Iophysical Properties of Membrane-Active Peptides Based on Micelle Modeling

Biophysical Properties of Membrane-Active Peptides Based on Micelle Modeling: A CaseStudy of Cell-Penetrating and Antimicrobial PeptidesQian Wang,Gongyi Hong,,Glenn R. Johnson,|Ruth Pachter,and Margaret S. Cheung*,Department of Physics, UniVersity of Houston, Houston, Texas, United States; Air Force Research Laboratory,Materials & Manufacturing Directorate, Wright-Patterson Air Force Base, Dayton, Ohio, United States;General Dynamics Information Technology, Inc., Dayton, Ohio, United States; and ARFL/RX, Tyndall AirForce Base, Tyndall, Florida, United StatesReceiVed: July 25, 2010; ReVised Manuscript ReceiVed: September 25, 2010Weinvestigatedthemolecularmechanismsofshortpeptidesinteractingwithmembrane-mimeticsystems.Three short peptides were selected for this study: penetratin as a cell-penetrating peptide (CPP), and temporinA and KSL as antimicrobial peptides (AMP). We investigated the detailed interactions of the peptides withdodecylphosphocholine (DPC) and sodium dodecyl sulfate (SDS) micelles, and the subsequent peptide insertionbasedonfreeenergycalculationsbyusingall-atomisticmoleculardynamicssimulationswiththeunitedatom force eld and explicit solvent models. First, we found that the free energy barrier to insertion for thethree peptides is dependent on the chemical composition of the micelles. Because of the favorable electrostaticinteractions between the peptides and the headgroups of lipids, the insertion barrier into an SDS micelle isless than a DPC micelle. Second, the peptides secondary structures may play a key role in their binding andinsertion ability, particularly for amphiphilic peptides such as penetratin and KSL. The secondary structureswith a stronger ability to bind with and insert into micelles are the ones that account for a smaller surfacearea of hydrophobic core, thus offering a possible criterion for peptide design with specic functionalities.1. IntroductionDuring the past decade, a large number of short membrane-active peptides, typically composed of 12-45 amino acids, wereshown to pass across cell membranes with different lyticactivities. Cell-penetratingpeptides(CPP)1present lowlyticactivitieswhereasantimicrobial peptides(AMP)2canrapidlykillmicrobeswithoutexertingtoxicityagainstthehost.Theysharesimilarphysicochemical propertiesandtheirmodesofinteraction depend on the target plasma membrane. However,littleis knownabout aquantitative, physical picturethat isrelated to the biological function of these short peptides (e.g.,structure-activity relationship). In this work, we investigate thebiophysical properties of three membrane-active peptides (i.e.,aCPPpenetratin3andtwoAMPstemporinA3andKSL4)bystudyingtheirstructuresanddynamicswheninteractingwithtwo chemically distinct membrane-mimetic micelles (DPC andSDS) using all-atomistic molecular dynamics simulations.PenetratinbelongstoacategoryofCPPsthat havedrawnmuchattentionbecausetheirabilityofinternalizationallowsnavigation into the nucleus of living cells without disruption ofplasma membranes. It was derived from the third helix of thehomeodomain of Antennapedia and its mechanism of translo-cation is energy-independent and receptor-independent,5indicat-ing that the uptake of this peptide does not involve endocytosis,although some have suggested a different mechanismthatrequiresendocytosis;6however, either doesnot ruleout thepossibility of the other. Nevertheless, it is worthwhile toinvestigate the relationship between the physicochemical proper-ties and the structural characteristics of penetratin. Usingmicelles as biomimetic modeling systems for biological mem-branes, a molecular understanding of penetratin interacting withlipidmoleculeswithout destroyingtheoverall structureofamicelle will give useful insights into the development ofpharmaceuticals for the purpose of noninvasive drug entry intoa cell.The structural characteristics of penetratin in the presence ofmembrane-mimetic materials such as phospholipid vesicles andmicelles were studied using a variety of spectroscopic methods7suchas circular dichroism(CD), uorescence, andnuclearmagneticresonancespectroscopy(NMR). Theseexperimentssuggested that penetratin shows high plasticity in its structure,which can be either R-helices, -turns, or randomcoils,dependentonthephospholipidsofselectedmembranemodelsystems. AlthoughtherolesofArgandLysofpenetratininthe translocation were recently addressed from13C,31P, and19Fsolid-state NMR,8the overall structural characteristics ofpenetratin that account for its permeability across plasmamembranes are still unclear.We also studied two AMPs, temporin Aand KSL, andinvestigated their interaction with different membrane-mimeticmodels. First, temporinAisanantimicrobialpeptide(AMP)from the temporin family derived from frog skins. Unlike theAMPs with bactericidal activities, the net charges in temporinsare typically lower, their peptide lengths are shorter, and theycan interact with both anionic and zwitterionic membranemodels.9Interestingly, temporinAandtemporinB10arealsoanantiparasiticAMPandcantarget anddamagetheplasmamembrane of Leishmania protozoa which adopts different lipidcompositions from bacteria. Second, KSL, a de novo designeddecapeptide composed of a few types of amino acids, includingcharged residues, has shown a broad spectrum of antibacterialactivities withlittlehemolyticactions.11Experiments onitsderivatives and analogues suggested that both the R-helical and* Corresponding author. E-mail: [email protected]. Tel.: 713-743-8358.University of Houston.Wright-Patterson Air Force Base.General Dynamics Information Technology, Inc.|Tyndall Air Force Base.J. Phys. Chem. B 2010, 114, 1372613735 1372610.1021/jp1069362 2010 American Chemical SocietyPublished on Web 10/12/2010-turnstructuresof KSLcanbethekeycharacteristicsthataccountforitsantibacterialactivities.12Themolecularmech-anism of typical AMPs was extensively studied2,13and a numberof antimicrobial peptide membrane permeationmechanismswereproposed. However, whetherthesemechanismsmaybeapplied to temporin A and KSL is still unknown. Nevertheless,it is of interest to explore the mechanism of their interactionswithmembrane-mimeticmicellesthat mayaccount for theirtoxicity at an early stage of peptide-membrane interaction.Here, we used molecular dynamics (MD) simulations to studytheinteractionbetweenthesethreepeptidesandmembrane-mimetic micelles. MDprovides an excellent approach tocomplement experimental ndingsof thestructure-functionrelationship of a peptide in the presence of different lipidmembranemodels.14All-atomisticMDsimulationswereper-formed on AMPs in lipid bilayers/micelles to probe their lyticactivities.15MDsimulationswerealsousedtoinvestigatethetranslocation properties of cell-penetrating peptides.16AlthoughMD is a powerful tool, it is often limited by the large size oflipid bilayers causing it to require large-scale computingresources. In contrast, theoretical models such as those describ-ing the association of cationic peptides with a lipid membrane17can provide a valuable method in computing a comprehensiveparameter phase space of lipid-peptide interaction. However,it oftenfailstodeliverspecicatomisticinformationthat iscrucial in the study of peptide-membrane interactions. In thisregard, we adopted membrane-mimetic micelles as a minimal-istsframeworktostudysomeimportant aspectsat anearlyand transient stage of the peptide binding and insertion process.DPC micelles are zwitterionic while SDS micelles are anionic.Despiteits simplicity, studies basedonthesemicelles haverevealed key interactions in peptide-membrane systems.18Moreover, detailed analysis of the roles of the secondarystructure and amino acid type, as described in this work, couldassist in peptide design for desired functionality.2. Methods2.1. SimulationDetails. Molecular dynamics simulationswere performed using the Gromacs 4.0.5 package.19TheGROMOS96 43a1 force eld,20an united atom force eld, wasimplemented for the simulations. These simulations wereperformed in a 9 nm periodic box with explicit water moleculesrepresented by the single-point charge (SPC) model. The typicalnumber of water molecules was about 22 000(1000. Inaddition, 0.02Msodiumchloride was addedtomatchtheconditions with the experiments on penetratin-membranesystems.21Sodium ions as countercation were added to the SDSmicelles such that the [Na+] reaches to 0.12 M in the systemswith SDS micelle. Sodium ions play an important role in chargeneutralization and remaining a reasonable ionic strength in themedium.ThefourthorderoftheparticlemeshEwald(PME)method was used to calculate the long-range electrostaticinteractions. The cutoff for the real space of Coulombicinteractions and pair list was set to 0.9 nm and the cutoff forvan der Waals interactions was set to 1.6 nm.22The Nose-Hoovermethod23wasusedtomaintainthetemperatureat300Kandthe Parrinello-Rahmanmethod24was usedtomaintainthepressure at 1.0 atm. Covalent bond length was constrained bythe LINCS method.252.2. PreparationofProtein-MicelleComputerModels.Penetratin. The protein structure of penetratin (RQIKIWFQN-RRMKWKK) was obtained from the protein data bank (PDBID: 1OMQ).26ItsN-terminuswascappedbyacetylationandthe C-terminus was capped by amidation to reect the conditionsin the experimental preparation.27Penetratin was energy-minimizedwiththesteepestdescentmethodandthenequili-brated at 500 K for 20 ns to produce randomized structures asinitial conditions. In order to investigate the role of the peptidessecondary structure in interacting with the micelles, we selecteda penetratin structure with the most helical content (PR), and apeptide structure withthe most beta content (P

) as initialcongurations(Figure1). ThepercentagesofeachdominantsecondaryconformationsinFigure1werecomputedbytheprogram STRIDE28from the ensemble.TemporinAand KSL. There are no solution structuresavailablefor temporinAandKSL; henceweproducedthecoordinatesofthetemporinA(FLPLIGRVLSGIL-NH2)andKSL (KKVVFKVKFK-NH2) peptides from their sequences byusing the xLEaP program in the AMBER9 package.29TemporinA and KSL were capped by amidation to reect the conditionsintheexperimentalpreparation.4,10Thesamestepsofenergyminimization as well as the conditions to produce randomizedinitial structures described above were applied on temporin AandKSL.Thestructuralcriteriaoftheselectionoftheinitialconditions, TR and T

, of temporin A are the same as penetratin(Figure1). However, for KSL, thereisnodominant helicalstructure; we chose a dominant turn shape (KT) and a dominant-strand structure (K

) (Figure 1).Micelle Models. Two micelle models were investigated. TheDPC micelle model with a radius of 2.5 nm, consisting of 65DPClipid molecules, was downloaded fromhttp://moose.bio.ucalgary.ca. The SDS micelle model with a radius of 2.35nm, consisting of 60 SDS lipid molecules, was obtained fromthe work of MacKerell.30The aggregation number 60 is closeto the experimental value of 63.31The topological les of themicelleswereproducedbyPRODRG.32DPCandSDSmol-ecules are represented in Figure S1 in the Supporting Informa-tion.2.3. Umbrella Sampling. The umbrella sampling (US)method33wasusedtocomputethefreeenergyof peptideinsertion as a function of the distance between the center ofFigure 1. Percentage of dominant secondary structures in the ensembleofapeptide:(A,B)penetratin,(C,D)temporinA,and(E,F)KSL.Biophysical Properties of Membrane-Active Peptides J. Phys. Chem. B, Vol. 114, No. 43, 2010 13727massofapeptide(e.g., penetratin)andthecenterofmassofamicelle(Dcom)at300K. TheinitialcongurationsforUSwerepreparedbypullingthepeptidesfromtheoutsideto the inside of a micelle using the initial conditions providedin Figure 1 for the two micelles, DPC (D) and SDS (S). Thepulling rate was controlled at 0.001 nm/ps. US was performedalongthedistancebetweenthecenterofmassofamicelleand a peptide from Dcom ) 2.5 nm to Dcom ) 0.5 nm with abinsizeof0.1nm(i.e., 21windows). Foreachwindow, a20ns molecular dynamics withharmonicconstraints wasperformed. The harmonic potential U(Dcom) was imposedbetweenthecenterofmassofapeptideandthecenterofmass of eachmicelle at a distance of Dcomwitha forceconstant of 1000 kJ/mol/nm2. The free energy as a functionofDcomwasobtainedbyreweighingthedensityofstates.34The error values on the free energy calculations wereestimatedbythebootstrapmethod.352.4. Kinetic Trajectories of a Peptide Binding to Micelles.Different initial congurations (i.e., PR, P

, TR, T

, KT, and K

)wererandomlydistributedaroundtheproximityofamicelle.Theminimumdistancesbetweenpeptidesandmicelleswereinthe range of 0.3-0.6nm. Once eachcongurationwaspositioned for initial binding, we also chose another twocongurationsasinitialconditionsbyrotatingthispeptidetoanother 90 and 180 against the peptides longitudinal axis. Apeptides longitudinal axis of each guration is dened by thetwo CR atoms in the following residues: PR, residue 1 and residue16; P

, residue 1 and residue 10; TR, residue 1 and residue 13;T

, residue 1 and residue 7; KT, residue 1 and residue 10; K

,residue 1 and residue 5. A 1 ns equilibration at T ) 300 K wasperformed, followed by a 20 ns simulation at a constant pressureof 1 atm and temperature 300 K. Each simulation under differentpeptideorientations was repeatedfour times usingdifferentrandom seed numbers; therefore, a total of 12 trajectories startingfromthe same initial conformations were producedfor thecalculation of temporal averages. A shell with a radius of 3.5nmlocatedatthecenterofmassofamicelleisdenedasacollision zone. For any trajectory, whenever any part of a peptideentersthecollisionzonethedatawascollectedfor bindinganalysis.3. Results3.1. Kinetic Analysis of Penetratins Binding to Micelles.Dependence on the secondary structures and the type of micellesfor the mechanismof penetratin binding to micelles wasinvestigatedbyall-atomisticmoleculardynamicssimulationswith explicit water models. Several trajectories of 20 nssimulations, which differ by the initial positions and secondarystructures of penetratin, were collected for the kinetic analysisofpenetratinsbindingtoaDPCmicelleasshowninFigure2A, and a SDS micelle as shown in Figure 2B.There are two interesting features from the kinetic analysisthat underpin the importance of penetratins secondary structureformicellebinding.First,theoveralltemporaldecayofDcomfor an R-helical dominant structure (PR) is much faster than a-strandinthepresenceofaDPCmicelle(Figure2A). Wetted the curve in Figure 2A with a single exponential (Table1)andshowedthatthebindingrateforPRwas5.45 10-3ps-1, about 2-fold greater than P

(2.91 10-3ps-1). Interest-ingly, the same feature that favors an R-helical dominantstructure for binding was observed for SDS micelles in Figure2B.ThebindingrateofPRwasttedto3.67 10-3ps-1inTable 1 and it is 3-fold greater than P

(1.33 10-3ps-1).Second, both parts A and B of Figure 2 demonstrate that after10 ns the distance between the center of mass of a peptide andthe center of mass of a micelle (Dcom) begins to plateau.However, theseprolesplateauat adifferent valueofDcom,where Dcom is much shorter for PR than for P

in the presenceof DPC micelles. Dcom of PR remains at 1.9 nm and Dcom forP

remains at 2.1 nm.3.2. Penetratins Structural Analysis uponBindingtoMicelles. In section 3.1 we identied the importance ofsecondary structures as initial conditions for penetratins bindingto micelles. We evaluated the participation of each residue inthebindingprocessofpenetratinbycomparingtheaveragedminimal distance (Dmin) between a residue and any lipidmolecule in a micelle at the last 2 ns of the simulation. If theDmin of a residue is less than 0.35 nm, which is approximatelythesizeofawatermolecule,weconsideredthisresidueasabinding residue that makes essential contact with a micelle. ForPRinthepresenceofaDPCmicelle, chargedresidues(i.e.,Arg1, Arg11, Lys15) and hydrophobic residues (i.e., Trp6, Phe7,Trp14) were binding residues (Figure 3A), indicating that boththeelectrostaticandhydrophobicinteractionswereimportanttobinding. However, for P

, onlychargedresidues (Arg1,Arg10, andArg11) weremost important tobindinginthepresenceofaDPCmicelle(Figure3B). InthepresenceofaSDS micelle, for PR, charged residues (i.e., Lys13, Lys15 andLys16) and hydrophobic residues (i.e., Trp14) are important tobinding(Figure 3C). For P

, chargedresidues (i.e., Lys13,Lys15, and Lys16) are important to binding as shown in Figure3D.Interestingly, there are a signicant number of bindingresidues, both PR and P

, located at the C-terminus of penetratin,indicating a tilted structure necessary for binding a SDS micelle.Particularly for PR, there is a relatively good agreement betweenour results and a prior NMR study in which the position of theC-terminusofahelicalpenetratinispositioneddeeperinsidethe micelle than its N-terminus.7bIn contrast to SDS, when inthepresenceof theDPCmicelle, thedistancesbetweentheresiduesattheN-terminusandthesurfaceofaDPCmicelleare nearly the same as the distances between the residues at theC-terminus and the surfaces of a DPC micelle, indicating thatpenetratin approaches a DPC micelle without a tilted preference.3.3. Mechanism of Penetratins Binding to DPC Micelles.Because differences between PR and P

in penetratins bindingmechanismare most evident wheninteractingwitha DPCmicelleasseenfromanalysisinpriorsections, wetookthissystemasanexampletoinvestigatethedrivingforcesofPRandP

forbinding. TemporallyaveragedelectrostaticenergyandvanderWaals(VDW)potential energyasafunctionoftime are plotted in Figure 4, A and B, respectively. The proleof VDW potential energy for PR decays faster than that of P

,whilethereislittledifferenceintheelectrostaticinteractionsfor both PRand P

, indicating that the LJ interaction isresponsible for a greater binding rate of PR than P

as observedinFigure2andTable1. Followingthediscussioninsection3.2onthe analysis of bindingresidues, this drivingVDWpotential energy for PR is attributed to the interactions betweenits hydrophobic binding residues (Trp6, Phe7, and Trp14) andazwitterionicDPCmicelle, whiletherearenohydrophobicresidue as binding residues for P.The role of these three hydrophobic residues in PRs and P

sbinding to a DPCmicelle is illustrated in a typical 20-nstrajectory in Figure 5. For PR, at t ) 1 ns, Trp6 (yellow) makescontact to a DPC micelle. At t ) 10 ns, Trp6 and Arg10 (red)insertintothemicelle, whileTrp6, Phe7(orange)andTrp14(orange) form a hydrophobic core. The hydrophobic surface areaofthishydrophobiccoreinvolvingTrp6,Phe6andTrp14is13728 J. Phys. Chem. B, Vol. 114, No. 43, 2010 Wang et al.1.74nm2.Att )20ns,thehydrophobiccoreresiduesTrp6,Phe7andTrp14aswellasArg10formacomplexandinsertthemselves under the surface of a DPC micelle. As a result, theorientation of lipid molecules of a DPC micelle is interruptedand tiled away froma radial direction. In contrast to thehydrophobicresiduesinPRthat areabletopackthemselvescloselyintoacore, andallowfavorablecation-interactionsbetween Trp6 and Arg10 as suggested in an experimentalstudy,36thesamehydrophobicresiduesinP

areunlikelytomake the same impact to binding when they are separated in arather extended conguration. The hydrophobic surface area ofP

involving Trp6, Phe7, and Trp14 is 2.15 nm2, increased by23.6% compared to that of PR.3.4. TemporinABinding to DPCandSDSMicelles.Temporally averaged Dcom between temporin A and the micellesisshowninFigure2C(DPCmicelle)and2D(SDSmicelle).For TRandT

, differencesintheprolesof Dcomregardingbinding to SDS and DPC micelles are small. After being ttedFigure 2. Temporally averaged distance (Dcom) between the center of mass of a peptide and the center of mass of a micelle (DPC or SDS) as afunction of time: (A) penetratin with a DPC micelle, (B) penetratin with a SDS micelle, (C) temporin A with a DPC micelle, (D) temporin A withan SDS micelle, (E) KSL with a DPC micelle, and (F) KSL with an SDS micelle. Error values shown are the standard error of the mean.TABLE 1: Fitted Single Exponential of Temporally Averaged Distance between the Center of Mass of Penetratin and theCenter of Mass of a Micelle (Dcom) as a Function of Time (t) in Figure 2 Using Dcom ) A0 exp(-A1t + A2)PDRPD

PSRPS

TDRTD

TSRTS

KDTKD

KSTKS

A0 (nm) 0.98 1.04 1.25 1.02 0.88 1.02 1.90 1.31 1.15 0.81 1.17 1.06A1 (10-3ps-1) 5.45 2.91 3.67 1.33 4.62 4.37 4.54 4.13 5.00 2.28 2.23 3.50A2 (nm) 1.91 2.14 1.82 1.74 1.87 1.84 1.52 1.60 1.93 2.02 1.62 1.77Biophysical Properties of Membrane-Active Peptides J. Phys. Chem. B, Vol. 114, No. 43, 2010 13729with a single exponential (Table 1), the binding rates of TR (4.62 10-3ps-1) and T

(4.37 10-3ps-1) are almost the same inthe presence of a DPC micelle. In the presence of a SDS micelle,the binding rate for TR was 4.54 10-3ps-1and for T

, 4.13 10-3ps-1. Dcom levels out at 1.9 nm for the DPC micelle and1.6 nm for the SDS micelle.The minimal distance (Dmin) between each residue of temporinA and the micelles was investigated in Figure S2 (see SupportingInformation). InthepresenceofaDPCmicelle, thebindingresidues of TR are hydrophobic: Phe1, Leu4, Leu9, indicatingthat the major driving force is the hydrophobic interaction. Thesame three residues were characterized for T

in binding witha DPC micelle. In the presence of a SDS micelle, thehydrophobic residues Phe1 and Gly11 were shown to beessential in binding for both TR and T

. These results demon-strate the importance of hydrophobic interaction over electro-staticinteractionsfor temporinAinbindingDPCandSDSmicelles because all the binding residues are hydrophobic.We further investigated the electrostatic interactions and theLJ potential energy between temporin A and a DPC micelle asa function of time in Figure 4, Cand Dfor TRand T

,respectively. There is little difference between TR and T

in bothpartsCandDofFigure4. Inaddition, inthepresenceofaSDS micelles (Figure S3 in the Supporting Information), theirdifferences are insignicant. In contrast to penetratin, thesecondary structure of temporin A makes little impact to bindingwitheitherDPCorSDSmicelles. Thismaybeattributedtothe chemical composition of temporin A, which is mostlycomposed of hydrophobic residues. Therefore, as long as thereisastronghydrophobicinteractionbetweentemporinAandthe micelles, the selection of secondary structures matters littleto binding.3.5. KSL Binding to DPC and SDS Micelles. TemporallyaveragedDcombetweenKSLandmicellesisshowninFigure2, part E (DPC micelles) and part F (SDS micelles). When ttedbyasingleexponential (Table1), inthepresenceofaDPCmicelle,thebindingrateforKTbindingis5.00 10-3ps-1,which is approximately 2-fold greater than that for K

(2.28 10-3ps-1), indicating that the overall temporal decay of KT ismuch faster than K

. Indeed, when only the LJ potential energyis plotted as a function of time for KT, its prole decays fasterthan that of K

(Figure 4F), while the proles of electrostaticinteractions in Figure 4E remain the same for both KT and K

.In contrast, in the presence of a SDS micelle, both the prolesofelectrostaticenergyandLJpotential energyremainindis-tinguishable (Figure S5 in the Supporting Information), indicat-ing that the secondary structure of KSL plays a lesser role inbinding to SDS micelles.WefurtherinvestigatedthebindingresiduesofKSLwheninteracting with a DPC micelle in Figure S4, A and B, and witha SDSmicelle in Figure S4, Cand D, in the SupportingInformation. The binding residues of KT are charged (i.e., Lys6,Lys8, Lys10) and hydrophobic (i.e., Val3, Phe5, Phe9), indicat-ingthat bothelectrostatic andhydrophobic interactions areimportantforKTtobindtoDPCmicelles(FigureS4A). ForK

in Figure S4B, the binding residues are charged (i.e., Lys1and Lys6) and hydrophobic (i.e., Val3 and Phe5). Phe9,however, isnot abindingresidueforK

, indicatingthat thehydrophobic interactionbetweenK

anda DPCmicelle isweaker than for KT. In the presence of a SDS micelle, for KT,thechargedresidues(i.e., Lys1, Lys6, Lys9)andthehydro-phobicresidues(i.e., Val3, Phe5, Phe9)arebindingresidues(as shown in Figure S4C), whereas for K

, the charged bindingFigure 3. Average minimal distance (Dmin) between each residue of a penetratin and a micelle in the last 2 ns of the kinetic binding simulations:(A) PR and DPC micelle, (B) P

and DPC micelle, (C) PR and SDS micelle, and (D) P

and SDS micelle.13730 J. Phys. Chem. B, Vol. 114, No. 43, 2010 Wang et al.residues are Lys1, Lys2 and Lys6, and the hydrophobic residuesare Val3 and Phe5, as shown in Figure S4D in the SupportingInformation.A typical 20 ns trajectory of KSL using two distinct secondarystructures as initial conditions to bind a DPCmicelle isillustrated in Figure 6. The two hydrophobic residues Phe5 andPhe9 are colored in yellow and orange, respectively, for visualguidance. At t ) 1 ns, the two residues in KT collapse and forma hydrophobic core attributed to an increase in the hydrophobicinteractions of KT. The hydrophobic surface area involving Phe5and Phe9 is 1.17 nm2. At t ) 10 ns, Phe5 in KT makes contactto the lipid molecules, interrupting the alignment of lipidmolecules. At t ) 20 ns, the hydrophobic core is inserted underthe surface of a DPC micelle. In contrast, these two hydrophobicresidues stayseparatedapart inK

andthestructureof K

remains at thesurfaceof aDPCmicelle. Thehydrophobicsurface area of the tworesidues of K

is 1.53nm2anditincreases by 30.8% when compared to that of KT.3.6. Free Energy of Peptide Insertion into Micelles. Thefreeenergyprolesasafunctionofthedistancebetweenthecenter of mass of the peptide and the center of mass of a micelle(Dcom) are shown in Figure 7. The free energy differences (Gb)betweenthefreeenergyminimum, oftenpositionednearthemicelle surface from Dcom) 1.5 nm to Dcom) 2.1 nm, and thefree energy near the center of mass of the micelle (Dcom) 0.7),can give a reasonable estimate on the free-energy cost of peptideinsertion inside a micelle. Despite the selection of peptide andsecondary structures, Gb ranges from 150 to 200 kJ/mol for aDPCmicelleandfrom50to125kJ/mol foraSDSmicelle.Suchaninsurmountablylarge free energybarrier makes aspontaneous peptide insertion nearly impossible on a nanosecondtime scale. It is unlikely that the systemcan reach thethermodynamicequilibriumpositions withinseveral tens ofnanoseconds whereas in the kinetic studies the simulation timestops when the trajectories reach plateau. As a result, the positionof free energy minima of peptide insertion is slightly off withrespect tothepositionwhenthetrajectoriesreachplateauinthe kinetic study.Interestingly, for eachpeptide, GbinaSDSmicelle(anegatively charged micelle) is about 75 kJ/mol less than that ina DPC micelle (a zwitterionic micelle) (Figure 7). In order toinvestigate the key interactions that account for such differencesin the free energy barrier, we measured the electrostatic energy(averaged over several trajectories as described in the Methodssection) between each peptide and the solvated water moleculesat thesurfaceandat thecenter of amicelleunder differentmicelle conditions as provided in Table 2. In the presence of aDPC micelle, the difference in the electrostatic energy betweenthecenterpositionandthesurfaceposition(EDPCC-S)isabout60-90 kJ/mol greater than that in the presence of a SDS micelleFigure4. Interactionbetweenapeptidewithdistinct secondarystructuresasinitial conditionsandaDPCmicelleasafunctionoftime: (A)penetratin, electrostatic energy, (B) penetratin, van der Waals potential energy, (C) temporin A, electrostatic energy, (D) temporin A, van der Waalspotential energy, (E) KSL, electrostatic energy, and (F) KSL, van der Waals potential energy.Biophysical Properties of Membrane-Active Peptides J. Phys. Chem. B, Vol. 114, No. 43, 2010 13731(ESDSC-S), whichisonthesameorderasthedifferenceinanoverall freeenergybarrier asmentionedabove(75kJ/mol).Thus, our results indicate that the electrostatic interaction fromthe surface charges of a SDS micelle can greatly reduce Gbby narrowing the difference in the peptide-water electrostaticenergybetweenthesurfaceandthecenterofaSDSmicelle.Next, thesecondarystructureofapeptidemayalsoaffectthe free energy of insertion, particularly on the position of freeenergyminima. ForpenetratininFigure7A, thefreeenergyminimum for PR insertion (Dcom < 1.8 nm) is closer to the centerofamicellethanforP

insertion(Dcom >2nm).Inaddition,Gb for PR is about 25 kJ/mol less than P

. These results are inagreement with the above binding simulations when DcombetweenPRandamicelleislessthanthatbetweenP

andamicelle after a 20 ns simulation. For temporin A, however, thereisnoclear trendinhowthesecondarystructureaffectstheFigure 5. Illustration of penetratin binding to a DPC micelle. The residue in yellow is Trp6, the residues in orange are Phe7 and Trp14, and theresidue in red is Arg10.Figure 6. Illustration of KSL binding to DPC micelle. The residue in yellow is Phe5 and the residue in orange is Phe9.13732 J. Phys. Chem. B, Vol. 114, No. 43, 2010 Wang et al.insertion of temporin A as illustrated in Figure 7B. For KSL,thefreeenergyminimumforKTinsertion(Dcom 2nm)(Figure7C).Inaddition, GbforKTisabout25kJ/molless than K. These results are in accord with the earlier bindingresults, for which Dcom of KT is less than that of K

at the nalstage in a 20 ns simulation.4. Discussion4.1. SecondaryStructureAffectstheAmphiphilicPep-tides Binding and Insertion Process. Our investigationdemonstrates that the selection of peptides secondary structuresmay play a key role in its ability of binding and insertion (nottoxicity), particularly for amphiphilic peptides such as penetratinand KSL. For example, PR binds faster and inserts deeper intoDPCandSDSmicellesthanP

. Theformof thesecondarystructuresdeterminesresidueinteractionsinapeptide, whichdictates its bindingandinsertionmechanism. Inpenetratin,formation of Trp6 and Arg10 through -cation interactions isenabled in an R-helical PR rather than in an extended strand ofP

. As a result, stronger -cation interactions may impact thealignment of lipidmolecules of a micelle. Inaddition, thehydrophobic core of PRinvolving Trp6, Phe7, and Trp14enhances the hydrophobic interactions between penetratin andthe micelle. Therefore, the free energy minimum of PR insertionis found to be under the micelle surface, while P

is found atthesurface. Oursimulationresultsemphasizetheimportanceof Arg and Trp, consistent with prior experimental studies onthe effect of Arg and Trp interactions on lipids, where the twotypes of amino acids form a complex36and remain at the surfaceof the lipid molecules through hydrogen bond formation withlipid head groups.37Although the types of amino acids important to binding werecharacterized, the experimental nding on the form of secondarystructures pertinent to binding and insertion is still unclear.7a,c-e,26,38InsomeexperimentsusingCD, ahelix-transformationofpenetratin was found to be necessary,7awhile based on anotherNMR experiment a -turn was important to penetratins insertionprocess.7eOur results on penetratin based on micelles may notappropriatelyaddresstheimportanceofpeptide-lipidratios,nor can micelles adequately represent lipids; nevertheless, oursimulations provide a thorough investigation of the biophysicalproperties of peptides when interacting with membrane-mimeticmaterials.4.2. Chemical Composition of a Micelle Affects theAmphiphilicPeptides BindingandInsertion. Our resultsshow that the electrostatic interaction between an amphiphilicpeptide (rich in Arg and Lys) and a micelle facilitates not onlythe kinetics of binding but also its insertion into micelles. BothintheCPPandintheAMPfamilies, thepositivelychargedresidues Arg and Lys were found to be important wheninteractingwithlipidsystems.8,39Inourstudy, weexaminedthe potential interactions and provided details on howthecharged residues bind better in negatively charged micelles. WeFigure 7. Free energy of peptide insertion as a function of the distance(Dcom) between the center of mass of a peptide and the center of mass ofmicelles (DPC or SDS): (A) penetratin, (B) temporin A, and (C) KSL.The error value, estimated by the bootstrap method, of each data point isless than 0.5 kJ/mol and too small to be shown in the gures.TABLE 2: Electrostatic Potential Energy Interaction (E) between a Peptide in Distinct Secondary Structures (shown in Figure1) and Its Solvated Water Molecules in the Presence of DPC (or SDS) MicellesaEDPCCEDPCSEDPCC-SESDSCESDSSESDSC-SPR-871 ( 6 -1107 ( 6 236 ( 12 -908 ( 5 -1067 ( 6 159 ( 11P

-873 ( 11 -1092 ( 5 219 ( 16 -927 ( 9 -1086 ( 8 159 ( 17TR-284 ( 4 -437 ( 4 -153 ( 8 -392 ( 8 -472 ( 10 -80 ( 18T

-264 ( 10 -437 ( 5 -173 ( 15 -388 ( 3 -471 ( 8 -103 ( 11KT-757 ( 9 -984 ( 8 227 ( 17 -757 ( 8 -900 ( 3 143 ( 11K

-760 ( 4 -999 ( 8 239 ( 12 -772 ( 6 -920 ( 5 148 ( 11aAsuperscriptSrepresentsthelocationofapeptideatthesurfaceofamicelleandasuperscriptCrepresentsthelocationofapeptideat thecenterofamicelle. Thedifferenceintheelectrostaticpotential energyinteraction(E)isalsopresented. Theunit isinkJ/mol. Errorbars are included.Biophysical Properties of Membrane-Active Peptides J. Phys. Chem. B, Vol. 114, No. 43, 2010 13733foundthat, whenatarget micelleisneutrallycharged(e.g.,DPC), it is unlikely for peptides to be found beneath the micellesurface where the hydrophobic tails of lipids are present, duetoa stronger electrostatic attractionbetweena peptide andsolvent moleculesthanvandeWaalsinteractionsbetweenapeptide and the hydrophobic lipid tails. However, when a targetmicelle is negatively charged such as a SDSmicelle, thehydrophilic phosphate headgroups of a micelle provide stabiliz-inginteractionswithapeptidethatattractspeptideinsertion.As a result, the free energy barrier Gb instead reduces in thepresence of SDS micelles.5. ConclusionWe have employed all-atom molecular dynamics simulationto investigate the inuence of the secondary structure of peptides(penetratin, temporin A, and KSL) and the chemical propertiesof target micellesonthebindingandinsertionprocess. Ourresultsdemonstratethat for penetratinanR-helical structurebetter facilitates the binding and insertion process than a -strandstructure in the presence of both DPC and SDS micelles. Fromcomparison of the free energy of peptide insertion for both typesof secondary structures, we suggest that the activity of penetratinfor insertion is likely to have a small hydrophobic surface area,whichmaybenet fromless disruptionof lipidorientationaroundthis hydrophobic core. For temporinA, there is nosecondary structure preference in both the binding and insertionprocesses. Becauseit is composedof ahighpercentageofhydrophobic residues, it appears to be irrelevant for temporinAtoadopt either of the secondarystructures tomaximizeamphiphilicity. For KSL, a turnshape, whichenhances itshydrophobicinteractiontoamicelle, ismorefavorablethanthe -strand structure. In addition to the selection of secondarystructures, the chemical composition of micelles inuences thefree energyminima of peptide insertion, mainlydrivenbyinteractions between a peptide and the hydrophilic head groups.Foranamphiphilicpeptidewithagivenset ofhydrophobicaminoacids, wheninteractingwithmembraneormembrane-mimeticstructures, itislikelytoadoptasecondarystructurewith a small hydrophobic surface area. We will further employbioinformatics studies on membrane-active AMPs and CPPs tocompare with existing experimental results.Acknowledgment. M.S.C. thanks the support partly from theTexas Center for Superconductivity at the University of Houston(TcSUH), UHGrants to Enhance and Advance Research(GEAR), and the Materials & Manufacturing Directorate, AirForce Research Laboratory. G.H. gratefully acknowledgessupport from the Defense Threat Reduction Agency (DTRA).Computations were partly supported by the Texas Learning andComputation Center (TLC2), Texas Advanced ComputingCenter (TACC), and2010IBMSharedUniversityResearch(SUR) Award on IBMs Power7 high performance cluster(BlueBioU) to Rice University as part of IBMs Smarter PlanetInitiatives in Life Science/Healthcare and in collaboration withthe Texas Medical Center partners, with additional contributionsfromIBM, CISCO, Qlogic, andAdaptiveComputing. Q.W.thanks Dr. Alexander Mackerell for sharing the SDS coordinateles and Dr. Hugh Nymeyer for the discussion on the interactioncutoffsinGromacs. M.S.C. thanksDr. HueyW. Huangforbringing penetratin to her attention.Supporting Information Available: Figure S1: The repre-sentation of DPC and SDS molecule. Figure S2: The averageminimal distance between temporinA and a micelle in kineticbinding simulations. Figure S3: Van der Waals potential energybetween temporinA and SDS micelle. Figure S4: The averageminimal distance between KSL and a micelle in kinetic bindingsimulations. Figure S5: Interaction between KSLand SDSmicelle. This material is available free of charge via the Internetat http://pubs.acs.org.References and Notes(1) Lindgren, M.; Hallbrink, M.; Prochiantz, A.; Langel, U. Cell-penetrating peptides. Trends Pharmacol. Sci. 2000, 21, 99103.(2) Yeaman, M. R.; Yount, N. Y. Mechanisms of antimicrobial peptideaction and resistance. Pharmacol. ReV. 2003, 55, 2755.(3) Derossi, D.; Chassaing, G.; Prochinatz, A. Trojanpeptides: thepenetratin system for intracellular delivery. Trends Cell Biol. 1998, 8, 8487.(4) Hong, S. Y.; Oh, J. E.; Kwon, M. Y.; Choi, M. J.; Lee, J. H.; Lee,B. L.; Moon, H. M.; Lee, K. H. Identication and characterization of novelantimicrobial decapeptides generated by combinatorial chemistry. Antimi-crob. Agents Chemother. 1998, 42, 25342541.(5) Derossi, D.; Calvet, S.; Trembleau, A.; Brunissen, A.; Chassaing,G.; Prochinatz, A. Cell internalization of the third helix of the Antennapediahomeodomain is receptor-independent. J. Biol. Chem. 1996, 271, 1818818193.(6) Drin, G.; Cottin, S.; Blanc, E.; Rees, A. R.; Temsamani, J. Studiesonthe internalizationmechanismof cationic cell-penetratingpeptides.J. Biol. Chem. 2003, 278, 3119231201.(7) (a) Magzoub, M.; Kilk, K.; Eriksson, L. E. G.; Langel, U.; Graslund,A. Interactionandstructureinductionofcell-penetratingpeptidesinthepresence of phospholipid vesicles. Biochim. Biophys. Acta 2001, 1512, 7789. (b)Lindberg, M.; Graslund, A.; Thepositionofthecell penetratingpeptidepenetratininSDSmicellesdeterminedby, N. M. R. FEBSLett.2001, 497, 3944. (c) Magzoub, M.; Eriksson, L. E. G.; Graslund, A.Conformational states of the cell-penetratin peptide penetratin wheninteracting with phospholipid vesicles: effects of surface charge and peptideconcentration. Biochim. Biophys. Acta2002, 1563, 5363. (d) Clayton,A. H. A.; Atcliffe, B. W.; Howlett, G. J.; Sawyer, W. H. Conformationandorientationofpenetratininphospholipidmembranes. J. PeptideSci.2006, 12, 233238. (e) Su, Y. C.; Mani, R.; Doherty, T.; A.J., W.; Hong,M. Reversible sheet-turn conformational change of a cell-penetrating peptidein lipid bilayers studied by solid-state NMR. J. Mol. Biol. 2008, 381, 11331144.(8) Su, Y. C.; Doherty, T.; Waring, A. J.; Ruchala, P.; Hong, M. Rolesofarginineandlysineresiduesinthetranslocationofacell-penetratingpeptide from 13C, 31P, and 19F Solid-State NMR. Biochemistry 2009, 48,45874595.(9) Mangoni, M. L.; Rinaldi, A. C.; Di Giulio, A.; Mignogna, G.; Bozzi,A.; Barra, D.; Simmaco, M. Structure-function relationships of temporins,small antimicrobial peptides from amphibian skin. Eur. J. Biochem. 2000,267, 14471454.(10) Mangoni, M. L.; Saugar, J. M.; Dellisanti, M.; Rivas, L. Temporins,small antimicrobial peptideswithleishmanicidal activity. J. Biol. Chem.2005, 280, 984990.(11) Concannon, S. P.; Crowe, T. D.; Abercrombie, J. J.; Molina, C. M.;Hou, P.; Sukumaran, D. K.; Raj, P. A.; Leung, K. P. Susceptibility of oralbacteria to an antimicrobial decapeptide. J. Med. Microbiol. 2003, 52, 10831093.(12) Oh, J. E.; Hong, S. Y.; K.H., L. Structure-activity relationship study:short antimicrobial peptides. J. Peptide Res. 1999, 53, 4146.(13) (a) Huang, H. W. Molecular mechanism of antimicrobial peptides:the origin of cooperativity. Biochim. Biophys. ActasBiomembr. 2006, 1758,12921302. (b) Shai, Y.; Orena, Z. From carpet mechanism to de-novodesigned diastereomeric cell-selective antimicrobial peptides. Peptides 2001,22, 16291641.(14) (a)Matyus,E.;Kandt,C.;Tieleman,D.P.Computersimulationof antimicrobial peptides. Curr. Med. Chem. 2007, 14, 27892798. (b)Biggin, P. C.; Sansom, M. S. P. Interactionsofalpha-heliceswithlipidbilayers:areviewofsimulationstudies. Biophys. Chem. 1999, 76, 161183.(15) LaRocca, P.;Biggin, P. C.;Tieleman, D. P.;Sansom, M. S. P.Simulationstudiesoftheinteractionofantimicrobial peptidesandlipidbilayers. Biochim. Biophys. Acta 1999, 1462, 185200.(16) (a)Herce, H. D.;Garcia, A. E. MoleculardynamicssimulationssuggestamechanismfortranslocationoftheHIV-1TATpeptideacrosslipid membranes. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 2080520810.(b) Yesylevskyy, S.; Marrink, S. J.; Mark, A. E. Alternative mechanismsfor the interaction of the cell-penetrating peptides penetratin and the TATpeptide with lipid bilayers. Biophys. J. 2009, 97, 4049.(17) Taheri-Araghi, S.; Ha, B. Y. Physical basis for membrane-chargeselectivityofcationicantimicrobial peptides. Phys. ReV. Lett. 2007, 98,168101.13734 J. Phys. Chem. B, Vol. 114, No. 43, 2010 Wang et al.(18) (a) Khandelia, K.; Kaznessis, Y. N. Molecular dynamics simulationsof helical antimicrobial peptides in SDS micelles: What do point mutationsachieve. Peptides 2005, 26, 20372049. (b) Kosol, S.; Zangger, K. Dynamicsand orientation of a cationic antimicrobial peptide in two membrane-mimeticsystems. J. Struct Biol. 2010, 170, 172179. (c) Balayssac, S.; Burlina, F.;Convert, O.; Bolbach, G.; Chassaing, G.; Lequin, O. Comparison ofpenetratinandotherhomeodomain-derivedcell-penetratingpeptides: in-teraction in a membrane-mimicking environment and cellular uptakeefciency. Biochemistry 2006, 45, 14081420. (d) Saviello, M. R.; Mal,S.; Campiglia, P.; Cavalli, A.; Grieco, P.; Novellino, E.; Carotenuto, A.Newinsightintothemechanismofactionofthetemporinantimicrobialpeptides. Biochemistry 2010, 49, 14771485. (e) Shang, B. Z.; Wang, Z. W.;Larson, R. G. Moleculardynamicssimulationofinteractionsbetweenasodium dodecyl sulfate micelle and a poly(ethylene oxide) Polymer. J. Phys.Chem. B 2008, 112, 28882900. (f) Wymore, T.; Wong, T. C. Moleculardynamics study of substance P peptides partitioned in a sodium dodecyl-sulfate micelle. Biophys. J. 1999, 76, 12131227.(19) (a) Berendsen, H. J. C.; van der Spoel, D.; van Drunen, R.GROMACS: A message-passing parallel molecular dynamics implementa-tion. Comput. Phys. Commun. 1995, 91, 4356. (b) Hess, B.; Kutzner, C.;van der Spoel, D.; Lindahl, E. GROMACS 4: algorithms for highly efcient,load-balanced, and scalable molecular simulation. J. Chem. Theory Comput.2008, 4, 435447.(20) Scott, W. R. P.;Hunenberger, P. H.;Tironi, I. G.;Mark, A. E.;Billeter, S. R.; Fennen, J.; Torda, A. E.; Huber, T.; Kruger, P.; vanGunsteren, W. F. The GROMOS biomolecular simulation program package.J. Phys. Chem. A 1999, 103, 35963607.(21) Binder, H.; Lindblom, G. Charge-dependent translocationoftheTrojanpeptidepenetratinacrosslipidmembranes. Biophys. J. 2003, 85,982995.(22) Daura,X.;MARK,A.E.;vanGunsteren,W.F.Parametrizationof aliphatic CHn united atoms of GROMOS96 force eld. J. Comput. Chem.1997, 19, 535547.(23) Hoover, W. G. Canonical dynamicssequilibriumphase-spacedistributions. Phys. ReV., A 1985, 31, 16951697.(24) Parrinello, M.; Rahman, A. Strain uctuations and elastic-constants.J. Chem. Phys. 1982, 76, 26622666.(25) Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M.LINCS:Alinearconstraintsloverformolecularsimulations. J. Comput.Chem. 1997, 18, 14631472.(26) Lindberg, M.; Biverstahl, H.; Graslund, A.; Maler, L. Structure andpositioning comparison of two variants of penetratin in two differentmembrane mimicking systems by NMR. Eur. J. Biochem. 2003, 270, 30553063.(27) Berlose, J. P.; Convert, O.; Derossi, D.; Brunissen, A.; Chassaing,G. Conformational and associative behaviours of the third helix ofantennapedia homeodomain in membrane-mimetic environments. Eur.J. Biochem. 2004, 242, 372386.(28) Heinig, M.; Frishman, D. STRIDE: awebserverforsecondarystructure assignment from known atomic coordinates of proteins. NucleicAcids Res. 2004, 32, 500502.(29) Pearlman, D. A.; Case, D. A.; Caldwell, J. W.; Ross, W. S.;Cheatham, T. E.; Debolt, S.; Ferguson, D.; Seibel, G.; Kollman, P. Amber,a package of computer-programs for applying molecular mechanics, normal-mode analysis, molecular-dynamics and free-energy calculations to simulatethe structural and energetic properties of molecules. Comput. Phys. Commun.1995, 91, 141.(30) MacKerell, A. D. Molecular dynamicssimulationanalysisof asodium dodecyl-sulfate micelle in aqueous solutionsDecreased uidity ofthe micelle hydrocarbon interior. J. Phys. Chem. 1995, 99, 18461855.(31) Croonen,Y.;Gelade,E.;Vanderzegel,M.;Vanderauweraer,M.;Vandendriessche, H.; Deschryver, F. C.; Almgren, M. Inuenceofsalt,detergent concentration, and temperature on the uorescence quenching of1-methylpyreneinsodiumdodecyl-sulfatewithmeta-dicyanobenzene. J.Phys. Chem. 1983, 87, 14261431.(32) Schuttelkopf, A. W.; van Aalten, D. M. PRODRG-a tool for high-throughput crystallography of protein-ligand complexes. Acta Crystallogr.2004, D 60, 13551363.(33) Roux, B. Thecalculationof thepotential of meanforceusingcomputer-simulations. Comput. Phys. Commun. 1995, 91, 275282.(34) Kumar, S.; Bouzida, D.; Swendsen, R. H.; Kollman, P. A.;Rosenberg, J. M. The weighted histogram analysis method for free-energycalculations on biomolecules.I. The method. J. Comput. Chem. 1992, 13,10111021.(35) Efron, B. Bootstrap methods: another look at the jackknife.Ann.Stat. 1979, 7, 126.(36) Lensink,M.F.;Christiaens,B.;Vandekerckhove,J.;Prochiantz,A.; Rosseneu, M. Penetratin-membrane association: W48/R52/W56 shieldthe peptide from the aqueous phase. Biophys. J. 2005, 88, 932952.(37) Schiffer, M.; Chang, C. H.; Stevens, F. J. The functions oftryptophan residues in membrane proteins. Protein Eng. 1992, 5, 213214.(38) Magzoub, M.; Eriksson, L. E. G.; Graslund, A. Comparison of theinteraction, positioning, structure induction and membrane perturbation ofcell-penetrating peptides and non-translocating variants with phospholipidvesicles. Biophys. Chem. 2003, 103, 271288.(39) Rinaldi, A. C.;DiGiulio, A.;Liberi, M.;Gualtieri, G.;Oratore,A.; Schinina, M. E.; Simmaco, M.; Bozzi, A. Effects of temporins onmoleculardynamicsandmembranepermeabilizationinlipidvesicles. J.Peptide Res. 2001, 58, 213220.JP1069362Biophysical Properties of Membrane-Active Peptides J. Phys. Chem. B, Vol. 114, No. 43, 2010 13735