DOI: 10.1021/la902660q 12235Langmuir 2009, 25(20), 12235–12242 Published on Web 09/16/2009

pubs.acs.org/Langmuir

© 2009 American Chemical Society

Cause and Effect of Melittin-Induced Pore Formation: A ComputationalApproach

Moutusi Manna and Chaitali Mukhopadhyay*

Department of Chemistry, University of Calcutta, 92, A. P. C. Road, Kolkata-700 009, India

Received February 4, 2009

Melittin embedded in a palmitoyl oleyl phosphatidylcholine bilayer at a high peptide/lipid ratio (1:30) was simulatedin the presence of explicit water and ions. The simulation results indicate the incipience of an ion-permeable water porethrough collectivemembrane perturbation by bound peptides. The positively charged residues ofmelittin not only act as“anchors” but also disrupt themembrane, leading to cell lysis. A detailed analysis of the lipid tail order parameter profiledepicts localized membrane perturbation. The lipids in the vicinity of the aqueous cavity adopt a tilted conformation,which allows local bilayer thinning. The prepore thus formed can be considered as themelittin-induced structural defectsin the bilayer membrane. Because of the strong cationic nature, the melittin-induced prepore exhibits selectivity towardanions over cations. As Cl- ions entered into the prepore, they are electrostatically entrapped by positively chargedresidues located at its wall. The confined motion of the Cl- ions in the membrane interior is obvious from calculateddiffusion coefficients. Moreover, reorientation of the local lipids occurs in such a way that few lipid heads along withpeptide helices can line the surface of the penetrating aqueous phase. The flipping of lipids argued in favor of melittin-induced toroidal pore over a barrel-stave mechanism. Thus, our result provides atomistic level details of the mechanismof membrane disruption by antimicrobial peptide melittin.

Introduction

A large group of membrane-active peptides, such as antimicro-bial peptides and toxins, are known to cause membrane damage,which galvanizes cell death. Such peptides are often used bynature in the defensive and offensive systems all across the plantand animal kingdoms. Antimicrobial peptides (AMPs), which areconsidered as “the native line of defense throughout nature”,show a high toxicity against both Gram-positive and Gram-negative bacteria as well as fungi, viruses, and mycobacteria,etc.1-4 Research in this field is of growing interest, as AMPs canact as potential alternatives to conventional antibiotics.AMPs aretypically small (∼10-50 residues), cationic, amphipathic peptides,known to permeate microbial cell walls, thus inducing leakage ofthe cellular components across the bilayer.5-11 Among the differ-entmechanisms proposed so far, themode of action ofAMPs canbe explained either by the formation of transmembrane (TM) ionpermeable pores (barrel-stave or toroidal pore model) or by the

surface binding of peptides (carpetmechanism) in a detergent-likemanner.5 In the barrel-stavemodel, the peptides span a nearly flatbilayer and aggregate to line the pore surface,10-12 whereas, in thetoroidal pore model, peptides perturb the local bilayer structure.The lipids in each leaflet sharply bend so that the lipid headgroups along with peptide helices line the aqueous channel.8,12,13

On the other hand, in the carpet mechanism, the peptides form acarpetlike monolayer on the bilayer surface, driven predomi-nantly by electrostatic interactions.8,14

Melittin is a naturally occurring AMP with pronouncedcytolytic potency. It is the principal toxic component of theEuropean honeybee venom, Apis mellifera.15 It is a highly basic,amphipathic hexacosa peptide (GIGAVLKVLTTGLPALISW-IKRKRQQ-CONH2), with a large hydrophobic region (residues1-20) and a stretch of predominantly hydrophilic amino acids(residues 21-26) at the carboxy terminal. Because of poor cellselectivity, it exhibits strong hemolytic activity against bothbacterial and mammalian cells.6,13,16-19 At a moderately highconcentration, melittin is known to cause micellization as well asmembrane fusion,20,21 in addition to voltage-dependent ionchannel formation across the planar lipid bilayer.22-24

*To whom correspondence should be addressed. Tel: 91-33-2350-8386.Fax: 91-33-2351-9755. E-mail: chaitalicu@yahoo.com or cmchem@caluniv.ac.in.(1) Chromek,M.; Slamov�a, Z.; Bergman, P.; Kov�acs, L.; Podrack�a, L; Ehr�en, I.;

H€okfelt, T.; Gudmundsson, G. H.; Gallo, R. L.; Agerberth, B.; Brauner, A. Nat.Med. 2006, 12, 636–641.(2) Hancock, R. E.; Scott, M. G. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 8856–

8861.(3) Li, M.; Lai, Y.; Villaruz, A. E.; Cha, D. J.; Sturdevant, D. E.; Otto, M. Proc.

Natl. Acad. Sci. 2007, 104, 9469–9474.(4) Zasloff, M. Nature 2002, 415, 389–395.(5) Brogden, K. A. Nat. Rev. Microbiol. 2005, 3, 238–250.(6) Lee, M.-T.; Hung, W.-C.; Chen, F.-Y.; Huang, H. W. Proc. Natl. Acad. Sci.

2008, 105, 5087–5092.(7) Dubovskii, P. V.; Volynsky, P. E.; Polyansky, A. A.; Karpunin, D. V.;

Chupin, V. V.; Efremov, R. G.; Arseniev, A. S. Biochemistry 2008, 47, 3525–3533.(8) Bond, P. J.; Parton, D. L.; Clark, J. F.; Sansom,M. S. P.Biophys. J. 2008, 95,

3802–3815.(9) Bringezu, F.; Wen, S.; Dante, S.; Hauss, T.; Majerowicz, M.; Waring, A.

Biochemistry 2007, 46, 5678–5686.(10) Langham, A. A.; Ahmad, A. S.; Kaznessis, Y. N. J. Am. Chem. Soc. 2008,

130, 4338–4346.(11) S�anchez-Martınez, S.; Huarte, N.; Maeso, R.; Madan, V.; Carrasco, L.;

Nieva, J. L. Biochemistry 2008, 47, 10731–10739.

(12) Yang, L.; Harroun, T. A.;Weiss, T.M.; Ding, L.; Huang, H.W.Biophys. J.2001, 81, 1475–1485.

(13) Allende, D.; Simon, S. A.; McIntosh, T. J. Biophys. J. 2005, 88, 1828–1837.(14) Papo, N.; Shai, Y. Biochemistry 2003, 42, 458–466.(15) Habermann, E. Science 1972, 177, 314–332.(16) Asthana, N.; Yadav, S. P.; Ghosh, J. K. J. Biol. Chem. 2004, 279, 55042–

55050.(17) Sharon, M.; Oren, Z.; Shai, Y.; Anglister, J. Biochemistry 1999, 38, 15305–

15316.(18) Lee, M.-T.; Chen, F.-Y.; Huang, H. W. Biochemistry 2004, 43, 3590–3599.(19) Matsuzaki, K.; Yoneyama, S.; Miyajima, K. Biophys. J. 1997, 73, 831–838.(20) Toraya, S.; Nagao, T.; Norisada, K.; Tuzi, S.; Saito, H.; Izumi, S.; Naito, A.

Biophys. J. 2005, 89, 3214–3222.(21) Naito, A.; Nagao, T.; Norisada, K.;Mizuno, T.; Tuzi, S.; Saito, H.Biophys.

J. 2000, 78, 2405–2417.(22) Tosteson, M. T.; Tosteson, D. C. Biophys. J. 1981, 36, 109–116.(23) Stankowski, S.; Pawlak, M.; Kaisheva, E.; Robert, C. H.; Schwarz, G.

Biochim. Biophys. Acta 1991, 1069, 77–86.(24) Becucci, L.; Guidelli, R. Langmuir 2007, 23, 5601–5608.

12236 DOI: 10.1021/la902660q Langmuir 2009, 25(20), 12235–12242

Article Manna and Mukhopadhyay

Melittin is mostly disordered as free monomer in solution;however, at high ionic strength, pH, or peptide concentration,monomers self-associate to form an R-helical tetramericaggregate.25-27 The tetramer is too stable, with a buried hydro-phobic and exposed hydrophilic face, to be inserted into thebilayer.28 In a lipidic environment,melittin adopts a bentR-helicalconformation (small hinge at Pro14) with segregated hydropho-bic and hydrophilic faces.29,30 The association of melittin withphospholipid bilayer initiates the lytic mechanism. Melittin isthought to lyse the membrane by disrupting the bilayer barrierproperty.31 Several studies reveal that the interaction is sensitiveto peptide concentration,6 lipid composition,13,32-34 ionicstrength,35 hydration level,36 and membrane potential.24 Theorientation of melittin helix inside the lipid bilayer also plays acrucial role in melittin-induced cell lysis.6 Depending upon the“physicochemical” conditions, melittin can either lie laterallyacross the membrane (“Wedge” model) or insert itself parallelto lipid normal.12 The two binding states can be interconvertedunder equilibrium conditions, which is directly related to lyticactivity. The previous experimental studies proposed that at apeptide concentration higher than a certain threshold concentra-tion, melittin changes from a surface-associated state to an insertedstate.18 Again, with protonation of the N-terminus, melittinadopts a transbilayer orientation.37 However, a recent articlededuced a more complicated oriental distribution of melittin.38

According to them, about three-fourths of melittin moleculesorient parallel to the bilayer surface with a slight tilt, while the restorient almost parallel to surface normal. Although there is noconsensus regarding its orientation, it is well-established thatmelittin can form a TM pore only in an inserted state, and inparallel orientations, peptides cause bilayer thinning in propor-tion to peptide concentration.6

Although a large number of studies have been undertaken todetermine the architecture of the pore induced byAMP, the pore-forming mechanism of melittin still remains controversial. Somestudies suggest the formation of a barrel-stave or toroidal-shapedpore, while others envisage a carpetmechanism.12-14,39,40Despitethe ambiguity, themelittin pore is thought to be due to the defectsin the bilayer, produced by the collective perturbation of amembrane structure by bound peptides.6,13 To gain betterinsight into the relevant lytic mechanism, we need atomisticlevel resolution of specific lipid-peptide interactions, details ofperturbations produced in the membrane bilayer, and also

environmental responses of surrounding ions and the aqueousmedium.

The objective of the presentwork is to investigate themoleculardetails of the pore-forming mechanism by venom toxin melittin,which is still in chaos. Baumgaertner et al.39 earlier investigatedthe stability of a hypotheticalmelittin pore consisting of amelittintetramer in a membrane bilayer. They found that as the poreexpanded, the initial tetrameric configuration decayed into astable trimer and one monomer. Now, instead of immersing apore structure, here, we have added four melittin monomers ina hydrated palmitoyl oleyl phosphatidylcholine (POPC) bilayer.In our system, the peptide to lipid (P/L) ratio, 1:30, is well abovethe critical peptide to lipid ratio of 1:62.18Melittins are inserted ina transbilayer orientation with their hydrophilic C-terminalprotruding out of the membrane. As revealed by previousexperiments, the conditions that we chose are necessary for thepore formation by cytolytic peptide melittin. Within the simula-tion time scale, we observe the formation of a prepore in the hostbilayer accomplished by a severe membrane perturbation. Theevidence that we have from the simulation shows preference forthe toroidal pore model over the barrel-stave mechanism. Wehave also noticed the movement of the chloride ion inside theprepore, which is complementary to the experimental finding ofanion selectivity of the melittin channel.22 To the best of ourknowledge, this is probably the first computational approach thatreports such spontaneous initiation of ion-permeable toroidal-shaped pore by hemolytic melittin, at a high peptide concentra-tion. Few additional simulations were run to study the influenceof the initial peptide position and the effect of the peptide chargestate in cell lysis (Supporting Information). When melittins aremore deeply inserted in the membrane, so their N-terminals reachnearly the lower interfacial region, we again observe the forma-tion of a water defect in the inner bilayer leaflet. The highly basicpeptide causes serious bilayer deformation, opening up a regionthroughwhichwater and ions can penetrate the bilayer.However,when they were placed further away from each other, eachisolated melittin individually exhibited cell lysis. Again, if Lys-7,which is exposed in the hydrophobic interior is made uncharged,we observe a drastic drop in water penetration, which indicatesthe importance of this specific amino acid residue or, broadly, thepeptide charge state in melittin-induced lysis. These results pointtoward the fact that the presence of highly basic peptides perturbsthe host bilayer, which ultimately leads to bilayer disruption.

Simulation Details

All molecular dynamics simulations were performed with theprogram CHARMM (Chemistry at Harvard MacromolecularMechanics)41 using the CHARMM27 parameter set,42 includingdihedral cross-term corrections (CMAP)43 for peptides andmodified TIP3 water models.44 The long-range electrostaticinteractions were treated via a Particle Mesh Ewald (PME)method using a Gaussian distribution width of k=0.34 A-1, areal space cutoff of 12 A, and fft grid points of 1 A in all directions,and a fifth order β-spline was used for the interpolation. TheLennard-Jones (LJ) potential was smoothly switched off bet-ween 10 and 12 A. The temperature and pressure of the sys-tem were kept constant at 300 K and 1 atm, respectively. The

(25) Terwilliger, T. C.; Eisenberg, D. J. Biol. Chem. 1982, 257, 6016–6022.(26) Hartings, M. R.; Gray, H. B.; Winkler, J. R. J. Phys. Chem. B 2008, 112,

3202–3207.(27) Qiu, W.; Zhang, L.; Kao, Y.-T.; Lu, W.; Li, T.; Kim, J.; Sollenberger,

G. M.; Wang, L.; Zhong, D. J. Phys. Chem. B 2005, 109, 16901–16910.(28) Hua, L.; Huang, X.; Liu, P.; Zhou, R.; Berne, B. J. J. Phys. Chem. B 2007,

111, 9069–9077.(29) Chatterjee, C.; Mukhopadhyay, C. Biochem. Biophys. Res. Commun. 2002,

292, 579–585.(30) Smith, R.; Separovic, F.; Milne, T. J.; Whittaker, A.; Bennett, F. M.;

Cornell, B. A.; Makriyannis, A. J. Mol. Biol. 1994, 241, 456–466.(31) Epand, R. M.; Vogel, H. J. Biochim. Biophys. Acta 1999, 1462, 11–28.(32) Raghuraman, H.; Chattopadhyay, A. Eur. Biophys. J. 2004, 33, 611–622.(33) Alakoskela, J.-M.; Sabatini, K.; Jiang, X.; Laitala, V.; Covey, D. F.;

Kinnunen, P. K. J. Langmuir 2008, 24, 830–836.(34) Wessman, P.; Str€omstedt, A. A.; Malmsten, M.; Edwards, K. Biophys. J.

2008, 95, 4324–4336.(35) Raghuraman, H.; Ganguly, S.; Chattopadhyay, A. Biophys. Chem. 2006,

124, 115–124.(36) Raghuraman, H.; Chattopadhyay, A. Langmuir 2003, 19, 10332–10341.(37) Bradshaw, J. P.; Dempsey, C. E.; Watts, A. Mol. Membr. Biol. 1994, 11,

79–86.(38) Chen, X.; Wang, J.; Boughton, A. P.; Kristalyn, C. B.; Chen, Z. J. Am.

Chem. Soc. 2007, 129, 1420–1427.(39) Lin, J.-H.; Baumgaertner, A. Biophys. J. 2000, 78, 1714–1724.(40) Ladokhin, A. S.; White, S. H. Biochim. Biophys. Acta, Biomembr. 2001,

1514, 253–260.

(41) Brooks, B.R.; Bruccoleri,R.E.;Olafson,B.D.; States,D. J.; Swaminathan, S.;Karplus, M. J. Comput. Chem. 1983, 4, 187–217.

(42) MacKerell, A. D. J. Phys. Chem. B 1998, 102, 3586–3616.(43) MacKerell, A.D.; Feig,M.; Brooks, C. L. J. Comput. Chem. 2004, 25, 1400–

1415.(44) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein,

M. L. J. Chem. Phys. 1983, 79, 926–935.

DOI: 10.1021/la902660q 12237Langmuir 2009, 25(20), 12235–12242

Manna and Mukhopadhyay Article

temperature was controlled by a Hoover thermostat with acoupling constant of 20000 kcal mol-1 ps-2 and pressure with apressure piston mass of 2000 amu. Periodic boundary conditions(PBC) were applied in all three directions using a tetragonallattice. The leapfrog Verlet integrator with a time step of 1 fs wasused to solve Newton’s equations of motion. Nonbonded andimage lists were updated every 20 integration steps. The overallrotational and translational motions of the system were removedevery 500 integration steps. All covalent bonds involving hydro-gen were constrained with the SHAKE algorithm.Pure POPC Bilayer. The model bilayer containing 128

POPC lipids (64 lipids in each leaflet) was constructed followingthe method as reported in refs 45 and 46. The starting configura-tion for a phospholipid system was prepared from randomselection of lipids from a pre-equilibrated, prehydrated set andthen placing them in a bilayer. The short contacts between heavyatoms were reduced through systematic rotations (around thez-axis) and translations (in the xy plane) of the lipids. The bilayerwas then solvated, and the overlappingwaters were deletedwithin2.6 A from the lipid molecules. The simulation cell dimension wasinitially set to allow 64 A2 area per lipid. Short steepest descent(SD) minimizations, followed by long adapted basis Newton-Raphson (ABNR) minimization, were performed to remove badatom contacts. The pure bilayer was then simulated with NPTensemble, as it is the most natural choice of an ensemble inmembrane simulation. As the POPC lipid has a low gel-to-fluidtransition temperature (Tm ≈ -5 �C or 268 K),47,48 simulationswere performed at 300 K to maintain bilayer fluidity.

After 10 ns ofNPT simulation, the bilayer shrinks considerablyalong the X-Y plane, resulting in an area per lipid of ∼56 A2,which is significantly smaller than the experimentally reportedvalue of 68.3 A2 (for an area per lipid vs time plot, see theSupporting Information,Figure S1).49As a consequence, the lipidtails became highly ordered (Figure S2 of the Supporting In-formation), indicating a liquid-to-gel phase transition of lipidbilayer. A similar trend of the decreasing area per lipid, usingCHARMM NPT simulation protocol, had been reported pre-viously in the literature.50 To alleviate this problem of CHARMMbilayer simulation, NPAT (constant number of atoms, pressure,cross-sectional area, and temperature) or NγPT (constant num-ber of atoms, pressure, surface tension, and temperature) ensem-bles are typically recommended, although NγPT is preferred forsimulating membrane embedded proteins/peptides.51 In the pre-sent simulation, to expand the area per lipid to its desired value,we then switched to theNγPT ensemble,45,52 with γ=50 dyn/cm.After the desired area per lipid was reached, the final coordinateset was then taken to initiate the additional NγPT simulation, atγ=20 dyn/cm. The choice of the optimal γ value was based onrecent simulation studies of the liquid DPPC membrane, wherethe surface tensions of 20 and 24.5 dyn/cm gave the best agree-ment with the known area per lipid.52,53 After a 10 ns productionrun, the simulation box dimensions became 65.02�65.01�81 A3,

with an area per lipid (Figure S3 of the Supporting Information)of ÆAæ= 66 ( 0.5 A2. The result agrees well with the area perPOPC, 63.5 A2 at 310 K and 66.5 A2 at 303 K, obtained fromrecent simulation studies (using AMBER and GROMACS,respectively).54,55

Combined System. The melittin atomic coordinate wasretrieved from the Protein Data Bank (PDB code: 2mlt). Fourmelittin monomers (each adjacent pair is separated by two lipidmolecules) were inserted vertically into the pre-equilibrated (finalcoordinate of 10 ns NγPT run) bilayer (Figure 1a), following theprocedure asdescribed in refs 56and 57. In thepresent simulation,melittin spans only ∼2/3 of the complete bilayer width; thus, theN-terminal remains free into the low-density tail region of thelower layer. For each melittin, two lipid molecules were removedonly from the upper leaflet, resulting in a P/L ratio of 1:30. Alloverlapping water molecules within 2.6 A of the peptide wereremoved. On the basis of the theoretical study56 and experimentalobservation,58 Trp19 was placed near the carbonyl group, whichis considered to be its highly energetically favorable interactionsite. As a result, the charged C-terminal moiety was nicelypositioned into the lipid/water interface, inserting the predomi-nately hydrophobic N-terminal segment into the bilayer hydro-phobic core. Two charged residues of the TM segment, that is,Lys7 and the protonatedN-terminus of all the fourmelittins, wereso oriented that they face each other and also to the center ofbilayer. To maintain electrical neutrality of the system, 4�6 Cl-

counterions were added and additional sodium chloride wasadded to produce the physical salt concentration of 150 mM.Then, a series of minimizations were performed to relax thesystem.52 First, the system was subjected to SD algorithm withfixed constraints on the peptide followed byABNRminimizationwith harmonic constraints applied on the peptide backbone.Finally, peptides were released, and the fully unconstrainedsystemwas then subjected to energyminimization (SD), followedby a production run, 15 ns long, using the NγPT ensemble, withγ=20 dyn/cm.59 A few control simulations were performed tostudy the effects of the initial peptide position and the peptidecharge state on cell lysis (see the Supporting Information), usingthe same simulation protocol. A list of total simulations per-formed is given in the Supporting Information (Table S1.)

Results and Discussion

Binding of melittin causes sizable disruption of the host POPCbilayer, leading to the initiation of an ion-permeable water pore.In this section, we focus on peptide conformation, details ofpeptide-membrane interaction, differential effects of the peptideon the membrane structure and dynamics, membrane perme-ability for water and ions, and the architecture of the preporeformed.Peptide Conformation. Peptide conformational studies are

necessary to gauge the influence of the lipid matrix on thestructure and dynamics of the bound peptide. In the presentsimulation, the conformation of melittin remained stable over theentire simulation time scale. Thus, in a bilayer, melittin retains itsR-helical structure (Figure S4 of the Supporting Information),with a time average bend angle of ÆΩæ ≈ 145 ( 10� (Figure S5 of

(45) Skibinsky, A.; Venable, R. M.; Pastor, R. W. Biophys. J. 2005, 89, 4111–4121.(46) Mondal, S.; Mukhopadhyay, C. Langmuir 2008, 24, 10298–10305.(47) Litman, B. J.; Lewis, E. N.; Levin, I. W. Biochemistry 1991, 30, 313–319.(48) Leekumjorn, S.; Sum, A. K. J. Phys. Chem. B 2007, 111, 6026–6033.(49) Kucerka, N.; Tristram-Nagle, S.; Nagle, J. F. J. Membr. Biol. 2005, 208,

193–202.(50) Jensen, M. Ø.; Mouritsen, O. G.; Peters, G. H. Biophys. J. 2004, 86, 3556–

3575.(51) Feller, S. E.; Zhang, Y.; Pastor, R. W. J. Chem. Phys. 1995, 103, 10267–

10276.(52) Dolan, E. A.; Venable, R. M.; Pastor, R. W.; Brooks, B. R. Biophys. J.

2002, 82, 2317–2325.(53) Cournia, Z.; Ullmann, G. M.; Smith, J. C. J. Phys. Chem. B 2007, 111,

1786–1801.

(54) R�og, T.; Murzyn, K.; Pasenkiewicz-Gierula, M. Acta Biochim. Pol. 2003,50, 789–798.

(55) Pandit, S. A.; Chiu, S.-W.; Jakobsson, E.; Grama, A.; Scott, H. L. Biophys.J. 2007, 92, 920–927.

(56) Bachar, M.; Becker, O. M. Biophys. J. 2000, 78, 1359–1375.(57) Kandt, C.; Ash, W. L.; Tieleman, D. P. Methods 2007, 41, 475–488.(58) Vogel, H.; J€ahnig, F. Biophys. J. 1986, 50, 573–582.(59) Gullingsrud, J.; Babakhani, A.; McCammon, J. A. Mol. Simul. 2006, 32,

831–838.

12238 DOI: 10.1021/la902660q Langmuir 2009, 25(20), 12235–12242

Article Manna and Mukhopadhyay

the Supporting Information). A very similar bend angle of ÆΩæ≈134( 20�was reported by Baumgaertner et al. for a melittin poreimmersed in the POPC bilayer.39 On the basis of a solid-stateNMR study, Naito et al.21 proposed that the kink angle betweenthe N- and the C-terminal helical rods of melittin in the lipidbilayer is ∼140 or ∼160�. The detail of the conformational driftwas obtained by computing the root-mean-square deviation(rmsd) for CR atoms of the peptide backbone relative to thestarting point (Figure S6 of the Supporting Information).60,61 Foreach peptide, the rmsd rises initially and closely approachesconvergence after about 8 ns. The rmsd values range from 2.7 to3.5 A (last 7 ns), indicating a small structural fluctuation relativeto the starting point, that is, stability of the helical structure in thebilayer environment.Solvation of Peptide: Peptide-Lipid and Peptide-Water

Interactions. Lipid-protein interactions are crucial for manycellular processes, includingmembrane trafficking, transport, andsignal transductions.62 Here, we have calculated the interactionenergies of melittin with surrounding lipid and water moleculesand also the partition of the total interaction energy into electro-static and van der Waals (vdW) terms (Figure 2), as in refs 63and 64. In the distribution curve for the peptide-lipid interaction(Figure 2a), a peak for the electrostatic contribution permelittin islocated at ∼-190 kcal/mol, whereas that for the vdW contribu-tion is located at∼-160 kcal/mol. The result is in close proximitywith results obtained for the interaction of single PG1 peptidewith a mixed POPC/POPG bilayer.63 The source of the columbicinteraction is hydrogen bonding between hydrophilic amino acidresidues of melittin with negatively charged lipid head groups,whereas the vdW contribution is the outcome of a nonbonded

interaction between the hydrophobic TM segment ofmelittin andthe lipid acyl tails. These interactions lead to the formation of asupramolecular lipid-peptide complex.

For a more focused examination of environment response tomelittin, we further calculated the number of contacts and also thenumber of hydrogenbonds ofpeptideswith surrounding lipid andsolvent molecules, as described in the literature.60,65-68 First, wecalculated the average number ofwater, phosphate, choline, ester,and acyl groupheavy atomswithin 4 A of the peptide heavy atoms(Figure 3), to understand themicroenvironment ofmelittin.65 Thepositively charged residues Lys21-Arg22-Lys23-Arg24 of the C-ter-minal region, nicely positioned at the energetically favorablewater/lipid interface, are highly solvated by water and negatively

Figure 1. (a) Initial and (b) final snapshot of the system, with melittin in magenta, POPC in silver, and water in green. The headgroupphosphate atoms and chloride ions are highlighted as silver and red van derWaals spheres, respectively. The image rendering was done withVMD.87 The figure represents considerable water penetration followed by reorientation of lipids, especially from the lower leaflet. Fewchloride ions are shown to enter into the membrane core, along with water.

Figure 2. Probability distribution of the interaction energy ofmelittin (per monomer) with surrounding (a) lipid and (b) watermolecules. The total interaction energy is partitioned into theelectrostatic and vdW terms.

(60) Psachoulia, E.; Sansom, M. S. P. Biochemistry 2008, 47, 4211–4220.(61) Jard�on-Valadez, E,; Ulloa-Aguirre, A.; Pi~neiro, �A. J. Phys. Chem. B 2008,

112, 10704–10713.(62) Cho, W.; Stahelin, R. V.Annu. Rev. Biophys. Biomol. Struct. 2005, 34, 119–

151.(63) Jang, H.; Ma, B.; Woolf, T. B.; Nussinov, R. Biophys. J. 2006, 91, 2848–

2859.(64) Cordomı, A.; Perez, J. J. J. Phys. Chem. B 2007, 111, 7052–7063.

(65) Gorfe, A. A.; Babakhani, A.; McCammon, J. A. J. Am. Chem. Soc. 2007,129, 12280–12286.

(66) Lorenz, C. D.; Faraudo, J.; Travesset, A. Langmuir 2008, 24, 1654–1658.(67) Wee, C. L.; Balali-Mood, K.; Gavaghan, D.; Sansom, M. S. P. Biophys. J.

2008, 95, 1649–1657.(68) Deol, S. S.; Bond, P. J.; Domene, C.; Sansom,M. S. P. Biophys. J. 2004, 87,

3737–3749.

DOI: 10.1021/la902660q 12239Langmuir 2009, 25(20), 12235–12242

Manna and Mukhopadhyay Article

charged lipid head groups. The polar Ser18-Trp19 residues exhibita more complex solvation pattern. They are exposed to bothnonpolar side chains and water by expelling their hydroxyl andindole side chains into the aqueous phase. This also indicates thatthe bound peptides enhance the water penetration into thehydrophilic-hydrophobic interfacial region of the phsopholipidbilayer. As expected, the hydrophobic TM segment exhibitsa large number of contacts with the lipid tails. However, fewN-terminal residues, especially Lys7 and protonated N-terminalGly1, are shown to be highly solvated by water molecules. Theapparent discrepancy arises due to the unfavorable partitioning ofpositively charged residues into the lipid tail region. The systemthen undergoes necessary adjustments and drags water by dis-rupting membranes. As explained by Becker,69 the dipole asso-ciated with Lys7 acts as an electrostatic “beacon”, steering waterpenetration from the extracellular side of the membrane. Rouxet al. proposed a very similar solvation pattern of melittin.70 Ascompared to our result, they observed a lesser extent of wateraround Lys7, in accord with the parallel orientation of melittin inthe plane of the outer leaflet.

To elucidate the electrostatic interaction,we then calculated thenumber of hydrogen bonds (H-bond) shared by peptide residues.The criteria that we chose for H-bonding are as follows: Thehydrogen-acceptor distance was e2.5 A, and the donor-hydrogen-acceptor angle was e60�.71 As depicted in Figure 4,the C-terminal charged residues tightly interact with the head-group phosphate and “lock” the peptide into its trans-membraneorientationwithin the bilayer, whereas themembrane-buriedLys7

and charged N-terminal show a high H-bonding propensity forwater oxygen. The N-terminal hydrophobic Ile2-Gly3-Ala4 re-sidues exhibit dual contribution, H-bonding with water via abackbone amide nitrogen (resulting in the reduction in the helicalcontent of these residues, as reflected from the secondary struc-ture analysis, Figure S4 of the Supporting Information) andnonbonded interaction via hydrophobic side chains. For visualperception, Figure S7 of the Supporting Information representsthe snapshots of few peptide-lipid supramolecular complexes.Figures 3 and 4 exhibit a large number of water penetration in thebilayer hydrophobic core, due to melittin-induced cell lysis,12,69

and also indicate the role of specific amino acid residues inmembrane disruption.Membrane-Perturbing Effects of Melittin. Lipid Tail

Order. The perturbation produced by melittin on the averagestructure of the bilayer interior can be characterized using theC-H bond orientational order parameter, SCH

SCH ¼ 0:5Æ3 cos2 θ-1æ

where θ is the angle between the C-H bond vector and themembrane normal and the angular brackets indicate averagingover time and over lipids. To investigate the effect of the preporeon local lipid perturbation, lipids are divided into two categories:local (those are within 10 A of peptides) and bulk (all others)lipids. The marked asymmetry of the system is reflected in theorder parameters profile of the acyl chains of the two bilayerleaflets (Figure 5).

In the case of a lower layer, a drastic drop inSCH values of locallipids indicates a localized but strong bilayer perturbation by theloosely bound melittin N-terminals. As compared to the purebilayer, bulk lipids in the lower layer exhibit a striking increase inorder parameter values. The local bilayer disorder is alsoobservedfor the upper layer lipids but to a lesser extent. The SCH value ofpure POPC bilayer agrees well with the previous findings.48,63

Another interesting observation is that the lipid tail ends becomemore ordered in the presence of peptide. The earlier theoreticalstudies indicate a very similar membrane perturbing effect bymelittin56 and other membrane-active peptides.50,63,65

Thus, our result suggests that melittin strongly influences thestructure and dynamics of the phospholipid membrane. In ourcase, the extracellular layer is less affected by the embeddedpeptide, whereas there is an increased level of disorder andstructural deformation of lower-layer phospholipids in the im-mediate vicinity of the peptide.As a result, cell lysis starts from theinner side of the bilayer. In a recent article, Dufourc et al.72 haveinvestigated the pore formation by an AMP cateslytin on theDMPC bilayer. The association of the charged peptide with theouter leaflet induces an electric field (ca. 0.1 V/nm, on the basis ofelectrophysiological measurement) inside the membrane, roughlyperpendicular to the average plane of the bilayer. Because of theasymmetric charge distribution, the water defect initiates onlyfrom the inner bilayer leaflet.

Membrane Thinning and Lipid Orientation. To study theimpact of melittin on bilayer thickness, we plot the ensemble

Figure 3. Contribution from the main components of the mem-brane system (water oxygens, choline, phosphate, and ester polarheadgroups and acyl chain carbons) to the solvation of theindividual residues of melittin (per monomer). The data in thisfigure are averaged over the last 7 ns.

Figure 4. Time-averaged melittin-POPC and melittin-waterH-bonds. In this figure, the last 7 ns of data has been used.

(69) Bachar, M.; Becker, O. M. J. Chem. Phys. 1999, 111, 8672–8685.(70) Bern�eche, S.; Nina, M.; Roux, B. Biophys. J. 1998, 75, 1603–1618.(71) Zhao, W.; R�og, T.; Gurtovenko, A. A.; Vattulainen, I.; Karttunen, M.

Biophys. J. 2007, 92, 1114–1124.(72) Jean-Franc-ois, F.; Elezgaray, J.; Berson, P.; Vacher, P.; Dufourc, E. J.

Biophys. J. 2008, 95, 5748–5756.

12240 DOI: 10.1021/la902660q Langmuir 2009, 25(20), 12235–12242

Article Manna and Mukhopadhyay

average (last 7 ns) position of 120 PO4- groups along the bilayer

normal, as a function of radial distance from the center of theprepore (Figure 6) (which is defined as the center of the penetrat-ing aqueous phase in the bilayer hydrophobic core, bound by fourmelittins). The peptides induce an asymmetry in the POPC head-group’s distribution. In the case of the upper layer, no significantchange in surface corrugation is observed. However, in the lowerlayer, the loosely bound N-terminals from melittin monomersperturb the regular lamellar structure of the lipid head groups.Membrane thinning is apparent from the thickness around∼30 Aat the prepore, as compared to its original value of ∼38 A awayfrom the prepore (the bilayer thickness of the pure POPC bilayeris ∼37.5 A; Figure S8 of the Supporting Information). The localbilayer thinning by melittin6,56,70 and other membrane-activeproteins/peptides10,63 have been reported previously by severalgroups.

As depicted in Figure 7, lipids closer to the prepore adopted amore tilted conformation. A considerable reorientation takesplace for lower-layer lipids. Few lipids in the lower layer changedthe orientation of their acyl chains with respect to the membranenormal from parallel to perpendicular. The superposition of lipidconformations, Figure S9 of the Supporting Information, sup-ports this fact. The reorientation allows the hydrophobic lipidtails to avoid being in contact with the penetrating aqueous phase.The flipped lipid heads are now translocated from the membranesurface to the surface of the aqueous channel. The rapid flip-flopof membrane lipids near melittin pore was demonstratedearlier.39,73 Such bending of bilayer leaflets is responsible for thelocalized membrane thinning. It has been earlier suggested thatthe presence of the peptide in a PCbilayer should alter the averageorientation of the headgroup dipoles.74,75 Although in the presentsimulation, the PC headgroup of upper layer lipids remain

roughly parallel to the membrane surface,48 some extent ofperturbation of the polar lipid head in the lower leaflet is obviousfrom a broader headgroup distribution (Figure S10 of theSupporting Information).Membrane Permeability. Water Penetration. Because

TM pore formation is central to many biological processes, it isan area of intense research.76-80 Recent theoretical studies haveexamined the water penetration through bilayer deformation byhighly charged peptides.81 In the present simulation, we havealready examined an enormous amount of water penetration intothe membrane core. The extent of water penetration can bevisualized from the atom density distribution along the bilayernormal.82 Figure S11 of the Supporting Information shows the

Figure 5. Order parameters, SCH, for the palmitoyl (left column) and oleoyl (right column) chains for the different lipid categories, for theupper (upper panel) and the lower (lower panel) layer lipids and that of the pure POPC bilayer.

Figure 6. Time average positions of PO4- head groups along the

bilayer normal as a function of distance from the prepore.

(73) Fattal, E.; Nir, S.; Parente, R. A.; Szoka, F. C., Jr. Biochemistry 1994, 33,6721–6731.(74) Shepherd, C.M.; Schaus, K. A.; Vogel, H. J.; Juffer, A. H. Biophys. J. 2001,

80, 579–596.(75) Kuchinka, E.; Seelig, J. Biochemistry 1989, 28, 4216–4221.

(76) Thøgersen, L.; Schiøtt, B.; Vosegaard, T.; Nielsen, N. C.; Tajkhorshid, E.Biophys. J. 2008, 95, 4337–4347.

(77) Illya, G.; Deserno, M. Biophys. J. 2008, 95, 4163–4173.(78) Mani, R.; Cady, S. D.; Tang, M.; Waring, A. J.; Lehrer, R. I.; Hong, M.

Proc. Natl. Acad. Sci. 2006, 103, 16242–16247.(79) Gkeka, P.; Sarkisov, L. J. Phys. Chem. B 2009, 113, 6–8.(80) Notman, R.; Anwar, J.; Briels, W. J.; Noro,M. G.; Otter, W. K. d. Biophys.

J. 2008, 95, 4763–4771.(81) Denning, E. J.; Woolf, T. B. Biophys. J. 2008, 95, 3161–3173.(82) Mondal, S.; Mukhopadhyay, C. Chem. Phys. Lett. 2007, 439, 166–170.

DOI: 10.1021/la902660q 12241Langmuir 2009, 25(20), 12235–12242

Manna and Mukhopadhyay Article

average distribution (last 7 ns) of water, along with the distribu-tion of water, carbonyl, and phosphate, at the starting point. Thebroadening of the water peak indicates water penetration fromboth leaflets of the membrane, although major lysis takes placefrom the lower layer. The prepore thus formed is of a well-definedsize6,19 (∼25 A in diameter, as reflected from the representation ofthe prepore in X-Y plane; Figure S12 of the Supporting In-formation), capable of ion permeation, rather than the hypothe-tical single-file water pore efficient for proton transfer, althoughthe pore width is not the same all across the bilayer. In a studyof melittin-induced hemolysis, Katsu et al.83 showed that thepores of 1.3 and 2.4 nm are formed at a melittin concentration of∼0.2 and 0.8 μM, respectively.

To get a quantitative estimation of water penetration, we thencalculated the number of penetrating water molecules (NW) as afunction of time. Figure S13 of the Supporting Informationindicates a sharp rise in water penetration followed by a gradualincrease. The result agrees well with the rapid water penetrationobserved by Roux et al.,70 immediately after the protonation ofthe N-terminus of melittin. Former experimental37 and theore-tical studies69 estimated the number of penetrating water mole-cules in the hydrophobic region of themembrane for eachmelittinas ∼20-30. The pore size as well as the number of penetratingwaters increase with an increase in the P/L ratio.19 Baumgaertneret al. investigated the stability of a hypothetical melittin poreconstructed by amelittin tetramer in aPOPCbilayer.39Accordingto their result, the number of water molecules inside the pore is∼400 at the end of the simulation. A smaller number of penetrat-ing water molecules, approximately ∼240, is observed in thepresent simulation (almost after 8 ns).

Ion Penetration. Although the melittin-induced pore is well-documented in the literature, only scarce information is availablefor ion conductance.22,23 The present study has explored themovement of chloride ion inside the prepore; however, no suchevidence is observed for sodium ions. The selectivity of theprepore for anions over cations is presumably due to its strongcationic nature. To monitor the movement of chloride ions, wehave plotted the Z-coordinates of Cl- ions as a function of time.Figure 8 shows that with the progress of time, few Cl- ionspenetrate deep into the membrane and are entrapped by thepositively charged N-terminus and Lys7 residues of melittin. Twoevents are observed where Cl- ions enter and exit the preporefrom the same side.Our result is compatiblewith themovement ofCl- ions in an octameric PG-1 pore, as described in ref 10. FigureS14 of the Supporting Information shows the radial distribution

function (RDFs) of various nitrogen (both backbone and sidechain) atoms of the TM segment of melittin with chloride ions,those that have entered into the aqueous cavity. As depicted inFigure S14 of the Supporting Information, mainly the positivelycharged Lys7 and N-terminus of melittin govern the electrostaticinteraction.

To evaluate the chloride occupancy of the prepore, we havecalculated the number of penetrating Cl- ions as a function oftime. Figure S15 of the Supporting Information reveals that onaverage∼5-6Cl- ions reside inside the prepore. Themultiple ionoccupancy is supportive to the formation of a prepore with well-defined size, in spite of a single-file water pore. Tosteson et al.22

earlier conjectured that four melittin monomers are needed toform the channel and at least four gating charges, that is, onecharge per melittin monomer, is needed to compensate thepositive charge on Lys7. Thus, Lys7 plays a crucial role in aniontransport through the melittin pore. Moreover, the electrostaticrepulsion between the positively charged residues protects theprepore from hydrophobic collapse.

Furthermore, we also investigate the diffusive nature of chlor-ide ions in bulk as well as in the interior of the prepore bycalculating the diffusion coefficient of Cl- ions along the mem-brane normal as a function of the distance from and positionalong the axis of the prepore10,84 (Figure 9). The confinedmotionof Cl- ions near the center of the prepore is due to their strongelectrostatic bonding with Lys7. As a result, they exhibit slowerdynamics as compared to bulk Cl- ion. The calculated diffusioncoefficient is comparable with the former theoretical finding.10

Altogether, our simulation data provide compelling evidence ofthe preference of melittin pore toward anions over cations andalso present a qualitative scenario of the movement of chlorideions in the pore interior. Thus, our simulation results pointtoward the initiation of an ion-permeable water pore in theabsence of TM potential19 (movies showing the initiation of celllysis by melittin and the chloride ion penetration into membranecore are in the Supporting Information).

Architecture of the Prepore.Despite intense research, there isan ongoing debate about the structure of the melittin pore. Fewstudies suggest the barrel-stave mechanism, while others invoke atoroidal (“worm-hole”) pore model. These two TM pores areactually differentiated on the basis of lipid orientation near thepore.12 In the barrel-stavemodel, peptides stand upright around acentral lumen. However, in the toroidal pore model, both of the

Figure 7. Time average lipid tilt angle as a function of radialdistance from the prepore (last 7 ns). The average tilt angledistribution of the pure POPC bilayer is shown in the inset. Thelipid tilt angle is the angle between the saturated lipid tail (fromcarbon numbers 2 to 16) and the membrane normal. Figure 8. Timeprofiles of the z-coordinate of representative chlor-

ide ions. The blue and yellow curves correspond to the Z-positionof Lys-7 and N-terminal Gly-1 respectively, as a function of time.

(83) Katsu, T.; Ninomiya, C.; Kuroko, M.; Kobayashi, H.; Hirota, T.; Fujita,Y. Biochim. Biophys. Acta 1988, 939, 57–63.

(84) Makarov, V. A.; Feig, M.; Andrews, B. K.; Pettitt, B. M. Biophys. J. 1998,75, 150–158.

12242 DOI: 10.1021/la902660q Langmuir 2009, 25(20), 12235–12242

Article Manna and Mukhopadhyay

bilayer leaflets sharply bend in the fashion of a toroidal hole, sothat the pore is lined by both peptide and lipid headgroups. Oursimulation results exhibit the tilting of lipid near the aqueouscavity. Few lipids with their lipid acyl tails undergo orientationaltransition from parallel to perpendicular wrt lipid normal. Thepore-lining lipid heads, which are originally on the membranesurface, are now flipped to line the prepore, in support of thetoroidal pore model. Few theoretical as well as experimentalstudies examined the reorientation of the lipid during the forma-tion of a toroidal-shaped pore previously.39,73,85 It has beensuggested that toroidal pore formation would be favored by thepresence of lipids with a positive curvature but opposed by lipidswith a negative curvature.13 On the basis of oriented circulardichroism (OCD) and neutron scattering studies, Huang et al.made a case study on the melittin pore.12 They were able tocrystallize the melittin pore at low temperature and low humidity.The properties of melittin pore are closely similar to those ofmagainin (toroidal pore) but unlike those of alamethicin (barrelstave pore).12 The single channel conductance of themelittin poremeasured by Sansom et al. is similar to magainin, which supportsthe toroidal model.86 Again, any mismatch between the hydro-phobic length of membrane and the peptide favors the toroidalpore over the barrel-stave pore,13 which is why in the latter casethe bilayer remains almost flat. In our simulation, the N-terminusof melittin in the lower half of the leaflet causes serious localdeformation and ultimately leads to membrane disruption. So,the evidence thatwe are able to gather fromour simulation arguedin favor of themelittin-induced toroidal pore over the barrel-stavemechanism.

Conclusion

The present work addresses the mechanism of action of AMPmelittin on the bilayer membrane. The cytolytic activity ofmelittin arises due to the defects introduced in the membraneby a highly cationic peptide. The present simulation indicates the

formation of a prepore through collective membrane perturba-tion by bound peptide. The analysis of the “lipid-peptide”interaction energy and also the partition of total interactionenergy into electrostatic and vdW terms will help to shed lighton the factors that confer the stability and activity of melittin inthe membrane. A detailed analysis of the average number ofcontacts and also the number of hydrogen bonds shared by thepeptide with its surrounding environment enables us to explorethe active role of specific amino acid residues in cell lysis. Thecharged residues of melittin not only act as “anchors” but alsodisrupt the membrane by dragging water from both sides, leadingto the initiation of a water-filled pore, which is also supported bythe surrounding lipid matrix. The presence of aqueous cavitycauses localized but strong bilayer perturbation. The tilting oflipid nearer to the prepore allows local bilayer thinning. Theflipped lipid head is now able to line the surface of the penetratingaqueous phase. From the conformation of lipids in the vicinity ofthe prepore, it can be classified as toroidal-shaped. Moreover, weare also able to capture the movement of chloride ions inside theaqueous cavity, although no such evidence is observed for sodiumions. The chloride ions entered into the prepore due to strongelectrostatic attraction by charged residues located at its wall. Wealso investigate the diffusive nature of chloride ions in bulk aswellas in the prepore interior. The slower diffusion coefficientobtained in the later case points toward the confined motion ofCl- ions near the center of the prepore. The selectivity of thecationic prepore toward anion is compatible with an experimentalapproach. Thus, we are able to produce themolecularmechanismrelated to the initiation of the ion-permeable water pore by AMPmelittin.

Acknowledgment. We are thankful to the Department ofChemistry, University of Calcutta, and the UPE project for thecomputational facilities.M.M. is thankful to CSIR, India, for thefellowship through CSIR-NET. This work is also supported bythe “Council for Scientific and Industrial Research” [No. 01-(2035)/06/EMR-II], Government of India.

Supporting Information Available: More data regardingthe equilibration of the pure POPC bilayer in NPT (area perlipid and order parameter profile) and NγPT (area per lipidand atom density distribution) ensembles, conformationalstudies of melittin (secondary structure analysis, interhelicalbend angle, and backbone rmsd), snapshots of lipid-peptidecomplexes, lipid conformations away and near the center ofthe prepore, bilayer structure of the melittin/lipid complexsystem (headgroup P-N vector and atom density dis-tribution), representation of the prepore structure in theX-Y plane, membrane permeability (water and ion penetra-tion,RDFofCL-), and results of the additional simulations.Two movies showing the initiation of cell lysis by melittinand the chloride ion penetration into membrane core.This material is available free of charge via the Internet athttp://pubs.acs.org.

Figure 9. Diffusion coefficient of Cl- ions along the z-axis, as afunction of the distance from and the position along the axis of theprepore (r,z).

(85) Matsuzaki, K.; Murase, O.; Tokuda, H.; Fujii, N.; Miyajima, K. Biochem-istry 1996, 35, 11361–11368.(86) Sansom, M. P. Prog. Biophys. Mol. Biol. 1991, 55, 139–215.(87) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 33–38.