Coupling molecular dynamics simulations with experiments for the rational design of indolicidin-analogous antimicrobial peptides.

Abstract

Antimicrobial peptides (AMPs) have attracted much interest in recent years because of their potential use as new-generation antibiotics. Indolicidin (IL) is a 13-residue cationic AMP that is effective against a broad spectrum of bacteria, fungi, and even viruses. Unfortunately, its high hemolytic activity retards its clinical applications. In this study, we adopted molecular dynamics (MD) simulations as an aid toward the rational design of IL analogues exhibiting high antimicrobial activity but low hemolysis. We employed long-timescale, multi-trajectory all-atom MD simulations to investigate the interactions of the peptide IL with model membranes. The lipid bilayer formed by the zwitterionic 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) was chosen as the model erythrocyte membrane; lipid bilayers formed from a mixture of POPC and the negatively charged 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol were chosen to model bacterial membranes. MD simulations with a total simulation time of up to 4 micros revealed the mechanisms of the processes of IL adsorption onto and insertion into the membranes. The packing order of these lipid bilayers presumably correlated to the membrane stability upon IL adsorption and insertion. We used the degree of local membrane thinning and the reduction in the order parameter of the acyl chains of the lipids to characterize the membrane stability. The order of the mixed 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol/POPC lipid bilayer reduced significantly upon the adsorption of IL. On the other hand, although the order of the pure-POPC lipid bilayer was perturbed slightly during the adsorption stage, the value was reduced more dramatically upon the insertion of IL into the membrane's hydrophobic region. The results imply that enhancing IL adsorption on the microbial membrane may amplify its antimicrobial activity, while the degree of hemolysis may be reduced through inhibition of IL insertion into the hydrophobic region of the erythrocyte membrane. In addition, through simulations, we identified the amino acids that are most responsible for the adsorption onto or insertion into the two model membranes. Positive charges are critical to the peptide's adsorption, whereas the presence of hydrophobic Trp8 and Trp9 leads to its deeper insertion. Combining the hypothetical relationships between the membrane disordering and the antimicrobial and hemolytical activities with the simulated results, we designed three new IL-analogous peptides: IL-K7 (Pro7-->Lys), IL-F89 (Trp8 and Trp9-->Phe), and IL-K7F89 (Pro7-->Lys; Trp8 and Trp9-->Phe). The hemolytic activity of IL-F89 is considerably lower than that of IL, whereas the antimicrobial activity of IL-K7 is greatly enhanced. In particular, the de novo peptide IL-K7F89 exhibits higher antimicrobial activity against Escherichia coli; its hemolytic activity decreased to only 10% of that of IL. Our simulated and experimental results correlated well. This approach-coupling MD simulations with experimental design-is a useful strategy toward the rational design of AMPs for potential therapeutic use.

Full text

Page 1

Coupling Molecular Dynamics Simulations
with Experiments for the Rational Design of
Indolicidin-Analogous Antimicrobial Peptides
Ching-Wei Tsai1, Ning-Yi Hsu1, Chang-Hsu Wang1, Chia-Yu Lu1,
Yung Chang2, Hui-Hsu Gavin Tsai3⁎ and Rouh-Chyu Ruaan1⁎
1Department of Chemical and
Materials Engineering, National
Central University, Jhong-Li
City, Tao-Yuan County 32001,
Taiwan, ROC
2Department of Chemical
Engineering, Chung Yuan
Christian University, Jhong-Li
City, Tao-Yuan County 32001,
Taiwan, ROC
3Department of Chemistry,
National Central University,
Jhong-Li City, Tao-Yuan County
32001, Taiwan, ROC
Received 10 March 2009;
received in revised form
8 June 2009;
accepted 27 June 2009
Available online
2 July 2009
Antimicrobial peptides (AMPs) have attracted much interest in recent years
because of their potential use as new-generation antibiotics. Indolicidin (IL)
is a 13-residue cationic AMP that is effective against a broad spectrum of
bacteria, fungi, and even viruses. Unfortunately, its high hemolytic activity
retards its clinical applications. In this study, we adopted molecular dyna-
mics (MD) simulations as an aid toward the rational design of IL analogues
exhibiting high antimicrobial activity but low hemolysis. We employed long-
timescale, multi-trajectory all-atom MD simulations to investigate the
interactions of the peptide IL with model membranes. The lipid bilayer
formed by thezwitterionic1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine
(POPC) was chosen as the model erythrocyte membrane; lipid bilayers
formed from a mixture of POPC and the negatively charged 1-palmitoyl-2-
oleoyl-sn-glycero-3-phosphoglycerol were chosen to model bacterial mem-
branes.MDsimulationswitha total simulation timeofupto4 μsrevealedthe
mechanisms of the processes of IL adsorption onto and insertion into the
membranes. The packing order of these lipid bilayers presumably correlated
to the membrane stability upon IL adsorption and insertion. We used the
degree of local membrane thinning and the reduction in the order parameter
of the acyl chains of the lipids to characterize the membrane stability. The
order of the mixed 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol/
POPC lipid bilayer reduced significantly upon the adsorption of IL. On the
other hand, although the order of the pure-POPC lipid bilayer was perturbed
slightly during the adsorption stage, the value was reduced more drama-
tically upon the insertion of IL into the membrane's hydrophobic region. The
results imply that enhancing IL adsorption on the microbial membrane may
amplify its antimicrobial activity, while the degree of hemolysis may be
reduced through inhibition of IL insertion into the hydrophobic region of the
erythrocyte membrane. In addition, through simulations, we identified the
aminoacidsthataremostresponsiblefortheadsorptionontoorinsertioninto
the two model membranes. Positive charges are critical to the peptide's
adsorption, whereas the presence of hydrophobic Trp8 and Trp9 leads to its
deeper insertion. Combining the hypothetical relationships between the
membrane disordering and the antimicrobial and hemolytical activities with
the simulated results, we designed three new IL-analogous peptides: IL-K7
(Pro7→Lys), IL-F89 (Trp8 and Trp9→Phe), and IL-K7F89 (Pro7→Lys; Trp8
and Trp9→Phe). The hemolytic activity of IL-F89 is considerably lower than
that of IL, whereas the antimicrobial activity of IL-K7 is greatly enhanced. In
particular, the de novo peptide IL-K7F89 exhibits higher antimicrobial
activity against Escherichia coli; its hemolytic activity decreased to only 10% of
*Corresponding authors. E-mail addresses: hhtsai@cc.ncu.edu.tw; ruaan@cc.ncu.edu.tw.
Abbreviations used: MD, molecular dynamics; AMP, antimicrobial peptide; IL, indolicidin; POPC, 1-palmitoyl-2-
oleoyl-sn-glycero-3-phosphocholine; POPG, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol; MHC, minimum
hemolytic concentration; MIC, minimum inhibitory concentration; DPC, dodecylphosphocholine; PG-1, protegrin-1; PC,
phosphatidylcholine; PG, phosphatidylglycerol.
doi:10.1016/j.jmb.2009.06.071J. Mol. Biol. (2009) 392, 837–854
Available online at www.sciencedirect.com
0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.

Page 2

that of IL. Our simulated and experimental results correlated well. This
approach—coupling MD simulations with experimental design—is a useful
strategy toward the rational design of AMPs for potential therapeutic use.
© 2009 Elsevier Ltd. All rights reserved.
Edited by J. Bowie
Keywords:
simulation; rational drug design
antimicrobialpeptide;indolicidin;moleculardynamics
Introduction
Antimicrobial peptides (AMPs) are host-defense
molecules produced by living organisms1,2(for an
AMP database, see Refs. 3 and 4; this database has
been updated recently and is accessible via the
Internet†). They are good candidates for the devel-
opment of clinical peptide drugs because of their
rapid action, broad spectrum, and low antibiotic
resistance. Most of the naturally occurring AMPs
are, however, somewhat toxic to mammals, thereby
hindering their direct therapeutic applications.
Current research in this field is focused primarily
on engineering the naturally occurring AMPs to
reduce their toxicity and enhance their antimicrobial
activity; that is, it is generally aimed at increasing
their therapeutic applications. Nevertheless, the
high cost of synthesizing longer peptides limits
progress in this field. Indolicidin (IL) is an AMP
derived from the cytoplasmic granules of bovine
neutrophils.5It is effective against a broad spectrum
of microorganisms, including Gram-positive and
Gram-negative bacteria and fungi;6,7in addition, it
is potent against herpes, Junin, and human immu-
nodeficiency viruses.8,9This 13-amino-acid-long
peptide (H-ILPWKWPWWPWRR-NH2) is rich in
hydrophobic Trp, cationic Arg/Lys, and structure-
hindering Pro residues. The high percentage of Pro
residues results in IL exhibiting a disordered struc-
ture in solution.10The structures of IL bound to
zwitterionic dodecylphosphocholine (DPC) and
anionicsodium dodecylsulfate micelles, determined
using NMR spectroscopy techniques,11are some-
what different: In DPC, the backbone structure is
extended with two half-turns at residues Lys5 and
Trp8, giving aboat shape;in sodiumdodecylsulfate,
the backbone structure is also extended, but in this
case, between residues 5 and 11. Such slight
variations in IL structures on different micelle types
have also been suggested through CD-based ana-
lyses. Fluorescence and CD spectra revealed similar
structures for IL bound with unilamellar phospho-
lipid vesicles and with detergent micelles.11IL is
generally classified in the subset of cationic AMPs
that are enriched in particular amino acids (e.g., Pro,
Arg, ;Trp) and not in terms of the common subsets of
secondary structure.12The short amino acid length
of IL makes it an attractive lead peptide for the
design of antimicrobial agents, allowing the rapid
and economical synthesis of its analogues.
Several investigations have focused on the engi-
neering of analogous IL peptides and attempts to
find IL analogues suitable for therapeutic use.10,13–17
In particular, Staubitz et al. systemically engineered
IL analogueswith (i)anAla scan of each amino acid ;
(ii) amino acid truncations from each terminus; (iii)
modified terminal functions; (iv) specific amino acid
exchanges of aromatic, charged, and structural resi-
dues; (v) retro-sequences and (vi) inverso and
retroinverso amino acids; they examined the anti-
microbial and hemolytic activities of each of these
engineered IL analogues.15They found that the IL
analogues in which a single amino acid had been
mutated to an Ala residue exhibited no obvious
changes in their minimum hemolytic concentrations
(MHCs) against human erythrocytes, but there were
various effects on their minimum inhibitory concen-
trations (MICs) against M. luteus. For example, the
analogues possessing reduced cationic charge or
featuring a Trp unit mutated to an Ala residue
generally exhibiting reduced MICs. Although a
single mutation to Ala did not affect the MHC,
when all of the Trp residues were mutated to Phe
residues (providing the AMP IL-F), the MHC
increased significantly but the MIC was unaffected.
Although this approach successfully suggested a
lowuehemolytic IL analogStaphylococcus, IL-F, was
discovered, this approach is time-consuming and the
cost of peptide synthesis is high. Nevertheless, IL
analogues simultaneously exhibiting lower hemoly-
tic and higher antimicrobial activities than those of
the parent have not yet been obtained.
Another approach toward peptide design is
through the understanding of the antibiotic and
hemolytic mechanisms of IL and other AMPs. In the
past two decades, several studies have been per-
formed to investigate the mechanisms of action of
IL. It has been demonstrated that the mode of IL
action leading to cell death is caused by the inter-
action of IL with the cytoplasm membrane. IL is
capable of crossing the outer membrane of Gram-
negative bacteria via a self-promoted uptake path-
way, killing bacteria either through disruption of the
cytoplasmic membrane10,18or through channel for-
mation.18,19Some structurally diverse AMPs exhibit
diverse mechanisms of action against Gram-positive
bacteria (e.g., that is S. aureus);20they do not involve
only membrane permeabilization. Moreover, some
studies have suggested the limited potential of IL to
kill cells solely through membrane disruption.19,21
Cell death can also be caused by peptide penetration
through the cytoplasmic membrane and subsequent
binding with such intracellular targets such as
†http://aps.unmc.edu/AP/main.html
838
Improvement of Indolicidin Activity by MD Simulation

Page 3

DNA,22,23Ca2+-calmodulin,24and phospholipase
A2.25To clearly demonstrate the possible mecha-
nisms, large unilamellar vesicles formed from the
phospholipids 1-palmitoyl-2-oleoyl-sn-glycero-3-
phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-
glycero-3-phosphoglycerol (POPG), and to clearly
demonstrate the possible mechanisms-(R)-dipalmi-
toylphosphatidic acid have often been used as
membrane modelsl. The membrane permeation of
IL through transient pore formation on the surface
of L-(R)-dipalmitoylphosphatidic acid phospholi-
pids, via membrane lysis, has been observed using
in situ atomic force microscopy-techniques.26IL also
disrupts the lipid bilayer of phosphocholine lipo-
somes in a salt-dependent matter.27Ladokhin et al.
demonstrated that leakage of the vesicles was
induced upon their interaction with IL; this leakage
was more effective with POPG vesicles than with
POPC vesicles.28Recently, Shaw et al. reported their
observations of membrane thinning, due to inter-
digitation of the inner and outer leaflet of the
membrane, upon interaction with IL;29they also
observed this behavior in molecular dynamics
(MD) simulations.30Other MD simulations have
been performed to study how the side chains inter-
actions of IL influence its structure near the mem-
brane.31Although several mechanisms of action of
IL have been proposed, the true mechanism
remains unclear. Nevertheless, it is clear that direct
contact between IL and the cell membrane is the
first—and mandatory—step. Therefore, the degree
of membrane perturbation after IL contact should
be related to the biological activity of IL.
Multi-scale computer simulations are useful
tools in the field of AMP research, providing
critical atomic and thermodynamic32,33informa-
tion relating to AMP-membrane interactions.
Langham et al. successfully simulated the sponta-
neous toroidal pore formation of the magainin
MG-H2 peptide within a model phospholipid
membrane using all-atom MD simulation.34
Sayyed-Ahmad and Kaznessis simulated a prote-
grin-1 octamer pore, proposed from NMR spectro-
scopy experiments, in a lipid bilayer composed of
POPE and POPG lipids, they discovered a pore
more close resembling the barrel-stave model than
the toroidal-pore model.35The pore was ion-
selective, allowing negatively charged chloride
ions to pass through. Computing-resource-saving
continuum models, mainly modeling effects of
solvent and biological membrane molecules in
terms of dielectric constants, provide another
approach to understand the mechanisms of AMP
action; for example, Sayyed-Ahmad and Kaznessis
solved the PoissonrBoltzmann equation (modeling
electostatic interactions) and solvent accessible
surface area (modeling non-polar interactions) to
investigate the most probable orientation of PG-1
in dilauroylphosphatidylcholine (DLPC) lipid
bilayers.36They determined that the most favor-
able orientations had a tilting angle of 19°, in
qualitative agreement with that determined using
solid-state NMR spectroscopy.
To improve potential pharmacological applica-
tions of IL-analogous peptides, they must be engi-
neered to differentiate, to a certain extent, between
their microbial targets and the host cells. Previous
studies into the design of IL-analogous peptides
have been aimed at understanding the roles of each
amino acid of IL in the mechanism of action and to
further improve their potency.10,13–17These IL ana-
logues have, however, been designed mainly based
on the properties of the amino acid sequence. Some
studies have addressed peptide–lipid interactions at
a structural level.11,37In the absence of atomic-
detail structure-based information relating to IL-
lipid bilayer interactions, it has been difficult to
engineer peptides exhibiting a degree of differen-
tiation between microbial and host cells. Our
complementary strategy toward the design of IL-
analogous AMPs was first to perform long-time-
scale and multi-trajectory MD simulations with
atomistic details. Such MD simulations allowed us
to study the thermal motions of the peptide, the
membrane, and the solvent molecules by consider-
ing their interactions at the atomic scale, thereby
providing opportunities to observe—and, further-
more, to analyze—the detailed interactions between
IL and lipid bilayer. We employed a pure zwitter-
ionic POPC lipid bilayer, the major component in
erythrocyte membranes, to mimic the cell mem-
branes of mammals; on the other hand, we used a
lipid bilayer comprising 25% negatively charged
POPG lipids and 75% zwitterionic POPC lipids—a
membrane component similar to that of Escherichia
coli—to mimic the cell membranes of microorgan-
isms. We simulated the IL–membrane association
processes in a spontaneous manner. Not only did
the IL peptide adsorb onto the membrane surface
from the aqueous phase, it also inserted into the
hydrophobic region of the membrane when mon-
itored at atomic and picosecond time resolution. We
recorded the peptide–membrane interaction process
and analyzed the driving forces of peptide adsorp-
tion and further insertion. After association with IL,
the degrees of perturbation of both membranes were
identified in terms of the degree of local membrane
thinning and the order parameter of the membrane's
aliphatic chains. The differences between the dis-
ordering of these two lipid bilayers could therefore
berecognized.Thecriticalaminoacidsresponsibleto
IL–membrane association and membrane disorder-
ing could also be identified through association
process monitoring and driving force analysis. We
assumed that membrane disordering after IL asso-
ciation would be closely related to cell death and
expected that we could differentiate the possible
causes of microbial death from those of erythrocytes.
On the basis of our MD simulation results, we
designed and synthesized new IL analogues that we
suspected would possess high potency against
microbes while causing minimal hemolysis. The
antimicrobial and hemolytic activities of the
designed peptides were then determined experi-
mentally to test the accuracy of our MD simulations
and subsequent membrane perturbation analyses.
839
Improvement of Indolicidin Activity by MD Simulation

Page 4

Results
IL–POPC lipid bilayer association
Simulations revealed that the peptide IL diffused
spontaneously from the aqueous phase to the water–
lipid head group interface and that some of its
residues inserted into the hydrophobic region of the
lipid bilayer. Four simulations initiated from diffe-
rentconfigurationsprovidedsimilarmechanismsfor
the IL–POPC lipid bilayer association, although
some different behavior (e.g., kinetics) was evident.
Taking the first simulation system (labeled IL-PC-I)
as an example, Fig. 1 presents the characteristic
processes of IL–POPC lipid bilayer association in
snapshots. First, IL in the aqueous phase gradually
diffused to the water lipid interfacial region (Fig. 1a).
At t=ca 82 ns, IL began to anchor onto the surface of
the POPC lipid bilayer through the formation of salt
bridges between IL and the head groups of the lipid
bilayer (Fig. 1b). Subsequently, at t=ca 120 ns, the
central hydrophobic residues Trp8 and Trp9 began
to immerse into the lipid bilayer and then enter into
the hydrophobic region of the lipids (Fig. 1c). At the
end of simulation (t=600 ns), most of the Trp
residues, as well as some of other hydrophobic resi-
dues (e.g., Ile1 and Leu2), had inserted into the
hydrophobic region of the lipid bilayer (Fig. 1d).
Snapshots of the other three IL–POPC simulation
systems are provided in the Supplementary Data.
The snapshots suggest that the electrostatic inter-
actions between IL and the lipid bilayer play impor-
tant roles in driving and/or stabilizing the IL unit on
the membrane; therefore, we analyzed them further
in terms of the shortest distance (dSB) between the
positively charged groups of IL and the negatively
charged phosphate groups of the POPC lipids as a
functionofthesimulationtime(Fig.2).Att=ca80ns,
the stable salt bridges located at the N-terminus and
R13 at the C-terminus were formed almost simulta-
neously, as indicated by their short distances dSB(ca
1.6 Å). Stable salt bridgesat K5 and R12 were formed
later. Although R13 was the first to form a stable salt
bridge, repeated breakage and reformation of this
salt bridge occurred during the late stages of
simulation. The first contact point does not necessa-
rily act as the anchor in this dynamic process; for
example, in Fig. 2, the first contact point is the R13-
NH2
and unstable. In contrast, we could consider the
stable contact point, rather than the first contact
point, as an anchor in the dynamic process (vide
infra). Moreover, we analyzed the interactions
between each hydrophobic residue and the POPC
membrane in terms of the z-coordinate (labeled z)
measured from the center of d mass of the residue's
sidechaintothecenterofthelipidbilayer(Fig.3).We
observed that contacts between the hydrophobic
residues and the hydrophobic region of the lipid
bilayer occurred after the formation of stable salt
bridges (e.g., at t=ca 80 ns). After t=ca 150 ns, all of
the hydrophobic residues, except for Trp11, gradu-
+∷PO4
−salt bridge, which is, however, transient
ally immersed into the low-dielectric hydrophobic
region of the lipid bilayer and then some of them, in
particular, Trp8 and Trp9, were inserted stably into
the hydrophobic region. The values of dSBand z of
the other three IL–POPC simulation systems are
provided in the Supplementary Data.
IL–mixed POPG/POPC lipid bilayer association
Figure 4 presents, as snapshots, the characteristic
steps of the IL−mixed POPG/POPC lipid bilayer
association process for the first simulation system
(denoted IL-PG/PC-I). The statistical ensemble
results of the four MD simulations are presented in
the next section. The IL unit quickly approached the
lipid surface at a simulation time of ca 1 ns (Fig. 4a).
At t=ca 15 ns, IL was adsorbed stably on the surface
of the mixed POPG/POPC lipid bilayer through salt
bridge interactions (Fig. 4b). Subsequently, some of
the hydrophobic residues of IL gradually entered
the hydrophobic region of the lipid bilayer (Fig. 4c).
Compared with the simulation of the POPC system,
the IL unit resided close to the interfacial region
between the head group and the acyl chains of the
lipid bilayer at the end of the simulation, but some of
the hydrophobic residues were inserted stably into
the hydrophobic region of the lipid bilayer (Fig. 4d).
Figures 5 and 6 present a quantitative view of the
IL–mixed POPG/POPC lipid bilayer association
process in terms of the distance dSBand the coor-
dinate z, respectively, of each residue side chain as a
function of simulation time. In Fig. 5, we observe
that the salt bridges that formed between IL and the
lipids were stable, except for that at Lys5, which
exhibited relatively large fluctuations. In Fig. 6, in
contrast to the results obtained for the POPC system,
in which the Trp residues were positioned within
the hydrophobic regions of lipids, we observe that
some of the Trp residues were located at the water–
lipid interface. Radial distribution analysis of the
phosphorus atoms of the lipids revealed no obvious
phase segregation and lipid redistribution (data not
shown).
Definitions of adsorption and stable insertion
stages
To characterize the membrane disordering mecha-
nism of IL, we classified the IL–lipid bilayer associa-
tion processes in terms of their interactions into
three distinct stages: non-association, adsorption,
and stable insertion. Through analysis of the salt
bridge formation and hydrophobic interactions, we
found that hydrophobic contact always occurred
after salt bridge formation. Therefore, the non-
association stage was defined in the time period
from t=0 to the time immediately prior to the for-
mation of at least two salt bridges. During this stage,
IL and the lipid bilayer did not interact to any
considerable degree. In the adsorption stage, we
considered IL to be adsorbed in the water–lipid
bilayer interfacial region. This stage was defined as
the time course between the moment at which the
840
Improvement of Indolicidin Activity by MD Simulation

Page 5

Fig. 1. Snapshots presenting the association processes of the IL peptide and the pure-POPC lipid bilayer. These snapshots were taken from the first simulation system. For IL,
the hydrophobic residues are white, the cationic residues and groups are blue, and the proline residues are orange; its backbone is presented in ribbon form. The water molecules
are simply presented by their oxygen atoms (pink). The larger red balls represent the phosphorus atoms of the POPC lipid head groups. The acyl chains of the aggregated POPC
lipids are colored gray and presented in line form. The salt bridges between IL and the head groups of the lipids are highlighted in circles. These snapshots were prepared using
VMD software.38
841
Improvement of Indolicidin Activity by MD Simulation