General aspects of peptide selectivity towards lipid bilayers andcell membranes studied by variation of the structural parameters of

amphipathic helical model peptides

Margitta Dathe a;*, Jana Meyer a, Michael Beyermann a, Bjo«rn Maul a,Christian Hoischen b, Michael Bienert a

a Research Institute of Molecular Pharmacology, Robert-Roessle-Str. 10, D-13125 Berlin, Germanyb Institute of Molecular Biotechnology, Beutenbergstr. 11, D-07745 Jena, Germany

Received 29 May 2001; received in revised form 14 September 2001; accepted 21 September 2001

Abstract

Model compounds of modified hydrophobicity (H), hydrophobic moment (W) and angle subtended by charged residues (x)were synthesized to define the general roles of structural motifs of cationic helical peptides for membrane activity andselectivity. The peptide sets were based on a highly hydrophobic, non-selective KLA model peptide with high antimicrobialand hemolytic activity. Variation of the investigated parameters was found to be a suitable method for modifying peptideselectivity towards either neutral or highly negatively charged lipid bilayers. H and W influenced selectivity preferentially viamodification of activity on 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) bilayers, while the size of thepolar/hydrophobic angle affected the activity against 1-palmitoyl-2-oleoylphosphatidyl-DL-glycerol (POPG). The influence ofthe parameters on the activity determining step was modest in both lipid systems and the activity profiles were the result ofthe parameters' influence on the second less pronounced permeabilization step. Thus, the activity towards POPC vesicles wasdetermined by the high permeabilizing efficiency, however, changes in the structural parameters preferentially influenced therelatively moderate affinity. In contrast, intensive peptide accumulation via electrostatic interactions was sufficient for thedestabilization of highly negatively charged POPG lipid membranes, but changes in the activity profile, as revealed by themodification of x, seem to be preferentially caused by variation of the low permeabilizing efficiency. The parameters provedvery effective also in modifying antimicrobial and hemolytic activity. However, their influence on cell selectivity was limited.A threshold value of hydrophobicity seems to exist which restricted the activity modifying potential of W and x on both lipidbilayers and cell membranes. ß 2002 Elsevier Science B.V. All rights reserved

Keywords: Antimicrobial peptides; Model peptides; Hydrophobicity; Hydrophobic moment; Liposomes; Permeability; Protoplasts

0005-2736 / 02 / $ ^ see front matter ß 2002 Elsevier Science B.V. All rights reservedPII: S 0 0 0 5 - 2 7 3 6 ( 0 1 ) 0 0 4 2 9 - 1

Abbreviations: BHI, brain heart infusion; CD, circular dichroism; CFU, colony forming units; EDTA, ethylenediaminetetraaceticacid; LB, Luria broth; LUVs, large unilamellar vesicles ; MIC, minimal inhibitory concentration; OD, optical density; POPC, 1-palmi-toyl-2-oleoyl-sn-glycero-3-phosphatidylcholine; POPG, 1-palmitoyl-2-oleoylphosphatidyl-DL-glycerol ; RP-HPLC, reversed phase-highperformance liquid chromatography; SMM, sucrose, maleic acid, MgCl solution; SMMPA, sucrose, maleic acid, MgCl, protein andantibiotica medium; SUVs, small unilamellar vesicles; TFE, 2,2,2-tri£uoroethanol ; Tris, tris(hydroxymethyl)aminomethane

* Corresponding author. Fax: +49-30-94-793-159. E-mail address: dathe@fmp-berlin.de (M. Dathe).

BBAMEM 78190 14-12-01

Biochimica et Biophysica Acta 1558 (2002) 171^186www.bba-direct.com

1. Introduction

The widespread increase of bacterial resistance to-wards many conventional antibiotics has resulted inan intensive search for alternative antimicrobialagents and new target sites. Initially discovered as adefence system in invertebrates and vertebrates, anti-microbial peptides are attracting increasing interestas potential therapeutics [1,2]. Unlike classical anti-biotics which must penetrate the target cell, the prin-cipal mode of action of peptides involves perturba-tion and permeabilization of the cell membrane. Thismechanism confers activity towards a broad spec-trum of microbial cells, but is also responsible forundesired lytic activity against mammalian cellssuch as erythrocytes (for reviews, see [3,4]). Muchevidence implies that peptide-induced membrane per-meabilization is the result of interaction of the pep-tides with the lipid matrix of the cell envelope. Theinduction of a speci¢c amphipathic, often helicalstructure upon interaction with the bilayer surfacehas been established as a requirement for lytic activ-ity [5]. The lipid matrix of the membrane provides aunique environment for binding of such peptides.Additionally, most of the known antimicrobial pep-tides bear cationic amino acid residues and interactpreferentially with negatively charged membranes.The high amount of anionic lipids in prokaryoticmembranes and their absence in the neutral lipidmatrix of erythrocytes may account for the antimi-crobial activity and selectivity of many cationic pep-tides. But, unlike magainins [6], dermaseptins [7]from frog skin and the insect cecropins [8] whichare selective for bacteria, the bee venom melittin [9]and the neurotoxin pardaxin [10] are lytic to bothbacterial and mammalian cells. The speci¢city isalso mimicked in model liposome studies. Magainins[11] and insect cecropins [12] induce leakage prefer-entially from acidic lipid vesicles. Melittin [13] andpardaxin [14] permeabilize neutral and negativelycharged vesicles. Among the helical antimicrobialpeptides, magainin analogs [15] and dermaseptin se-quences [16] have been suggested as promising can-didates for the development of potential antimicro-bial therapeutic agents. Since a better understandingof the interplay between membrane properties andthe peptides' physico-chemical properties may pro-vide the basis for the design of compounds for di-

rected interaction, many studies have been devotedto the interactions of peptides with model mem-branes and their biological relevance (for review see[17,18]). The activity modulating role of individualstructural parameters, namely peptide helicity (K),hydrophobicity (H), hydrophobic moment (W), theangle subtended by charged residues (x) as well asthe total peptide charge (Q) was demonstrated [18^22]. We showed that the strengthening of hydropho-bic peptide properties favored the lytic e¡ect on theneutral lipid bilayer of red blood cells and reducedantimicrobial selectivity, since the interaction withnegatively charged lipids, characteristic for prokary-otic cell membranes, was only modestly a¡ected. Re-duction of the parameters K, H, W and x actuallyenhanced antimicrobial selectivity but was accompa-nied by a pronounced decrease of activity.

Besides the characterization of new antibacterialpeptides and the design of analogs of natural pep-tides, model peptides composed of a limited numberof di¡erent amino acid residues have been studiedextensively in order to understand the general aspectsof peptide^lipid interaction, [19,23^28]. Many of theearlier studies simultaneously modi¢ed several pa-rameters, making it di¤cult to distinguish the con-tribution of each individual motif to the overall ef-fect.

We investigated sets of model peptides with indi-vidually modi¢ed H, W and x, while the other pa-rameters were conserved. The basic KLA compoundhad pronounced antimicrobial as well as high hemo-lytic activity. The study was made to determinewhether the characteristics described for the activityand selectivity of the antimicrobial magainin 2 amideare generally valid. We examined whether modula-tion of H, W and the size of the polar (x) and hydro-phobic (8) helix surfaces would be su¤cient to con-fer membrane selectivity on the non-selective parentpeptide. Structural studies of bilayer-associatedpeptides and the determination of peptide a¤nityand permeabilizing e¤ciency on highly negativelycharged and electrically neutral lipid bilayers wereused to elucidate the driving forces in the bilayerpermeabilization process. Pure, rather than mixedlipid systems were used because in the latter case itis di¤cult to distinguish whether the peptide interactswith the initially provided mixture or with domainslocally enriched in one of the components. Addition-

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186172

ally, although the outer membrane of Gram-negativebacteria is characterized by an even higher negativecharge density the pure 1-palmitoyl-2-oleoylphospha-tidyl-DL-glycerol (POPG) bilayer was used to mimicits charge properties [29,30].

The results demonstrate that modulation of theparameters was a suitable method for inducing spe-ci¢c interaction with electrically neutral or negativelycharged lipid bilayers. The magnitude of a parameterwas found to be related to the modifying potential ofthe motifs. Thus, the high hydrophobicity of the pep-tide dominated the e¡ect on neutral vesicles and re-stricted the activity modulating potential of W and x.The relation of peptide bilayer interactions to thebiological e¡ects emphasized the activity modifyingrole of the parameters and showed that the activitiesobserved on neutral and charged bilayers superim-pose on the complex membranes of biological cells.As a consequence, the activity was modi¢ed by var-iation of the parameters but their selectivity in£uenc-ing potential was limited.

2. Materials and methods

2.1. Materials

The lipids 1-palmitoyl-2-oleoyl-sn-glycero-3-phos-phatidylcholine (POPC) and POPG were purchasedfrom Avanti Polar Lipids (Alabaster, AL, USA).Calcein was obtained from Fluka Chemie (Neu-Ulm, Germany), and 2,2,2-tri£uoroethanol (TFE)from Aldrich-Chemie (Steinheim, Germany). Tris-(hydroxymethyl)aminomethane (Tris), maleic acid,ethylenediaminetetraacetic acid (EDTA), MgCl2and other chemicals were from Merck (Germany).Luria broth (LB) was from Gibco BRL (UK), su-crose and bovine serum albumin from Sigma(USA), antibiotica medium 3 from Difco (USA), ly-sozyme from Roanal (Hungary), and brain heart in-fusion (BHI) from Becton (USA). Sucrose was fromRoth (Germany), penicillin G and yeast extract fromServa (Germany).

2.2. Peptide synthesis and characterization

Model peptides were synthesized automatically bythe solid phase method using standard Fmoc chem-

istry in the continuous £ow mode on a MilliGen9050 (Millipore, USA) peptide synthesizer [31]. Thepeptides were puri¢ed by preparative reversed phasehigh performance liquid chromatography (RP-HPLC) to give ¢nal products more than 95% pureby HPLC analysis. The peptides were further char-acterized by matrix assisted laser desorption/ioniza-tion mass spectrometry (MALDI-II, Kratos, Man-chester, UK) and quantitative amino acid analysis(LC 3000, Biotronik-Eppendorf, Germany). Chro-matographic characterization was performed on aShimadzu LC-10A gradient HPLC system. Runswere carried out on a PolyEncap-7 A 300 (250U4.0mm i.d.) column (Bischo¡ Analysentechnik, Ger-many) using a Shimadzu LC-M10A gradient HPLCsystem with a diode array detector operating at 220nm. The sample concentration was 1 mg/ml peptidein eluent A. Mobile phase A was 0.1% tri£uoroaceticacid in water and B was 0.1% tri£uoroacetic acid in80% acetonitrile/20% water (v/v). The retention times(tR) of the peptides were determined using a lineargradient of 5^95% B over 40 min at 22³C. The pre-cision of tR was þ 0.1 min.

Peptide hydrophobicity (H), hydrophobic moment(W) and hydrophobicity of the non-polar helix surface(Hhd) were calculated using the Eisenberg consensusscale for hydrophobicity [32].

2.3. Preparation of small and large unilamellarvesicles

Small unilamellar vesicles (SUVs) were preparedby drying the lipid under high vacuum, suspendingthe ¢lm by vortex mixing in bu¡er (10 mM Tris, 154mM NaCl, 0.1 mM EDTA, pH 7.4) to a ¢nal lipidconcentration of about 30 mM and sonicating thesuspension (under nitrogen, in ice water) for 25min using a titanium tip ultra sonicator. Dynamiclight scattering measurements (N4 Plus, Coulter Cor-poration, USA) con¢rmed the existence of a mainpopulation of POPC and POPG vesicles (morethan 95% mass content) with a mean diameter of45 þ 3 nm (polydispersity index 0.3). Calcein contain-ing large unilamellar vesicles (LUVs) were preparedby vortexing the lipid in dye bu¡er solution (70 mMcalcein, 10 mM Tris, 0.1 mM EDTA, pH 7.4). Thesuspension was freeze-thawed in liquid nitrogen forsix cycles and extruded (Lipex Biomembranes, Can-

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 173

ada) through polycarbonate ¢lters [33] (six timesthrough two stacked 0.4 Wm pore size ¢lters andeight times through two stacked 0.1 Wm pore size¢lters). Untrapped calcein was removed using a mini-column centrifugation method [34]. A plastic syringe(1 ml volume, plugged with a ¢lter pad) mounted ina centrifugation tube was ¢lled with hydrated Sepha-dex G-50 gel. After spinning at 2000 rpm for 3 minthe gel column had dried and parted from the sidesof the syringe. 500 Wl of the vesicle suspension wasdropped onto the gel bed and the liposomes wereeluted by centrifugation at 2000 rpm for 3 min.The mean diameter of the vesicles was determinedto be 93 þ 1 (polydispersity index 0.08). Lipid con-centration was determined by phosphorus analysis[35].

2.4. Circular dichroism measurements

Circular dichroism (CD) measurements of 1035 Mpeptide solutions in TFE/bu¡er (10 mM Tris, 154mM NaF, pH 7.4) (1/1, v/v) and freshly preparedSUV suspensions were carried out on a J-720 spec-trometer (Jasco, Japan) at room temperature. Minorcontributions of CD and circular di¡erential scatter-ing of the SUVs were eliminated by subtracting thelipid spectra of the corresponding peptide-free sus-pensions. The amount of helix was calculated fromthe mean residue ellipticity at 222 nm [36]. The errorin helicity was 9 5%.

2.5. Dye e¥ux measurements

LUV suspension (10 Wl) was injected into cuvettescontaining 2.5 ml of stirred peptide solutions ofdi¡erent concentration. Calcein release from vesicleswas monitored £uorimetrically by measuring thedecrease in self-quenching (excitation at 490 nm,emission at 520 nm) after 1 min at room temper-ature on an LS 50B spectro£uorimeter (Perkin El-mer, Germany). The £uorescence intensity corre-sponding to 100% release was determined after theaddition of Triton X-100 (100 Wl, 10% v/v in water)[30]. The concentration of half maximal dye release(EC50) was determined from dose^response curves.Dye release, F (%), as a function of bound peptideper lipid, r, was determined as described [11]. Dose^response curves were determined at three lipid con-

centrations, usually 12, 36 and 120 WM. Plotting thetotal peptide concentration cp as a function of lipidconcentration cl for a given F (%) results in astraight line. According to the mass conservationlaw cp = rUcl+cf , the slope of the curves gives thedegree of binding r = cb/cl and the intercept with thecp axis describes the concentration of free peptidecf .

2.6. Peptide binding

Binding isotherms were determined from thechange of the CD of peptide solutions (three di¡erentconcentrations between 5U1035 and 2U1036 M)after adding di¡erent amounts of SUVs. For the de-termination of the binding isotherms, the relationsF =3222(p)33222 and F = Fr(cb/cp) = Fr(cl/cp)Urwith r = cb/cl were used. F is the relative CD signal,3222(p) the ellipticity at 222 nm in the absence oflipid, 3222 the measured ellipticity in the presenceof lipid, Fr is F of the completely lipid-bound pep-tide, cb is the concentration of lipid-bound peptide, cl

is the lipid concentration and cp is the total peptideconcentration. Binding isotherms are derived fromthese equations and the mass conservation law [37].Binding isotherms for peptide interaction with POPGvesicles were estimated from dye release experimentsusing the procedure described above [11]. With r andcf binding isotherms were constructed and the appar-ent binding constant, Kapp could be derived from theinitial slope of the curves.

2.7. Hemolytic assay

The hemolytic activity of the peptides was deter-mined using human red blood cells (BlutspendedienstDeutsches Rotes Kreuz, Berlin, Germany) as de-scribed previously [30]. In brief, the suspensions (10mM Tris, 150 mM NaCl, pH 7.4) containing thepeptide and 2.3U108 cells/ml were incubated for 30min at 37³C. After cooling in ice water and centrifu-gation, an aliquot of the supernatant was dilutedwith 0.5% NH4OH and the optical density was mea-sured at 540 nm (Lambda 9 spectrophotometer, Per-kin Elmer, Germany). Peptide concentrations caus-ing 50% hemolysis (EC50) were derived from thedose^response curves. Values determined in repeatexperiments di¡ered by less than 5%.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186174

2.8. Antibacterial activity

Gram-negative Escherichia coli (DH5K strain) andGram-positive Bacillus subtilis (PY 22 strain) wereused to test the antibacterial activity of the peptides.Bacteria were cultivated in LB at 37³C with shakingat 180 rpm. The inoculum was prepared from midlogarithmic phase cultures (OD600 = 0.5). Aliquots ofthe cell suspensions were added to the wells of amicrotiter plate containing 50 Wl peptide solutionsof di¡erent concentrations. The ¢nal concentrationof bacteria in the wells was 1.25U106 colony formingunits (CFU)/ml. The ¢nal concentrations of peptidesranged from 0.04 to 80 WM in two-fold dilutions.Peptides were tested in duplicate. The microtiterplates were incubated overnight at 37³C with gentleshaking. The absorbance was read at 600 nm (Auto-reader EL 311, Bio-Tek Instruments, USA). Theminimum inhibitory concentration (MIC) is de¢nedas the lowest concentration of peptide at which therewas no change in optical density.

Bacterial protoplasts were prepared as follows. Analiquot of an overnight culture of E. coli was furthercultivated in LB to an OD600 = 0.8 at 37³C with gen-tle shaking at 120 rpm. 50 Wl cell suspension wascentrifuged for 10 min at 3000Ug and 4³C. The pel-let was resuspended in 2.5 ml ice cold sucrose solu-tion (10% in 50 mM Tris, 10 mM MgCl2, pH 8.0)and mixed with 0.5 ml lysozyme (5 mg/ml in 50 mMTris, pH 8.0). After 5 min at 4³C 1 ml EDTA (25mM in 50 mM Tris, pH 8.0) was added. After gentleshaking for 5 min at 4³C 1 ml Tris bu¡er (50 mM,pH 8.0) was added. After shaking (80 rpm) for an-other 15 min at 37³C 2 ml 10% sucrose solution wasadded and the suspension was centrifuged for 20 minat 2000Ug. The pelleted E. coli protoplasts were re-suspended in 10% sucrose solution for further use.Protoplasts of B. subtilis were prepared by cultiva-tion of the cells to an OD600 = 0.8. After centrifuga-tion of 50 ml cell suspension for 10 min at 3000Ugand 4³C the pellet was resuspended in 5 ml SMMPAsolution and mixed with 5 ml lysozyme (4 mg/mlSMMPA). SMMPA solution was prepared from49.5 ml SMM solution (consisting of 2 ml 1 Mmaleic acid, pH 6.4; 2 ml 1 M MgCl2 ; 25 ml 2 Msucrose; 21 ml H2O), 36 ml antibiotica medium 3 and4.5 ml bovine serum albumin solution (50 mg/ml

H2O). After incubation for 120 min at 37³C andgentle shaking the suspension was centrifuged for15 min at 3000Ug and the pellet was resuspendedin 2.5 ml SMMPA. Incubation, centrifugation andresuspension were repeated once. The concentrationof protoplasts was estimated by cell counting underthe microscope using a Neubauer blood countingchamber. In order to determine peptide-induced lysisaliquots of the protoplast suspension were pipettedinto cuvettes containing di¡erent concentrations ofpeptide in the corresponding medium. The ¢nal cellcontent was 2U108 protoplasts/ml. The peptide con-centration varied between 0.25 and 64 WM. After10 min at room temperature the optical densityof the samples of B. subtilis protoplasts was mea-sured at 600 nm (Lambda 9, Perkin Elmer). Peptideslysed protoplasts in a concentration-dependent man-ner. The EC50 was derived from dose^responsecurves as the peptide concentration causing halfmaximal reduction of optical density. Peptide-in-duced lysis of E. coli protoplasts was followed bymeasuring the scattering intensity at 475 nm usinga £uorescence spectrophotometer (LS 50B, PerkinElmer, Germany). The EC50 of lysis was derivedfrom dose^response curves as the concentration caus-ing half maximal reduction of the scattering inten-sity.

E. coli W1655 F+ cells (LWF+) were cultivated inBHI (containing 100 U/l penicillin G). B. subtilis 170(L170) grew in BHI (containing 1% yeast extract, 3%sucrose and 100 U/l penicillin G). Bacteria suspendedin 5 ml cultivation medium were incubated overnightat 37³C and at 180 rpm. An aliquot of the culturewas diluted in 50 ml medium and further cultivatedto an OD550 of 0.8 corresponding to 7.8U108 LWF+cells/ml and 3.2U108 L170 cells/ml. After dilution,the inoculum was prepared from an OD550 = 0.4cell suspension. 180 Wl cell suspension was added towells of a microtiter plate containing 20 Wl peptidesolution. The ¢nal peptide concentration ranged be-tween 0.5 and 100 WM, the cell concentration in thewells was 3.5U108 LWF+ cells/ml and 1.4U108

L170 cells/ml. After incubation overnight at 37³Cand shaking at 180 rpm the absorbance was readat 600 nm (Microplate Autoreader EL 311, Bio-Tek Instruments, USA) and the MIC was deter-mined.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 175

3. Results

3.1. Peptide design and structural properties

The model peptides were designed on the basis ofcriteria derived for natural antimicrobial peptides:cationic charge and the ability to from an amphi-pathic helix. The basic sequence, KLA1, consists of18 amino acid residues including ¢ve lysine residuesand one tryptophan in addition to leucine and ala-nine (Fig. 1). Sets of peptides of identical charge andsize but of modi¢ed hydrophobicity (H-set), changedhydrophobic moment (W-set) and varying angle sub-tended by charged (x)/hydrophobic (8) residues (x-set) (Table 1) were designed on the basis of twoprinciples: changes in the position of individual ami-no acid residues and minor residue exchange. TheEisenberg consensus scale of hydrophobicity wasused for calculating the structural parameters [32].H is the mean residue hydrophobicity calculated asthe sum of the hydrophobicities of the individualresidues and W is their vector sum. The peptides of

the H-set were developed by reducing the number ofleucine and increasing the number of alanine residues(compare KLA1 vs. KLA 2 vs. KLA3). As the resultof amino acid exchange the total peptide hydropho-bicity varied between 30.025 and 30.087 and thehydrophobicity of the non-polar helix surface (Hhd)decreased from 0.389 to 0.302. Hhd was calculated asmean residue hydrophobicity of the amino acid res-idues on the helix surface described by the angle 8.The hydrophobic moment of KLA2 (0.33) was en-hanced by changing the position of selected alanineand leucine residues (compare KLA2 vs. KLA12).Starting from the most hydrophobic KLA1 addition-al amino acid substitutions were necessary to reducethe hydrophobic moment from 0.33 to 0.28 (compareKLA1 vs. KLA11). H, Hhd and 8 were conserved inthe two W-sets. Peptides of the x-set were designedexclusively by changing the position of the ¢ve cat-ionic, six alanine and six leucine residues. Althoughthe total hydrophobicity was unchanged these mod-i¢cation resulted in di¡erences in Hhd, which thus laybetween 0.363 and 0.389 for the peptides with a large

Fig. 1. Helical wheel projection of KLA model peptides of modi¢ed hydrophobicity (H), hydrophobic moment (W) and angle sub-tended by polar (x)/hydrophobic (8) amino acid residues.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186176

hydrophobic domain (280³s8s 240³) and between0.43 and 0.47 for peptides with 220³s8s 180³(Table 1).

The linear peptides exhibited high conformational£exibility in aqueous solution, but assumed a helicalconformation in the presence of structure inducingTFE as con¢rmed by CD spectroscopic studies (Ta-ble 1). Following Lehrman et al. [38], we take thehelicity in the TFE environment as a measure ofthe ability of peptides to form a helix. Comparablyhigh helicities were determined for vesicle-boundpeptides. Conditions of binding were derived fromthe CD spectroscopic titration experiments presentedin Section 3.3.

For potentially amphipathic peptides, it has beenshown that the retention time (tR) in RP-HPLC isrelated to the helicity of peptides bound to the hy-drophobic stationary phase [39,40]. The tR of pep-tides of the H-set correlated well with the helix prob-ability, but considering all peptides, Fig. 2A revealsno correlation between tR and K in TFE (correlationcoe¤cient 30.008). An improved relationship be-tween tR and the structural properties of the peptideswas reached when taking into consideration Hhd. Thecorrelation coe¤cient is 0.64 for tR = f(Hhd) and 0.60for the regression presented in Fig. 2B. Propertiesrelated to helicity such as the size of the hydrophobic

surface and Hhd determine the interaction of the pep-tides with the RP-HPLC stationary phase. A pro-nounced H which is independent of the peptide con-formation, however, may restrict their bindingmodulating potential.

3.2. Membrane permeabilizing activity

Peptides of modi¢ed H and W displayed minor dif-ferences in their permeabilizing activity on negativelycharged POPG vesicles (Fig. 3). The peptide concen-tration required to induce half maximal £uorescencedequenching (EC50) ranged between about 0.1 and0.5 WM. In contrast, the activity modifying e¡ect ofthe angle subtended by charged residues on POPGbilayers was much more pronounced. Increase of thesize of the polar helix surface connected with a de-crease of the hydrophobic surface area distinctly im-paired the permeabilizing e¡ect. On neutral POPCvesicles, H was the most e¡ective activity modifyingfactor, W was less e¡ective and the in£uence of theangle x was negligible (Fig. 3).

Changes in the activity pro¢le in£uenced the bi-layer selectivity. KLA1 was slightly selective to neu-tral lipid membranes (Fig. 3). With decreasing H theselectivity shifted towards negatively charged bilayersas result of conserved activity on POPG but drasti-

Table 1Properties of KLA peptides

Code H W x/8 (³) Hhd tR (min) K (%)

TFE POPG cl/cp = 240 POPC cl/cp = 500

KLA1 30.025 0.329 80/280 0.389 20.4 73 54 62KLA2 30.0561 0.329 0.345 17.8 68 46 25b

KLA3 30.0872 0.329 0.302 13.1 59 56 ^c

KLA11 30.0267 0.284 0.386 19.9 69 ^a 53KLA12 30.0561 0.391 0.345 20.4 67 54 66KLA80 30.025 0.32 80/280 0.389 19.7 61 68 68KLA100 30.025 0.300 100/260 0.377 19.4 60 68 70KLA120 30.025 0.295 120/240 0.363 21.1 59 58 69KLA140 30.025 0.299 140/220 0.43 22.6 62 67 78KLA160 30.025 0.297 160/200 0.45 21.6 60 59 63KLA180 30.025 0.291 180/180 0.475 22.3 62 54 57

Characteristics of KLA model peptides: hydrophobicity (H), hydrophobic moment (W) and angle subtended by cationic (x)/hydropho-bic (8) residues, hydrophobicity (per residue) of the hydrophobic surface (Hhd), retention time (tR) in RP-HPLC and amount of helix(K) in TFE/bu¡er mixture (1/1 v/v), bu¡ered POPG, POPC SUV suspension at a lipid to peptide molar ratio cl/cp. The peptide con-centration, cp was 1035 M.aThe sample was turbid at all cl/cp.bAll values except this represent the conformation of vesicle-bound peptides.cThe peptide did not bind to POPC vesicles.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 177

cally reduced activity on POPC vesicles. The selectiv-ity modifying e¡ect of W was the higher the lower thepeptide hydrophobicity (compare KLA1, KAL11with KLA2, KLA12). On analyzing the peptides ofthe x-set it became obvious that the selectivity forPOPC bilayers is distinctly enhanced for analogswith a large x and small hydrophobic domain(KLA140, KLA160, KLA180). Here, the increasedPOPC speci¢city is mainly caused by a pronouncedreduction of peptide activity against the negativelycharged POPG vesicles.

3.3. Binding and permeabilizing e¤ciency

Modi¢cation of membrane activity and selectivityof the peptides can be due to variations in membranea¤nity as well as changes in the ability of the boundpeptide fraction to disturb the bilayer structure. CDspectroscopic studies of the peptides with POPGSUVs revealed the high a¤nity of all peptides tonegatively charged membranes. The spectra of a giv-en analog (peptide concentration 1035 M) were al-most identical at POPG concentrations between

Fig. 2. Relationship between the RP-HPLC retention times (tR) of KLA peptides and helical content (K) (A) and the product of Kand the hydrophobicity of the non-polar helix surface (Hhd) (B). K was determined in a TFE/bu¡er (1/1 v/v) mixture. The bu¡er was10 mM Tris, 154 mM NaCl, 0,1 mM EDTA, pH 7.4. Hhd was calculated as the mean residue hydrophobicity of the amino acid resi-dues covering the hydrophobic helix surface de¢ned by 8.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186178

2.5U1034 and 5U1033 M and re£ect maximal hel-icity as result of binding at a low lipid/peptide ratioof 25 (data not shown). Binding studies of peptidesof the x-set using a dye release assay con¢rmed thehigh a¤nity. This method is suitable if membranebinding correlates directly with membrane leakage[41]. While the apparent binding constants (Kapp) ofpeptides with a small polar domain were greater than200 000 M31, the Kapp of peptides with 80³9x9 120³ranged between 130 000 and 180 000 M31 (Table 2).

Peptide a¤nity to the neutral POPC bilayer wasmuch lower and strongly in£uenced by H and W (Fig.4). The Kapp values of KLA1 and KLA2 were deter-mined to be 5000 and 500 M31, respectively, whilethe most hydrophilic KLA3 did not bind at all. Sim-ilarly, binding to POPC vesicles was reduced withreduction of W in the order KLA1EKLA11 andKLA12sKLA2. The binding constants of analogswith 806x6 120 were found to be about 5000 M31

while peptides with 1406x6 180 bound to POPC

Fig. 3. Concentration of half maximal dequenching of calcein £uorescence (EC50) induced by KLA model peptides for negativelycharged POPG (black bars) and neutral POPC (white bars) LUVs. The lipid concentration, cl, was 12 WM in bu¡er (10 mM Tris, 154mM NaCl, 0.1 mM EDTA, pH 7.4).

Table 2Activity, a¤nity and e¤ciency of peptides of modi¢ed x on lipid bilayers

Peptide EC50 (WM) Kapp (1/M) r�F�50%�

POPC POPG POPC POPG POPC POPG

KLA80 0.03 0.09 5100 250 000 0.0010 0.007KLA100 0.051 0.10 n.d. 200 000 0.0007 0.008KLA120 0.023 0.26 5500 220 000 n.d. 0.016KLA140 0.044 1.0 7500 130 000 n.d. 0.044KLA160 0.070 0.60 7500 180 000 0.0023 0.034KLA180 0.042 0.71 n.d. 130 000 0.0011 0.036

Activity of KLA peptides of modi¢ed angle subtended by charged residues (x) on neutral POPC and negatively charged POPGvesicles and their a¤nities and permeabilizing e¤ciencies: half maximal concentration (EC50) of peptide induced dye release (see alsoFig. 3), CD spectroscopically derived apparent binding constants (Kapp) for the binding to POPC SUVs (see also Fig. 4) and toPOPG vesicles as determined by dye release assay. The ratio of bound peptide per lipid inducing half maximal dye from lipid vesicles(r�F�50%�) is a measure of the bilayer permeabilizing e¤ciency of the bound peptides. n.d., not determined.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 179

with a Kapp of about 7500 M31 (Table 2). The a¤n-ity seems to correlate with Hhd which is less than 0.39for the ¢rst peptide group and over 0.43 for thesecond (Tables 1 and 2).

Peptide a¤nity to POPC correlated qualitativelywith the bilayer permeabilizing activity. The di¡er-ences in the binding constants of the peptides ofdi¡erent H and W do not, however, explain themuch more pronounced changes in activity. Thus,while reduction of H resulted in a 10 times reduceda¤nity (compare KLA1 vs. KLA2) the activity de-creased by a factor of 92. Comparably, reduction ofthe hydrophobic moment (KLA11 vs. KLA1) re-duced Kapp by a factor of about 3 but the activitywas 8 times lower. Furthermore, although the bind-ing constant of KLA1 and peptides of the x-set wasmany times higher on POPG than on POPC bilayers,the peptides were much more active against POPCvesicles.

Fig. 5 and Table 2 illustrating the relationship be-tween dye release (F) and the amount of bound pep-

tide per lipid (r) demonstrate the parameter-depen-dent permeabilizing e¤ciency of the peptides. Theability of the peptides to permeabilize neutralPOPC bilayers is high and H, W and x exhibit onlya slight modulating e¡ect. Binding of ¢ve KLA1 and11 KLA11 molecules per 10 000 POPC molecules in-duced an initial dye release of 50%, demonstratingthe modest variability of the permeabilizing e¤ciencywith reduction of the hydrophobic moment. No cor-relation was found between x and r. Half maximaldequenching of dye £uorescence was induced by thebinding of 7^23 peptides per 10 000 lipid molecules(Fig. 5, Table 2).

In contrast, on highly negatively charged POPGbilayers the high a¤nity was o¡set by a low permea-bilizing e¤ciency. The ability of KLA1 to permeabi-lize the POPG membrane (rF�50% = 0.012) was morethan 20 times lower than their ability to disrupt neu-tral POPC bilayers. Peptides of modi¢ed x con¢rmthe distinct di¡erences (Table 2). Additionally, a dif-ferentiation in e¤ciency on POPG was found be-

Fig. 4. Binding isotherms for the interaction of KLA modelpeptides with POPC SUVs. The degree of binding, r (mol pep-tide bound per mol lipid) and the free peptide concentration, cf

were determined by CD spectroscopic titration experiments (seeSection 2).

Fig. 5. Relationship between peptide-induced dye release (de-quenching of calcein £uorescence, F) from POPC LUVs andthe molar ratio of bound peptide per lipid (r). The r value at agiven F is a measure for the permeabilizing e¤ciency of thepeptides.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186180

tween peptides with a small (809x9 120) and largepolar surface area (1409x9 180). Binding of 70^160 molecules of the ¢rst group but of 340^440 mol-ecules of the latter per 10 000 lipid molecules wasnecessary to induce half maximal £uorescence de-quenching of POPG vesicle entrapped dye.

3.4. Biological activity

The investigated parameters are modulators of theantimicrobial as well as hemolytic activity (Fig.6A,B). The peptide activity towards red blood cellsdecreased with reduction of H and W while there is

no correlation with x. The activity pro¢les werecomparable to those of dye release from POPCvesicles (compare Fig. 6A and 3). Decrease of Hled also to a pronounced reduction of the antimicro-bial activity against both E. coli and B. subtilis (Fig.6B). W slightly in£uenced the activity against E. coli.The e¡ect was more pronounced for peptides of re-duced H (KLA2, KLA12). The growth of E. coli wasalso sensitive to changes in the size of the polar helixsurface while the activity against B. subtilis appearedto be less x-dependent. To further elucidate the pep-tide e¡ect on bacterial cells the lytic activity of thestructurally modi¢ed peptides on protoplasts and cell

Fig. 6. Biological activity of KLA model peptides. (A) Concentration for half maximal lysis (EC50) of human red blood cells(2.3U108 cells/ml). Values obtained in repeat determinations di¡ered by less than 5%. (B) MIC for the growth of E. coli (black bars)and B. subtilis (gray bars) (1.25U106 CFU/ml).

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 181

wall-less L-forms of E. coli and B. subtilis was deter-mined. Removal of the cell wall of E. coli was ex-pected to eliminate the e¡ect caused by the highlynegatively charged cell envelope and thus to changethe activity pattern. The results are summarized inTable 3. Decrease of H distinctly reduced the activityon protoplasts, LWF+ and L170 cells. W became ane¡ective activity modulating parameter in the case ofless hydrophobic KLA peptides (KLA2, KLA12).No common activity pro¢le was found for peptidesof the x-set, but KLA120 seemed to be the mostactive analog on protoplasts and the L-form cells.

4. Discussion

This study demonstrates that modulation of thepeptide parameters H, W and x was su¤cient to con-fer speci¢c interaction with either electrically neutralor negatively charged lipid bilayers.

Increase of the parameters H, W and x enhancedthe speci¢city for neutral POPC bilayers, but thephysical basis was di¡erent. For peptides of the H-and W-sets, the selectivity increase resulted from anenhanced activity against the neutral bilayer. Hydro-phobic interactions between the non-polar surface of

the peptide helix and the lipid acyl chains of thebilayer have been suggested to be responsible forthe pronounced permeabilizing e¤ciency [30]. But,while the in£uence of H and W on the activity deter-mining permeabilizing e¤ciency was modest, the ac-tivity pro¢les of the peptide sets correlated well withthe a¤nity to POPC bilayers and the retention be-havior on the hydrophobic HPLC stationary phase.Thus, the pronounced changes in activity re£ect var-iations in the low binding a¤nity of the analogs tothe neutral bilayer.

In contrast, the modulated selectivity of peptidesof the x-set is based on activity changes againsthighly negatively charged POPG bilayers. Interest-ingly, two peptide classes could be distinguished:high a¤nity and moderate bilayer permeabilizingability characterize the peptides with a small x do-main while peptides with xv140³ are somewhat lessa¤ne and much less e¤cient. As shown for severalcationic peptides, the e¡ect on highly negativelycharged POPG bilayers is determined by high bind-ing via electrostatic interactions [30,42] o¡set by alow permeabilizing e¤ciency which is mediated byhydrophobic interactions. The study of the x-setshows that the properties of both the charged andthe hydrophobic helix domain are important for

Table 3Peptide activity against bacterial protoplasts and L-form cells

Peptide Modi¢ed parameter E. coli B. subtilis

Protoplasts EC50 (WM) LWF+ MIC (WM) Protoplasts EC50 (WM) L170 MIC (WM)

HKLA3 30.087 ^ s 80 ^ 13.5KLA2 30.056 67.3 80 7.2 13.1KLA1 30.025 1.8 10 0.4 1.6

WKLA2 0.329 67.3 80 7.2 13.1KLA12 0.391 1.9 5.0 0.6 1.6KLA11 0.284 1.7 n.d. 0.4 1.8KLA1 0.329 1.8 10 0.4 1.6

xKLA80 80 1.7 10 0.3 2.3KLA100 100 1.4 10 0.4 3.6KLA120 120 1.1 2.5 0.4 1.0KLA140 140 1.8 n.d. 0.8 1.9KLA160 160 3.9 5.0 0.9 2.3KLA180 180 2.3 2.5 1.1 3.4Cells/ml 2U108 7.8U108 2U108 3.5U108

Antibacterial activity of KLA peptides: the MIC for bacterial growth against E. coli derived LWF+ cells, B. subtilis derived L170 cellsand the half maximal concentration for lysis of E. coli and B. subtilis protoplasts (EC50).

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186182

KLA peptide e¡ects on POPG bilayers. Since thetotal peptide charge was conserved, the reductionof electrostatic attraction must be associated withreduction of the cationic charge density of the polarhelix surface. A¤nity di¡erences of other peptides tonegatively charged lipid bilayers might also be re-lated to variations in their charge topology [21].The distinctly reduced permeabilizing e¤ciency ap-pears to be related to the decrease in the size ofthe hydrophobic helix surface. Even a pronouncedincrease of Hhd could not compensate for the re-duced hydrophobic peptide^lipid interactions causedby a smaller 8 and reduced insertion. The resultsrevealed a general principle of activity modi¢cation:depending on the lipid system, one of the two deter-minants of the bilayer permeabilization step, eitherbinding or e¤ciency, dominates the activity, but itsmodulation is caused by the in£uence of the struc-tural parameters on the second, less pronounced de-terminant.

The second important outcome of the studies isthe observation that the activity modifying potentialof the structural motifs was limited and connectedwith the magnitude of the parameters. The high val-ue of hydrophobicity of the model peptides restrictedthe activity modulating potential of W and x. Thresh-old values of hydrophobicity dictating peptide prop-erties and behavior in lipid bilayers have also beendescribed for magainin peptides of modi¢ed totalcharge [43] and are known for transmembrane pep-tide sequences [44].

Our studies result in a model of peptide^lipid in-teraction which explains activity as function of thenumber of surface accumulated peptide moleculesand their depth of insertion (Fig. 7). The high activ-ity of the peptides against the POPG bilayer (Fig.7A) derives from the concentrated surface accumula-tion. Insertion into the acyl chain region is inhibitedbecause of electrostatic binding in the lipid headgroup region. Resultant expansion of the outer lipidlayer induces tension between the outer and innerlea£ets which may be released by rupture of the bi-layer [4]. A lipid exchange between the layers viatoroidal pores [45] in highly negatively charged bi-layers has not yet been described. Decrease of thecharge density of the polar helix surface reducesbinding and the size-reduced hydrophobic helix sur-face inhibits bilayer insertion, thus reducing activity.

On neutral POPC bilayers (Fig. 7B) peptides, evenwhen weakly bound, may penetrate deeply into thehydrophobic membrane region. The high permeabi-lizing e¤ciency of the hydrophobic peptides points todrastic disturbance of the lipid bilayer arrangement,possibly by the formation of large holes as result ofthe release of peptide^lipid complexes. KLA peptide-induced enhancement of the phase transition temper-ature of bilayers composed of lipids with an intrinsicnegative curvature strain (phosphatidylethanol-amine; data not shown), as also described for mag-ainin sequences [21], point to the induction of pos-itive curvature strain in £at POPC bilayers. Positivecurvature strain may promote bilayer disruption byformation of micelles. Thus, melittin induces a con-vex structure in phosphatidylcholine bilayers [46,47]and stimulates the release of lipids from erythrocytes[48].

It has been proposed that the main target for thekilling of bacteria by antimicrobial peptides is thecytoplasmic membrane [49]. On disturbance of thelipid matrix the transmembrane potential and thepH gradient are destroyed, the osmotic regulation

Fig. 7. Cartoon illustrating the di¡erent interactions of KLAmodel peptides with highly negatively charged POPG (A) andneutral POPC bilayers (B). (A) Electrostatic interactions be-tween cationic peptide residues and the anionic lipid headgroups anchor the helix in the bilayer surface. The sizes of thepolar/hydrophobic domains modify the depth of insertion. (B)The highly hydrophobic peptide inserts deeply into the lipidacyl chain region thus disturbing the arrangement of a largenumber of molecules and inducing the release of peptide^lipidmicelles.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 183

is a¡ected and respiration is uncoupled [50^52].Comparable lytic e¡ects are observed with erythro-cytes [48].

The membrane of red blood cells is composed ofelectrically neutral zwitterionic lipids. Actually, theactivity pro¢les of the investigated peptides corre-lated well with the e¡ects on POPC bilayers. Thehighest variability in the hemolytic e¡ect was inducedby modi¢cation of H, but a high H restricted theactivity modulating potential of W and x.

The envelope of Gram-negative bacteria has acomplex structure. To reach the inner target mem-brane the cationic peptides must cross the outer wallwhich they apparently do by utilizing the `self pro-moted uptake' pathway [53]. The increase in activitytowards Gram-negative bacteria is largely consistentwith the demonstrated increase in binding to thehighly negatively charged lipopolysaccharide andsubsequent outer membrane permeabilization [54].However, the lipid composition of the inner mem-brane of E. coli is dominated by neutral phosphati-dylethanolamine. On this level, membrane damage isdetermined by hydrophobic interactions. This isdocumented by the corresponding activity pro¢leson neutral bilayers and on protoplasts and LWF+cells. The di¡erences in the susceptibility of E. coliprotoplasts and cultured LWF+ cells might bebased on slight di¡erences in the respective mem-brane properties. LWF+ cells were derived fromthe E. coli K12 strain [55], protoplasts were preparedfrom the DH5K strain. It is well known that themembrane lipid composition in bacteria is in£uencedby the physiological state of the cell and by variousgrowth factors [56]. Additionally, a comparison ofthe lipid composition of the LWF+ membrane andthe cytoplasmic membrane of the walled parentstrain showed that the contents of negatively chargedphosphatidylglycerols were 8% and 17%, respectively[57]. Electrostatic interaction has also been suggestedto contribute to the high activity of the peptidesagainst the Gram-positive cell strains [19]. Compara-ble activities towards B. subtilis and the protoplasts,presented in this study, imply that the negativecharges in the murein envelope [58] are of reducedimportance. Thus, the low MIC values againstB. subtilis might be explained by the increased a¤n-ity to the negatively charged membrane lipids.Additionally, hydrophobic interactions are present,

consistent with the observation that the activitypro¢le corresponds to that of neutral lipid bilayers.

In summary, we have shown that the importanceof the peptide structural parameters derives fromtheir di¡erent roles in peptide interaction with neu-tral and negatively charged membranes. Coulombicforces are responsible for the pronounced a¤nitywhich dominates the activity on highly negativelycharged lipid bilayers. Hydrophobic interactionsdetermine the permeabilizing e¤ciency which is re-sponsible for the drastic e¡ect of KLA peptideson neutral lipid bilayers. However, the variabilityof peptide activity on the two lipid systems investi-gated seems to be based on the parameter's in£uenceon the second, less dominating step of the bilayerpermeabilization process. Furthermore, improvementof the peptide selectivity for neutral lipid bilayersresults from either increased activity against POPCor reduced activity against POPG bilayers. The ¢rstis dominated by the properties of the hydrophobichelix surface, the latter depends on the chargedensity of the polar and on the size of the hydro-phobic helix domains. A high hydrophobicity re-duces the activity modifying potential of the otherinvestigated parameters. On the complex membranesof biological cells, parameter-dependent peptide ef-fects on charged and neutral lipid bilayers becomesuperimposed. As a consequence, the activity ismodi¢ed by variation of the parameters but theirselectivity in£uencing potential on cells is substan-tially reduced.

Acknowledgements

The authors acknowledge the excellent technicalassistance of Heike Nikolenko and thank AnneKlose and Dagmar Smettan for peptide synthesisand characterization. John Dickson is thanked forcritically reading the manuscript.

References

[1] R.E. Hancock, R. Lehrer, Trends Biotechnol. 16 (1998) 82^88.

[2] R.E. Hancock, M.G. Scott, Proc. Natl. Acad. Sci. USA 97(2000) 8856^8862.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186184

[3] R.M. Epand, H.J. Vogel, Biochim. Biophys. Acta 1462(1999) 11^28.

[4] Y. Shai, Biochim. Biophys. Acta 1462 (1999) 55^70.[5] E.T. Kaiser, F.J. Kezdy, Annu. Rev. Biophys. Biophys.

Chem. 16 (1987) 561^581.[6] M. Zaslo¡, Proc. Natl. Acad. Sci. USA 84 (1987) 5449^5453.[7] A. Mor, V.H. Nguyen, A. Delfour, D. Migliore Samour, P.

Nicolas, Biochemistry 30 (1991) 8824^8830.[8] H. Steiner, D. Hultmark, A. Engstrom, H. Bennich, H.G.

Boman, Nature 292 (1981) 246^248.[9] E. Habermann, J. Jentsch, Hoppe Seylers Z. Physiol. Chem.

348 (1967) 37^50.[10] Y. Shai, J. Fox, C. Caratsch, Y.L. Shih, C. Edwards, P.

Lazarocici, FEBS Lett. 242 (1988) 161^166.[11] K. Matsuzaki, M. Harada, T. Handa, S. Funakoshi, N. Fu-

jii, H. Yajima, K. Miyajima, Biochim. Biophys. Acta 981(1989) 130^134.

[12] E. Gazit, A. Boman, H.G. Boman, Y. Shai, Biochemistry 34(1995) 11479^11488.

[13] T. Benachir, M. La£eur, Biochim. Biophys. Acta 1235 (1995)452^460.

[14] Y. Pouny, Y. Shai, Biochemistry 31 (1992) 9482^9490.[15] W.L. Maloy, U.P. Kari, Biopolymers 37 (1995) 105^122.[16] A. Mor, Drug Dev. Res. 50 (2000) 440^447.[17] N. Sitaram, R. Nagaraj, Biochim. Biophys. Acta 1462 (1999)

29^54.[18] S.E. Blondelle, K. Lohner, M. Aguilar, Biochim. Biophys.

Acta 1462 (1999) 89^108.[19] M. Dathe, T. Wieprecht, H. Nikolenko, L. Handel, W.L.

Maloy, D.L. MacDonald, M. Beyermann, M. Bienert,FEBS Lett. 403 (1997) 208^212.

[20] T. Wieprecht, M. Dathe, M. Beyermann, E. Krause, W.L.Maloy, D.L. MacDonald, M. Bienert, Biochemistry 36(1997) 6124^6132.

[21] T. Wieprecht, M. Dathe, R.M. Epand, M. Beyermann, E.Krause, W.L. Maloy, D.L. MacDonald, M. Bienert, Bio-chemistry 36 (1997) 12869^12880.

[22] T. Wieprecht, M. Dathe, E. Krause, M. Beyermann, W.L.Maloy, D.L. MacDonald, M. Bienert, FEBS Lett. 417 (1997)135^140.

[23] S.E. Blondelle, R.A. Houghten, Biochemistry 31 (1992)12688^12694.

[24] T. Kiyota, S. Lee, G. Sugihara, Biochemistry 35 (1996)13196^13204.

[25] M.M. Javadpour, M.M. Juban, W.C. Lo, S.M. Bishop, J.B.Alberty, S.M. Cowell, C.L. Becker, M.L. McLaughlin,J. Med. Chem. 39 (1996) 3107^3113.

[26] I. Cornut, K. Buttner, J.L. Dasseux, J. Dufourcq, FEBSLett. 349 (1994) 29^33.

[27] S. Castano, I. Cornut, K. Buttner, J.L. Dasseux, J. Du-fourcq, Biochim. Biophys. Acta 1416 (1999) 161^175.

[28] N. Uematsu, K. Matsuzaki, Biophys. J. 79 (2000) 2075^2083.

[29] F.R. Rana, E.A. Macias, C.M. Sultany, M.C. Modzrakow-ski, J. Blazyk, Biochemistry 30 (1991) 5858^5866.

[30] M. Dathe, M. Schumann, T. Wieprecht, A. Winkler, M.Beyermann, E. Krause, K. Matsuzaki, O. Murase, M. Bie-nert, Biochemistry 35 (1996) 12612^12622.

[31] M. Beyermann, M. Bienert, Tetrahedron Lett. 33 (1992)3745^3748.

[32] D. Eisenberg, Annu. Rev. Biochem. 53 (1984) 595^623.[33] M.J. Hope, M.B. Bally, G. Webb, P.R. Cullis, Biochim.

Biophys. Acta 812 (1985) 55^65.[34] R.B.C. New, Liposomes, A Practical Approach, Oxford

University Press, New York, 1990, pp. 103^161.[35] C.J.F. Bo«ttcher, C.M. Van Gent, C. Pries, Anal. Chim. Acta

24 (1961) 203^204.[36] Y.H. Chen, J.T. Yang, H.M. Martinez, Biochemisty 11

(1972) 4120^4131.[37] G. Schwarz, G. Beschiaschvili, Biochim. Biophys. Acta 979

(1989) 82^90.[38] S.R. Lehrman, J.L. Tuls, M. Lund, Biochemistry 29 (1990)

5590^5596.[39] R.A. Houghten, S.T. DeGraw, J. Chromatogr. 386 (1987)

223^228.[40] E. Krause, M. Beyermann, M. Dathe, S. Rothemund, M.

Bienert, Anal. Chem. 67 (1995) 252^258.[41] I. Ben-Efraim, Y. Kliger, C. Hermesh, Y. Shai, J. Mol. Biol.

285 (1999) 609^625.[42] Z. Oren, J. Hong, Y. Shai, J. Biol. Chem. 272 (1997) 14643^

14649.[43] M. Dathe, H. Nikolenko, J. Meyer, M. Beyermann, M. Bie-

nert, FEBS Lett. 501 (2001) 146^159.[44] L.P. Liu, S.C. Li, N.K. Goto, C.M. Deber, Biopolymers 39

(1996) 465^470.[45] K. Matsuzaki, O. Murase, N. Fujii, K. Miyajima, Biochem-

istry 35 (1996) 11361^11368.[46] A.M. Batenburg, B. De Kruij¡, Biosci. Rep. 8 (1988) 299^

307.[47] J. Dufourcq, J.F. Faucon, G. Fourche, J.L. Dasseux, M.

LeMaire, T. Gulik-Krzywicki, Biochim. Biophys. Acta 859(1986) 33^48.

[48] T. Katsu, M. Kuroko, T. Morikawa, K. Sanchika, Y. Fujita,H. Yamamura, M. Uda, Biochim. Biophys. Acta 983 (1989)135^141.

[49] R.E.W. Hancock, D.S. Chapple, Antimicrob. Agents Che-mother. 43 (1999) 1317^1323.

[50] H.V. Westerho¡, D. Juretic, R.W. Hendler, M. Zaslo¡,Proc. Natl. Acad. Sci. USA 86 (1989) 6597^6601.

[51] D. Juretic, R.W. Hendler, F. Kamp, W.S. Caughey, M.Zaslo¡, H.V. Westerho¡, Biochemistry 33 (1994) 4562^4570.

[52] M. Hugosson, D. Andreu, H.G. Boman, E. Glaser, Eur. J.Biochem. 223 (1994) 1027^1033.

[53] R.E. Hancock, Lancet 349 (1997) 418^422.[54] K. Matsuzaki, K. Sugishita, M. Harada, N. Fujii, K. Miya-

jima, Biochim. Biophys. Acta 1327 (1997) 119^130.[55] E. Schuhmann, U. Taubeneck, Z. Allg. Mikrobiol. 9 (1969)

297^313.[56] G. Cevc, Phospholipid Handbock, Marcel Dekker, New

York, 1993.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186 185

[57] K. Gura, Charakterisierung der Membranlipide aus stabilenProtoplastentyp-L-Formen und N-Formen von Escherichiacoli, Proteus mirabilis und Streptomyces hygroscopicus, The-sis, Mathematisch-Naturwissenschaftlich-Technische Fakul-ta«t der Martin-Luther-Universita«t, Halle-Wittenberg, 1998.

[58] C.R. Harwood, R.D. Coxon, I.C. Hancock, in: C.R. Har-wood, S.M. Cutting (Eds.), Molecular Biological Methodsfor Bacillus, John Wiley and Sons, Chichester, 1990, pp.327^390.

BBAMEM 78190 14-12-01

M. Dathe et al. / Biochimica et Biophysica Acta 1558 (2002) 171^186186