How lipids influence the mode of action of membrane-active peptides

Available online at www.sciencedirect.com

1768 (2007) 2586–2595www.elsevier.com/locate/bbamem

Biochimica et Biophysica Acta

How lipids influence the mode of action of membrane-active peptides

E. Sevcsik, G. Pabst, A. Jilek 1, K. Lohner ⁎

Institute of Biophysics and Nanosystems Research, Austrian Academy of Sciences, Schmiedlstrasse 6, 8042 Graz, Austria

Received 25 April 2007; received in revised form 1 June 2007; accepted 14 June 2007Available online 23 June 2007

Abstract

The human, multifunctional peptide LL-37 causes membrane disruption by distinctly different mechanisms strongly dependent on the nature ofthe membrane lipid composition, varying not only with lipid headgroup charge but also with hydrocarbon chain length. Specifically, LL-37 induces apeptide-associated quasi-interdigitated phase in negatively charged phosphatidylglycerol (PG) model membranes, where the hydrocarbon chains areshielded from water by the peptide. In turn, LL-37 leads to a disintegration of the lamellar organization of zwitterionic dipalmitoyl-phosphatidylcholine (DPPC) into disk-like micelles. Interestingly, interdigitation was also observed for the longer-chain C18 and C20 PCs. This dualbehavior of LL-37 can be attributed to a balance between electrostatic interactions reflected in different penetration depths of the peptide andhydrocarbon chain length. Thus, our observations indicate that there is a tight coupling between the peptide properties and those of the lipid bilayer,which needs to be considered in studies of lipid/peptide interaction. Very similar effects were also observed for melittin and the frog skin peptidePGLa. Therefore, we propose a phase diagram showing different lipid/peptide arrangements as a function of hydrocarbon chain length and LL-37concentration and suggest that this phase diagram is generally applicable to membrane-active peptides localized parallel to the membrane surface.© 2007 Elsevier B.V. All rights reserved.

Keywords: LL-37; Antimicrobial peptides; Interdigitated bilayer; Micellar disks

1. Introduction

Given the rapid emergence of antibiotic resistant bacteriastrains, the development of alternatives to conventional anti-biotics with a different mode of action has become an im-perative. Antimicrobial peptides (AMPs), effector molecules ofthe innate immune system, which confer a first line of defenseagainst invading pathogens, promise to be a solution to thisproblem [1–3]. The main advantage of this class of substances,when considering bacterial resistance, is their mode of action:The majority of AMPs kills their target cells by rapid destructionof their membrane integrity [4–6]. Several models have been

Abbreviations: DMPC, dimyristoyl-phosphatidylcholine; DMPG, dimyris-toyl-phosphatidylglycerol; DPPC, dipalmitoyl-phosphatidylcholine; DPPG,dipalmitoyl-phosphatidylglycerol; DSPC, distearoyl-phosphatidylcholine;DSPG, distearoyl-phosphatidylglycerol; DAPC, diarachidoyl-phosphatidylcho-line; DSA, doxyl stearic acid; SWAXS, small- and wide-angle X-ray scattering;DSC, differential scanning calorimetry; ESR, electron spin resonance⁎ Corresponding author. Tel.: +43 316 4120 323; fax: +43 316 4120 390.E-mail address: [email protected] (K. Lohner).

1 Present address: Institute of Organic Chemistry, Johannes Kepler University,A-4040 Linz, Austria.

0005-2736/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.bbamem.2007.06.015

developed to explain membrane disruption or permeabilizationof AMPs. Thereby the focus of debate has mainly been on theformation of pores [7,8] and on the carpet-model [9], while lessattention has been paid to alternative mechanisms such asformation of defects (charge segregation) or non-lamellarstructures, which strongly depend on the membrane lipid cons-tituents [10]. The present study further emphasizes the importantrole of the nature of the membrane lipids revealing differentmechanisms of membrane perturbation exhibited by one and thesame peptide.

The human peptide LL-37, a member of the cathelicidinpeptide family, is considered to play an important role in theimmune defense participating in the inflammatory process byseveral means. The peptide was shown to be chemotactic andangiogenic, to act on cancer cells, virus and fungi and to haveLPS-binding properties (for a review see [11]). In addition, theability of LL-37 to interact with phospholipid membranes is aninherent property by which the peptide can exert its antibioticand cytolytic activity. The biophysical properties of LL-37 havebeen studied in various buffer and model membrane systems[11]. LL-37 is a 37-amino acid peptide, 16 of which are chargedat physiological pH, 11 positive and 5 negative, resulting in a net

2587E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

charge of +6. Assuming a random coil structure in pure water,the peptide adopts an amphipathic α-helical structure withapolar and polar side chains roughly equally distributed fromresidues 11 to 32 in aqueous solution at physiological pH andionic strength that is stable over a wide range of anion con-centrations as well as in the presence of lipids. The N-terminalregion consists of stretches of hydrophilic and, to a greaterextent, hydrophobic residues. Li et al. [12] also found a helicalregion from residues 3 to 7 in a 12-amino acid N-terminalfragment of LL-37. The peptide oligomerizes in solution and itwas shown that its antibacterial activity is dependent on thehelical structure and aggregation [13,14]. A feature that dis-tinguishes LL-37 from most other antimicrobial peptides, whichact selectively on bacterial cells, is its cytotoxicity [13].

Negatively charged phosphatidylglycerols (PGs) and zwitter-ionic phosphatidylcholines (PCs) are widely used in modelmembrane studies to mimic bacterial and mammalian cell mem-branes, respectively [15]. The peptide LL-37was shown to bind toand permeate effectively both zwitterionic PC and charged PC/PSvesicles [14]. In contrast, Neville et al. [16] found LL-37 to insertand disrupt only dipalmitoyl-PG (DPPG) monolayers, withvirtually no effect on dipalmitoyl-PC (DPPC) monolayers.However, these inconsistencies may be explained by the differentexperimental approach. ATR-FTIR and 31P NMR spectroscopydata of oriented bilayers indicate that LL-37 (up to 5 mol%) liesroughly parallel to the bilayer plane in PC, PC/cholesterol and PC/PG multilayers and does not penetrate deeply into thehydrophobic core of the membrane suggesting a non-pore carpetmechanism [14,17]. A detergent-like mechanism, i.e. micelli-zation was rated unlikely since rapidly tumbling membranefragments were not observed in NMR experiments [17].However, in extension to these studies our observations on PGand PC model membranes clearly demonstrate that the peptideLL-37 does not act by one single molecular mechanism, butcauses membrane disruption by distinctly different mechanismsstrongly dependent not only on the nature of the lipid headgroupbut also on the hydrocarbon chain length. Similar observations onthe antimicrobial peptide PGLa and the lytic peptide melittinsuggest a general principle for the action of membrane-activepeptides, which are localized parallel to the membrane surface.

2. Materials and methods

2.1. Materials

LL-37 was purchased from Peptide Specialty Laboratories GmbH (Heidel-berg, Germany). PGLa was purchased from Neo-MPS (San Diego, CA) andmelittin was purchased from Sigma (St. Louis, MO). Peptide purity was N95%.Phospholipids were purchased from Avanti Polar Lipids (Alabaster, AL) andused without further purification. Stock solutions of all lipids were prepared inchloroform/methanol (9:1 v/v) and stored at −18 °C. Purity (N99%) was checkedbefore and after experiments by thin layer chromatography showing a single spotusing CHCl3/CH3OH/NH3 (25% in water) 65/35/5 (v/v) as mobile phase anddetection with the phosphorus-sensitive reagent molybdenic acid.

2.2. Preparation of liposomes

Appropriate amounts of phospholipid and peptide stock solutions weremixed and dried under a stream of nitrogen, stored in vacuum overnight and

dispersed in sodium phosphate buffered saline (PBS, 20 mM, 130 mMNaCl, pH7.4). Hydration at a temperature above the phase transition of the respectivephospholipids was intermitted by vigorous vortex mixing. For ESR experi-ments, the required amount of 16-doxyl stearic acid (16-DSA) spin label (Sigma,St. Louis, MO) dissolved in ethanol was added before solvent evaporation (lipid/DSA molar ratio 150:1).

2.3. Small- and wide-angle X-ray scattering (SWAXS)

X-ray scattering experiments were performedwith a SWAXcamera equippedwith two linear, position-sensitive detectors (HECUS X-ray systems, Graz,Austria) allowing simultaneous sampling of diffraction data in the small- and thewide-angle region. The camera was mounted on a sealed-tube X-ray generator(Seifert, Ahrensburg, Germany), which was operated at 2 kW. CuKα radiation(λ=1.542 Å) was selected using a Ni filter in combination with a pulse heightdiscriminator. After equilibrating the samples for 10 min at the respectivetemperature, diffractograms for the small- and wide-angle regime were recordedwith exposure times of 3600 s each. Sample concentration was 50 mg/mL.Automatic temperature control was provided by a programmable Peltier unit.TLC performed prior to and after X-ray experiments did not reveal any signs ofX-ray induced sample degradation. Angular calibration of the scatteredintensities in the small-angle region was performed with silver stearate and inthe wide-angle region with p-bromo-benzoic acid. Background corrected datawere evaluated in terms of a global analysis technique [18]. The scatteredintensity of non-interacting particles is modeled as

IðqÞ ¼ jFðqÞj2SðqÞ; ð1Þwhere the form factorF(q) describes themodulation of the electron density by thebilayer and the structure factor S(q) the nature of the crystalline lattice. q=4π sin(θ)/λ is the modulus of the scattering wave vector. In the case of positionallycorrelated bilayers in the gel phase the paracrystalline theory [19], whichconsiders the bilayers to be rigid sheets that exhibit stacking disorder, isapplicable [18,20]. Within this theory S(q) is described by

SðqÞ ¼ N þ 2XN�1

k¼1

ðN � kÞcosðkqdÞexpð�k2q2r2PT=2Þ; ð2Þ

where N gives the average number of bilayers per scattering domain, d thelamellar repeat between adjacent bilayers and σPT is a measure for statis-tical fluctuations of bilayer separation. In the absence of positional correlationsS(q)=1.

2.3.1. Bilayer modelsBilayers are modeled as sheets of infinite extent and F(q) is simply given by

the Fourier transform of the variation of the electron density normal to the bilayersurface. As described in detail previously [21], the electron density profile ofnon-interdigitated bilayers is given by the summation of three Gaussiandistributions, two describing the electron dense headgroup region at ±zH andone at the center of the bilayer (z=0) accounting for the trough at the methylterminus of the hydrocarbon chains.

qðzÞ ¼ exp t� ðz� zHÞ2=2r2H bþ exp t� ðzþ zHÞ2=2r2H b� qrexpð�z2=2r2CÞ; ð3Þ

whereσH,C are the widths of the head and chain Gaussian peaks, respectively andρr the relative amplitude of the hydrocarbon chain peak. In the case ofinterdigitated bilayers, the hydrocarbon chain peak is replaced by a broad trough[22,23]. This is accounted for in a model description of the electron densityprofile similar to the one applied by Wiener et al. [24] by

qðzÞ ¼ qGðzÞ þ qCðzÞ þ qBðzÞ; ð4Þwhere

qGðzÞ ¼ exp t� ðz� zHÞ2=2r2H bþ exp t� ðzþ zHÞ2=2r2H b; ð5Þis the contribution from the headgroups,

qCðzÞ ¼ 0 for jzjN zH � rHqir for jzjV zH � rH

�ð6Þ

Fig. 1. (A) Electron density profile models for non-interdigitated bilayers (solidline) and interdigitated bilayers (dashed line), where Z is the distance from thebilayer center in Å and ρ the electron density in arbitrary units (a.u.). The highdensity peaks indicate the location of the phosphates of the polar headgroups; thelow density region in the center of the profiles corresponds to the hydrocarbonchains. (B) Schematic of the disk model.

2588 E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

models the constant electron density in the headgroup region and

qB zð Þ ¼ �qir2cos � p

rHz� zHð Þ þ qir

2

� �zH � rHbjzjVzH

0 zHbjzjVzH � rH

8<: ð7Þ

gives a smooth bridge between ρri and the electron density of the solvent, which is

set equal to zero. Fig. 1A shows an example of the two electron density profiles.The form factor of F(q) can be calculated analytically by

F qð Þ ¼ 1q

Z d=2

�d=2q zð Þcos qzð Þdq ð8Þ

using either Eqs. (3) or (4).If the system is a mixture of an interdigitated and a non-interdigitated phase,

the observed scattered intensity is modeled as a linear combination

IðqÞ ¼ ð1� /iÞIniðqÞ þ /iIiðqÞ ð9Þ

of the intensity from non-interdigitated bilayers Ini(q) and interdigitated bilayersIi(q), where ϕi is the fraction of bilayers in the interdigitated phase. From thefitting parameters we define the phosphate-phosphate distance dPP=2zH as ameasure for the membrane thickness.

2.3.2. Disk modelThe form factor of disk-like micelles is modeled analogous to Nieh et al.

[25], i.e.,

jFðqÞj2 ¼ PðqÞ ¼Z p=2

0g2ðq; aÞsina da; ð10Þ

where

g q; að Þ ¼ qcore � qheadð ÞVcoresinðx1Þx1

J1ðy1Þy1

þ qcoreVdisksinðx2Þx2

J1ðy2Þy2

: ð11Þρcore and ρhead are average electron densities of the hydrophobic core andhydrophilic heads, Vcore and Vdisk the volume of the hydrocarbon core and thedisk, respectively. J1 denotes the first order Bessel function and

x1 ¼ qL2cosa;

x2 ¼ qðL=2þ tÞcosa;

y1 ¼ qRsina;

y2 ¼ qðRþ tÞsina

with L being the thickness of the hydrocarbon core, t the thickness of thehydrophilic layer and R the disk radius (Fig. 1B). The form factor is obtained bynumerical integration of Eq. (10). Additionally, we account for polydispersedisk-radii by incorporating the Schultz distribution function

f rð Þ ¼p�2=p2 r

hR0i� �ð1�p2Þ=p2

exp � rp2hR0i

� �

hR0iCðp�2Þ ; ð12Þ

where Γ is the Gamma function, p=σ/⟨R0⟩ and σ2 is the variance of R and ⟨R0⟩

represents the average disk radius.

2.3.3. WAXS data analysisFrom the wide-angle peaks we calculate the average area per hydrocarbon

chain for tilted bilayers as [26]

AC ¼ 4p2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiq220q

211 � q420=4

p ; ð13Þ

and for the LβIq phase, where chains pack on a hexagonal lattice, we apply [27]

AC ¼ 8p2ffiffiffiffiffiffiffiffiffi3q211

p : ð14Þ

qhk corresponds to the position of the hydrocarbon reflections in the WAXSregime with the Miller indices h and k.

2.4. Electron spin resonance (ESR)

ESR spectra were recorded on a X-band ECS106 spectrometer (Bruker,Rheinstetten, Germany) cooled with liquid nitrogen. Measurement conditionswere: 3390G center field, 100G sweepwidth, 1 Gmodulation amplitude, 50 kHzmodulation frequency and 10 mW microwave power. A sweep time of 21 s andaccumulation of 10 scans yielding acceptable signal to noise ratio were usedthroughout the experiments. Measurements were carried out at 9 °C. Spectralparameters were determined using the Win-EPR Software. The maximumhyperfine splitting, AII, of the ESR spectra was used as a measure of the motionalrestriction of the spin label. Sample concentration was 30 mg/mL.

2.5. Differential scanning calorimetry (DSC)

DSC studies were performed with a Microcal VP-DSC high-sensitivitydifferential scanning calorimeter (Microcal, Northampton, MA). Scans wererecorded at a constant rate of 30 °C/h. Sample concentration was 1 mg/mL. Dataanalysis was done using Microcal's Origin software. Calorimetric enthalpieswere calculated by integrating the peak areas after baseline adjustment andnormalization to the mass of phospholipid. The phase transition temperature wasdefined as the temperature at the peak maximum.

2.6. Negative staining electron microscopy

Lipid suspensions were diluted to 0.2 mg/mL, 1 drop was incubated for 1 minon a pioloform-coated 100 mesh copper grid and stained with either 1% (w/v)phosphotungstic acid or 1% (w/v) uranyl acetate for 1 min, dabbed and air-dried.Images were recorded on a Tecnai G212 (FEI Company, Eindhoven, Netherlands)equippedwith a Gatan BioscanCCD camera (Pleasanton, CA) at 100 kVand 23 °C.

3. Results

3.1. X-ray scattering experiments

The effect of LL-37 on the structural parameters of modelmembranes was studied by small- and wide-angle X-ray

Fig. 2. SAXS pattern of pure DPPG (A), DPPG with 2 mol% LL-37 (B) and4 mol% LL-37 (C). Data were recorded at 25 °C. Solid lines show full-q-rangefits. WAXS pattern are shown in the middle panel. The electron density profilescalculated from the fits are shown at the right.

2589E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

scattering. Data presented were analyzed by applying a full-q-range, global fitting model [18,20]. At 25 °C, the SAXS patternof pure dipalmitoyl-PG (DPPG) liposomes is characteristic forpositionally weakly correlated bilayers of oligolamellar vesicles.A lamellar d-spacing of 110 Å and a phosphate-phosphatedistance (dPP) of 44.6 Å were calculated (Fig. 2A, Table 1). Thecorresponding WAXS data are characteristic for the tiltedlamellar (Lβ′) phase [28].

In the presence of 4 mol% LL-37 the SAXS pattern of pureDPPG is replaced by diffuse scattering indicating a loss ofpositional correlations between the bilayers (Fig. 2C). Analysisof the scattering data gives dPP=31.1 Å, which is considerablysmaller than in the absence of the peptide. We were further ableto fit the data with an electron density profile model ρ(Z), whichexhibits a broad, ill-defined trough in contrast to pure DPPG,whose ρ(Z) exhibits a pronounced minimum at the center of thebilayer. This suggests the formation of an interdigitatedstructure, which is in agreement with WAXS data (Fig. 2C)showing a single symmetric peak, characteristic for hexagonallypacked all-trans lipid acyl chains oriented perpendicular to the

Table 1Phosphate–phosphate distances in Å of PG and PC bilayers in the gel phase inthe presence of different concentrations of LL-37

Mol% LL-37

0 2 4

DMPG 41.1 n.d. 40.4DPPG 44.6 46.3/31.2 ⁎ 31.1DSPG 48.1 n.d. 32.4DMPC 39.5 n.d. 39.6DPPC 43.4 44.1 44.0DSPC 47.0 n.d. 48.8/34.0 ⁎

DAPC 51.0 n.d. 53.3/35.5 ⁎

Data were recorded at (Tm—15) °C.⁎ LβIq/Lβ′ phase coexistence.

bilayer plane. However, the width of the WAXS peak is aboutthree times larger than the sharp (2,0) reflection in pure DPPGindicating a reduced correlation length originating from adecrease of packing order. The interdigitated bilayer ismaintained up to the gel-to-fluid phase transition temperature(Tm). In the fluid (Lα) phase, dPP increases to 35.7 Å, which isstill noticeably thinner than pure DPPG bilayers (dPP=38.1 Å).At 2 mol% LL-37 the SAXS pattern shows additional diffusescattering around q=0.2 Å−1 indicating the presence of asecond coexisting bilayer structure (Fig. 2B). Application of afitting model for coexisting interdigitated and non-interdigitatedbilayers yields dPP=46.3 Å for the Lβ′ and dPP=31.2 Å for theinterdigitated phase, respectively. The effect of LL-37 on thelonger-chain distearoyl-PG (DSPG) is comparable to DPPG, aswe only find a single interdigitated phase at 4 mol% peptide(Table 1).

The SAXS pattern of pure dipalmitoyl-PC (DPPC) showscharacteristic Bragg peaks up to the 4th order indicatingmultilamellar vesicles with a lamellar repeat d of 65.7 Å at25 °C corresponding to the Lβ′-phase (Fig. 3A). The fit yieldsdPP=43.4 Å, which is in good agreement with the valuepublished by Nagle and coworkers [29] at 20 °C, consideringthe usual increase of dPP in the gel phase with temperature. Inthe presence of 4 mol% peptide, the multilamellar arrangementof DPPC is completely destroyed. Analysis of X-ray data

Fig. 3. SAXS pattern of pure DPPC at 25 °C (A), DPPC with 4 mol% LL-37 at25 °C (B) and 50 °C (C). Solid lines show full-q-range fits with q being thescattering vector q=(4π sin θ)/λ. Below the scattering pattern, sketches of thecorresponding macroscopic phases are shown: Multilamellar vesicles (A), disk-like micelles (B) and extended bilayers (C).

Fig. 4. Effect of LL-37 on DPPG and DPPC bilayers. (A) ESR spectra of a 16-DSA spin label in vesicles of DPPG with 4 (I) and 2 mol% LL-37 (II) and DPPCwith 4 mol% LL-37 (III). The corresponding spectra of the pure lipids are shownin grey. T=9 °C; total scan width 100 G. (B) Thermograms of DPPG with 4 (I)and 2 mol% LL-37 (II) and DPPC with 4 mol% LL-37 (III) recorded at a scanrate of 30 °C/h. The corresponding thermograms of the pure lipids are shown ingrey. For clarity, traces are offset.

2590 E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

reveals that the diffuse scattering arises from discoidal particleswith a diameter of 270 Å and a bilayer thickness of 55 Å.Considering the amphipathic structure of LL-37, the most likelyarrangement of lipids and peptides in the discoidal particleswould be a lipid disk surrounded by a ring of peptides, assketched in Fig. 3B. Upon heating above the gel-to-fluid phasetransition temperature, the SAXS pattern displays again a smallpeak at d=74 Å (Fig. 3C), indicating that the disks have fusedinto a positionally weakly correlated bilayer structure with a dPPof 37.4 Å, which is similar to that of pure DPPC (38.2 Å). Thelarger d-value (as compared to 66 Å for pure DPPC) furthersuggests that the peptide helices are located parallel to thebilayer surface yielding a net-repulsive interaction due to thepositive surface charges of LL-37. This peptide location is inagreement with earlier studies on PC/cholesterol and PC/PGmixtures [14,17].

LL-37 has a comparable effect on dimyristoyl-PC (DMPC)(data not shown), but formation of disk-like micelles was notobserved for the longer-chain distearoyl-PC (DSPC) anddiarachidoyl-PC (DAPC) (Table 1). In contrast, data analysisstrongly suggests a coexistence of interdigitated and non-interdigitated lipid domains, quite similar to the situationobserved for DPPG with 2 mol% LL-37 (Table 1).

3.2. Electron spin resonance spectroscopy

A major difference between interdigitated and non-inter-digitated bilayers is the mobility of the hydrocarbon chains nearthe terminal methyl group. In order to obtain additional evidencefor the formation of interdigitated phases in the presence of LL-37, we have therefore performed ESR experiments using 16-doxyl stearic acid (DSA) spin labels. The spectrum of DPPG at4 mol% peptide clearly shows an increase of the maximum outerhyperfine splitting, 2AII (Fig. 4A). The high AII value of 30.6 Gcompared to the value of 24.8 G for pure DPPG indicates that thespin label is more motionally restricted than in the pure lipidbehaving rather like a label located much closer to the carboxylgroup than a label near the terminal methyl group. ESR spectrasuggesting a coexistence of two bilayer structures with differentorder parameters were found for DPPGwith 2mol%LL-37 (Fig.4A), DSPC with 4 mol% LL-37 and DAPC with 4 mol% LL-37(data not shown) supporting our assumptions from SAXS dataanalysis. In contrast, the spectra of DPPC with and withoutpeptide are comparable.

3.3. Differential scanning calorimetry

The heat capacity function of pure DPPG shown in Fig. 4B ischaracterized by a symmetric peak at 40.7 °C corresponding tothe highly cooperative main phase transition (Pβ′→Lα) and apre-transition related to the formation of a ripple phase(Lβ′→Pβ′) at 32.8 °C, consistent with previously publisheddata [30]. At 4 mol% LL-37, Tm is increased to 42.6 °C due to astabilization of the gel phase in consequence of interdigitation.At 2 mol% LL-37, two overlapping transitions can be observed,probably arising from a phase separation into interdigitated andnon-interdigitated domains, as already assumed from SAXS and

ESR data. A similar situation is found for DAPC with 4 mol%LL-37 (data not shown). No significant change in the maintransition enthalpy was noticed. However, in the presence of4 mol% LL-37 the sharp peak of the DPPCmain phase transitionis replaced by a very broad transition with a significantlyreduced enthalpy characteristic for systems in which smallparticles – e.g. disk-like micelles – have been generated.

3.4. Negative staining electron microscopy

Fig. 5A shows DPPC multilamellar vesicles of a wide sizedistribution stained with uranyl acetate. In the presence of 4 mol% LL-37 the multilamellar arrangement is completely disin-tegrated and small particles exhibiting a disk-like structure arecreated. The disks are better observable in Fig. 5C showing thesame sample stained with phosphotungstic acid, where therouleaux-like appearance of the DPPC/LL-37 disks is an artifactof the negative staining technique [31]. The disk thickness isestimated to be around 50 Å, the diameters range from 250 to300 Å, which is in accordance with SAXS data analysis.

4. Discussion

Electrostatic effects are regarded as the main reason for theselectivity of antimicrobial peptides towards bacteria, which iscertainly the case as far as peptide accumulation on themembrane surface is concerned. Thus, while the importance oflipid headgroup charge has taken heed in most model membraneexperiments, other structural parameters like lipid molecularshape and chain length and the consequences that arise for the

Fig. 5. Negative staining electron microscopy images of pure DPPC (A) andwith 4 mol% LL-37 stained with uranyl acetate (B) and DPPC with 4 mol% LL-37 stained with phosphotungstic acid (C) at 23 °C.

Fig. 6. Cartoon illustrating the coexistence of a peptide-associated LβIq phasewith a Lβ′ phase. The Lβ′ phase is suggested to be inclined with respect to theinterdigitated phase in order to minimize the defect zone as suggested in [51].For simplification, the LL-37 helix is drawn without side chains. We tentativelyindicate different lipid headgroup orientations in the different phases asdescribed in the text.

2591E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

interaction with peptides often are still not sufficientlyconsidered [10]. In our study we clearly show that the differentmembrane-perturbing mechanisms of these peptides are notsolely based on charge effects but also are closely related to theinherent physicochemical properties of the lipids. Further, wepropose a tentative phase diagram, which accounts for thecoupling of the lipid properties to the macroscopic phases foundin the presence of LL-37 (see below).

4.1. Structural rearrangements of PG and PC bilayers

As concluded from a drastically reduced bilayer thickness inassociation with a characteristic electron density profile and anincreased order parameter at the terminal methyl group of thehydrocarbon chains, LL-37 induces the formation of aninterdigitated bilayer structure in DPPG and DSPG. It isinteresting to note that large, amphipathic cations like acetylcho-line [32] and polymyxin B [33] have been shown to give rise tohydrocarbon chain interdigitation in DPPG vesicles as well. Inthe present case and given the spectroscopic evidences [14,17]the most likely organization of an amphipathic peptide like LL-37 is parallel to the bilayer plane shielding the tail ends of thelipids from the interfacial water, thereby forming a quasi-

interdigitated structure LβIq (Fig. 6). From WAXS data an areaper lipid of 40 Å2 was calculated. This value is significantlysmaller as compared to the Lβ′ phase (46 Å

2) and thus indicatesthat the headgroups tilt away from an orientation parallel to themembrane surface allowing the tighter hydrocarbon chainpacking of the interdigitated phase as sketched in Fig. 6.

The lateral area occupied by the peptide in the membrane canonly be roughly estimated, as no knowledge exists on the spacerequirements of the unstructured C- and N-terminal parts, whichinclude 6 and 10 amino acids, respectively. As a firstapproximation, assuming a full helical structure, Neville et al.[16] suggested the molecular area of LL-37 to be around 550 Å2.However, this estimation did not include side chains. Whentaking into account space for the side chains, the helix diameterwould increase to about 16 Å as can be deduced from data of aclass A amphipathic helix [34]. Consequently, the total areawould amount to 880 Å2. These estimated areas yield a molarratio of 14 to maximally 22 lipid molecules per LL-37suggesting that not all lipids are taking part in the formation ofthe quasi-interdigitated structure at 4mol% peptide. However, aswe are able to fit the SAXS data of DPPG and DSPG in thepresence of 4 mol% LL-37 with only a single bilayer model weconclude that LL-37 is able to induce an interdigitated phasewithin the non-peptide associated, pure lipid phase. Theunderlying reason for this is most likely that the pure lipids fillthe space between the peptide/lipid domains in order to avoidenergetically unfavorable line defects along the boundaries ofquasi-interdigitated and non-interdigitated domains. Addition-ally, longer chain lipids like DSPG, which are prone to adopt aninterdigitated structure by themselves, as will be describedbelow, will further promote such an arrangement [35]. Atintermediate concentrations (2 mol% LL-37), the perturbationimposed by the peptide is not sufficient to convert the non-peptide associated lipids completely into interdigitated bilayersand the LβIq phase coexists with a Lβ′ phase. We find a similarsituation for the long-chain zwitterionic lipids DSPC and DAPCat 4 mol% LL-37.

Quite in contrast to the effect observed on PGs and longchain PCs, addition of LL-37 to DPPC gives rise to the

Table 2Macroscopic phases observed for different lipids in the presence of LL-37,PGLa and melittin

2 mol%LL-37

4 mol%LL-37

4 mol%Melittin

8 mol%Melittin

4 mol%PGLa

8 mol%PGLa

DMPC M/L M Ma n.d. L LDPPC M/L M Mb Mb L LDAPC n.d. I/L I/L (Mb, c) I/L I/L I/LDMPG L L n.d. L L n.d.DPPG I/L I I/L I/L I/L I/LDSPG I/L I I/L I I/L I/L

Data were recorded at (Tm—15) °C. L (lamellar non-interdigitated), M (micellardisks), I (interdigitated).a From [53], ((Tm—11) °C).b From [52], (6.7 mol% melittin, (Tm—11) °C).c Only after temperature quench from Lα.

2592 E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

formation of disk-like micelles with diameters around 270 Åand a bilayer thickness of about 55 Å. Disk-like micelles with acomparable diameter but a smaller thickness (49 Å) were alsofound for DMPC with 4 mol% LL-37. The appearance ofdiscoidal particles has already been described for DMPC in thepresence of melittin [36], staphylococcal δ-lysin [37] andapolipoproteins [38]. Formation of disk-like micelles in thepresence of LL-37 is not in contradiction to results obtained byHenzler-Wildman et al. [17] using 31P NMR, because theirexperiments were performed in the fluid phase, where we alsoobserve weakly correlated bilayer structures (Fig. 4C).

4.2. Rationale for the two different mechanisms of membraneperturbation by LL-37

Upon membrane contact, an amphipathic peptide like LL-37,with its polar and apolar side chains roughly equally distributed,will position itself in the interface [11,13], thereby occupyingspace in the headgroup region of the outer leaflet and,consequently, leading to voids in the hydrophobic core of themembrane, an energetically unfavorable situation. One canenvision that the lipids in the fluid phase with a high degree offlexibility will fill these voids – up to a certain point – by e.g.increased trans-gauche isomerization of the neighboring lipidsand/or by moving the interior leaflet towards the peptides as putforward by Huang and co-workers [39]. In the gel phase, themore rigid chains cannot compensate the voids created below thepeptide helices, which eventually leads to a structural rearrange-ment of the bilayer. With LL-37, the system seems to have twopossibilities to deal with the inability of the gel phase bilayer toaccommodate the peptide: formation of a quasi-interdigitatedphase (Fig. 6), as observed for DPPG, DSPG, DSPC and DAPCor formation of disk-like micelles, as seen for DMPC and DPPC.

When comparing the C16 phospholipids DPPG and DPPC,the most obvious parameter leading to the different membrane-perturbing mechanisms is electrostatic attraction. Thus, onepossibility to explain the different modes of action of LL-37 is

Fig. 7. Schematic phase diagram proposing different lipid/peptide arrangementsfound for LL-37/phospholipid mixtures as a function of hydrocarbon chainlength and peptide concentration. L (lamellar non-interdigitated), M (micellardisks), I (interdigitated), PG (phosphatidylglycerol), PC (phosphatidylcholine).

that electrostatic interactions between the cationic peptide andanionic lipid headgroups inhibits deeper insertion into the acylchain region [40,41]. Hence, LL-37 will be anchored in theheadgroup region of PGs and create voids below the peptidehelices, which cannot be compensated by the hydrocarbonchains in the gel phase. The system consequently minimizes itsenergy by moving the interior leaflet towards the peptidesforming a LβIq phase. The ability to induce hydrocarbon chaininterdigitation has also been proposed for e.g. short chainalcohols, glycerol, Tris, anesthetics, polymyxin B, choline andacetylcholine [22,32,33,42–45]. Although the precise require-ments for a molecule to cause the formation of an interdigitatedphase are not known, it seems that the occupation of space in theheadgroup region is a major prerequisite, whereby the moleculesmust not extend too far into the hydrophobic core [22].

In turn, cationic peptides may penetrate much deeper into theinterior of neutral bilayers, such as DPPC. If the double diameterof an α-helix is comparable to the thickness of the hydrophobiccore of the membrane (∼38 Å for DPPC and ∼33 Å for DMPCas compared to an estimated helix diameter of LL-37 of∼16 Å),the peptidemay change to a position inside the hydrophobic corewith its apolar side facing the hydrocarbon chains analogous tothe “bicycle-tire” model suggested for melittin [46] andapolipoprotein E [47]. Such disk-like micelles are observed forDPPC and DMPC in the presence of LL-37 as sketched in Fig.3B.Minor differences between the double helix diameter and thehydrophobic core of the bilayer can be compensated by theflexibility of the peptide itself and/or by curvature of the lipidbilayer in the vicinity of the peptide. In bilayers of longer-chainPCs, disk-formation is obviously more difficult, as parallelarranged LL-37 cannot match anymore the hydrophobic core ofthe bilayer and thus a quasi-interdigitated structure becomesenergetically more favorable.

The balance between peptide penetration depth and hydro-carbon core thickness certainly is a major reason for thedifferent modes of action of LL-37. However, there is anotherfactor to consider: the lipids own propensity to forminterdigitated phases that facilitates the conversion into such astructure. Sun et al. [48] have reported the coexistence of Lβ′

and LβI phases in long-chain (C24) PC. Interestingly, an analogscenario has been most recently found in our laboratory for

Fig. 8. Representative examples for SAXS patterns of lipids in the presence ofPGLa or melittin. DPPG with 4 mol% PGLa (A), DSPG with 8 mol% melittin(B) and DAPC with 4 mol% melittin (C). Data were recorded at (Tm—15) °C.Solid lines show full-q-range fits with q being the scattering vector q=(4π sinθ)/λ. The electron density profiles calculated from the fits are shown in theinsets, where Z is the distance from the bilayer center in Å and ρ the electrondensity in arbitrary units (a.u.).

2593E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

DSPG, which exhibits small amounts of an interdigitated phase[35]. The tendency to interdigitate is based on the packingfrustration arising from the mismatch between the stericrequirements of the lipid head groups and the acyl chains andis additionally promoted in long-chain lipids, since the van derWaals energy gain upon interdigitation is directly related tochain length [49]. Although PGs have smaller headgroups thanPCs, electrostatic repulsion between the negatively chargedheadgroups makes them more prone to interdigitation, which isreflected in the occurrence of an interdigitated phase at shorterchain length. At even shorter acyl chain lengths, the lamellararrangement of PGs becomes unstable. For dicapryl-PG (C10) amixture of micellar, lamellar and hexagonal-like arrangementswas reported at 150 mM NaCl, whereas PC builds stablebilayers with acyl chains ≥C9 [50]. The importance of thehydrocarbon chain length is also reflected in the interaction ofLL-37 with dimyristoyl-PG (DMPG, C14), where we findstacks of non-interdigitated bilayers with a lamellar repeat of50 Å (Table 1).

4.3. Introduction of a phase diagram and its general applicability

Fig. 7 shows a tentative phase diagram depicting the lipid/peptide arrangements found for phospholipid/LL-37 mixtures

as a function of peptide concentration and lipid chain length.While the phase diagrams of the lipid/peptide mixtures arequalitatively quite comparable, the respective phase boundariesof PCs/LL-37 are shifted towards longer chain lengths ascompared to PGs/LL-37. We speculated that this behaviormight not be restricted to LL-37 but apply to other α-helical,amphipathic peptides as well. Indeed, we observe phasecoexistence of quasi-interdigitated and non-interdigitatedphases in DSPG, DPPG and DAPC in the presence of 4 mol% PGLa or melittin, which consist of 21 and 26 amino acidresidues, respectively (Table 2, Fig. 8). A fully interdigitatedphase was found only for melittin at 8 mol%. Differences in theability and extent of the induction of a quasi-interdigitatedphase and disk-like micelles as seen for these peptides can beattributed to e.g. different helix topology and length, hydro-phobic angle and charge distribution. We thus have strongevidence that the phase diagram shown in Fig. 7 is in principleapplicable for other lipid/peptide mixtures, as well, as long asthe amphipathic peptides adopts a spatial orientation parallel tothe membrane surface.

In summary, we clearly demonstrate that the mechanism ofmembrane disruption by membrane-active peptides is stronglydependent on the nature of the lipids and can vary not onlywith lipid headgroup charge but also with hydrocarbon chainlength. The differences in the mode of interaction of thepeptides investigated with the various phospholipids, namelyinterdigitation and micellization, are first and foremostapparent in the gel phase. It seems only plausible that thevery different ways of membrane perturbation observed in thegel phase must be reflected in some form in the fluid phase,however, this phase presents a more complex situation.Although further systematic studies will be necessary toaddress this point, the decrease of membrane thickness in thefluid phase observed in this study for LL-37 in the case of gelphase interdigitation indeed provides a first indication that theeffects also propagate into the fluid phase. This would also bein accordance with reports of membrane thinning for a numberof other antimicrobial peptides [39]. The observed decrease of2–3 Å was explained by Huang to be due to an average valueover local dimple deformations, which share structuralsimilarity with the quasi-interdigitated lipid–peptide arrange-ment, i.e. bending of the opposite monolayer towards thepeptide. Independent from the fluid phase, the very differentlipid/peptide arrangements observed in the gel phase stronglysuggest that the mode of interaction of these membrane-activepeptides is influenced by the physicochemical properties of thelipids resulting in different perturbation of the membranebilayer.

Acknowledgements

We thank E. Ingolic for the EM images and R. Prassl forexpert technical assistance with the ESR spectroscopy. Thisstudy has been supported by grant P 15657 of the Fonds zurFörderung der wissenschaftlichen Forschung in Österreich andby the Alois Sonnleitner Stiftung, Österreichische Akademieder Wissenschaften.

2594 E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

References

[1] H.G. Boman, Antibacterial peptides: key components needed in immunity,Cell 65 (1991) 205–207.

[2] M. Zasloff, Antimicrobial peptides of multicellular organisms, Nature 415(2002) 389–395.

[3] T. Ganz, Defensins and host defense, Science 286 (1999) 420–421.[4] D.Wade, A. Boman, B.Wahlin, C.M. Drain, D. Andreu, H.G. Boman, R.B.

Merrifield, All-D amino acid-containing channel-forming antibioticpeptides, Proc. Natl. Acad. Sci. U. S. A. 87 (1990) 4761–4765.

[5] K. Lohner, R.M. Epand, Membrane interactions of hemolytic and anti-bacterial peptides, Advances in Physical Chemistry, Jai Press Inc.,Greenwich, Conn., 1997, pp. 53–66.

[6] Y. Shai, Mode of action of membrane active antimicrobial peptides,Biopolymers 66 (2002) 236–248.

[7] H.W. Huang, Action of antimicrobial peptides: two-state model,Biochemistry 39 (2000) 8347–8352.

[8] K. Matsuzaki, O. Murase, N. Fujii, K. Miyajima, An antimicrobal peptide,magainin 2, induced rapid flip-flop of phospholipids coupled with poreformation and peptide translocation, Biochemistry 35 (1996) 11361–11368.

[9] Y. Shai, Mechanism of the binding, insertion and destabilization ofphospholipid bilayer membranes by alpha-helical antimicrobial and cellnon-selective membrane-lytic peptides, Biochim. Biophys. Acta 1462(1999) 55–70.

[10] K. Lohner, S.E. Blondelle, Molecular mechanisms of membraneperturbation by antimicrobial peptides and the use of biophysical studiesin the design of novel peptide antibiotics, Comb. Chem. High TroughputScreen. 8 (2005) 241–256.

[11] U.H. Durr, U.S. Sudheendra, A. Ramamoorthy, LL-37, the only humanmember of the cathelicidin family of antimicrobial peptides, Biochim.Biophys. Acta 1758 (2006) 1408–1425.

[12] X. Li, Y. Li, H. Han, D.W. Miller, G. Wang, Solution structures of humanLL-37 fragments and NMR-based identification of a minimal membrane-targeting antimicrobial and anticancer region, J. Am. Chem. Soc. 128(2006) 5776–5785.

[13] J. Johansson, G.H. Gudmundsson, M.E. Rottenberg, K.D. Berndt, B.Agerberth, Conformation-dependent antibacterial activity of the naturallyoccurring human peptide LL-37, J. Biol. Chem. 273 (1998) 3718–3724.

[14] Z. Oren, J.C. Lerman, G.H. Gudmundsson, B. Agerberth, Y. Shai,Structure and organization of the human antimicrobial peptide LL-37 inphospholipid membranes: relevance to the molecular basis for its non-cell-selective activity, Biochem. J. 341 (Pt. 3) (1999) 501–513.

[15] K. Lohner, The role of membrane lipid composition in cell targeting ofantimicrobial peptides, in: K. Lohner (Ed.), Development of NovelAntimicrobial Agents: Emerging Strategies, Horizon Scientific Press,Wymondham, Norfolk, UK, 2001, pp. 149–165.

[16] F. Neville, M. Cahuzac, O. Konovalov, Y. Ishitsuka, K.Y. Lee, I.Kuzmenko, G.M. Kale, D. Gidalevitz, Lipid headgroup discriminationby antimicrobial peptide LL-37: insight into mechanism of action,Biophys. J. 90 (2006) 1275–1287.

[17] K.A. Henzler Wildman, D.K. Lee, A. Ramamoorthy, Mechanism of lipidbilayer disruption by the human antimicrobial peptide, LL-37, Biochem-istry 42 (2003) 6545–6558.

[18] G. Pabst, Global properties of biomimetic membranes: perspectives onmolecular features, Biophys. Rev. Lett. 1 (2006) 57–84.

[19] A. Guiner, X-ray Diffraction, Freeman, San Francisco, 1963.[20] G. Pabst, R. Koschuch, B. Pozo-Navas, M. Rappolt, K. Lohner, P.

Laggner, Structural analysis of weakly ordered membrane stacks, J. Appl.Cryst. 36 (2003) 1378–1388.

[21] G. Pabst, M. Rappolt, H. Amenitsch, P. Laggner, Structural informationfrom multilamellar liposomes at full hydration: full q-range fitting withhigh quality X-ray data, Phys. Rev., E 62 (2000) 4000–4009.

[22] T.J. McIntosh, R.V. McDaniel, S.A. Simon, Induction of an interdigitatedgel phase in fully hydrated phosphatidylcholine bilayers, Biochim.Biophys. Acta 731 (1983) 109–114.

[23] I. Winter, G. Pabst, M. Rappolt, K. Lohner, Refined structure of 1,2-diacyl-P-O-ethylphosphatidylcholine bilayer membranes, Chem. Phys. Lipids112 (2001) 137–150.

[24] M.C. Wiener, R.M. Suter, J.F. Nagle, Structure of the fully hydrated gelphase of dipalmitoylphosphatidylcholine, Biophys. J. 55 (1989) 315–325.

[25] M.P. Nieh, C.J. Glinka, S. Krueger, R.S. Prosser, J. Katsaras, SANS studyof the structural phases of magnetically alignable lanthanide-dopedphospholipid mixtures, Langmuir 17 (2001) 2629–2638.

[26] S. Tristram-Nagle, R. Zhang, R.M. Suter, C.R. Worthington, W.J. Sun, J.F.Nagle, Measurement of chain tilt angle in fully hydrated bilayers of gelphase lecithins, Biophys. J. 64 (1993) 1097–1109.

[27] T.J. McIntosh, S.A. Simon, Area per molecule and distribution of water infully hydrated dilauroylphosphatidylethanolamine bilayers, Biochemistry25 (1986) 4948–4952.

[28] A. Tardieu, V. Luzzati, F.C. Reman, Structure and polymorphism of thehydrocarbon chains of lipids: a study of lecithin–water phases, J. Mol.Biol. 75 (1973) 711–733.

[29] J.F. Nagle, S. Tristram-Nagle, Structure of lipid bilayers, Biochim.Biophys. Acta 1469 (2000) 159–195.

[30] P.M. Abuja, A. Zenz, M. Trabi, D.J. Craik, K. Lohner, The cyclicantimicrobial peptide RTD-1 induces stabilized lipid-peptide domainsmore efficiently than its open-chain analogue, FEBS Lett. 566 (2004)301–306.

[31] T.M. Forte, A.V. Nichols, E.L. Gong, R.I. Levy, S. Lux, Electronmicroscopic study on reassembly of plasma high density apoprotein withvarious lipids, Biochim. Biophys. Acta 248 (1971) 381–386.

[32] J.L. Ranck, J.F. Tocanne, Choline and acetylcholine induce interdigitationof hydrocarbon chains in dipalmitoylphosphatidylglycerol lamellar phasewith stiff chains, FEBS Lett. 143 (1982) 171–174.

[33] J.L. Ranck, J.F. Tocanne, Polymyxin B induces interdigitation indipalmitoylphosphatidylglycerol lamellar phase with stiff hydrocarbonchains, FEBS Lett. 143 (1982) 175–178.

[34] K. Hristova, W.C. Wimley, V.K. Mishra, G.M. Anantharamiah, J.P.Segrest, S.H. White, An amphipathic alpha-helix at a membrane interface:a structural study using a novel X-ray diffraction method, J. Mol. Biol. 290(1999) 99–117.

[35] G. Pabst, S. Danner, S. Karmakar, G. Deutsch, V.A. Raghunathan, On thepropensity of phosphatidylglycerols to form interdigitated phases,Biophys. J. 93 (2007) 513–525.

[36] J. Dufourcq, J.-F. Faucon, G. Fourche, J.-L. Dasseux, M. le Maire, T.Gulik-Krzywicki, Morphological changes of phosphatidylcholine bilayersinduced by melittin: vesicularization, fusion, discoidal particles, Biochim.Biophys. Acta 859 (1986) 33–48.

[37] K. Lohner, E. Staudegger, E.J. Prenner, R.N. Lewis, M. Kriechbaum, G.Degovics, R.N. McElhaney, Effect of staphylococcal delta-lysin on thethermotropic phase behavior and vesicle morphology of dimyristoylpho-sphatidylcholine lipid bilayer model membranes. Differential scanningcalorimetric, 31P nuclear magnetic resonance and Fourier transforminfrared spectroscopic, and X-ray diffraction studies, Biochemistry 38(1999) 16514–16528.

[38] D. Atkinson, D.M. Small, Recombinant lipoproteins: implications forstructure and assembly of native lipoproteins, Annu. Rev. Biophys.Biophys. Chem. 15 (1986) 403–456.

[39] H.W. Huang, Molecular mechanism of antimicrobial peptides: the origin ofcooperativity, Biochim. Biophys. Acta 1758 (2006) 1292–1302.

[40] K.A. Henzler-Wildman, G.V. Martinez, M.F. Brown, A. Ramamoorthy,Perturbation of the hydrophobic core of lipid bilayers by the humanantimicrobial peptide LL-37, Biochemistry 43 (2004) 8459–8469.

[41] M. Dathe, J. Meyer, M. Beyermann, B. Maul, C. Hoischen, M. Bienert,General aspects of peptide selectivity towards lipid bilayers and cellmembranes studied by variation of the structural parameters of amphi-pathic helical model peptides, Biochim. Biophys. Acta 1558 (2002)171–186.

[42] T. Hata, H. Matsuki, S. Kaneshina, Effect of local anesthetics on the bilayermembrane of dipalmitoylphosphatidylcholine: interdigitation of lipidbilayer and vesicle-micelle transition, Biophys. Chem. 87 (2000) 25–36.

[43] E.S. Rowe, Comparative effects of short chain alcohols on lipid phasetransitions, Alcohol 2 (1985) 173–176.

[44] D.A.Wilkinson, D.A. Tirrell, A.B. Turek, T.J. McIntosh, Tris buffer causesacyl chain interdigitation in phosphatidylglycerol, Biochim. Biophys. Acta905 (1987) 447–453.

2595E. Sevcsik et al. / Biochimica et Biophysica Acta 1768 (2007) 2586–2595

[45] J.M. Boggs, G. Rangaraj, Phase transitions and fatty acid spin labelbehavior in interdigitated lipid phases induced by glycerol and polymyxin,Biochim. Biophys. Acta 816 (1985) 221–233.

[46] C.E. Dempsey, The actions of melittin on membranes, Biochim. Biophys.Acta 1031 (1990) 143–161.

[47] V. Raussens, C.A. Fisher, E. Goormaghtigh, R.O. Ryan, J.M. Ruysschaert,The low density lipoprotein receptor active conformation of apolipoproteinE. Helix organization in n-terminal domain-phospholipid disc particles,J. Biol. Chem. 273 (1998) 25825–25830.

[48] W.J. Sun, S. Tristram-Nagle, R.M. Suter, J.F. Nagle, Anomalous phasebehavior of long chain saturated lecithin bilayers, Biochim. Biophys. Acta1279 (1996) 17–24.

[49] J.F. Nagle, D.A. Wilkinson, Lecithin bilayers. Density measurement andmolecular interactions, Biophys. J. 23 (1978) 159–175.

[50] J.H. Kleinschmidt, L.K. Tamm, Structural transitions in short-chain lipidassemblies studied by (31)P-NMR spectroscopy, Biophys. J. 83 (2002)994–1003.

[51] P. Laggner, K. Lohner, G. Degovics, K. Muller, A. Schuster, Structure andthermodynamics of the dihexadecylphosphatidylcholine– water system,Chem. Phys. Lipids 44 (1987) 31–60.

[52] J.-F. Faucon, J.-M. Bonmatin, J. Dufourcq, E.J. Dufourc, Acyl chain lengthdependence in the stability of melittin–phosphatidylcholine complexes. Alight scattering and 31P-NMR study, Biochim. Biophys. Acta 1234 (1995)235–243.

[53] C.E. Dempsey, A. Watts, A deuterium and phosphorus-31 nuclearmagnetic resonance study of the interaction of melittin with dimyristoyl-phosphatidylcholine bilayers and the effects of contaminating phopholi-pase A2, Biochemistry 26 (1987) 5803–5811.