3.44	2.11	0.32
Average ± Standard deviation	3.44 ± 0.16	2.14 ± 0.11	0.33 ± 0.03
p-value (no treatment vs. MSI-78)	0.05	0.19	0.0008
2.5 mM EDTA+30 % MSI-78 SK-49	3.23	2.43	0.32
2.5 mM EDTA+30 % MSI-78 SK-54Average ± Standard deviation	3.733.48 ± 0.35	2.522.45 ± 0.06	0.340.33 ± 0.01
9.0 mM EDTA+30 % MSI-78 SK-55	3.16	1.86	0.38
9.0 mM EDTA+30 % MSI-78 SK-67	3.09	1.82	0.41
Average ± Standard deviation	3.12 ± 0.05	1.84 ± 0.03	0.39 ± 0.02
p-value (9.0 mM EDTA alone vs 9.0 mM EDTA+30 % MSI-78)	0.01	0.03	0.002
p-value (30 % MSI-78 vs. 9.0 mM EDTA+ 30% MSI-78)	0.07	0.02	0.07
30 % TP2 SK-71	3.46	2.08	0.30
30 % TP2 SK-75	3.47	2.10	0.29
Average ± Standard deviation	3.46 ± 0.01	2.09 ± 0.01	0.29 ± 0.00
p-value (30 % MSI-78 vs. 30 % TP2)	0.87	0.59	0.11
9.0 mM EDTA + 30 %TP2 SK-72	4.13	2.89	0.25
9.0 mM EDTA +30 %TP2 SK-77	3.53	2.11	0.25
Average ± Standard deviation	3.83 ± 0.42	2.50 ± 0.55	0.25 ± 0.00
p-value (9.0 mM EDTA + 30%MSI-78 vs. 9.0 mM EDTA+ 30 % TP2)	0.14	0.23	0.01
p-value (30 % TP2 vs. 9.0 mM EDTA + 30 % TP2)	0.35	0.40	0.03

Fig. 2.

$M_1$, $M_2$, and ∆₂ values for all the spectra with averages and standard deviations. P-values for ANOVA comparison of treatment groups are indicated for all comparisons where the p-value is 0.07 or smaller.

3.2. Effects of LPS destabilization on cell wall staining, lipid chain order, and permeability

Once the reproducibility of the ²H NMR spectra of E. coli was confirmed, we sought to apply an EDTA treatment to gently disrupt the LPS without killing the cells. We (Section 2.5) and others [55] have found that 2.5 mM and 9.0 mM EDTA concentrations bracket the range of gentle LPS disruption. Thus, we chose to characterize EDTA's effects on bacterial acyl chain orientational order, using deuterium NMR, at the lower, 2.5 mM, and upper, 9.0 mM, extremes of this EDTA concentration range. Representative spectra shown in Fig. 3A are of E. coli in the absence of EDTA treatment (black), after treatment with 2.5 mM EDTA (blue), and after treatment with 9.0 mM EDTA (red). Additional spectra from independently prepared duplicate (for 2.5 mM EDTA) and triplicate (for 9.0 mM EDTA) samples are provided in Supplementary materials (Figs. S1 and S2). The NMR spectra of the EDTA-treated cells are very similar to those of untreated cells. NMR observations were complemented by flow cytometry (Fig. 3B) and light microscopy of Gram-stained cells (Fig. 3C). Flow cytometry showed that bacteria treated with either 2.5 mM or 9.0 mM EDTA had only ∼3 % of the cells PI-permeable.

In contrast to the flow cytometry observations, Gram-staining did show a clear difference between the effects of the two EDTA treatments (Fig. 3C). Treatment with 9.0 mM EDTA led to the destabilization of the LPS layer as seen by the change in Gram staining colour from pink to purple and a change in cell shape from rod-like to spherical. Taking the results of flow cytometry and microscopy together, it appears that treatment with 9.0 mM EDTA disrupts LPS without killing the cells. To ensure we examined bacteria with a definite change in LPS, as well as with a less extreme treatment, both 2.5 and 9.0 mM EDTA treatments were employed in the remainder of the work.

Given that the Gram-staining shows the cells are strongly affected by 9.0 mM EDTA (Fig. F3C), it is useful to further analyse the NMR spectra (Fig. 3A, S1, S2) that, from inspection, appear to show little difference between untreated and EDTA-treated cells. Specifically, it is seen that the ∆₂ parameters calculated from the spectra (Table 1 and Fig. 2) increase slightly with 9.0 mM EDTA treatment. Statistics carried out with the ∆₂ values from the three control spectra and three 9.0 mM EDTA spectra indicate a p-value of 0.09 for the comparison, suggesting there may be a small change in the lipid acyl chain order distribution along the acyl chain following EDTA treatment. Nevertheless, it is striking that while EDTA treatment leads to drastic changes in overall cell shape (Fig. 3C), there is effectively little to no observable change in the ²H NMR spectra that characterize the orientational order of bacterial lipid-acyl chains. Thus, at least at the level at which ²H NMR reports, i.e. lipid acyl chain motions on the 10^-5 s timescale, what happens in the cation-binding region of LPS does not have a substantial effect in the bilayer interior.

3.3. LPS disruption alters the lipid chain disordering effect of AMP treatment

Since AMPs must traverse the carbohydrates of the LPS layer before reaching the lipid acyl chains, it was important to find out if LPS disruption would sensitize the cells to AMP-induced acyl chain disruption. To address this question, we used ²H NMR spectroscopy and flow cytometry to examine the effects of treating E. coli with EDTA to destabilize the LPS layer prior to treatment using 30 % MSI-78 and then compared the results with the effects of AMP treatment alone. Fig. 4 compares the effects of 30 % MSI-78 on ²H NMR spectra and flow cytometry of E. coli with and without prior treatment using 9.0 mM EDTA.

Fig. 4.

LPS pre-treatment increases the susceptibility of bacteria to AMP as judged by ²H NMR and flow cytometry. A: ²H NMR spectra of deuterium-enriched E. coli that are untreated (black, SK-53); after treatment with 30 % MSI-78 (green, SK-47) (% by dry weight of bacteria), and after treatment with 9 mM EDTA followed by 30 % MSI-78 (purple, SK-55). The top panel shows the overlaid spectra. Dashed lines at ∼±12.5 kHz are shown to facilitate the comparison of the spectra. The inset showing overlaid spectra is included to better show spectral changes around the center of the spectra. Each spectrum is obtained from 110,000 scans recorded over 12 h at 37 °C in a 600 MHz NMR spectrometer and normalized by area. Only selected spectra are shown. Additional spectra are shown in Figs S4 and S5. B: Cell count vs. PI fluorescence intensity for treated E. coli cells treated with 30 % MSI-78 alone and 9.0 mM EDTA + MSI-78, as probed by flow cytometry with PI. Peptide amounts are given as a % weight by dry weight of bacteria to facilitate calculation of AMP: lipid ratios and allow comparisons with other types of AMP studies [41].

Flow cytometry (Fig. 4B) showed that treatment with 30 % MSI-78 alone permeabilized ∼68 % of cells while treatment with 9.0 mM EDTA followed by 30 % MSI-78 permeabilized 92 % of cells i.e., more permeabilization than for cells treated with MSI-78 alone. The flow cytometry experiments were repeated with independently prepared samples and gave similar results (data not shown). Thus, as judged by cell permeabilization, disruption of LPS does sensitize the cells to MSI-78. Additionally, 9.0 mM EDTA + MSI-78 shows an increase in the side scatter spread of the cells (Supplementary data) suggesting there is an increase in intracellular granularity.

Unlike the flow cytometry results, which indirectly report on the state of the membrane by indicating if the dye can reach DNA inside the cells or not, ²H NMR spectra report directly on the lipid acyl chains. Representative ²H NMR spectra shown in Fig. 4A include the spectra for E. coli with no treatment, E. coli treated with MSI-78 only, and cells treated with 9.0 mM EDTA followed by MSI-78. Additional spectra from independently prepared duplicate samples are provided in the Supplementary materials as Figs. S3, S4 and S5.

Firstly, we confirm that the effect on the bacterial membrane of 30 % MSI-78 alone is similar to that reported from the bacterial membrane of 30 % MSI-78 alone and similar to that of previous studies [41,42]. Specifically, the ²H NMR observations shown in Fig. 4A demonstrate that MSI-78 induced substantial changes in the E. coli spectra, including decreased intensity at the ∼±12.5 edges and increased intensity at smaller splittings (∼<5 kHz). These ²H NMR observations are consistent with an MSI-78-induced decrease in bacterial membrane lipid acyl chain orientational order, i.e., an increased angular amplitude of lipid chain motions. We also looked closely at the isotropic signals in the center of the spectra for the sample treated with AMP alone and pre-treated with EDTA+AMP; there was no substantial increase in the isotropic signal between these samples.

To address how pre-treatment with EDTA to destabilize the LPS affects the bacteria's sensitivity to AMP-induced lipid-acyl chain disruption, we next compare the spectra for cells treated with MSI-78 alone to the spectra of cells pre-treated with EDTA before AMP treatment. Compared to treatment with MSI-78 alone, spectra of cells treated with EDTA before MSI-78 displayed less intensity at ∼±12.5 kHz and substantially more intensity in the ∼±2.5 kHz range of the spectrum (Fig. 4A inset). This observation is consistent with there being more AMP-induced lipid acyl chain disruption in LPS-destabilized cells. Hence, intact LPS does appear to prevent some membrane disruption by MSI-78.

To compare, more quantitatively, the effects of the AMP alone with those of EDTA-pre treatment + MSI-78, we performed a statistical analysis of $M_1$, $M_2$, and ∆₂ values. These comparisons are shown in Table 1 and Fig. 2. Starting with $M_1$, treating with 9.0 mM EDTA alone has no statistically significant change in the average $M_1$, although $M_1$ did tend to increase from 4.09 × 10⁴ s^-1 for untreated cells to 4.35 ×10 ⁴ s^-1 for cells treated with 9.0 mM EDTA (p-value 0.40). Interestingly, treatment with EDTA has the opposite effect on $M_1$ to that of treatment with AMP alone; $M_1$was significantly (p-value 0.05) smaller for the spectra of cells treated with AMP (3.44 × 10⁴ s⁻¹) than for the spectra of untreated cells. This observation is in keeping with previous studies (e.g., [41,42]) that noted that treatment with AMP leads to a decrease in average chain orientational order in model lipids. As expected from visual inspection of the spectra in Fig. 4A, treatment with 9.0 mM EDTA and 30 % AMP together results in $M_1$ being reduced even more (3.12 × 10⁴ s⁻¹) than for cells treated with AMP alone (p-value 0.07).

Similar to the observations for$M_1$, treatment with 9.0 mM EDTA alone does not make a significant change in the observed value of $M_2$. The average $M_2$ increases slightly from 2.61 × 10⁹ s⁻² for the spectra of untreated cells to 2.73 × 10⁹ s⁻² for the spectra of cells treated with 9.0 mM EDTA (p-value 0.28). Treatment with AMP moves $M_2$ in the opposite direction to that seen for treatment with EDTA, i.e., to a lower value (2.14 × 10⁹ s⁻²) although the difference between no treatment and AMP treatment is not statistically significant (p-value 0.19). With 9.0 mM EDTA followed by MSI-78 treatment, the average value of $M_2$ decreased even more to 1.84 × 10⁹ s⁻². This value of $M_2$ was significantly different from both the $M_2$ values for cells treated with MSI-78 alone (p-value 0.03) and cells treated with EDTA alone (p-value 0.02).

It is important to note that while treatment with EDTA tends to increase the spectral moments slightly, AMP does the opposite. With the joint EDTA + AMP treatment, the spectral moments decrease even more than for AMP alone, implying there is a synergistic effect between the two treatments. That is to say, the EDTA followed by AMP treatment decreases acyl chain orientational order more than the sum of the effects of EDTA or AMP applied separately. For$M_1$, the comparison between AMP alone and EDTA followed by AMP is on the edge of statistical significance with a p-value of 0.07, while the $M_2$ comparison is more significant with a p-value of 0.02. Importantly, since EDTA and AMP alone have opposite effects on$M_1\text{and}{M}_2$, the effects of AMP and EDTA are not additive rather the effect of EDTA treatment on AMP-induced reduction in $M_1\text{and}{M}_2$ is synergistic.

Moving to the ∆₂ values, which reflect the spectral shape, the average ∆₂ value increases significantly from 0.14 for untreated cells to 0.33 for AMP-treated cells (p-value 0.0008). With the 9.0 mM EDTA followed by MSI-78 treatment, ∆₂ values increased slightly more to 0.39 (p-value for AMP alone vs EDTA + MSI-78 is 0.07). While the effects of EDTA and MSI-78 on average chain order, as indicated by $M_1$, $\text{and}{M}_2$, were synergistic, the effects on ∆₂ were more additive in nature, i.e., both EDTA and MSI-78 tended to increase the relative mean squared deviation of the quadrupolar splittings, consistent with both treatments altering the distribution of orientational order parameters along the acyl chains.

At this point, it is helpful to examine more closely what the parameter ∆₂ means. AMPs tend to decrease spectral intensity at splittings corresponding to the plateau region of the order parameter profile, i.e. close to the lipid headgroups of model membranes [56]. ∆₂ reflects the shape of the distribution in ²H splittings [59] and since in unperturbed membranes, a large fraction of the observed splittings, come from the plateau region and have similar values, the distribution of splittings peaked near the upper range of the observed splittings, and ∆₂ is relatively small. In the presence of AMP, the distribution of splittings becomes flattened (the order parameter profile becomes more linear) and ∆₂ increases. In a typical liquid crystalline bilayer, packing constraints near the headgroup end of the acyl chains lead to the persistence of higher orientational order through the plateau region segments [57]. Interaction with AMPs appears to relax these constraints in a way that removes the prominent spectral edges characteristic of the plateau in the orientational order parameter. The resulting spectrum appears to reflect a more uniform change in orientational order parameter with position along the acyl chains. This probably reflects an AMP-induced increase in the amplitude of fluctuations in bacterial membrane bilayer shape and thickness, likely due to effects on lipid headgroup packing. Such changes are consistent with MSI-78 acting via a toroidal pore model, but from this set of data it is not possible to rule out other membrane-interaction modes that would cause the same changes to the order parameter profile.

In summary, $M_1$, $M_2$ ∆₂ values for samples treated with MSI-78 alone vs. pretreated with EDTA followed by MSI-78 are all different, with confidence values of 93 %, 98 % and 93 %, respectively, demonstrating that intact LPS helps reduces lipid acyl chain disruption by MSI-78.

3.4. The CPP TP2 causes a similar change in ²H NMR spectra of E. coli as the AMP MSI-78

It is thought-provoking to compare the lipid acyl chain disorder induced by MSI-78, which permeabilizes cells to a fluorescent dye, with the effect of a CPP, a peptide that has been shown to cross the membrane without making it permeable to dye [52]. In our hands, flow cytometry confirmed Wimley et al.’s findings [52] that bacteria were not dye-permeable after treatment with TP2. So, we were quite surprised to find that TP2’s effects on the ²H NMR bacteria were similar to those of MSI-78, making cells permeable to the dye.

Representative spectra shown in Fig. 5A present the ²H NMR spectra of E. coli with no treatment and E. coli treated with TP2 alone. Fig 5B compares spectra for bacteria treated with TP2 alone and with 9.0 mM EDTA followed by TP2. Duplicate spectra from independently prepared samples are provided in Supplementary material Fig. S6. While the NMR spectra show little if any difference between the effects of TP2 and MSI-78 on lipid chain order, flow cytometry confirms that, unlike MSI-78, TP2 does not permeabilize the cells to PI (Fig. 5C).

Having shown that pre-treatment to destabilize the LPS affects the bacteria's sensitivity to lipid acyl chain disruption by MSI-78, we asked if destabilization of the LPS layer also affects acyl chain disruption induced by TP2 (Fig. 5B). Looking at the spectra, any differences were hard to discern, so we examined the ∆₂ values calculated from these spectra.

As for MSI-78, there was a significant increase in ∆₂ with TP2 treatment (Table 1, Fig. 2). Interestingly, unlike for MSI-78, the ∆₂ value of the spectrum resulting from the pre-treatment of 9.0 mM EDTA followed by TP2 was smaller, 0.25, than for TP2 alone, 0.29 (p-value 0.03). In other words, while intact LPS reduced membrane disruption by MSI-78, it appears to increase membrane disruption by TP2. Overall, MSI-78 has a stronger effect on LPS-disrupted cells than TP2, with ∆₂ values of 0.39 and 0.25, respectively, and this was a statistically significant difference (p-value 0.01).

4. Discussion

Our group and others [41], [42], [43], [44], [45] have deuterated membrane phospholipid acyl chains in Gram-negative bacteria E. coli. It is interesting to compare the $M_1$, $M_2$, and ∆₂ values reported here to corresponding values reported for earlier studies. However, we expect the values to be altered by the E. coli strain chosen and the growth protocol used.

The previous studies have been on wild-type E. coli prepared using the protocol described here [42,43,45] and E. coli deuterated by taking advantage of a mutation that suppresses total fatty acid synthesis [41]. The $M_1$, $M_2$, and ∆₂ for untreated cells found in this study are within the range observed for untreated E. coli in all of these studies. Similarly, the increase in ∆₂ with 30 % MSI-78 in the present study with E. coli JM109s, of 0.16, is comparable to the 0.10 increase seen for the same amount of MSI-78 in the earlier study with mutated E. coli.

The interaction between the negative charges of Lipid A and the divalent ions Mg²⁺ and Ca²⁺ is essential for the stability of the LPS membrane. We found that EDTA drastically alters the gram-staining of the bacteria and changes them from rod-shaped to spherical-shaped without killing them (Fig. 3B and C). It is thus rather striking that despite these dramatic changes to the bacteria, the lipid chain order remains largely unchanged by treatment with 9.0 mM EDTA (Fig. 3B). The dramatic changes in the carbohydrate region of the LPS and cell shape appear to leave the lipid acyl chains of the bacterial membrane largely unperturbed. Additionally, it should be noted that while treatment with EDTA destabilizes the LPS by chelating the stabilizing cations, it can be expected that other changes to the bacteria will occur, including changes in gene expression, as bacteria attempt to adapt to the LPS disruption.

The central question posed in the current work is the role of the LPS layer in either protecting bacteria from AMPs or in potentiating their activity. As expanded on in the Introduction, if some of the AMP gets bound up in the LPS layer and doesn't reach the membrane, that could explain why AMP:L ratios needed to see activity are higher in studies with bacteria compared to studies of model membrane permeabilization [20], [21], [22].

In this work, we show that LPS destabilization by EDTA before treatment with the AMP MSI-78 does indeed increase the ability of MSI-78 to perturb the lipid acyl chains. Thus, at least for MSI-78, the LPS layer does help reduce the peptide-induced membrane disruption. This result can be seen by comparing the NMR spectra of membrane-deuterated bacteria treated with MSI-78 alone to spectra from bacteria first treated with EDTA to destabilize the LPS followed by AMP treatment (Fig. 4) and can be more simply appreciated from the parameters calculated from the NMR spectra, $M_1$, $M_2$ and ∆₂ (Fig. 2). In particular, $M_1$ and $M_2$ both decrease more in the doubly treated cells than in those treated with AMP alone, indicating more AMP-induced acyl chain disruption. ∆₂, which is sensitive to the shape of the spectra and is expected to increase if the AMP alters the distribution of orientational order along the acyl chain, does increase more in the doubly treated cells than for AMP treatment alone.

This finding has the potential to explain unanswered questions about how AMPs interact with bacteria. It has long been known that it takes a greater AMP:L ratio to permeabilize whole bacteria than to permeabilize model membranes [[13], [14], [15],17,18,21], but this difference is not understood. We show that LPS, present in bacteria but not model membranes, reduces the membrane disruption caused by AMP MSI-78. LPS protection from AMP membrane disruption would explain why model lipid bilayers are more prone to disruption by AMP than are bilayers in whole bacteria. Furthermore, since efforts to optimize AMPs as drugs often rely solely on optimizing the AMP's lipid-disrupting activities, consideration of other interactions like AMP-LPS interactions may prove helpful in AMP-based drug design.

We find it fascinating that while one peptide, MSI-78, permeabilizes the bilayer to a fluorescent dye and the other, TP2, does not (Fig. 4B, 5B) [52], the two peptides have much the same effect at the level of the ²H NMR spectra (Fig. 2). Possible explanations for why the ²H NMR spectra are not sensitive to the difference in mechanism between these two peptide types might relate to the timescale of the peptide-induced lipid perturbations (which our NMR data probe on the ∼100 us timescale) or the diameter of the defects induced by the peptides. Whatever the explanation, it is intriguing to find such a fundamental property, i.e. acyl chain disruption, shared by two peptides that have different effects on membrane permeabilization.

Another thought-provoking aspect of comparing the lipid interactions of MSI-78 and TP2 was the effect of LPS destabilization on the peptide-induced lipid disruptions. It was quite unexpected that intact LPS appeared to protect bacteria from MSI-78 but sensitize bacteria to TP2, as judged by the ∆₂ values calculated from the NMR spectra (Fig. 2). It is unclear why this is. Still, one possibility might be the difference in the sequence of the two peptides. For example, TP2 gets its positive charge from arginine residues, while MSI-78 has no arginines but many lysines. Additionally, TP2 is a much shorter peptide than MSI-78 and, MSI-78 carries more ring residues in its sequence. Further study is required to understand if this is a general difference between AMPs and CPPs, or if it is specific to MSI-78 and TP2, as well as to understand the mechanism for the difference. In any case, the result further underlines the need to consider interactions beyond peptide-lipid interactions in understanding the mechanisms by which surface-active peptides affect bacteria.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.