Lipid Segregation Explains Selective Toxicity of a Series of Fragments Derived from the Human Cathelicidin LL-37

Abstract

The only human cathelicidin, the 37-residue peptide LL-37, exhibits antimicrobial activity against both gram-positive and gram-negative bacteria. We studied the ability of several fragments of LL-37, exhibiting different antimicrobial activities, to interact with membranes whose compositions mimic the cytoplasmic membranes of gram-positive or of gram-negative bacteria. These fragments are as follows: KR-12, the smallest active segment of LL-37, with the sequence KRIVQRIKDFLR, which exhibits antimicrobial activity only against gram-negative bacteria; a slightly smaller peptide, RI-10, missing the two cationic residues at the N and C termini of KR-12, which has been shown not to have any antimicrobial activity; a longer peptide, GF-17, which shows antimicrobial activity against gram-positive as well as gram-negative bacteria; and GF-17D3, with 3 d-amino-acid residues, which is also selective only for gram-negative bacteria. Those fragments with the capacity to cluster anionic lipids away from zwitterionic lipids in a membrane exhibit selective toxicity toward bacteria containing zwitterionic as well as anionic lipids in their cytoplasmic membranes but not toward bacteria with only anionic lipids. This finding allows for the prediction of the bacterial-species selectivity of certain agents and paves the way for designing new antimicrobials targeted specifically toward gram-negative bacteria.

The increasing emergence of bacterial strains that are resistant to traditional antibiotics makes the development of a new generation of antimicrobial agents an urgent matter. Naturally occurring antimicrobial peptides (there are more than 1,450 entries in the Antimicrobial Peptide Database) (33) are promising candidates because of their potency and low toxicity. These compounds are found in all life forms, indicating their important role as key defense molecules of the innate immune systems of nearly all life forms. Defensins and cathelicidins are the two major families of host defense peptides found in mammals. In humans, the peptide LL-37 represents the only member of the cathelicidin family that has been identified. LL-37 has an important role in protecting humans from infections. Patients lacking LL-37 are found to be more prone to infections (23), as are LL-37 knockout mice (19). LL-37 is found at lower concentrations in the airways of individuals with cystic fibrosis who are more prone to infection (5). In vitro, this human cathelicidin peptide shows antimicrobial activity against gram-positive bacteria, gram-negative bacteria, viruses, and fungi (30, 34). A variety of approaches were applied in identifying fragments of LL-37 that retain antimicrobial activity (4, 13, 18, 25, 26). In a recent study, a smaller fragment of LL-37, corresponding to residues 18 to 29 (designated KR-12), was found to be the smallest LL-37 fragment that retained activity against gram-negative bacteria (32), despite having only 12 of the 37 residues of the native peptide. KR-12 is similar to the core peptide identified by structural analysis of LL-37, using total correlated spectroscopy by trimming nonessential regions (12). The identification of short, active peptides from LL-37 facilitates peptide synthesis and structure-activity studies. It may also provide insights into novel mechanisms of action of antimicrobial agents. In this work, we evaluate the antimicrobial activities of a panel of LL-37-derived peptides against the gram-negative bacterium Escherichia coli (K-12) as well as two strains of the gram-positive bacteria Staphylococcus aureus UAMS-1 (methicillin [meticillin]-sensitive S. aureus [MSSA]) and USA300 (methicillin-resistant S. aureus [MRSA]). These two species of bacteria have very different lipid compositions in their membranes (Table 1), and these lipid compositions have a profound influence on the relative toxicities of different peptides. We measured the interactions of these peptides with membranes having lipid compositions corresponding to those of these two bacterial species. Our work not only provides evidence for a novel mechanism of action but also explains the specificities of certain agents toward different species of bacteria.

TABLE 1.

Lipid compositions of the bacterial strains used

Bacterial species	Lipid composition (%)
	PE	PG	CL
S. aureus	0	58	42
E. coli	80	15	5

MATERIALS AND METHODS

Materials.

The phospholipids 1-palmitoyl-2-oleoyl phosphatidylethanolamine (POPE), dioleoylphosphatidylglycerol (DOPG), tetraoleoyl cardiolipin (TOCL), and dioleoylphosphatidic acid (DOPA) were purchased from Avanti Polar Lipids (Alabaster, AL). The 8-aminonaphthalene-1,3,6-trisulfonic acid (ANTS) and p-xylene-bis-pyridinium bromide (DPX) were purchased from Invitrogen. The peptides with C-terminal amidation were synthesized and purified to >95% by Genemed Synthesis, Inc. (San Antonio, TX). The sequences of the peptides used are summarized in Table 2.

TABLE 2.

Peptides used in this study

Peptide	Sequencea
LL-37	LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES
RI-10	RIVQRIKDFL
KR-12	KRIVQRIKDFLR
GF-17	GFKRIVQRIKDFLRNLV
GF-17D3	GFKRiVQRiKDFlRNLV

^aAll peptides, except LL-37, are amidated at their carboxyl termini. In the cases of GF-17 and GF-17D3, a glycine is added at the N terminus, presumably to stabilize the peptide. Lowercase letters indicate a d-residue.

Antimicrobial-activity assays.

The antibacterial activities of LL-37 fragments were determined using the standard broth microdilution method as described previously (31). The strains of bacteria chosen for these assays were E. coli K-12, S. aureus UAMS-1 (antibiotic sensitive), and S. aureus USA300 (antibiotic resistant). In brief, a 5-ml culture was grown overnight. A fresh culture (5 ml) was inoculated with a small aliquot of the overnight culture and incubated at 37°C (250 rpm) until the bacterial growth reached the logarithm stage, as indicated by the rate of increase of the optical density. The optical density was approximately 0.5 when the bacteria were harvested. The culture was then diluted, and 90-μl aliquots were placed into a 96-well microplate with ∼10⁶ CFU per well. A series of peptide solutions (10 μl) at twofold dilutions were added. The assays were performed in triplet for each peptide. The plate was incubated at 37°C overnight (∼16 h) and read on an Ultra microplate reader at 630 nm (BioTek Instruments). The MIC is defined as the lowest peptide concentration that fully inhibits the bacterial growth.

Preparation of lipid films.

Lipid mixtures were prepared from stock solutions in 2:1 (vol/vol) chloroform-methanol, and appropriate amounts were mixed in a glass tube. The solvent was evaporated under a stream of nitrogen gas to deposit the lipid as a film on the wall of the tube. Final traces of solvent were removed in a vacuum chamber attached to a liquid nitrogen trap, and the chamber was evacuated for about 3 hours.

DSC.

Differential scanning calorimetry (DSC) measurements were made in a Nano II differential scanning calorimeter (Calorimeter Sciences Corporation, Lindon, UT). Lipid films were hydrated at room temperature with peptide solutions in PIPES buffer {20 mM PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid)], 140 mM NaCl, 1 mM EDTA, pH 7.4} or with PIPES buffer alone, vortexed vigorously to make multilamellar vesicles (MLVs), and then degassed before being loaded into the sample cell of the calorimeter. Controls using peptide solutions in the sample cell in the absence of lipid showed no transition. Degassed PIPES buffer was placed in the reference cell. The concentration of phospholipids in the samples was maintained at 2.5 mg/ml and the lipid-to-peptide molar ratio at 10. The cell volume was 0.34 ml. Samples were equilibrated in the calorimeter at 0°C, and successive heating and cooling scans were carried out between 0 and 35°C at a scan rate of 1.0°C/min, with a delay of 5 min between sequential scans in a series to allow for thermal equilibration. The resulting curves were analyzed by using the fitting program provided by Microcal Inc. (Northampton, MA) and plotted with Origin version 5.0.

Leakage of aqueous contents.

Lipid films were hydrated with a solution containing 12.5 mM ANTS, 45 mM DPX, and 10 mM HEPES, adjusted to pH 7.4. A solution of NaCl was added to adjust the final osmolarity to 300 mosM/liter using a freezing point osmometer (model 3300; Advanced Instruments, Inc., Norwood, MA). The lipid suspensions were then freeze-thawed 5 times and extruded 10 times through two stacked polycarbonate filters having pores 100 nm in diameter, to make visually transparent solutions of large unilamellar vesicles (LUV) with a 100-nm diameter. The LUV were put through a 20- by 1.5-cm column of Sephadex G-75 (Pharmacia, Uppsala, Sweden) preequilibrated with a buffer containing 10 mM HEPES, 1 mM EDTA, and NaCl adjusted to 300 mosM/liter and pH 7.4 (HEPES buffer). LUV were collected in the void volume, and the concentration of lipid was determined by phosphate analysis (1).

An aliquot of LUV was placed in a quartz cuvette containing HEPES buffer at 37°C to give a final lipid concentration of 50 μM, and fluorescence was recorded for a few seconds. Then an aliquot of peptide solution was added, and the increase of fluorescence was monitored for 3 to 4 min. To determine 100% release of entrapped fluorophore, 20 μl of 20% Lubrol XT was added at the end of the leakage measurement. Excitation was set at 360 nm and emission at 530 nm, with a 500-nm-cutoff filter in the emission monochromator and a band pass of 8 nm for excitation and 4 nm for emission. Measurements were performed at 37°C in an Aminco Bowman SLM-II spectrofluorimeter.

³¹P magic angle spinning-nuclear magnetic resonance (³¹P MAS-NMR).

Lipid films were prepared as described above. Dried films were kept under argon gas at −20°C if not used immediately. Films were hydrated with buffer or with a solution of peptide in 20 mM PIPES, 1 mM EDTA, 150 mM NaCl, pH 7.40, to give a lipid-to-peptide ratio of 10. The lipid film was suspended as MLVs by vigorous vortexing. The MLVs were then pelleted by centrifugation in an Eppendorf centrifuge at room temperature.

The solid-state NMR spectra were recorded on a Bruker Avance 500 spectrometer equipped with a standard-bore 11.7-Tesla magnet, giving a 202.45-MHz frequency for ³¹P (500.12 MHz ¹H base frequency). The spectrometer was equipped with a 4-mm broadband tuneable MAS probe, and the spectra were recorded as a function of temperature. Sample temperature was controlled using a Bruker model BVT-3000 controller coupled to a liquid nitrogen heat exchanger for low-temperature experiments. Rotor temperature was calibrated by several methods, and at a 5-kHz spinning speed, the sample temperature was always found to be within 2°C of the bearing temperature for the temperature range used in this work. Because the corrected temperatures were within error of the bearing temperature, the data are reported as a function of the bearing temperature. MAS was controlled using a Bruker model H2620 pneumatic MAS controller. The chemical-shift scale was referenced to an external phosphoric acid standard.

The ³¹P-NMR spectra were recorded over a 25-kHz sweep width. Raw spectra were recorded using 14,000 data points (acquisition time, less than 0.3 s), and the data were processed with zero filling to 16,000 data points and with exponential multiplication (line broadening = 3 Hz). The Bruker Topspin solids line shape analysis routine (version 2.1) was used to simulate each of the spectra. The observed peak maxima for the isotropic peaks were at the same chemical shift as those for the calculated isotropic chemical shifts.

RESULTS

MICs.

Several peptides related to LL-37 were studied (Table 2). The peptide KR-12 was previously found to be the smallest fragment active against E. coli (32). We found that this peptide is inactive against S. aureus UAMS-1 and USA300, which are antibiotic-susceptible and -resistant strains, respectively (Table 3). A longer fragment, GF-17, is active against these two strains of S. aureus as well as against E. coli K-12 (Table 3). Incorporation of three d-amino acids into the peptide at positions 20, 24, and 28 of GF-17 (as numbered for LL-37) led to GF-17D3, which is active against E. coli (12) but not the two strains of S. aureus (Table 3). Full-length LL-37 was active against E. coli K-12 but showed weak activity against both strains of S. aureus. Previous studies had indicated potent activity of LL-37 against most strains of S. aureus (27, 28), with MICs in the lower micromolar range and comparable only to the MICs for E. coli. However, a recent study of clinical isolates of S. aureus has found a broad range of susceptibilities to LL-37 in both MSSA and MRSA strains (21). One of the factors contributing to the resistance of some strains of S. aureus to LL-37 is the presence of proteolytic enzymes (25). One of these enzymes has been identified as a glutamylendopeptidase that cleaves LL-37 between Glu¹⁶ and Phe¹⁷. This enzyme thus cleaves LL-37 but not GF-17. The N-terminal cleavage product of LL-37 in this reaction is FK-21, which is active against S. aureus. It is possible, however, that FK-21 is proteolyzed more rapidly in these resistant strains of S. aureus than is GF-17, with its N-terminal Gly-protecting group. Regardless of what the mechanism is for, the greater resistance of the strains of S. aureus used in this work to LL-37 is not likely to be caused by differences in the lipid compositions of the membranes of these strains.

TABLE 3.

Antimicrobial activities of LL-37 and segments of this peptide

Peptide	MIC (μM) for:
	E. coli K-12	S. aureus UAMS-1 (MSSA)	S. aureus USA300 (MRSA)
KR-12	66	>264	>264
GF-17D3	32	>256	>256
GF-17	14	7	7
LL-37	16	>126a	>126a

^aStrongly dependent on a strain of S. aureus. See the text.

Because several of the peptides did not fully inhibit the growth of S. aureus at the concentrations used (Table 3), we also present the peptide concentrations required for 30% inhibition of growth (Table 4) so as to distinguish the relative antimicrobial activities for these cases. By this criterion, it is clear that LL-37 has activity against both strains of S. aureus.

TABLE 4.

Inhibition of growth by LL-37 and segments of this peptide

Peptide	Concn (M) that inhibited by 30% the growth of:
	S. aureus UAMS-1 (MSSA)	S. aureus USA300 (MRSA)
KR-12	16	32
GF-17D3	32	64
LL-37	8	8

DSC.

The interaction of these peptides with phospholipid bilayers was evaluated by means of DSC. Advantages of DSC for studying phase separation are that it does not require the use of bulky spectroscopic probes and it can detect the presence of domains that are too small to image by microscopy methods (9). To provide insight into the mechanisms of action of these LL-37-derived peptides, we have used simplified lipid systems to mimic the cytoplasmic membranes of bacteria by producing mixtures with two components representative of the major lipids present in bacteria but having uniform acyl chain compositions that will exhibit phase transition behavior. These include lipid mixtures representing bacteria that have a high content of phosphatidylethanolamine (PE), such as most gram-negative bacteria, e.g., E. coli, whose membrane composition is PE-phosphatidylglycerol (PG)-cardiolipin (CL) (80:15:5) (16). The lipid compositions contained either one of the anionic lipids (CL or PG) found in the membrane of E. coli in a lipid mixture with PE. In addition, it has been found that a pgsA-null Escherichia coli strain, UE54, contains phosphatidic acid and N-acyl-phosphatidylethanolamine as the major anionic phospholipids in place of PG and CL (15). We therefore also tested mixtures of the principal anionic lipid of this mutant strain, i.e., phosphatidic acid, as a mixture with PE to determine if the lipids would exhibit the same behavior in this mutant. The compositions tested were 75:25 POPE-TOCL, 75:25 POPE-DOPG, and 75:25 POPE-DOPA. We also employed a lipid mixture devoid of PE, representing most gram-positive bacteria, e.g., S. aureus, whose membrane composition is PG-CL (60:40) (3). For this bacterium, a binary combination of lipids, DMPG-TOCL (75:25), was used for DSC. All of these binary combinations of lipids have one lipid component in the liquid-crystalline phase above 0°C (i.e., TOCL, DOPG, or DOPA) and another component exhibiting a transition from the gel phase to the liquid-crystalline phase close to 25°C as a pure component (i.e., POPE or DMPG). For the mixtures tested in this work, these combinations of lipids exhibit a phase transition around 14 to 18°C (Table 5). This phase transition has a calorimetric enthalpy (ΔH_o) in the range of 2 to 4 kcal/mol and a van't Hoff enthalpy (ΔH*) of 60 to 100 kcal/mol, indicative of a transition of moderate cooperativity. This enables us to detect lateral-phase separation of lipids induced by cationic peptides or proteins. The preferential association of the cationic peptide with the negatively charged lipid moves this lipid away from the mixture and leaves the second lipid component in purer form, allowing its phase transition to move closer to that of the pure zwitterionic lipid. We demonstrate the sensitivity of the phase transition behavior of these lipid mixtures to the ratio of lipids in the mixture (Fig. 1).

TABLE 5.

DSC parameters of lipid mixtures^a

Lipid composition (75:25)	First heating scan			First cooling scan
	Tm (°C)	ΔHo (kcal/mol)	ΔH* (kcal/mol)	Tm (°C)	ΔHo (kcal/mol)	ΔH* (kcal/mol)
POPE-TOCL	14.6	2.3	80	12.8	2.4	100
POPE-DOPG	15.4b	2.7	65	15.7	2.7	66
DMPG-TOCL	14.4	2.0	70	13.5	2.05	60
POPE-DOPA	17.8	3.9	100	16.3	3.9	100

^aT_m, melting temperature; H_o, calorimetric enthalpy; H*, van't Hoff enthalpy.

^bThere was a second component in the heating scan at 18°C with an enthalpy of 0.9 kcal/mol.

FIG. 1.

Phase transitions of lipid mixtures by DSC. Lipid concentration, 2.5 mg/ml; scan rate, 1°C/min. The top four curves in each panel are heating scans, while the bottom four are cooling scans. Numbers refer to the mol% of TOCL or DOPG in the mixture. Phase transition temperatures of pure lipid components were 25°C for POPE and DMPG and below 0°C for the others.

The lipid mixture POPE-TOCL (75:25) has a gel-phase to liquid-crystalline-phase transition centered at about 15°C (Fig. 2). The gel- to liquid-crystalline-phase transition temperatures of the individual lipids that comprise this mixture are 25°C for POPE and below 0°C for TOCL. Addition of LL-37 to this membrane composition has little effect on the transition temperature in the DSC heating curve, although the transition temperature range is somewhat broadened. However, LL-37 does lower the transition temperature on cooling (Fig. 2). The DSC indicates that there is no, or very little, lateral-phase separation of these lipids induced by LL-37 since there is no evidence for a domain that is enriched in POPE that would have a phase transition temperature shifted to higher temperatures in both the heating and cooling scans. The cause for the hysteresis in the phase transition in the presence of LL-37 can be explained by a greater solubility of LL-37 in the liquid-crystalline phase than in the gel phase. Thus, the peptide incorporates into the membrane at higher temperatures and is present in the membrane at temperatures around that of the phase transition during the cooling scans but is partially displaced from the membrane at the end of the cooling scan, resulting in the peptide having less of an effect on the phase transition measured in the heating mode. Therefore, this system exhibits hysteresis because the peptide is slow to redissolve in the lipid at low temperatures. The peptides KR-12, GF-17D3, and GF-17 produce effects on the transitions of this lipid mixture that are similar to each other and different from the effects of LL-37 and of RI-10. On the other hand, KR-12, GF-17D3, and GF-17 show a component with a higher transition temperature, seen in both heating and cooling curves (Fig. 2). This is consistent with these peptides binding to TOCL and clustering this anionic lipid, thus leaving the remaining lipid enriched in POPE. The peptide RI-10, which is inactive against bacteria, has almost no effect on the phase transition properties of this lipid mixture, in either the heating or the cooling scans (Fig. 2).

FIG. 2.

Phase separation of POPE-TOCL mixtures determined by DSC. DSC carried out with 2.5 mg/ml POPE-TOCL (75:25) at a scan rate of 1.0°/min in the absence and presence of LL-37, RI-10, KR-12, GF-17D3, or GF-17 in 20 mM PIPES, pH 7.4 (0.14 M NaCl, 1 mM EDTA). The lipid/peptide molar ratio was 10. Odd numbers are heating scans, and even numbers are cooling scans. Labels 1′ and 2′ refer to heating and cooling scans, respectively, of the lipid in the absence of peptide.

Similar observations were made with POPE-DOPG (75:25). This lipid composition also mimics the cytoplasmic membrane composition of E. coli (Table 1). The mixture also has a broad phase transition centered at 15°C (Fig. 3). However, another component is present at 18°C, indicating incomplete miscibility by these two lipids, which is seen at all mole fractions of DOPG (Fig. 1). The phase transition temperature of DOPG is below 0°C. Again, LL-37 shows a decrease in the phase transition on cooling, while KR-12 and GF-17D3 induce the formation of a component enriched in POPE (Fig. 3). KR-12 and GF-17D3 are capable then not only of forming domains in mixed systems but also of rearranging existing ones. With this system, RI-10 again does not cause any change in the phase transition behavior of the lipid mixture.

FIG. 3.

Phase separation of POPE-DOPG mixtures determined by DSC. DSC carried out with 2.5 mg/ml POPE-DOPG (75:25) at a scan rate of 1.0°/min in the absence and presence of LL-37, RI-10, KR-12, or GF-17D3 in 20 mM PIPES, pH 7.4 (0.14 M NaCl, 1 mM EDTA). The lipid/peptide molar ratio was 10. Odd numbers are heating scans, and even numbers are cooling scans. Labels 1′ and 2′ refer to heating and cooling scans, respectively, of the lipid in the absence of peptide. Note the change in the scale of the y axis for the LL-37 panel.

With POPE-DOPA (75:25), we mimic the cytoplasmic membrane composition of a mutant strain of E. coli, UE54. The mixture has a broad phase transition centered at 17.8°C (Fig. 4). The phase transition temperature of DOPA is below 0°C. KR-12 induces the formation of a component enriched in POPE (Fig. 4).

FIG. 4.

Phase separation of POPE-DOPA mixtures determined by DSC. DSC carried out with 2.5 mg/ml POPE-DOPA (75:25) at a scan rate of 1.0°/min in the absence and presence of KR-12 in 20 mM PIPES, pH 7.4 (0.14 M NaCl, 1 mM EDTA). The lipid/peptide molar ratio was 10. Odd numbers are heating scans, and even numbers are cooling scans. Labels 1′ and 2′ refer to heating and cooling scans, respectively, of the lipid in the absence of peptide.

We also measured the transition properties of a mixture of two anionic lipids, DMPG and TOCL (75:25), alone and in the presence of one of the peptides LL-37, GF-17D3, and GF-17 (Fig. 5). Figure 1 shows that these two lipids have a high degree of miscibility at all compositions studied. LL-37 and GF-17 were selected because they exhibit antimicrobial activity against gram-positive bacteria that are largely devoid of zwitterionic lipids. However, these peptides, as well as GF-17D3, had rather modest effects on the phase transition of this lipid mixture, suggesting that their action against these gram-positive bacteria was by a mechanism different from lateral-phase separation.

FIG. 5.

Peptide effects on DMPG-TOCL mixtures determined by DSC. DSC carried out with 2.5 mg/ml DMPG-TOCL (75:25) at a scan rate of 1.0°/min in the absence and presence of LL-37, GF-17D3, or GF-17 in 20 mM PIPES, pH 7.4 (0.14 M NaCl, 1 mM EDTA). The lipid/peptide molar ratio was 10. Odd numbers are heating scans, and even numbers are cooling scans. Labels 1′ and 2′ refer to heating and cooling scans, respectively, of the lipid in the absence of peptide.

We have summarized the extents of the shifts in the phase transition temperature upon the addition of the various peptides to allow facile comparison among the various systems studied (Table 6). In general, LL-37 is the only peptide that exhibits a large hysteresis between heating and cooling curves, although this effect is smaller with the anionic lipid mixture DMPG-TOCL (75:25). This could be a consequence of greater aggregation of LL-37 with increased temperature in the presence of zwitterionic lipids. It is known that LL-37 is aggregated in the presence of zwitterionic lipids but is monomeric with anionic lipids (20). RI-10 shows the smallest effects of any of the peptides in the two systems in which it was studied. The other three peptides, KR-12, GF-17D3, and GF-17, promote a significant shift of the phase transition to higher temperatures in all systems except the anionic lipid mixture DMPG-TOCL (75:25).

TABLE 6.

Shifts in melting temperatures of major DSC peaks

Peptidea	Lipid composition (75:25)	ΔTemp(s) (°C)b
		Heating scan	Cooling scan
LL-37	POPE-TOCL	+0.6	−4.8
KR-12		+3.0	+1.9, +4.8c
GF-17D3		+3.5	+0.7, +4.1c
GF-17		+3.0	−1.8, +3.6c
RI-10		+0.3	−0.3
LL-37	POPE-DOPG	+0.6	−6.1
KR-12		+4.6	+2.3
GF-17D3		+3	+0.6
RI-10		+1.8	−0.7
LL-37	DMPG-TOCL	+0.6	−0.9
GF-17D3		0	−0.9
GF-17		0	−1.2
KR-12	POPE-DOPA	+2.0	+0.3, +1.9, +3.2d

^aThe peptide/lipid ratio was kept at 10 for all cases.

^bShift from the melting temperature of the pure lipid mixture.

^cThere were two components in the cooling curve.

^dThere were three components in the cooling curve.

Leakage.

For the liposome permeabilization studies, we used an assay based on the relief of DPX quenching of ANTS as a result of dilution. The lipid components were chosen so that they would remain in the fluid, liquid-crystalline phase at room temperature. We again used mixtures to mimic the cytoplasmic membrane lipid compositions of certain bacteria. Since phase behavior was not being measured, we could use a somewhat more complex mixture of three lipids, POPE-DOPG-TOCL (80:15:5), to more exactly mimic E. coli cytoplasmic membranes. No release of entrapped components was observed with liposomes made of this lipid mixture at any concentration up to 40 μM RI-10, KR-12, GF-17D3, or GF-17.

A mixture composed of DOPG-TOCL (60:40) was used to mimic the lipid composition of the membrane of S. aureus. LL-37 was very potent in inducting leakage. In the results presented, LL-37 is at half the peptide concentration of the fragments (Fig. 6). Little or no leakage was observed with KR-12 or RI-10. With GF-17D3 and GF-17, leakage exhibited a very different behavior, which we highlight at a high peptide concentration (40 μM). While release was very rapid for GF-17, with GF-17D3, there was a considerable lag time. At lower concentrations, no release was seen with GF-17D3, while GF-17 was still active (not shown).

FIG. 6.

Leakage of aqueous contents from vesicles (ANTS-DPX assay). Experiments were carried out at 37°C with 10 mM HEPES buffer, pH 7.4 (0.14 M NaCl, 1 mM EDTA). Peptide (40 μM) was added at zero time, except for LL-37, for which the data for 20 μM peptide is presented. Fifty micromolar DOPG-TOCL (60:40) as LUV was used. Lubrol LX was added to obtain a maximum value (100% leakage) of vesicle permeabilization. Curve 1, LL-37 (20 μM); curve 2, GF-17 (40 μM); curve 3, GF-17D3 (40 μM); curve 4, KR-12 (40 μM); curve 5, RI-10 (40 μM).

³¹P MAS-NMR.

We have analyzed the temperature dependence of the ³¹P MAS-NMR spectra of a mixture of TOCL and POPE (25:75). An advantage of ³¹P MAS-NMR for this system is that the properties of each of the lipid components in the mixture and their alteration in the presence of a peptide can be monitored separately.

We can compare the chemical shift of each of the components with those of the individual pure lipids as well as those of the same lipid mixture in the absence of peptide (Table 7). Both the TOCL and POPE peaks move closer to each other in the lipid mixture than those of the individual components. This likely reflects the fact that the two lipids are miscible, and the changes in the chemical shift reflect the change in the environment of each of the lipids as a consequence of the presence of the second lipid. Addition of the peptide KR-12, GF-17D3, or GF-17 causes the POPE peak to have a chemical shift close to that of pure POPE, in accord with the results of DSC, suggesting the formation of a phase highly enriched in POPE. The TOCL peak shifts further from the resonance position of pure TOCL when one of the peptides is added. We suggest that this is a consequence of neutralization of the negative charge of the TOCL by binding to the cationic peptide, resulting in the TOCL peak attaining a value more similar to that of zwitterionic lipids. This is confirmed by the observations discussed below.

TABLE 7.

Isotropic chemical shifts

Temp (°C)	Isotropic chemical shift (ppm)
	TOCL peak					POPE peak
	TOCL	POPE-TOCL	POPE-TOCL KR-12	POPE-TOCL GF-17	POPE-TOCL GF-17D3	POPE	POPE-TOCL	POPE-TOCL KR-12	POPE-TOCL GF-17	POPE-TOCL GF-17D3
−10	0.666
−5		0.590	0.27	0.41	0.43		0.081	−0.17	−0.10	−0.08
0	0.640	0.553	0.29	0.53	0.29	−0.143		−0.19	−0.10	−0.13
5		0.521	0.29	0.43	0.32		0.085	−0.15	−0.11	−0.09
10	0.636	0.486	0.22	0.42	0.37	−0.143	0.068	−0.09	−0.04	−0.06
15		0.482	0.28	0.34	0.33		0.072	−0.07	−0.05	−0.06
20	0.618	0.479	0.28	0.34	0.31	−0.136	0.073	−0.07	−0.05	−0.06
25		0.466	0.26	0.31	0.30	−0.033	0.063	−0.07	−0.05	−0.06
30	0.598	0.470	0.26	0.30	0.30	−0.019	0.065	−0.07	−0.05	−0.05

The chemical shift anisotropy (CSA) is a measure of the motional averaging of the spatial orientations of the phosphate in the head group within the lifetime of the excited ³¹P spin state. Less motional averaging results in a higher value for the CSA. This can be either because of a low rate of diffusion of the lipid or because of the geometry of the lipid phase. In the case of POPE or of the lipid mixture, the CSA increases at low temperatures because of the formation of a gel phase (Table 8). The peptides narrow the CSAs of both lipid components, suggesting that they fluidize the membrane. Even the phase enriched in POPE in the presence of peptide does not have a gel-state-like CSA at low temperature, indicating that this phase also contains some peptide and/or TOCL.

TABLE 8.

CSA results

Temp (°C)	CSA result (ppm)
	TOCL peak					POPE peak
	TOCL	POPE-TOCL	POPE-TOCL KR-12	POPE-TOCL GF-17	POPE-TOCL GF-17D3	POPE	POPE-TOCL	POPE-TOCL KR-12	POPE-TOCL GF-17	POPE-TOCL GF-17D3
−10	35.40
−5		49.53	33.1	27.1	36.8		54.24	38.2	35.7	39.2
0	33.42	43.14	33.9	23.1	32.7	55.83	49.94	34.4	33.6	37.3
5		38.57	26.0	23.0	29.0		40.25	32.5	30.4	32.7
10	33.05	35.63	26.5	24.9	25.6	54.42	37.01	27.7	28.4	28.9
15		32.27	24.5	22.8	24.4		34.17	25.6	25.1	25.7
20	30.48	34.05	23.7	23.4	22.1	51.50	34.08	23.6	24.9	25.7
25		33.47	24.0	21.4	22.1	39.36	34.16	24.5	23.9	25.5
30	28.29	33.11	22.9	21.2	22.7	38.99	33.57	21.8	23.9	24.1

The line width is a measure of the rate of motion and is very different from the CSA, which is determined by the extent of motion. The peptides significantly increase the line width of the TOCL peak compared to that with either TOCL alone or the POPE-TOCL mixture (Table 9). This demonstrates that the peptides slow certain motions of TOCL at all temperatures. In contrast, the peptides have relatively little effect on the line width of the POPE peak compared with that produced with pure POPE (Table 9), indicating that the peptides do not affect the motional properties of this lipid component.

TABLE 9.

³¹P MAS-NMR line widths

Temp (°C)	31P MAS-NMR line width (Hz)
	TOCL peak					POPE peak
	TOCL	POPE-TOCL	POPE-TOCL KR-12	POPE-TOCL GF-17	POPE-TOCL GF-17D3	POPE	POPE-TOCL	POPE-TOCL KR-12	POPE-TOCL GF-17	POPE-TOCL GF-17D3
−10	70.93
−5		69.48	252.6	297.4	130.1		125.50	114.8	154.5	162.3
0	54.90	59.97	212.3	230.3	161.8	124.29	80.67	97.8	121.9	92.3
5		49.80	136.3	106.6	170.7		50.46	100.7	121.5	107.1
10	47.21	41.75	117.1	57.3	91.3	99.72	36.36	68.8	84.3	72.2
15		35.41	69.2	71.9	68.4		34.22	45.8	53.5	43.1
20	46.57	35.87	71.3	66.8	57.3	138.05	33.56	48.7	47.9	40.7
25		34.64	63.7	55.3	49.3	42.24	33.89	50.9	47.3	39.1
30	41.36	33.73	68.8	52.9	47.4	42.08	35.17	47.2	44.3	34.8

The present ³¹P MAS-NMR study is similar to one previously applied to an acyl-oligolysine (11). Both studies demonstrate the ability of these very different antimicrobial agents to segregate anionic and zwitterionic lipids. By comparison, the present set of small peptides, KR-12, GF-17D3, and GF-17, are somewhat more effective in segregating lipids, as observed by NMR. With these three peptides, the TOCL peak in the lipid mixture is shifted more and has an even greater increase in line width, while the POPE peak is closer to the chemical shift of pure POPE. The CSA of the lipids is narrower with these peptides, indicating that they have more effect in fluidizing the bilayer and/or decreasing the average size of the MLV.

DISCUSSION

Antimicrobial peptides function by a variety of mechanisms. In general, these agents have not evolved to be specific against a single target. One of the factors that appears to contribute to the bactericidal effects of these agents is their ability to perturb membranes by reorganizing the domain structure of the membrane (2, 10, 11). We have previously shown that some peptides with unnatural amino acids (6, 7) as well as a lipopeptide (11) function by this mechanism. Evidence that this mechanism plays an important role in the actions of some antimicrobial peptides is that it allows the prediction of which bacterial species will be most susceptible to agents that act in this fashion (10, 11). The results of the present investigation indicate that the smallest active antimicrobial fragment of LL-37 (i.e., KR-12) functions by inducing lateral-phase separation. KR-12 is toxic to E. coli, an organism with both anionic and zwitterionic lipids in which cationic peptides can induce the segregation of these two kinds of lipids. In contrast, S. aureus is comprised largely of anionic lipids, and therefore a cationic peptide cannot induce phase segregation, and KR-12 (as well as some of the other peptides studied) has low toxicity to this organism. This is an example of a segment of a natural defense peptide that can act by segregating anionic lipids. The longer peptide GF-17 also can induce lateral-phase separation in mixed zwitterionic and anionic membranes, but it also promotes the permeabilization of anionic membranes. Thus, GF-17 can function by a combination of at least two membrane-related mechanisms and as a consequence is active against a broad range of bacteria. LL-37 is not capable of segregating lipid; however, it is potent in perturbing membrane organization and inducing membrane permeabilization. It is found that it is also toxic to a range of bacteria by mechanisms other than lateral-phase separation. GF-17D3 exhibits membrane properties similar to those of KR-12 and is also selectively toxic to bacteria containing anionic and zwitterionic membrane lipids. The shortest peptide, RI-10, has little effect either on lipid segregation or on membrane permeabilization. Thus, there is a strong correlation between the behavior of these peptides with model membranes and their toxicity to different bacterial species.

Although LL-37 contains within its sequence the segment with the residues of KR-12, LL-37 is not effective in promoting lateral-phase separation. We have suggested that a property required for promoting lateral-phase separation is a conformational flexibility that allows the peptide to adopt its conformation so that the distances between positive charges match those between the negative charges on the lipid (10). It appears that the amphipathic helix formed by LL-37 is too rigid to effectively fulfill this requirement. This is supported by the heteronuclear nuclear Overhauser effect data of LL-37 in complex with anionic micelles (12), which indicate that only the C-terminal region (residues 33 to 37) is truly mobile on the ps-ns time scale and that it does not participate in membrane binding. In addition, it is possible that some cationic side chains, which can interact with PG in shortened peptides, are not available to interact with anionic lipids due to the formation of intramolecular salt bridges with acidic residues to stabilize the long helix of LL-37 covering residues 2 to 31 (32). Interestingly the mitochondrial creatine kinase, a larger protein with a specific oligomeric structure, is capable of inducing lateral-phase separation (8), with clustering of anionic lipids. However, the segment of this protein that interacts with anionic lipids is situated on flexible, relatively mobile segments that are poorly resolved in the electron density maps of the X-ray structures (24). Taken together, we propose that it is necessary that key cationic side chains not be “locked” in the protein structure so that they are available for interactions with anionic lipids, leading to lipid domain formation.

Antimicrobial agents can kill bacteria by binding to intracellular targets as well as affecting properties of the cell membrane. However, for most of these agents, the membrane plays some role, either directly as the drug target or indirectly as a barrier that determines how much of the agent reaches the intracellular target. Mechanisms related to membrane damage have received much attention. It has been shown that peptides forming clusters of amphipathic helices can stabilize the formation of transient pores, as has been suggested for other antimicrobial peptides (14), or permeabilize via a “carpet” model that has been suggested specifically for LL-37 (20). The structures of LL-37 in zwitterionic micelles (22) and in anionic micelles (32) have been determined by NMR. LL-37 forms the most stable amphipathic helix, with insertion of some of the aromatic side chains into the lipid (32). This peptide is also the most potent in inducing membrane permeabilization (Fig. 6), likely by a pore or carpet mechanism. This is in agreement with a recent report of the potent liposome lytic activity of this peptide (17). Furthermore, the same study also showed that, compared with the human form of LL-37, orangutan LL-37 has higher helicity and greater lytic potency, while the orthologue from rhesus macaques has less helicity and is less lytic (17). These results show a clear relationship between helix-forming ability and lytic potency. This also provides an explanation for why LL-37 is toxic to gram-negative bacteria with a high PE content, even though it does not segregate anionic and zwitterionic lipids.

A shortened version of LL-37, the GF-17 fragment, is expected to have a shorter helix and it is indeed less lytic (Fig. 6). When three residues of this peptide are replaced by d-amino acids to form GF-17D3, the resulting peptide has lower hydrophobicity (12) and was determined to have less helicity because of the mixture of l- and d-residues. A still shorter peptide, KR-12, is partly helical, as determined by NMR in PG micelles (32); however, it has been found by NMR to have a frayed structure at the amino terminus (32). The removal of the critical hydrophobic residue F17 also caused the disruption of the hydrophobic cluster involving this aromatic ring. This deletion again led to a less hydrophobic peptide. Hence, these two peptides, KR-12 and GF-17D3, cannot form pores as readily, as is shown by the fact that they do not lyse anionic liposomes. In addition, the still shorter, inactive peptide RI-10 cannot induce either leakage of anionic liposomes or separation between anionic and zwitterionic lipids. Key and flexible cationic side chains are indispensable for these activities, because removal of the terminal K18 and R29 from KR-12 disables the PG-binding capability of the resulting peptides (e.g., RI-10 in Fig. 2), thereby disrupting antibacterial activity (32).

The activities of KR-12, GF-17, and GF-17D3 against E. coli can be attributed in part to their equal numbers of cationic side chains. These peptides contain three arginines and two lysine residues (Table 2). This cationic-residue-rich domain of human LL-37 covered by the sequence of KR-12 is responsible for the induction of lipid domain formation by attracting anionic lipids such as PG and CL. Indeed, Arg-PG interactions have been detected by solution NMR for these peptides (12, 29, 32), indicating that the side chain of arginines can approach PG head groups to ∼5 Å.

The strong correlation between the ability of certain peptides to induce lipid clustering and their toxicity only to bacteria that contain both anionic and zwitterionic lipids in their membranes suggests that this phenomenon contributes to the antimicrobial activities of only certain antimicrobial agents. However, the exact mechanism by which the segregation of membrane lipids in bacteria leads to toxicity is not known, although several processes may contribute to this novel antimicrobial mechanism. Clustering of lipids will result in boundaries between these clusters and the remainder of the membrane. Such a boundary will have a line tension that will cause the formation of defects and lower the resistance of the membrane. A minor defect may allow increased permeability only when the membrane is under stress. A larger defect may allow for the dissipation of the electrochemical gradient. A much larger defect would allow the passage of larger polar molecules and approach the properties of a pore. Clustering of anionic lipids by the cationic antimicrobial agents could also remove these lipids from existing bacterial domains, damaging the biological functioning of these domains. One should not view these mechanisms as alternatives but rather as a continuum and combination of mechanisms.

In summary, KR-12 is effective in segregating anionic from zwitterionic lipids. This is in accord with its more specific toxicity to bacteria that contain both anionic and zwitterionic lipids. GF-17, in contrast, in addition to segregating anionic and zwitterionic lipids, can also induce leakage of membrane bilayers composed only of anionic lipids (Fig. 6). Hence, GF-17 has a broader range of toxicities (Table 3). However, a still less helical peptide, GF-17D3, acts on membranes almost exclusively by lipid segregation and does not induce permeabilization. This peptide, like KR-12, is also specific for bacteria containing zwitterionic lipid. A slightly shorter peptide, RI-10, has no significant effects on model membranes and is devoid of significant antimicrobial activity. Our work demonstrates that the smallest and cationic-residue-rich segment (KR-12) of a natural antimicrobial peptide, LL-37, can function by clustering anionic lipids. This finding is in accord with the known antimicrobial specificity of KR-12, which differs significantly from the wide-spectrum activity of its parent molecule.