Báo cáo khoa học: Contribution of a central proline in model amphipathic a-helical peptides to self-association, interaction with phospholipids, and antimicrobial mode of action ppt

To determine the role of Pro in structure, antibiotic activity, and interaction with phospholipids, we generated a ser-ies of model amphipathic a-helical peptides with different chain le

a-helical peptides to self-association, interaction with

phospholipids, and antimicrobial mode of action

Sung-Tae Yang1, Ju Yeon Lee1, Hyun-Jin Kim1, Young-Jae Eu1, Song Yub Shin2,

Kyung-Soo Hahm2and Jae Il Kim1

1 Department of Life Science, Gwangju Institute of Science and Technology, Korea

2 Department of Bio-Materials, Graduate School and Research Center for Proteineous Materials, Chosun University, Gwangju, Korea

Antimicrobial peptides are produced as components of

the innate immune system by a wide variety of insects,

amphibians, and mammals, including humans [1–4] In

recent decades, the structures and functions of many antimicrobial peptides have been extensively studied to elucidate their mode of action Typically, antimicrobial

Keywords

aggregation; amphipathic helix; antimicrobial

peptides; membrane depolarization; proline

Correspondence

J Kim, Department of Life Science,

Gwangju Institute of Science and

Technology, Gwangju 500-712, Korea

Fax: +82 62 970 2484

Tel: +82 62 970 2494

E-mail: jikim@gist.ac.kr

(Received 9 February 2006, revised 28 June

2006, accepted 5 July 2006)

doi:10.1111/j.1742-4658.2006.05407.x

Model amphipathic peptides have been widely used as a tool to determine the structural and biological properties that control the interaction of pep-tides with membranes Here, we have focused on the role of a central Pro

in membrane-active peptides To determine the role of Pro in structure, antibiotic activity, and interaction with phospholipids, we generated a ser-ies of model amphipathic a-helical peptides with different chain lengths and containing or lacking a single central Pro CD studies showed that Pro-free peptides (PFPs) formed stable a-helical structures even in aqueous buffer through self-association, whereas Pro-containing peptides (PCPs) had random coil structures In contrast, in trifluoroethanol or SDS mi-celles, both PFPs and PCPs adopted highly ordered a-helical structures, although relatively lower helical contents were observed for the PCPs than the PFPs This structural consequence indicates that a central Pro residue limits the formation of highly helical aggregates in aqueous buffer and cau-ses a partial distortion of the stable a-helix in membrane-mimetic environ-ments With regard to antibiotic activity, PCPs had a 2–8-fold higher antibacterial activity and significantly reduced hemolytic activity compared with PFPs In membrane depolarization assays, PCPs passed rapidly across the peptidoglycan layer and immediately dissipated the membrane potential

in Staphylococcus aureus, whereas PFPs had a greatly reduced ability Fluorescence studies indicated that, although PFPs had strong binding affinity for both zwitterionic and anionic liposomes, PCPs interacted weakly with zwitterionic liposomes and strongly with anionic liposomes The selective membrane interaction of PCPs with negatively charged phospholipids may explain their antibacterial selectivity The difference in mode of action between PCPs and PFPs was further supported by kinetic analysis of surface plasmon resonance data The possible role of the increased local backbone distortion or flexibility introduced by the proline residue in the antimicrobial mode of action is discussed

Abbreviations

DiSC 3 (5), 3,3¢-dipropylthiadicarbocyanine iodide; PamOlePtdCho, oleoylphosphatidylcholine; PamOlePtdGro, 1-palmitoyl-2-oleoylphosphatidylglycerol; PCPs, proline-containing peptides; PFPs, proline-free peptides; SPR, surface plasmon resonance.

peptides contain multiple basic amino acids and

am-phipathic structures with clusters of hydrophobic and

hydrophilic residues [5–9] Although their precise

mechanism of action is not yet fully understood, it is

widely accepted that cationic antimicrobial peptides

interact with negatively charged bacterial membranes

by electrostatic interactions and then cause cell death

by permeabilizing cell membranes by forming

barrel-stave or toroidal pores [10–14] or by disrupting the

membrane via a ‘carpet’ mechanism [15–17] It is also

known that in some cases peptides inhibit the

macro-molecular synthesis by penentrating into the bacterial

cytoplasm followed by DNA⁄ RNA binding without

causing membrane permeabilization [18–21] Some

antimicrobial peptides can lyse not only microbial but

also eukaryotic cells [22] This activity against

eukary-otic cells should be eliminated so that the antimicrobial

peptides can be used therapeutically Thus,

consider-able attention has been focused on the design of new

antimicrobial peptides with good selectivity for

bacter-ial cells

Structure–function studies of antimicrobial peptides

have shown that a number of variables modulate

antibiotic activity, including chain length, helical

pro-pensity, amphipathicity, net positive charge,

hydro-phobicity, and hydrophobic moment [23–29] A variety

of model amphipathic a-helical peptides and artificial

membranes have been used to analyze the molecular

structure–function relationships, understand the

gen-eral aspects of peptide–lipid interactions, and

deter-mine the variables that control cell selectivity For

example, to investigate the effect of hydrophobic–

hydrophilic balance on biological and membrane-lytic

activities, Kiyota et al [30] synthesized five 18-residue

model peptides composed of nonpolar (Leu) and basic

(Lys) residues of varying hydrophobic–hydrophilic

balance In addition, to determine the proper chain

length for potent antimicrobial peptides, Blondelle &

Houghten [31] prepared a series of 8–22-residue model

amphipathic peptides comprising Leu and Lys Also,

Papo et al [32] generated several short model peptides

and their diastereomeric analogs to study the structural

and functional effects of d-amino acids in amphipathic

a-helices These systematic analyses have helped to

clarify the characteristics needed for the design of

potent, selective antimicrobial peptides as antibiotics

The presence of Pro residues in a-helices generally

cre-ates a bend or kink in the peptide backbone because of

the lack of an amide proton, which normally provides a

hydrogen bond donor, and they are commonly found

within the amphipathic a-helices of antimicrobial

pep-tides Pro residues in amphipathic a-helical peptides

have been the focus of extensive research because they

are functionally important in peptide–lipid interactions For example, several recent studies have investigated the effect of Pro substitutions on the biological activity and structure of naturally occurring antimicrobial peptides such as PMAP-23 [33], melittin [34], gaegurin [35], tri-trpticin [36], and maculatin [37] These studies revealed that the replacement of a Pro with an Ala maintained or decreased the antimicrobial activity but significantly increased the hemolytic activity In addition, Oh et al [38] reported that a cecropin A–magainin II hybrid pep-tide and its analog P2, which have amphipathic a-helical structures with a central hinge region due to the presence of Gly or Pro, have potent and selective antimi-crobial activity We also reported that the replacement

of a Pro with Leu or Ala in the hybrid analog P18 decreases its antibacterial activity and increases its hemolytic activity [39]

In general, introduction of a Pro near the central region of a-helical antimicrobial peptides reduces the a-helical structure This partial disruption of the struc-ture appears to contribute to selective cytotoxicity On the other hand, Pro residues are also found in ion-channel-forming peptides Alamethicin, for example, has a Pro-kink helical structure which is important for its insertion into lipid bilayers Once in the lipid bilay-ers, they form transmembrane helices that contribute

to transmembrane pores or voltage-induced channels [40,41] In addition, statistical analysis of transmem-brane helices has established the significance of Pro-containing motifs in transmembrane a-helices [42], and several studies have investigated the structural and dynamic role of Pro residues in transmembrane helices [43] Although there is growing evidence that the Pro residues largely contribute to the ability of antimicro-bial peptides to kill various types of microantimicro-bial cells and to form transmembrane helices, the role of the internal kink induced by Pro in amphipathic a-helices has not been systematically studied, and the kinetic significance of this structure remains unknown

Here, we have systematically examined the role of a central Pro residue by using model 17–25-residue am-phipathic a-helical peptides that either contain or lack

a Pro residue We also applied biosensor technology

to distinguish the kinetics of membrane binding by Pro-containing peptides (PCPs) and Pro-free peptides (PFPs) We found that the synthetic PCPs have much more potent antibacterial activity and significantly reduced hemolytic activity than the PFPs In addition, the PCPs were able to selectively bind and strongly permeabilize negatively charged liposomes We further discuss the role of the helix–bend–helix structure induced by a central Pro residue in the mechanism of selective antimicrobial activity

Peptide design

To investigate the influence of a central Pro on the

biological activity, structure, membrane binding, and

membrane-disrupting activity of antimicrobial

pep-tides, we generated amphipathic a-helical peptides with

different chain lengths (17, 21, and 25 residues) and

containing or lacking a Pro residue The model

pep-tides are composed of repeats of hydrophobic (Leu)

and basic (Lys) residues to create perfect amphipathic

a-helices A single Trp residue was introduced in

posi-tion 2 of these peptides to allow fluorescent

deter-mination of their concentration and peptide–lipid

interactions The Pro-free peptides (PFPs) included

M17, M21, and M25, and their counterpart central

Pro-containing peptides (PCPs) were M17P, M21P,

and M25P, respectively (Table 1)

Comparison of antimicrobial and hemolytic

activities of the peptides

The model amphipathic a-helical peptides were studied

for their ability to inhibit the growth of Gram-negative

and Gram-positive bacteria as well as for their

cyto-toxicity against human erythrocytes The minimal

inhibitory concentrations for the peptides against

bacteria are summarized in Table 2, and the dose–

response relationship of the hemolytic activity is

depic-ted in Fig 1 As shown in Table 2, the different chain

length of the peptides did not significantly affect their

activity toward both Gram-negative and Gram-positive

bacteria These results suggest that a long chain length

is not required for improved antibacterial activity

Interestingly, compared with PFPs, PCPs had 

2–8-fold greater antibacterial activities As shown in Fig 1,

however, PFPs (M17, M21, and M25) were relatively

strongly hemolytic (63%, 65%, and 75% at 50 lm,

respectively), whereas PCPs (M17P, M21P, and M25P)

displayed significantly reduced hemolytic activity (4%, 21%, and 11% at 50 lm, respectively) These data suggest that introduction of Pro residues at a central position improved the peptide selectivity for bacterial versus mammalian cells

Structural analysis of the peptides

CD spectroscopy was used to monitor the secondary structure of the peptides The CD spectra of peptides were collected in 50 mm sodium phosphate buffer⁄ 50% trifluoroethanol⁄ 30 mm SDS micelles, PamOlePtd-Cho⁄ PamOlePtdGro (1 : 1) liposomes or PamOlePtd-Cho liposomes (Fig 2) The CD spectra of all of the synthetic peptides dissolved in water in the absence of salt showed that they were mainly random coils (data not shown) In buffer (50 mm sodium phosphate buffer, pH 7.2), however, PFPs (Fig 2A, filled

Table 1 Amino-acid sequences and molecular masses of the

model peptides Observed mass was from Kratos Kompact MALDI

TOF MS.

Peptide Sequence

Mass Calculated Observed M25 KWKKLLKKLLKLLKKLLKKLKKLLK-NH 2 3114.3 3115.2

M25P KWKKLLKKLLKLPKKLLKKLKKLLK-NH23098.2 3098.8

M21 KWKKLLKKLLKLLKKLLKKLK-NH2 2631.6 2631.9

M21P KWKKLLKKLLPLLKKLLKKLK-NH 2 2600.5 2601.4

Table 2 Minimal inhibitory concentration (l M ) for the peptides Results indicate the range of three independent experiments, each performed in triplicate.

Bacterial strain

Peptide

Fig 1 Dose–response curves of hemolytic activity of the peptides toward human erythrocytes Hemolysis assays were carried out for the following peptides: M25 (d), M25P (s), M21 (.), M21P (,), M17 (n), and M17P (h) Results represent the means of duplicate measurements from three independent assays.

symbols) exhibited typical a-helical CD spectra, with

minimal mean residue molar ellipticity values at 208

and 222 nm, whereas the CD spectra of PCPs

(Fig 2A, empty symbols) had a negative band below

200 nm, indicating a lack of ordered structure This

result supports the idea that proline is an effective

a-helix breaker, as previously reported for globular

proteins [44] As expected, the CD spectra of all of the

peptides indicated a-helix structures in the presence of

trifluoroethanol (Fig 2B) or SDS micelles (Fig 2C),

but there was little difference in the helical contents

between PCPs and PFPs These results suggest that

PCPs have a partially distorted helix structure with a

kink around the central Pro in membrane-mimetic

environments Interestingly, the shape of PFP spectra

in the presence of PamOlePtdCho⁄ PamOlePtdGro

(1 : 1) liposomes is apparently different from that in

the presence of SDS or trifluoroethanol (Fig 2D) This

may point to strong aggregation in this type of

mem-brane, which may correlate with the increased

cytotox-icity of these peptides In PamOlePtdCho liposomes

(Fig 2E), PCPs had no distinct secondary structure,

but there was a weak shoulder in their spectra,

com-pared with aqueous solution, suggesting that some

interaction does occur In contrast, PFPs adopted

a-helical structures, indicating that PFPs can strongly

interact with zwitterionic liposomes

Next, to determine in detail the effect of salt on

the conformational transition from a random coil to

an a-helix, the CD spectra were recorded as a

func-tion of the NaCl concentrafunc-tion from 0 to 100 mm at

a constant peptide concentration (Fig 3) The results

for peptide M21 are shown as an example in

Fig 3A The CD spectra of M21 in the presence of

various NaCl concentrations exhibited an isodichroic

point at 203 nm, indicating a two-state equilibrium between a random coil and an a-helix In pure water, M17 and M25 also became more a-helical as the NaCl concentration was increased (Fig 3B) Helix formation by the PFPs appears to be accom-panied by self-association In addition, the ratio of ellipticity values at 222⁄ 208 nm is close to 1 in buf-fer, which this is taken to indicate aggregation [45] The CD spectra of PCPs (M17P, M21P, and M25P), however, did not change as the NaCl concentration was increased These results suggest that the presence

of a kink induced by a Pro residue in amphipathic a-helices is essential for maintaining them as mono-mers in aqueous solution

Peptide-induced dye leakage from liposomes

We next measured the membrane-disrupting abilities

of the peptides by examining calcein leakage from neg-atively charged PamOlePtdCho⁄ PamOlePtdGro (1 : 1)

or zwitterionic PamOlePtdCho liposomes Upon addition of the peptides to the liposomes, the entrapped calcein (70 mm) was released into the buf-fer by lysis This relieves self-quenching of the dye within the liposomes, increasing the fluorescence intensity Relative lytic efficiencies were determined

by comparing the effects of the peptides with those

of Triton X-100, which corresponds to the total fluorescence Dose responses of peptide-induced calc-ein release from the PamOlePtdCho⁄ PamOlePtdGro (1 : 1) and PamOlePtdCho liposomes are shown in Fig 4 Compared with PCPs, PFPs released as much

or slightly more calcein from the PamOlePtdCho⁄ PamOlePtdGro (1 : 1) liposomes All of the amphi-pathic peptides caused an almost total disruption of

Fig 2 CD spectra of the model peptides under various conditions CD spectra were obtained at 25 C in (A) 50 m M sodium phosphate buf-fer (pH 7.2), (B) 50% trifluoroethanol, (C) 30 m M SDS micelles, (D) PamOlePtdGro ⁄ PamOlePtdCho (1 : 1) liposomes, or (E) PamOlePtdCho liposomes and in the presence of the following peptides at 25 l M concentration: M25 (d), M25P (s), M21 (.), M21P (,), M17 (n), and M17P (h).

the PamOlePtdCho⁄ PamOlePtdGro (1 : 1) liposomes

at 1 : 20 molar ratio of peptide to liposome In

con-trast, the PFPs (M17, M21, and M25) caused

relat-ively large calcein leakage (57%, 60%, and 66%,

respectively) from PamOlePtdCho liposomes at a

peptide to liposome molar ratio of 1 : 10, whereas

the PCPs showed a relatively reduced ability to reduce PamOlePtdCho membranes These results agree well with those from analysis of hemolysis, and they indicate that introduction of Pro into amphi-pathic a-helical peptides confers the ability to selec-tively disrupt anionic versus zwitterionic liposomes

Fig 3 CD spectra of M21 and [.]222for the model peptides at various NaCl concentrations (A) CD spectra were recorded as a function of the NaCl concentration (from 0 to 50 m M at increments of 5 m M ) for 25 l M peptide M21 at 25 C (B) Plot of [h] 222 versus NaCl concentra-tion (0–100 m M ) for the following peptides at 25 l M concentration: M25 (d), M25P (s), M21 (.), M21P (,), M17 (n), and M17P (h).

Fig 4 Calcein leakage as a function of molar ratio of peptide to lipid Calcein-containing (A) PamOlePtdCho ⁄ PamOlePtdGro (1 : 1) or (B) PamOlePtdCho liposomes at 25 C were mixed with the following peptides: M25 (d), M25P (s), M21 (.), M21P (,), M17 (n), or M17P (h) Results represent the means of three independent experiments.

Of the PCPs, M17P and M25P showed negligible

cyto-toxicity against human red blood cells and a relatively

weak ability to disrupt artificial neutral liposomes,

whereas M21P had moderate cytolytic activity In the

case of M17P and M25P, Pro replaced the central Leu

of the hydrophobic helix face (based on an

amphipath-ic helamphipath-ical wheel diagram), whereas in M21P, it replaced

the central Lys of the hydrophilic helix face The

mod-erate cytotoxicity of M21P suggests that placement of

Pro in the hydrophobic face of amphipathic a-helical

peptides is more effective than placement in the

hydro-philic region for generating peptides with selectivity for

bacterial versus red blood cells

Tryptophan fluorescence

To study the interaction of PCPs and PFPs with

membranes, we next examined changes in Trp

fluor-escence in pure water, aqueous buffer, or anionic

PamOlePtdCho⁄ PamOlePtdGro (1 : 1) or zwitterionic

PamOlePtdCho liposomes As the fluorescence

emis-sion characteristics of the Trp are sensitive to its

immediate environment, it can be used to monitor

the binding of peptides to membranes All the

pep-tides listed in Table 1 have a single Trp residue at

position 2 The corresponding maximum emission

wavelength (kmax) is plotted as a function of the

lipid⁄ peptide molar ratio in Fig 5, and the kmax

val-ues of the peptides at lipid⁄ peptide molar ratio of

50 : 1 are shown in Table 3 In Tris⁄ HCl buffer, the

kmax values for PCPs were  352 nm, indicating that

Trp residues are fully exposed to a hydrophilic

envi-ronment In contrast, the kmax value of the Trp

resi-due in PFPs was  343 nm, indicating that Trp was surrounded by a hydrophobic environment through self-association of the peptides in buffer Addition of PamOlePtdCho⁄ PamOlePtdGro (1 : 1) liposomes with both PFPs and PCPs results in a large blue shift (22–

24 nm) in the kmax and an increase in the fluores-cence quantum yield for all the peptides, indicating that the peptides strongly bind to negatively charged membranes With zwitterionic PamOlePtdCho lipo-somes, there was an almost constant kmax around

349 nm for M17P and M25P at all lipid⁄ peptide ratios and a small blue shift (9 nm) for M21P, indica-ting a lack of binding to PCPs In contrast, the three model PFPs displayed a large blue shift (18–19 nm) These results indicate that PCPs interact weakly with zwitterionic phospholipids but strongly with anionic phospholipids

Fig 5 k max of tryptophan fluorescence as a function of the lipid ⁄ peptide ratio Fluorescence spectra were recorded at increasing concentra-tions of (A) PamOlePtdCho ⁄ PamOlePtdGro (1 : 1) or (B) PamOlePtdCho liposomes in Tris ⁄ HCl buffer (pH 7.4) at 25 C and at 3 l M of the following peptides: M25 (d), M25P (s), M21 (.), M21P (,), M17 (n), or M17P (h) The excitation wavelength was 280 nm, the excitation band width was 5 nm, and the emission band width was 3 nm Results represent the means of three independent experiments.

Table 3 k max (nm) of tryptophan fluorescence for the peptides and

K SV in the presence of liposomes Assays were carried out in Tris ⁄ HCl buffer or in the presence of PamOlePtdCho ⁄ PamOlePtd-Gro (1 : 1) or PamOlePtdCho liposomes at a lipid ⁄ peptide molar ratio of 50 : 1.

Peptide Pure water

Tris ⁄ HCl buffer

PtdCho ⁄ PtdGro

PamOle-PtdCho

K SV ( M )1)

PtdCho ⁄ PtdGro

PamOle-PtdCho

Quenching of the intrinsic fluorescence

by acrylamide

To compare the membrane-integrated state of PCPs

and PFPs following their interaction with negatively

charged PamOlePtdCho⁄ PamOlePtdGro (1 : 1) or

neutral PamOlePtdCho liposomes, we next performed

a fluorescence quenching experiment using the neutral

fluorescence quencher acrylamide This quencher can

approach Trp more easily when the peptide is free in

solution than when it is bound to model membranes

Stern-Volmer plots for fluorescence quenching of Trp

by acrylamide in the presence of PamOlePtdCho⁄

PamOlePtdGro (1 : 1) or PamOlePtdCho liposomes

are depicted in Fig 6, and the apparent KSVvalues are

shown in Table 3 The Trp fluorescence intensity for

PFPs decreased in a similar concentration-dependent

manner for both types of liposome after the addition

of acrylamide, indicating that PFPs are buried in both anionic and neutral liposomes The tendency of PFPs

to self-associate appears to affect their nonselective interaction However, quenching of Trp fluorescence of PCPs is less efficient with PamOlePtdCho⁄ PamOlePtd-Gro (1 : 1) than with PamOlePtdCho vesicles, suggest-ing that the Trp residue of PCPs penetrates more efficiently into the hydrophobic core of negatively charged bilayers than zwitterionic bilayers

Membrane depolarization by PFPs and PCPs

It is widely believed that many membrane-active anti-microbial peptides pass through the peptidoglycan layer and then kill the target micro-organism by inter-acting with and permeabilizing the cytoplasmic mem-brane To further study this hypothesis, we examined the ability of PFPs and PCPs to depolarize the mem-brane using the memmem-brane-potential-sensitive dye 3,3¢-dipropylthiadicarbocyanine iodide [DiSC3(5)] (Fig 7) Upon addition to a suspension of S aureus, the fluor-escence of DiSC3(5) (first arrow) is strongly quenched and quickly stabilized Addition of peptides (second arrow) increased the fluorescence caused by membrane depolarization, and subsequent addition of gramicidin

D (third arrow) fully disrupted the membrane poten-tial Interestingly, PCPs almost completely dissipated the membrane potential at 0.3 lm, but self-associated PFPs showed a largely reduced ability to cause mem-brane depolarization In addition, all PCPs caused an immediate increase in fluorescence intensity, indicating rapid membrane depolarization, whereas the PFPs caused a gradual increase in the fluorescence These results suggest that the self-association of PFPs, which have less potent antimicrobial activity, interferes with their passage across the peptidoglycan layer

Analysis of binding using an surface plasmon resonance (SPR) biosensor

Finally, we used SPR to monitor the binding of PFPs and PCPs to PamOlePtdCho⁄ PamOlePtdGro (1 : 1) liposomes immobilized on an L1 sensor chip Figure 8 shows representative sensorgrams for the binding of M17 and M17P The sensorgrams for M25 and M21 were similar to that for M17, whereas the sensorgrams for M25P and M21P were similar to that for M17P (data not shown) Examination of the shape of the sensorgrams for M17 and M17P reveals significantly different binding kinetics In particular, the sensor-grams indicate that the initial association of M17P with the lipid surface starts as a very fast process

Fig 6 Stern-Volmer plots for the quenching of Trp fluorescence by

the peptides Quenching assays were carried out in the presence

of 150 l M of either (A) PamOlePtdGro ⁄ PamOlePtdGro or (B)

PamOlePtdCho liposomes and the following peptides at 3 l M

con-centration: M25 (d), M25P (s), M21 (.), M21P (,), M17 (n), or

M17P (h) Results represent the means of three independent

experiments.

compared with that of M17 In addition, whereas

M17P exhibited a distinct association and dissociation,

M17 had a very slow dissociation at a low peptide

concentration (less 20 lm) and failed to dissociate

from the liposomes at high concentrations (more

40 lm) When the sensorgrams are fitted using

differ-ent concdiffer-entrations of M17P, the two-state reaction

model fits better than the simple 1 : 1 Langmuir bind-ing model, suggestbind-ing that a two-step process mediates the interaction of the peptide with lipid bilayers How-ever, M17 had similarC2values in both fitting models Only peptide sensorgrams obtained at low peptide concentrations (2.5–20 lm) were used to calculate the association constants for M17 because the peptide was bound irreversibly to the lipid bilayers at high concen-trations The average values for the rate constants and affinity constants obtained from the two-state model analysis are listed in Table 4 There were striking dif-ferences between PFPs and PCPs in the association rate (ka1) for first step and the dissociation rate (kd2) for second step The observations with PFPs and PCPs seem to be in line with those of Zelezetsky et al [46], using different types of aggregating⁄ nonaggregating model peptides

Discussion

Membrane-active peptides mediate a wide range of biological events, including signal transduction, trans-port through the membrane, membrane fusion and lysis, ion channel formation, and antimicrobial def-ense These peptides exhibit a structural transition from an extended coil to a well-defined secondary structure upon binding to membrane surfaces Interac-tion of the peptides with membranes plays an import-ant role in many cellular processes In particular, Pro residues often appear in the central region of mem-brane-active peptides, and they may control the folding process and affect the membrane translocation or pen-etration [43,47–49]

Recently, model peptides have been intensively stud-ied as tools for determining the structural and biologi-cal properties of antimicrobial peptides In particular, model amphipathic a-helical peptides have been stud-ied extensively to identify general properties related to peptide–lipid interaction and their relationships with the biological activity of the peptides [28–32] In the present study, we carried out a systematic structure– activity study on a series of model peptides to deter-mine the role of a central Pro on the biological activ-ity, peptide structure, and interaction with membranes One interesting finding of this study was that intro-duction of a Pro in the middle position of the sequences of nonselective cytolytic peptides confers high selectivity for bacterial cells In particular, we found that the depolarization of bacterial membranes caused by PCPs is more potent and rapid than that caused by PFPs There was a direct correlation between the ability of the peptides to dissipate the membrane potential and their antimicrobial activity In

Fig 7 Kinetics of membrane depolarization of S aureus by PFPs

and PCPs DiSC3(5) was added to exponential-phase S aureus

cells once the fluorescence was stable, the peptides (0.3 l M ) were

added, and membrane depolarization was measured Gramicidin D

(0.22 n M ) was used to induce full collapse of the membrane

poten-tial The results are representative of two independent

experi-ments.

addition, Trp fluorescence measurements indicated that PFPs interacted nonselectively with negatively charged and zwitterionic liposomes, whereas PCPs bound strongly and selectively to anionic liposomes These results are consistent with the ability of the peptides to induce dye leakage preferentially from negatively charged lipid membranes The selective membrane interaction of Pro-containing peptides with negatively charged phospholipids may explain the selective anti-bacterial activity because zwitterionic phospholipids are the major constituent of the outer leaflet of red blood cells

Understanding the process of peptide folding in aqueous buffer or in membrane-mimetic environments

is critical for elucidating the mechanism of

antimicro-Fig 8 Sensorgrams for the binding of peptides to 1 : 1 PamOlePtdCho ⁄ PamOlePtdGro lipid bilayers Overlay of the experimental (solid line) and calculated (dotted line) sensorgrams using a two-state model (A and C) or a 1 : 1 Langmuir model (B and D) Lower plot, 5 l M ; upper plot, 20 l M Results are representative of two independent experiments.

Table 4 Kinetic interaction of the peptides with

PamOlePtd-Cho ⁄ PamOlePtdGro (1 : 1) lipid bilayers Association (k a1, ka2) and

dissociation (kd1,kd2) kinetic rate constants for the interaction of

PFPs and PCPs with PamOlePtdCho ⁄ PamOlePtdGro (1 : 1) were

determined by numerical integration using a two-state reaction

model The affinity constant (K) was determined as (ka1⁄ k d1 )

(k a2 ⁄ k d2 ).

Peptide

k a1

(1 ⁄ M s) kd1(1 ⁄ s) ka2(1 ⁄ s) kd2(1 ⁄ s) K (1 ⁄ M )

M25 601 1.41 · 10)2 3.49 · 10)2 1.64 · 10)6 9.07 · 10 8

M25P 4347 5.92 · 10)2 2.03 · 10)2 3.17 · 10)3 4.70 · 10 5

M21 598 1.33 · 10)2 2.23 · 10)2 2.55 · 10)6 3.93 · 10 8

M21P 4822 5.76 · 10)2 1.56 · 10)2 3.80 · 10)3 3.44 · 10 5

M17 634 1.02 · 10)2 2.95 · 10)2 1.96 · 10)6 9.35 · 10 8

M17P 4080 5.92 · 10)2 1.08 · 10)2 2.65 · 10)3 2.81 · 10 5

bial action CD spectra of the model amphipathic

pep-tides revealed that, in buffer, the central Pro residue

effectively disrupts the a-helical structure, but, in

mem-brane-mimetic environments, the Pro kept a-helical

structures, which means that Pro does not always

behave as a strong helix breaker in certain

surround-ings including membrane-mimetic environments These

findings with amphipathic a-helical peptides agree with

those reported by Li et al [50] for model

transmem-brane helical peptides

The CD and Trp fluorescence spectra of PFPs were

very sensitive to the salt concentration In buffer, PFPs

are thought to take on an a-helical structure because

of self-association An increase in ionic strength seems

to lead to a decrease in the electrostatic repulsive

for-ces between the positively charged residues because of

the presence of counterions In contrast, despite the

reduced electrostatic repulsion in the presence of a

high salt concentration, PCPs had unordered

struc-tures As suggested by Sansom & Weinstein [51], this

is presumably due to structural dynamics such as

twist-ing and kinktwist-ing induced by a central Pro residue

The aggregation of PFPs in buffer correlates with

the ability of the peptides to cause the lysis of human

red blood cells and zwitterionic liposomes In contrast,

the self association of PFPs appears to interfere with

their ability to cross the peptidoglycan layer and reach

the cytoplasmic membrane Therefore, PFPs are likely

to show less potent membrane depolarization and

greatly reduced antimicrobial activity Despite the

cytotoxicity of PFPs, however, it appears that the

structural stability and oligomeric form of PFPs in

the presence of a high NaCl concentration could be

useful for treating cystic fibrosis patients if their

anti-microbial versus hemolytic activity is optimized

Kinetic analysis of the sensorgram results suggests

that the binding of M17P to the lipid bilayer occurs by

a distinct two-step process: the peptides may first bind

to the lipid head groups via electrostatic interaction

and then insert further into the hydrophobic interior

of the membrane via hydrophobic interactions The

largest differences between M17 and M17P were

increases in the rate of association (ka1; Table 4) in the

first step and dissociation (kd2, Table 4) in the second

step These findings indicate, respectively, that the

central Pro of a-helical peptides is important for fast

electrostatic interaction with PtdCho⁄ PtdGro

mem-branes and that the Pro is important for effective

translocation across the membrane In addition, the

values of ka1⁄ kd1 (K1; initial binding) and ka2⁄ kd2 (K2;

insertion) correspond, respectively, to the affinity

con-stants for electrostatic and hydrophobic interaction of

the peptides with lipid bilayers As observed in other

amphipathic a-helical peptides such as magainin [52,53], the initial binding of M17P (K1¼ 6.8· 104m)1) was much faster than the following insertion step (K2¼ 4.0) This suggested that the elec-trostatic interaction is a crucial factor for M17P and is responsible for its selective cytotoxicity In contrast, for M17, the rate of the first step (K1¼ 6.2 · 104m)1) was similar to that of the second (K2 ¼ 1.5 · 104) The fact that the K2 value for M17 is much higher than that for M17P indicates that the affinity of the pep-tides for membranes is driven predominantly by hydro-phobic interactions This may explain the nonselective interaction of M17 with both zwitterionic and negat-ively charged membranes

Clarifying the structural aspects of the peptides that confer selective binding to negatively charged lipid membranes and identification of the driving forces for membrane partitioning are essential for understanding the mechanism of permeabilization and improving antimicrobial selectivity The interaction of PCPs with negatively charged membranes is thought to confer selective antimicrobial function, but the induction of plasma membrane leakage alone may not be sufficient

to explain the action of these peptides Our results indicate that Pro residues of amphipathic a-helical peptides may promote formation of a bent structure

by inducing the formation of a helix turn in mem-brane-mimetic environments The bending of PCPs is presumably to provide a membrane anchor after their initial interaction with the membrane surface The overall amphipathic helix of PCPs lies approximately parallel to the bilayer plane, so the bending potential may be the driving force for penetration of the N-ter-minus or C-terN-ter-minus of the peptides into the core of the bilayer In other words, partial conformational flexibility may be a prerequisite for import of the pep-tides into membranes or the cytosol For example, a single Pro residue has been found to be a key struc-tural factor for the penetration of cells by buforin II [20] Also, the Pro residue is thought to promote trans-location across lipid bilayers [21] Many signal peptides also contain a helix-breaking residue and adopt a dynamic helix–break–helix conformation, and this structural motif is thought to be important for the effi-cient initiation of translocation [54–56] In addition, all enveloped viruses enter cells by peptide-mediated mem-brane fusion The viral fusion peptides involved in this process interact with and destabilize the target mem-brane A common feature of many internal viral fusion peptides is the presence of a Pro near the center of their sequence, and it is known that the central Pro residue in fusion peptides is important for the forma-tion of their native structure as well as for the