| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Biophys J, March 2000, p. 1359-1375, Vol. 78, No. 3
School of Chemistry, Tel Aviv University, Ramat Aviv, Tel Aviv 69978, Israel
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
THE SIMULATION
RESULTS
DISCUSSION
REFERENCES
|
|---|
A molecular dynamics simulation of melittin in a hydrated dipalmitoylphosphatidylcholine (DPPC) bilayer was performed. The 19,000-atom system included a 72-DPPC phospholipid bilayer, a 26-amino acid peptide, and more than 3000 water molecules. The N-terminus of the peptide was protonated and embedded in the membrane in a transbilayer orientation perpendicular to the surface. The simulation results show that the peptide affects the lower (intracellular) layer of the bilayer more strongly than the upper (extracellular) layer. The simulation results can be interpreted as indicating an increased level of disorder and structural deformation for lower-layer phospholipids in the immediate vicinity of the peptide. This conclusion is supported by the calculated deuterium order parameters, the observed deformation at the intracellular interface, and an increase in fractional free volume. The upper layer was less affected by the embedded peptide, except for an acquired tilt relative to the bilayer normal. The effect of melittin on the surrounding membrane is localized to its immediate vicinity, and its asymmetry with respect to the two layers may result from the fact that it is not fully transmembranal. Melittin's hydrophilic C-terminus anchors it at the extracellular interface, leaving the N-terminus "loose" in the lower layer of the membrane. In general, the simulation supports a role for local deformation and water penetration in melittin-induced lysis. As for the peptide, like other membrane-embedded polypeptides, melittin adopts a significant 25o tilt relative to the membrane normal. This tilt is correlated with a comparable tilt of the lipids in the upper membrane layer. The peptide itself retains an overall helical structure throughout the simulation (with the exception of the three N-terminal residues), adopting a 30o intrahelical bend angle.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
THE SIMULATION
RESULTS
DISCUSSION
REFERENCES
|
|---|
Biological membranes consist of a lipid matrix (the lipid bilayer) in which molecules, such as proteins and cholesterol, are embedded. When a protein is incorporated into the lipid bilayer it may influence the membrane phase behavior, affecting the membrane dynamics and modifying its biological function. Thus the manner in which the presence of proteins in phospholipid bilayers affects the properties of the membrane has for a long time been the focus of intensive research. Of particular interest is the difference between "boundary" phospholipids, i.e., lipids in direct contact with the embedded protein, and the more distant "bulk" lipids.
Lipid bilayers composed of a single lipid species undergo several
different types of phase transitions (Geve and March, 1987
; Mouritsen,
1991; Mouritsen and Biltonen, 1993). The transition that is believed to
have a substantial effect on membrane function is the main gel-to-fluid
phase transition. This phase transition takes the membrane from a
lower-temperature gel phase, characterized by acyl-chain order, to a
higher-temperature fluid phase characterized by conformational disorder
and fast lateral diffusion. This phase behavior is further complicated
when the bilayer is composed of more than a single lipid species and
when integral membrane proteins are present. It is sensitive to
molecular interactions in the lipid bilayer system, to temporal
fluctuations of the lipids (as seen, for example, by fluorescence
experiments; Ruggiero and Hudson, 1989), as well as to the presence of heteromolecules.
The overall effect of incorporating integral proteins into membranes is
a significant change in the phase equilibria (Mouritsen and Biltonen,
1993), involving structural changes in the adjacent lipid molecules.
Common experimental probes to this effect are spectroscopic order
parameters, which refer to the conformational order of the acyl chains.
Early electron spin resonance studies indicated that on the
experiment's time scale, lipids around a protein are ordered in a way
that is different from the way in which bulk lipids are ordered (Jost
et al., 1973). This difference between the two types of lipids
disappears on the much longer time scale of NMR experiments (Deveaux
and Siegneuret, 1985).
Theoretical investigations of lipid-protein interactions were limited
for a long time to simplified models. Monte Carlo simulations of model
systems, composed of a schematic "cylindrical" protein in a
simplified carbohydrate monolayer, have shown very little effect of the
"protein" on the order parameter of the neighboring lipids (Scott,
1986). Similar simulations of cholesterol in the model monolayers
(Scott and Kalaskar, 1989) had indicated that the cholesterol affects
the upper portion of the chains, constraining their conformational
freedom, while making the lower termini of C-16 and C-18 chains a
little less ordered than bulk chains. The reduced order at the chain
termini was not observed with shorter C-14 chains. In a different study
the insertion of peptides into membranes was studied using Monte Carlo
and a simplified model (Milik and Skolnick, 1993). With the increase in
computer power, model membrane simulations were gradually replaced by
detailed all-atom molecular dynamics (MD) studies of phospholipid
bilayers. MD simulations of bilayer systems began with modest model
systems (van der Ploeg and Berendsen, 1982) and improved with time,
reaching various solvated bilayers of more than 100 phospholipids
(Alper and Stouch, 1995; Berger et al., 1997; Chiu et al., 1995;
Damodaran and Merz, 1994; Feller et al., 1997; Heller et al., 1993;
Lopez Cascales et al., 1996; Marrink et al., 1996, 1998; Pastor, 1994; Tieleman and Berendsen, 1996; Tieleman et al., 1997). These simulations reproduce experimental results as well as probe at an atomic level the
interactions between the bilayer and the surrounding water molecules.
Recently the first detailed all-atom MD simulations of peptides in a
phospholipid bilayer were reported. Damodaran et al. (1995) reported a
simulation of a small tripeptide in a 16 × 2 dipalmitoylphosphatidylcholine (DMPC) bilayer. Roux, Woolf, and collaborators have performed a series of studies on several
transmembrane helices embedded in a solvated DMPC bilayer. In these
simulations the peptides, gramicidin A (Woolf and Roux, 1996), a model
amphiphilic helix (Belohorcova et al., 1997), and individual helices
from bacteriorhodopsin (Woolf, 1997; Woolf, 1998), were surrounded by
8 × 2 or 12 × 2 DMPC phospholipids in a hexagonal boundary and capped by water molecules. In separate studies Shen et al. (1997)
simulated a single transmembranal polyalanine helix in a solvated
16 × 2 DMPC bilayer, and Tieleman et al. (1999) studied the
alamethicin channel-forming peptide in a
palmitoyloleoylphosphatidylcholine bilayer. The primary focus of these
studies was the effect of the membrane on the properties of the
embedded polypeptide. It is only in recent years, with the increase in
computer power, that the complementary question, i.e., the effect of
the protein on the embedding membrane, is being addressed. Very
recently an unprecedented large system, consisting of the large OmpF
porin trimer and a hydrated bilayer consisting of 318 palmitoyloleoylphosphatidylethanolamine (POPE) phospholipids was
simulated by Tieleman and Berendsen (1998). An in-depth analysis of the
protein's effect on the embedding phospholipids was recently reported
by Tieleman et. al (1998), who studied the proteins OmpF, influenza M2,
and alamethicin in a large POPE membrane.
The focus of the present study is the reciprocal effect of a
membrane-embedded peptide, exemplified by the peptide melittin, on the
properties of the host membrane and vice versa. The 26-amino acid
peptide melittin is the principal toxic component of the venom of the
honey bee (Terwilliger and Eisenberg, 1982). It spontaneously binds to
lipid bilayers and acts as a lytic agent (Vogel and Jähnig, 1986). Three mechanisms for the lytic activity of melittin have been
proposed (Dempsey, 1990): 1) lysis is due to melittin-induced formation
of ion-permeable water pores (Tosteson et al., 1985); 2) lysis results
from the perturbation of the lipid bilayer due to the presence of the
peptide in the headgroup region (DeGrado et al., 1982); 3) lysis occurs
as a result of melittin aggregation (Dufourc et al., 1986). It seems
that the first two mechanisms are best supported by the experimental
data, although to a lesser extent the third mechanism also has
experimental support. In fact, the first two mechanisms are probably
intimately connected, as water penetration can be helped by structural
perturbations at the bilayer's surface. For example, MD simulations
(Marrink and Berendsen, 1994) have shown that the rate of water
translocation through the bilayer is limited by the interfacial region
near the glycerol. Water penetration can increase by the formation of
defects at the interface, possibly by a membrane-embedded peptide.
In any case, regardless of the lysis mechanism assumed, the molecular
details of these mechanisms are still unclear. In particular, it is
unclear how lysis is related to the change in melittin's membrane-binding orientation. Several researchers have suggested that
partial translocation of melittin, from the initial binding orientation
parallel to the membrane surface to a transbilayer orientation in the
hydrophobic core, is involved in lysis (Berneche et al., 1998; Bradshaw
et al., 1994; Weaver et al., 1992). Based on neutron scattering
measurements at different bulk pH, Bradshaw et al. (1994) proposed that
melittin with an unprotonated N-terminus binds parallel to the membrane
surface, whereas melittin with a protonated N-terminus binds in a
transbilayer way. Interconversion between the two binding modes appears
to be possible under equilibrium conditions.
A characterization at the molecular level of the association of
melittin with membranes and its influence on the surrounding lipid
phase is necessary for a better understanding of the microscopic mechanism involved in the lytic event. Very recently, an MD simulation of melittin in a bilayer environment was reported by Berneche et al.
(1998). In that simulation melittin with an unprotonated N-terminus was
bound parallel to a DMPC membrane surface (although during the
simulation the N-terminus penetrated into the upper layer of the
membrane). Upon protonation of the N-terminus it was found that water
penetration into the bilayer increases significantly.
Unfortunately, interconversion between parallel and perpendicular
melittin orientations is not possible on the time scale of MD
simulations (even though the simulated membrane is in its L
phase).
Thus in the present study we have simulated the complementary scenario,
in which melittin is already embedded in the membrane adopting its
transbilayer orientation. In accord with neutron-scattering measurements (Bradshaw et al., 1994), the N terminus of the peptide in
the present simulation was protonated. In all, the simulated system
consists of the peptide melittin embedded in a 36 × 2 DPPC bilayer at its L phase (72 1,2-dipalmitoyl-3-sn-phosphatidylcholine phospholipids),
capped by a 40-Å water layer (more than 3000 water molecules), under
periodic boundary conditions. In principle, this system is large enough
to study the effect of the peptide on the properties of
non-nearest-neighbor phospholipids.
It is important to note that even at this so-called transbilayer
orientation melittin does not span the bilayer from side to side. In
fact, it spans no more than two-thirds of the bilayer's width. This
means that melittin is anchored to only one of the two membrane's
surfaces (the extracellular side), leaving the N-terminus free in the
membrane's hydrophobic core. In this respect, the melittin/membrane
system is significantly different from other peptide/membrane systems
studied by MD, in which the peptides fully span the membrane from one
side to the other and are anchored on both sides on the bilayer
(Belohorcova et al., 1997; Shen et al., 1997; Tieleman et al., 1999;
Woolf, 1997, 1998; Woolf and Roux, 1996).
| |
THE SIMULATION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
THE SIMULATION
RESULTS
DISCUSSION
REFERENCES
|
|---|
Melittin
Melittin is a 26-amino acid protein
(Gly-Ile-Gly-Ala-Val5- Leu-Lys-Val-Leu-Thr10-Thr-Gly-Leu-Pro-Ala15-Leu-Ile-Ser-Trp-Ile20-Lys-Arg-Lys-Arg-Gln25-Gln)
and is the principal toxic component of the venom of the honey bee
(Dempsey, 1990; Terwilliger and Eisenberg, 1982); it has a cationic
character. The 20 amino acids at the N-terminus part of melittin have a
predominantly hydrophobic character, whereas the six amino acids at the
C-terminus are very hydrophilic and basic. Melittin is soluble in both
water and methanol. In water it is either monomeric or tetrameric and
is shaped like a bent rod, in which the bend in the molecule is induced
by residue Pro14 (Pastore et al., 1989; Vogel and
Jähnig, 1986). In methanol melittin is monomeric and helical,
with a bend angle of ~20o, significantly
smaller than the 60o angle observed in water
(Bazzo et al., 1988). Various experimental studies have shown that
melittin is helical in a lipid environment. The orientation of melittin
in a phospholipid bilayer appears to be complex and is sensitive to
experimental conditions (Milik and Skolnick, 1993). When the peptide is
membrane bound, residue Trp19 is located
approximately at the height of the C1 atoms of the lipid hydrocarbon
chains (Vogel and Jähnig, 1986), leaving the six C-terminus amino
acids outside of the hydrophobic region, allowing them to interact with
the polar headgroups and the surrounding water. As described above, the
orientation of membrane-bound melittin is uncertain (Dempsey, 1990).
Several experimental results indicate that the peptide is oriented
roughly perpendicular to the membrane surface (Vogel, 1987; Vogel and
Jähnig, 1986), although there are also results (in the fluid
phase of the bilayer) that indicate a parallel orientation (Dempsey,
1990). Recent neutron scattering measurements with melittin at
different bulk pH (Bradshaw et al., 1994) suggest that melittin with
unprotonated N-terminus binds parallel to the membrane surface, whereas
melittin with a protonated N-terminus binds in a transbilayer way.
Support for the role of the perpendicular orientation in membrane lysis
is given by the requirement for a specific cationic C-terminal that
anchors the molecule in the vertical orientation (Habermann and
Kowallek, 1970; Manjunatha-Kini and Evans, 1989), as well as by the
fact that shortened N-terminal sequences are very poor lytic agents (Gevod and Birdi, 1984). The ambiguity regarding peptide orientation pertains also to computational studies that used simplified models to
represent the melittin-membrane system. In one study the preferred orientation of the helix was found to be roughly perpendicular to the
bilayer surface (Milik and Skolnick, 1993), whereas another study
indicated that melittin can access a wide range of orientations, from
perpendicular to almost parallel (Ducarme et al., 1998).
The initial melittin coordinates were taken from the x-ray structure of
Terwilliger and Eisenberg (protein data bank entry 2mlt; Terwilliger
and Eisenberg, 1982). To prepare the molecular model for interfacial
positioning at the membrane interface, the six C-terminal amino acids
were minimized separately, using a water dielectric constant (the rest
of the peptide was fixed), and the 20 N-terminal amino acids were
minimized separately, using a hydrophobic dielectric constant. The
CHARMM molecular mechanics program (Brooks et al., 1983) and force
field were used (MacKerell et al., 1998).
Pure phospholipid bilayer
In general, melittin's lysis activity strongly depends on
membrane composition. Membranes with longer hydrocarbon chains are less
affected by the lysis activity of melittin (Bradrick et al., 1995),
while bilayers of zwitterionic lipids are more affected compared to
bilayers of charged phospholipids (Monette and Lafleur, 1995). These
two considerations led us to simulate melittin in a
1,2-dipalmitoyl-3-sn-phosphatidylcholine (DPPC) bilayer.
DPPC is a zwitterionic phospholipid with a medium tail consisting of 16 carbons. It is of a length appropriate to the study of the effect of
melittin and yet is still relevant to realistic biological membranes,
which typically include 16-20- or 22-carbon-long phospholipids. In
addition, DPPC is one of the best studied phospholipids, by both
experiment and simulation, and has been termed the "benchmark" of
lipid simulations (Tieleman et al., 1997). Thus the well-characterized DPPC system sets a clear baseline for studying the effect of the embedded peptide on membrane properties.
The DPPC phospholipid bilayer model used in this study is based on the
model developed by Feller and Pastor (Feller et al., 1997). Molecular
dynamics simulations of this system under NPAT conditions, constant
normal pressure, and fixed surface area (Feller and Pastor, 1996;
Feller et al., 1997), involving periodic boundary conditions, were
shown to be reliable. These simulations agree with experimental
results, such as deuterium order parameter and electron density
profiles. Feller and Pastor's results also support the validity of the
empirical phospholipid force field included in the CHARMM 22b4
parameter set (MacKerell et al., 1998; Schlenkrich et al., 1996). The
pure membrane model includes 72 DPPC phospholipid molecules arranged in
a square 36 × 2 bilayer (with periodic boundary conditions)
corresponding to the biologically active L
phase of the membrane. The bilayer size is 47.6 Å × 47.6 Å × 68 Å,
set to allow the area per phospholipid headgroup to be 62.9 Å2. This value for the area per headgroup was
suggested by Nagle et al. (1996) and was shown by Feller and Pastor to
best reproduce experimental results (Feller et al., 1997). The bilayer
is flanked by a 30-Å layer of TIP3P water (using periodic boundary
conditions in the direction normal to the bilayer). A total of 2509 water molecules were included in the pure membrane simulation. The
initial conformation of the equilibrated bilayer was from Feller and
Pastor (Feller et al., 1997). An additional 100 ps of dynamics was
performed on the pure membrane model, to obtain a baseline to which the melittin/bilayer system will be compared.
The CHARMM molecular dynamics program (Brooks et al., 1983) and the
CHARMM all-atom phospholipid force field (Schlenkrich et al., 1996)
were used. The SHAKE algorithm was applied to all bonds involving
hydrogen atoms, and a 2-fs time step was used. The nonbonded
Lennard-Jones interactions were switched to zero over the region 12-14
Å. Ewald summation was used for the calculation of electrostatics
(Feller et al., 1996). The real space summation was truncated at 12 Å,
using 
= 0.210 Å
1. The
reciprocal space summation was truncated at
kxx = kyy = 6 and
kzz = 9. Three-dimensional periodic
boundary conditions were applied, and the cell length normal to the
membrane was allowed to adjust to maintain a constant normal pressure
of 1 atm, using a Langevin piston algorithm (Feller et al., 1995). A
mass of 500 amu and a collision frequency of 5 ps1 were used for the pressure piston.
The simulation temperature of 320 K, which was controlled by a
Nose-Hoover thermostat, is above the gel-liquid phase transition temperature for DPPC (314.5 K in multilamellar vesicles) (Koynova and
Caffrey, 1998). This temperature ensures that the system is at the
biologically relevant, fully hydrated liquid crystalline L phase of
the bilayer, yet is still low enough to ensure the stability of the
embedded peptide.
Combined melittin/bilayer system
Before we introduced the peptide into the membrane, the water layer had to be broadened to ensure full solvation of the peptide's C-terminus and to prevent an artificial interaction of the peptide with the lower membrane surface (through the periodic boundary conditions). For this end a 12-Å slab of water from the center of the water layer was copied and reintroduced next to the original slab, allowing a small overlap between the new layer and the original layer. The combined water layer was then minimized for 200 steepest descent steps, which was sufficient to relax most of the van der Waals overlaps. The excess water spreads uniformly during the equilibration process, resulting in a water layer of ~40 Å, with a uniform water density similar to that of the original system. Attempts to extend the water layer without an overlap region resulted in density fluctuations and cavity formation. At the end of this process the system included 3207 water molecules, and the dimensions of the extended system were 47.6 Å × 47.6 Å × 80 Å.
An important issue in modeling a combined peptide/membrane system is
the initial placement and orientation of the embedded peptide. In this
study, melittin was introduced into the membrane in a vertical
orientation, i.e., perpendicular to the membrane surface. This initial
placement means that the simulation will not address the process of
peptide insertion or the change in orientation from parallel to
transbilayer (these processes are too slow and cannot be studied on the
time scale of MD simulations). Rather we study the reciprocal effects
between peptide and membrane after the peptide has already been
embedded in the membrane in the transbilayer orientation. Following the
neutron-scattering measurements of Bradshaw et al. (1994), which
proposed that melittin with an unprotonated N-terminus binds parallel
to the membrane surface, whereas melittin with an protonated N-terminus
binds in a transbilayer way, the N-terminus of melittin in the present study was protonated. Practically, the melittin peptide was placed in
the cavity formed by removing two DPPC phospholipids from the center of
the upper layer. It was then rotated around its z axis to
find the optimal rotational orientation that would best fit the cavity.
The orientation with the least amount of bad contact, i.e., with the
lowest van der Waals repulsion, was selected as the starting
conformation. The vertical positioning was based on the experimental
observation (Vogel and Jähnig, 1986) that the peptide is
practically perpendicular to the bilayer plane with residue
Trp19 slightly above the plane defined by the C1
carbons of the lipid hydrocarbon chains. Thirty-six water molecules
were removed from the model because of overlaps with the peptide
C-terminus amino acids. Two hundred steepest descent minimization steps
of the peptide alone, followed by 2000 steepest descent steps of the combined model (peptide + phospholipids + water), were required to
relax the system and remove practically all bad contacts. At the end of
this process the peptide was positioned nicely in the membrane. No
significant cavities were formed because of the insertion process (as
indicated by the fractional free volume discussed below). Although the
embedded peptide also interacts with the lower layer of phospholipids,
no special treatment was assigned to it because the peptide penetration
is limited to the low-density region, near the membrane midplane, of
this layer (see the density profiles below). The method used in this
study for placing the peptide in the membrane resembles the method of
Shen et al. (1997), which was later adopted by Tieleman and Berendsen
(Tieleman and Berendsen, 1998; Tieleman et al., 1999). In both methods
the peptide is inserted into a precreated cavity in the membrane
(although in the present study there was no need for an additional
cylindrical repulsive force to expand the cavity). In the studies of
Roux, Woolf ,and collaborators, phospholipids were added around a
preexisting helix (Belohorcova et al., 1997; Woolf, 1997, 1998; Woolf
and Roux, 1996).
The complete combined model included the melittin peptide, 70 phospholipids, and 3171 water molecules, a total of 19,049 atoms all together. After the initial preparation of the model it was gradually heated from 100 K to 320 K over a 15-ps period. The combined peptide-membrane-water system was reequilibrated for 200 ps (recall that the pure membrane was already equilibrated), followed by a 300-ps production run at 320 K, totaling 500 ps of simulation. Conformation snapshots of the trajectory were saved every 0.5 ps throughout the simulation. All simulations were performed on an IBM SP2 super computer, at a rate of 70 min/ps, using 16 processors in parallel.
The length of the simulation was dictated by available computer power
and the assumed time scale of the processes of interest. Previous
simulations of membrane systems indicated that relaxation of perturbed
bilayers takes place on the 200-ps time scale (Feller et al., 1997), so
we expect the membrane to adjust to the presence of the peptide well
within the time scale of the simulation. Clearly, the simulation length
is not sufficient for the study of a possible orientational
reorganization of the helix from the perpendicular to the parallel
orientations. Significantly longer lime scales are required for such a study.
Fig. 1 shows a detailed view of the combined system after 500 ps of molecular dynamics (about half of the membrane phospholipids "in front" of the peptide were removed to obtain a clearer view).
|
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
THE SIMULATION
RESULTS
DISCUSSION
REFERENCES
|
|---|
Density profile of the combined system
Molecular density profiles present distributions of certain
molecular components, atom or chemical groups, along the axis perpendicular to the membrane surface. Previous studies (Tieleman et
al., 1997) showed that density profiles calculated for pure membrane
bilayers are characterized by three phenomena: 1) a rough membrane
surface characterized by a broad headgroup region, ~10-13 Å wide;
2) water penetration that reaches deep into the headgroup region and up
to the carbonyl groups of the membrane; 3) a wide distribution of the
CH3 segments, indicating that tails can fold back
on themselves to give considerable CH3 density
even far from the membrane midplane.
The density profiles in Fig. 2 are calculated for the combined membrane-peptide-water system. Unlike standard density profiles, which show the density at the level of the atoms (atoms/A3), the density plots of Fig. 2 are normalized and are calculated at the level of chemical groups. This means that the area under each density profile is normalized to one. These normalized density plots were chosen because they highlight the membrane interface (the phosphate group) and the membrane midsection (CH3 groups). Fig. 2 shows normalized density plots for key chemical groups in the bilayer system (phosphate, CH2, CH3, and water) averaged over the last 300 ps of the dynamics simulation. The atomic density of the embedded peptide is also shown. It should be noted that the values in the normalized density plots are quite different from those in standard density plots. For example, in this system the normalized density of bulk water is 0.023. Translated back to standard density, this normalized value is equal to the expected experimental density of bulk water (0.033 molecule/A3).
|
Fig. 2 indicates that the density of the different membrane components
in the combined system is similar to that of the pure membrane. In
particular, the width of the interface, as reflected by the width of
the density peak of the phosphorous atoms, is 12 Å, similar to the
values calculated for pure bilayers (10-13 Å) (Tieleman et al.,
1997). A slightly broader estimate of the surface corrugation is
obtained from the width of the peaks that correspond to the glycerol
carbon atoms (not shown in Fig. 2), which in the combined system is
~14 Å. Finally, the very broad CH3 peak
characteristic of pure bilayers is also observed in the combined system
(~14 Å on both sides of the membrane midplane). Determining the
average width of the membrane depends on the chemical group used to
define the membrane surface. A definition based on the peak density of
the glycerol carbon atoms (not shown in Fig. 2) yields an average
membrane width of 33 ± 2 Å, whereas a definition based on the
phosphate groups yields an average membrane width of 38 ± 2 Å (the standard deviation of each phosphate peak ±1 Å).
The peptide density profile is, of course, a unique feature of the combined system. Fig. 2 shows that the hydrophilic C-terminus of the peptide extends ~12 Å above the average membrane surface (defined by the phosphorous atoms). The hydrophobic N-terminus of the peptide, on the other hand, penetrates ~9 Å into the lower half of the bilayer. The large number of water molecules flanking the membrane is clearly seen.
Peptide structure
The reciprocal effect of the membrane on the peptide and of the peptide on the membrane are the focus of the present study. Although these effects are coupled, we shall first discuss the effects of the membrane on the structure of the peptide, following which the peptide's effect on the structural properties of the membrane will be presented. The structural properties of the embedded peptide are gauged through the following properties: the structural integrity of the peptide, its intrahelix bend angle, the tilt of the peptide relative to the membrane normal, and its vertical position relative to the membrane surface.
Structural integrity of the peptide
The issue of melittin's "structural integrity" corresponds to the question of whether its overall helical structure remains intact in the phospholipid bilayer. In the present work the helical structure is characterized simply by the peptide's backbone (
,
) dihedral
angles, which in an ordered -helix are in the vicinity of
60o. Fig. 3 shows
the values obtained for each of the peptide's backbone dihedral angles
during the 500 ps of the simulation (sampled every picosecond). As can
be seen, the majority of both (Fig. 3 A) and (Fig. 3
B) dihedral angles reflect a very stable -helical structure, with values in a 30o range around
65o for the dihedral angles and values in a
30o range around 45o for
the dihedral angle. The standard deviation of individual dihedral
angles was between ±7o and
±12o. The main exception to the structural
stability of the peptide is the three N-terminal residues
(Gly1-Ile2-Gly3),
whose backbone dihedral angles deviate from helical with higher than
standard fluctuations (in particular, Gly3, with
values of 17.2o ± 40o
and 71.0o ± 29o for the
and dihedral angles, respectively). This "unwinding" of the
flexible helix terminus is strongest in residues
Gly1 to Gly3, but can also
be seen to affect the dihedral angle of residues Ala4 (29.6o ± 12o) and Val5
(18.8o ± 17o). It is
also manifested in a increased level of fluctuations in the dihedral angle of residue Ala4
(±13o). On the time scale of the simulation,
this structural perturbation does not propagate further along the
helix. The flexibility and "unwinding" of the N-terminus may be
explained in two ways. The first is that the "unwinding" is due to
its location in the relatively low density and structureless interior
of the membrane, especially given the presence of two Gly residues in
this section of the peptide. A second explanation may be that the
"unwinding" is caused by the presence of penetrating water
molecules (Bachar and Becker, 1999). A similar loss of the -helical
character of the N-terminus segment Gly1 to
Val5 was observed in the study of Berneche et al.
(1998) after protonation of the N-terminus and penetration of water
into the bilayer (peptide orientation parallel to the membrane-solvent
interface). Further support for the water-induced "unwinding" can
be obtained from a previous computational study based on a simplified
model (Milik and Skolnick, 1993), which found that the helix is better
defined at the membrane-embedded N-terminus of melittin than the
surface-anchored C-terminus.
|
dihedral angle
of residues Thr10 and Gly12
(±12.2o and ±13.7o,
respectively) and, as expected, a different value for
Pro14 ( = 17.3o ± 10.7o). The stability of the bend region
agrees with the observation that the intrahelical bend angle fluctuates
very little, on the order of ±4o, around its
average value (see below). Fig. 3 also shows that the helical structure
at the peptide's C-terminus is kept intact throughout the simulation.
This stability is due to the many specific interactions that it forms
with the phospholipid headgroups.
Intrahelical bend
The membrane-embedded helix of melittin comprises two "rods" connected by a characteristic kink between residues Thr10 and Pro14. The intrahelical bend angle of melittin was determined to be 20o in methanol (Bazzo et al., 1988) but
60o in water (Terwilliger and Eisenberg, 1982).
Molecular dynamics simulations of melittin in vacuum also resulted in
an intrahelical bend angle of ~20o (with large
fluctuations) (Pastore et al., 1989).
In the present study the intrahelical bend angle is defined as the
angle between two axes, one characterizing the lower "rod" of the
helix (axis defined between the center of mass of the backbone atoms of
residues 5-8 and the center of mass of the backbone atoms of residues
7-10) and another axis characterizing the upper "rod" (a similarly
defined axis between the backbone center of mass of residues 14-17 and
17-20). Because the simulation started from the x-ray structure 2mlt
(Terwilliger and Eisenberg, 1982), the initial intrahelical bend angle
was 60o, as in water. However, within less than
10 ps of dynamics the intrahelical bend angle changed to
~30o and remained at that value throughout the
dynamic simulation. Fig. 4 shows the
intrahelical bend angle of melittin during the 500 ps of the dynamics
simulation. The average value of this bend angle, during the last 300 ps of the simulation, is 29.4o ± 4.1o. As expected, this bend angle is closer to
the value in methanol than to that in water. A similar bend angle of
35o ± 5o was reported by
Berneche et al. (1998) for melittin with an unprotonated N-terminus
bound parallel to the surface.
|
Helix tilt relative to membrane normal
An important property that characterizes membrane-embedded helices is their orientation relative to the membrane plane. In the case of melittin, both experimentally and theoretically, the peptide's orientation relative to the bilayer normal is not clear. As reviewed above, some experiments indicated a perpendicular orientation, while others indicated a parallel orientation. The results obtained from simplified computational models were also inconclusive, with one study pointing toward a perpendicular orientation and another indicating a broad range of possible orientations. This complex scenario probably indicates that melittin adsorbs to the membrane surface (parallel orientation) before its insertion into the membrane (perpendicular orientation). In the present study, the initial orientation of the peptide was set perpendicular to the membrane surface, to mimic the postinsertion conformation. During the course of the dynamics the peptide's orientation changed and it acquired a tilt relative to the membrane normal. This tilt is clearly visible in Fig. 1, which shows the final conformation of the system after 500 ps of dynamics. Fig. 5 depicts the tilt angle between the upper part of the membrane-embedded helix of melittin (defined as residues 14-21) and the normal to the membrane surface (z axis). The first change in orientation occurred very quickly during the initial heating of the system, when the peptide attained a small 10o tilt angle. After the initial change, the peptide's tilt fluctuates strongly and gradually increases over the rest of the simulation. The average tilt angle over the last 300 ps of the simulation is 21.8o ± 4.2o relative to the membrane normal. A slightly higher value of 23.9o ± 3.0o is obtained from averaging only over the last 200 ps of the simulation.
|
; Shen et al., 1997)
showed that the orientation of these helices, relative to the membrane
normal, fluctuates significantly, and they obtain a tilt angle of up to
30o. Clearly, much longer simulations are needed
before this short time scale tilt can be compared to experimental
results, which average the tilting process on much longer time scales.
Vertical position
An interesting question that may reflect on the quality of the simulation is whether the vertical position of the peptide has changed during the simulation. A large variation in this quantity could indicate that the initial vertical positioning of the peptide, with residue Trp19 near the plane defined by the phospholipid C1 atoms, was inappropriate. During the 500 ps of the simulation there was practically no change in the vertical position of the peptide relative to the average surface of the membrane nor any perceived change in the overall width on the membrane (although deformation of the surface at the intracellular side makes it hard to determine the exact width near the peptide; see below). Calculations of the average z coordinate of the topmost residue Gln26 showed no vertical drift and only relatively small fluctuations of ±0.7 Å around its average position. Even smaller fluctuations of ±0.5 Å were observed for residue Trp19, which also did not change its vertical position during the simulation. These results indicate that the vertical position of the peptide at the membrane interface is stable on the time scale of the simulation. No conclusions regarding the stability of this vertical position over longer time scales can be drawn.Membrane Structure
Membrane order parameter
Deuterium order parameters, obtained from NMR experiments, are often used to probe the structure of lipid bilayer membranes. The order parameter, SCD, is given by
|
(1) |
is the angle between the CD bond vector and the bilayer
normal, and the brackets denote an average over time over all of the
lipids (or over a subset of the membrane lipids). The value of
SCD quantifies the degree of
reorientation that occurs on the NMR time scale, i.e., how ordered the
molecules are and their average orientation with respect to their
bilayer normal. A vector undergoing isotropic rotation will have an
order parameter of zero. Typical values of
SCD for fluid-phase lipid bilayers range from ~ 0.2 near the headgroups to near zero in the
terminal methyl groups. Order parameters along the hydrocarbon chains
of a fully hydrated DPPC bilayer were determined experimentally at 50°C (Douliez et al., 1995; Seelig and Seelig, 1974). A useful comparison parameter is the average of the CH order parameters in the
plateau region between C4 and C8,
SCD
[4,8]. For DPPC
the experimentally determined value of this parameter is in the range
of 0.209-0.217. This value was reproduced in the MD simulations of
Feller et al. (1997) at 50°C
(SCD[4,8] = 0.215),
as well as in our simulation of a pure DPPC system
(SCD[4,8] = 0.209 ± 0.006). Individual fluctuations in the order parameter of
individual carbon atoms are on the order of 0.1.
Because melittin does not span the membrane from side to side, its
effect on the order parameter of phospholipid in the upper (i.e.,
extracellular) layer has to be studied separately from its effect on
the lower (i.e., intracellular) layer. Furthermore, to study the effect
of peptide proximity on the order parameter of the phospholipids, these
were grouped into three equal size "tiers" based on their distance
from the peptide. The first group includes the 11 phospholipids closest
to the peptide (the distance in the horizontal two dimensions
x and y, defined as the distance between the
peptide's center of mass and the phospholipid's center of mass, is
less than 18 Å). The second group includes 11 phospholipids that are
at an intermediate distance from the peptide (horizontal distances
between 18 Å and 23.5 Å). The third group includes the last 12 phospholipids, which are the farthest from the peptide (horizontal
distances over 23.5 Å). Separate order parameter profiles were
calculated for each group. It should be noted that in addition to the
advantages for the analysis, this partitioning has the disadvantage of
increasing the errors (due to the decrease in sample size).
Fig. 6 depicts the order parameter
profiles calculated for the phospholipids in the upper layer of the
membrane. Shown are the deuterium order parameters for the
phospholipids closest to the peptide in comparison to the order
parameters of phospholipids farther away (averaged over the second and
third groups) and to the order parameter of the pure DPPC membrane. The
order parameters are averaged over the last 300 ps of the simulation. A
dramatic effect is seen in the order parameter of the lipids closest to the peptide, which show a significant reduction in
SCD values, compared to phospholipids
far from the peptide, which exhibit an order parameter profile similar
to that of a pure bilayer. The average order parameter in the plateau
region, between C4 and C8, for the lipids closest to the peptide is
SCD[4,8] = 0.157 ± 0.009 (i.e., standard deviation of 0.009) compared to SCD[4,8] = 0.215 ± 0.006 for lipids in the second and third tiers, and
SCD[4,8] = 0.209 ± 0.006 for the pure bilayer. A qualitatively similar
reduction in order parameter near embedded peptides was observed in the
simulation of alamethicin, influenza M2, and OmpF embedded in a POPE
membrane (Tieleman et al., 1998). However, in that study the maximum
reduction in order parameter, which was in the presence of OmpF, was
~15%, while in the present study the reduction in order parameter
near the peptide is greater, almost 25%.
|
, suggests that
the decrease in order parameters close to an embedded peptide is due to
a change in the overall tilt of the lipids. They argue that the
observed 25-30o tilt can account for the
observed 15% decrease in order parameter in the case of OmpF in
DOPE. Because the lipid tilt in this study is similar to the above
value (see below), it is possible that also in the melittin/DPPC system
the tilts contribute ~15% to the reduction in order parameter. A
question still remains regarding the origin of the additional 10%
reduction in order parameter. As will be discussed below, both the rate
of dihedral transitions and the percentage of gauche defects
in the phospholipids closest to the peptide were similar to those for
more distant phospholipids, preventing a clear distinction between the
different scenarios. Order parameter values alone are not sufficient to
determine whether the upper layer of the bilayer exhibits a higher
level of disorder near the peptide. It is likely that significantly
longer simulations will be required to settle this question.
An interesting phenomenon is also observed in Fig. 6 at the tail end of
the acyl chains. As expected, carbons in the tail region are more
disordered than carbons at the plateau region closer to the headgroups.
However, in the presence of the peptide we find that for all
phospholipids, both close to and far from the peptide, the values of
SCD for the C13 to C15 are 15-20%
higher than the equivalent values in the pure DPPC bilayer. It seems that the presence of the peptide in the relatively unstructured low-density region, toward the membrane's midsection, slightly increases the order in that region (which is still disordered compared
to the plateau region). It is interesting to note that in this region,
carbons 13-15, the effect is seen throughout the three tiers of phospholipids.
The dip in the order parameter profile immediately behind the
headgroup, SCD = 0.15-0.17, seen in both pure
and peptide-embedded bilayers, is also observed in NMR experimental
data at 350 K, SCD = 0.16 (Marrink,
1994; Tieleman et al., 1997). It is likely that the observed low-order
parameter at the surface is a consequence of a reduced amount of motion
near the interface.
Although the peptide was inserted into the upper extracellular later of
the membrane, it also partially penetrates into the lower intracellular
layer. In fact, the density profiles depicted in Fig. 2 show that the
peptide penetrates as much as 9 Å into the lower layer and as such is
likely to affect the order in that part of the system as well. Fig.
7 shows the deuterium order parameter profiles of the phospholipids in the lower layer of the membrane, grouped according to their distance from the peptide. Also shown is the
order parameter profile of the pure membrane. With the exception of the
order parameter of lipids in the third tier, farthest from the peptide,
the deuterium order parameters in the lower layer are qualitatively
similar to those in the upper layer. That is, phospholipids closest to
the peptide show a dramatic drop in their order parameter values, while
lipids in the second tier exhibit an order parameter profile similar to
that of a pure bilayer. The corresponding plateau averages are
SCD[4,8] = 0.121 ± 0.019 for lipids in the first tier and
SCD[4,8] = 0.205 ± 0.016 for lipids in the second tier.
|
SCD[4,
8] = 0.121 ± 0.019 in the lower layer compared to
SCD[4,
8] = 0.157 ± 0.009 in the upper layer). This effect
is present despite the fact that the peptide penetrates only to the
tail region of these lipids. It is not in direct contact with any of
the methylene groups that comprise the "plateau" region. In fact,
the reduction in order parameter increases significantly for methylene
groups that are farthest from the peptide. The average value for
carbons C2-C6 in lipids at the first tier in the lower layer is
SCD[2,6] = 0.105 ± 0.014, almost 45% smaller than the equivalent value in the upper layer
SCD[2,6] = 0.151 ± 0.017. However, the average value for carbons C7-C12 is
SCD[7,12] = 0.144 ± 0.008, only 13% lower than the corresponding value in the upper layer,
SCD[7,12] = 0.162 ± 0.005. The fact that the methylene groups that exhibit
the strongest effect are not in contact with the peptide indicates that
the reduction in order parameter in this case is not due to restricted
motion in the vicinity of the peptide. These observations support the
conclusion that the observed reduction in
SCD may be attributed, to a
significant degree, to a real increase in disorder in the lower layer.
In addition, lipids in the lower layer adopt only a very small tilt angle relative to the membrane normal (see below), further reducing the
role of this tilt as a major source for the observed reduced order parameters.
Second, the increase in order toward the tail end of the chains,
which was observed in the upper layer, is less prominent in the lower
layer. It is seen mainly in the second tier of lipids. Finally, the
most striking peculiarity in the order parameter profiles of lower
layer is the order parameter associated with lipids in the third tier,
furthest from the peptide. While such lipids in the upper layer behave
as if they were part of a pure bilayer, the phospholipids furthest from
the peptide (the third tier) in the lower layer exhibit a surprising
increase in their order parameter, with a plateau average of
SCD[4,8] = 0.302 ± 0.04, indicating that they are more ordered than in a
pure bilayer. This, however, should be considered a nonphysical artifact of the simulation. While we took care to remove two
phospholipids from the upper layer to make room for the peptide, no
such preparation was made in the lower layer (expecting the peptide to
primarily penetrate the low-density interior of the membrane). A
consequence of this treatment is an increase in the density at
the lower half of the membrane, resulting in artificially
increased order near the simulation boundaries.
Finally, to allow comparison with experimental data, an overall order
parameter profile averaged over the whole membrane was calculated. The
calculated average plateau value for the whole membrane in the presence
of the peptide was
SCD[4,8] = 0.203. This average is ~3% smaller than the average plateau value
calculated from the simulation of the pure bilayer,
SCD[4,8] = 0.209. These results should be compared to the NMR experimental data
of Dufourc et al. (1986), which indicate that in the presence of
melittin the quadropolar splitting of a liquid crystalline DPPC bilayer
(at 61°C), which is proportional to the deuterium order parameter, is
reduced by ~12% (the effect in the gel phase at
41o was much smaller). The turnover effect at the
tail end of the acyl chain, seen in the simulations, was not observed
in those NMR experiments. Recently an experimental study by De Planque et al. (1998) of the effect of several transmembrane peptides on
various diacylphosphatidylcholine bilayers resulted in very small
effects of 17-19-residue -helices on the order parameter of DPPC at
51o (close to the simulation temperature of
47o).
Free volume in the membrane
Another indicator of the effect of the embedded peptide on the structural properties of the membrane is whether the fractional free volume in the membrane changes as a result of its presence. In general, the fractional free volume inside a biological membrane (in the functional L phase) is quite high, reflecting the rather fluid state
of this macromolecular ensemble. The "empty" free volume was
calculated by placing a grid over the system and counting the
percentage of grid points that lay outside of any atom's van der Waals
sphere. The grid density was approximately one point per 0.5 Å in the
x and y directions, and one point per 0.8 Å in the z direction. The fractional free volume calculated for
the hydrocarbon region of the pure DPPC bilayer model (excluding the headgroup region), averaged over 100 ps of molecular dynamics, is
54.0 ± 1%. This value agrees with those computed for a DPPC bilayer in a previous study by Marrink et al. (1996). Using a grid
density of about one point per 0.5 Å, they found an average free
volume of 40% at the ordered part of the membrane near the interface,
and close to 60% free volume at the membrane interior (in that study a
slightly finer grid was used).
Initially, embedding the peptide in the membrane did not change the
fractional free volumes, indicating that the embedding process did not
create artificial cavities in the membrane (cavities could have been
created by geometric mismatches between the peptide and the empty space
created after removing the two phospholipids from the center of the
upper layer). In fact, the same percentage of free volume was
maintained during the first 270 ps of the dynamic simulations. During
this period the average fractional free volume in the membrane's
hydrophobic region was 54.6 ± 1%, similar to that obtained in a
pure membrane. Fig. 8 compares the
fractional free volume in the lipid region of the bilayer for the last
300 ps of the combined melittin/membrane simulation and in a pure DPPC
membrane (for 100 ps of dynamics). At time t 
270 ps
there is a sudden increase in the fraction free volume inside the lipid region of the membrane. After this increase, and for the remainder of
the simulation, the fractional free volume inside the lipid region of
the membrane stabilizes around a new average value of 57.4 ± 1.4%, almost 3% higher than fractional free volume in the pure
membrane. Further analysis showed that the fractional free volume
increases in both the upper and lower layers of the membrane. However,
the increase in free volume in the lower layer was a little larger than
in the upper layer (by ~1%).
|
Lipid tilts relative to membrane normal
Fig. 1, which depicts the system's conformation at the end of the 500 ps of dynamics, shows the striking correlation between the peptide's tilt and the average tilt of the upper lipid layer. In the above discussion it was shown that in the course of the dynamic simulation the peptide acquires a 24o tilt relative to the membrane normal. Given the elongated nature of the lipid hydrocarbon chains, it is not surprising that the peptide tilt is correlated with the overall orientation of the lipid chains in the membrane. Fig. 9 shows the average tilt of the lipids in the upper and lower layers of the membrane as a function of time. The difference between the two layers is clear. Lipids in the lower layer, which is less perturbed by the peptide, are essentially perpendicular to the membrane surface, exhibiting an average tilt angle of 8o ± 2o. However, one cannot rule out the possibility that the limited tilt in the lower layer may be an artifact cased by the slightly higher density in the lower layer. Lipids in the upper layer, on the other hand, acquire a tilt angle similar to that of the peptide, averaging 24o ± 2o over the last 200 ps of the simulation. Overlaying the average tilt of the membrane and the tilt of the peptide (Fig. 9) shows that these two quantities are highly correlated throughout the simulation. In fact, the correlation between the two curves is such that the average difference between the two curves is only 3.8o ± 2.8o. The tilts observed for the upper layer are similar to the 25-30o tilts reported for other membranes with embedded peptides (Belohorcova et al., 1997; Shen et al., 1997;
Tieleman et al., 1998).
|
Lipid dihedral angles
As discussed above, the dramatic effect of the embedded peptide on the order parameter of the phospholipids closest to it is expected to be reflected in the dynamic properties of the acyl-chain dihedral angles. In biological membranes the lipid chains tend to adopt an extended trans conformation, although bends and turns involving local gauche conformations are known to occur (see, for example, the very broad CH3 peak in the systems density plot in Fig. 2). To check for the peptide's influence on the dynamics of lipid chain conformations, we compared the trans/gauche tendencies as well as dihedral transition rates for lipids close to the peptide and for lipids far from it. The results did not reveal a clear difference between the two groups, especially as there is a very large variance within each of them. For example, the percentage of gauche conformation observed for individual dihedral angles (at the central segment of the lipid chain, during the last 250 ps of the simulation) varied in both groups from 0% to ~55% of the time. Likewise, the rate of dihedral transitions was also characterized by a broad variance. Observed transition rates for individual dihedral angles (at the central segment of the lipid chain) varied from 0 to as much as 15 back-and-forth transitions during the 500 ps of the simulation. No clear difference between lipids close to and far from the peptide was observed. In some cases a high percentage of gauche was associated with many back-and-forth transitions between the two conformations (e.g., the dihedral angle defined by carbons C4-C5-C6-C7 of one of the close lipids shown in Fig. 10 A, which, all together, exhibits 41% gauche). In others, a similar value of gauche was associated with a single trans-gauche transition, following which the lipid maintained its new conformation (e.g., the dihedral angle defined by carbons C6-C7-C8-C9 of one of the close lipids shown in Fig. 10 B, which, all together, exhibits 56% gauche).
|
Surface corrugation
Common to all computer simulations of lipid bilayers is the significant corrugation of the membrane surface. The present study, in which the surface corrugation is between 12 Å (estimated from the width of the phosphorous density profile; Fig. 2) to 14 Å (estimated from the density peak of the glycerol carbons), is no exception to the rule. An interesting question is whether the presence of the embedded peptide effects increases or reduces the corrugation of the membrane surface. For the upper layer, in which the peptide was embedded, the results of the present simulations did not show any significant change in surface corrugation resulting from the presence of the peptide. Changes in the vertical position of the different phospholipids during the dynamic simulation were similar to those observed in the pure bilayer resulting from normal fluctuations. They could not be correlated with the presence of the peptide. The standard deviation in the average vertical position of the headgroups in the upper layer was similar in pure and peptide-hosting membrane models, with the latter exhibiting a slightly smaller degree of fluctuation. This picture changes for the lower layer of the membrane. A significant degree of deformation was observed in the lower lipid layer near the peptide's N-terminus. Most striking is the response of the phospholipid closest to the N-terminus, which moved deeper into the bilayer. Fig. 11 shows the position of the phosphate atom of this phospholipid as a function of time. It is seen that within the first 200 ps the headgroup of this phospholipid moved as much as 2 Å upward into the bilayer (for comparison, its fluctuations in the new location are only ±0.4 Å). Smaller deformations were observed in a few other phospholipids in that vicinity. This deformation correlates well with observed penetration of water from the intracellular side of the membrane toward the protonated N-terminus of the peptide. A detailed analysis of water penetration into the bilayer in the presence of melittin is presented elsewhere (Bachar and Becker, 1999). In general, water was observed to penetrate
from both sides of the membrane within 200 ps of the simulation. From
the extracellular side water penetrate toward the charged residue
Lys7, and from the intracellular side water
molecules penetrate toward the protonated N-terminus.
|
).
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
THE SIMULATION
RESULTS
DISCUSSION
REFERENCES
|
|---|
The molecular dynamics calculation presented here studies peptide-membrane interaction for the 26-amino acid peptide melittin (with a protonated N-terminus) embedded in a transbilayer orientation in a fully hydrated bilayer consisting of 72 DPPC phospholipids. The simulated membrane is large enough to study the effect of the embedded peptide on the structural properties of the host membrane. Indeed, the peptide's effect was demonstrated in many of the bilayer properties, including deuterium order parameters, fractional free volume, lipid tilts, and surface corrugation.
As reviewed above, one of the putative mechanisms for melittin-induced
lysis suggests that this phenomenon occurs through some sort of
deformation of the lipid bilayer. A goal of the present simulation was
to check this hypothesis and see whether the peptide induces membrane
disorder and deformation. A first conclusion from the simulation is
that the effect of the peptide on the membrane is local, limited to
those phospholipids in its immediate vicinity. This conclusion is in
line with the experimental observations of De Planque et al. (1998) and
the simulation results of Tieleman et al. (1998).
Interestingly, in this respect we find a difference between the two membrane layers. There is evidence supporting an increased level of disorder in the lower (intracellular) layer of the membrane in the vicinity of the perpendicularly oriented peptide. The increased disorder is indicated by 1) a significant reduction in order parameter close to the peptide's N-terminus, especially for methylene groups that are not in direct contact with the peptide, i.e., toward the intracellular interface; 2) a significant deformation of the intracellular interface just "below" the protonated N-terminus of the peptide; and 3) a preferential increase in the fractional free volume in the lower layer of the membrane (although the upper layer also exhibits an increase in fractional free volume). On the other hand, it seems that the upper, extracellular layer (into which melittin was inserted) is less effected by the embedded peptide. For this layer, however, the results are much less conclusive. The decrease in order parameter of lipids close to the peptide in the upper layer may result from restricted motion due to the peptide's presence or from the lipid's tilt rather than from an increased level of disorder. Furthermore, the decrease in order parameter was not accompanied by an increased rate in dihedral transitions nor by any significant surface deformation of the upper interface.
Alternatively phrased, it seems that the origin of the asymmetrical
effect of melittin on the surrounding membrane may be the fact that
melittin is not a transmembranal helix. Rather, it is
anchored only at the extracellular surface of the membrane, spanning no
more than two-thirds of the bilayer's width. As a result, melittin's
protonated N-terminus is loose in the lower layer of the membrane,
leading to an increased amount of disorder in lower-layer phospholipids
near the peptide. On the other hand, being anchored at the
extracellular interface, melittin has a smaller effect on the
structural properties of this layer. Recall that the effect of melittin
on the surrounding membrane, in both layers, was localized to its
immediate vicinity. In general, the simulation supports a role for
local deformation of the phospholipid bilayer in the mechanism of
melittin-induced lysis, possibly in relation to water penetration
(Bachar and Becker, 1999).
In this respect, melittin is different from the other membrane-embedded
peptides studied so far by MD simulations. In most other studies, the
embedded peptides were transmembranally anchored in both intra- and
extracellular sides of the membrane (Belohorcova et al., 1997; Shen et
al., 1997; Tieleman et al., 1998, 1999
