Tài liệu Báo cáo khoa học: Perturbation of membranes by the amyloid b-peptide – a molecular dynamics study pptx

Perturbation of membranes by the amyloid b-peptide –
a molecular dynamics study
Justin A. Lemkul and David R. Bevan
Department of Biochemistry, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA
Alzheimer’s disease is a neurodegenerative disorder in
which the hallmark symptoms include cognitive decline
and dementia [1]. Characteristic of this disorder is the
formation of extracellular amyloid fibrils, and intra-
cellular deposition of hyperphosphorylated tau [2].
Alzheimer’s disease is considered to affect approxi-
mately five million Americans, and this number is
expected to triple by the year 2050, according to recent
estimates from the Alzheimer’s Association. The num-
ber of Alzheimer’s patients worldwide has recently
been estimated at 20–25 million [3]. With the current
annual health care costs estimated at $100 billion in
the USA alone, the molecular basis for this disease is a
topic of intense scientific research.
According to the ‘amyloid hypothesis’, interactions
between the amyloid b-peptide (Ab) and other cellular
components, especially membranes, are considered to
give rise to the neurotoxicity observed in Alzheimer’s
disease. Ab is derived from the amyloid precursor
protein by sequential proteolytic cleavage by two
membrane-bound proteases, b- and c-secretase [2].
The length of the peptide is variable, ranging from 39
Keywords
Alzheimer’s; amyloid; membrane;
protein–lipid interactions; simulation
Correspondence
D. R. Bevan, Department of Biochemistry,

Virginia Polytechnic Institute and State
University, 201 Fralin Biotechnology Center,
Blacksburg, VA 24061, USA
Fax: +1 540 231 9070
Tel: +1 540 231 5040
E-mail:
Website: http://
www.bevanlab.biochem.vt.edu
(Received 19 February 2009, revised 24
March 2009, accepted 26 March 2009)
doi:10.1111/j.1742-4658.2009.07024.x
The etiology of Alzheimer’s disease is considered to be linked to interac-
tions between amyloid b-peptide (Ab) and neural cell membranes. Mem-
brane disruption and increased ion conductance have been observed
in vitro in the presence of Ab, and it is assumed that these same phenom-
ena occur in the brain of an individual afflicted with Alzheimer’s. The
effects of Ab on lipid behavior have been characterized experimentally, but
details are lacking regarding how Ab induces these effects. Simulations of
Ab in a bilayer environment can provide the resolution necessary to
explain how the peptide interacts with the surrounding lipids. In the pres-
ent study, we present an extensive analysis of lipid parameters for a model
dipalmitoylphosphatidylcholine bilayer in the presence of the 40-residue Ab
peptide (Ab40). The simulated systems examine the effects of the insertion
depth of the peptide, temperature, the protonation state of the peptide, and
ionic strength on the features of the lipid bilayer. The results show that
Ab40 is capable of disordering nearby lipids, as well as decreasing bilayer
thickness and area per lipid headgroup. These phenomena arise as a result
of the unfolding process of the peptide, which leads to a disordered,
extended conformation that is capable of extensive electrostatic and hydro-
gen-bonding interactions between the peptide and the lipid headgroups.

Comparisons are made using melittin-dipalmitoylphosphatidylcholine sys-
tems as positive controls of a membrane-disrupting peptide because these
systems have previously been characterized experimentally as well as in
molecular dynamics simulations.
Abbreviations
Ab, amyloid b-peptide; Ab40
,
40-residue alloform of the amyloid b-peptide; DMPC, dimyristoylphosphatidylcholine; DPPC,
dipalmitoylphosphatidylcholine; MD, molecular dynamics.
3060 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
to 43 amino acids, with the 40- and 42-residue allo-
forms being the most common. The peptide is consid-
ered to be partially embedded in the cell membrane
[4], but it can exit over time and accumulate in the
extracellular environment, giving rise to the neuritic
plaques observed in the brains of Alzheimer’s patients.
Because Ab is localized in the plasma membrane,
an analysis of the interactions between the peptide
and the membrane environment is crucial to under-
standing the exit pathway of the peptide and the
manner in which it disrupts membranes. There are
two proposed positions for Ab within the plasma
membrane, as determined by experiments performed
in vitro: one with Val24 at the membrane–water inter-
face, the other with Lys28 at the interface [4,5]. The
location of the peptide within the membrane may
affect the types of interactions that it has with the
surrounding lipid matrix and the pathway that it fol-
lows to exit from this environment. The use of
atomic-force microscopy has concluded that the

40-residue form of Ab (Ab40) is partially embedded
in model dimyristoylphosphatidylcholine (DMPC)
micelles [6]. Work conducted in vitro with Ab in the
presence of rat synaptic plasma membranes has
shown that monomeric Ab can intercalate into the
bilayer interior and lead to decreased bilayer thick-
ness [7]. The same study also concluded that Ab40
increased the fluidity of the lipids in the membrane,
in agreement with a previous study [8]. However, the
effects of Ab on lipid fluidity are contentious because
another study found that Ab decreased the fluidity of
the surrounding lipids [9].
In disturbing the integrity of the plasma membrane,
Ab promotes the increased leakage of ions, particularly
calcium, into the cell [10]. The disruption of calcium
homeostasis, and thus the promotion of neuronal
excitotoxicity, is considered to be a component of
Alzheimer’s disease. Perturbation of the plasma
membrane in the presence of Ab has been noted in sev-
eral studies [11,12]. Although a study by Kayed et al.
[11] concluded that permeabilization of the plasma
membrane was only caused by oligomeric A b, it was
also noted that monomeric and low-molecular weight
Ab species could incorporate into the membrane and
cause a reduction in the thickness of the bilayer, and this
observation was corroborated by Ambroggio et al. [12].
These investigators found that Ab42 could stably incor-
porate into the plasma membrane and reduce the cohe-
sive forces between surrounding lipids.
Perturbation of membranes has been associated with

other toxic peptides and proteins, most notably melit-
tin, a component of bee venom that is considered to
exert its toxic effect by associating with cell
membranes [13–15]. Model systems of melittin in
dipalmitoylphosphatidylcholine (DPPC) and DMPC
bilayers have been studied by molecular dynamics
(MD) simulations [16–18], demonstrating that melittin
interacts asymmetrically with the leaflets of the bilayer
and can draw water into the membrane. That is, the
peptide disorders the leaflet with which it interacts
most closely (i.e. the extracellular face), at the same
time as increasing lipid order in the cytofacial leaflet.
Simulations of melittin in DPPC lead to an interest-
ing comparison with the Ab-DPPC simulations
reported in the present study. Both melittin and Ab
are short, mostly helical peptides that are assumed to
be asymmetrically oriented with respect to the mem-
brane. Both are considered to cause some amount of
disorder on the surrounding lipid environment.
Because the interactions between melittin and lipids
have been well-characterized in previous MD simula-
tions, we used melittin–membrane systems as a basis
for interpreting the disruptive effect of Ab40 on its
surrounding lipid environment.
The success of applying MD to membrane protein
systems has been well documented, and simulations
have illustrated the conformational dynamics of pro-
teins embedded in membranes [19,20] as well as the
interactions between proteins and the surrounding
lipids [19,21,22]. A recent review has discussed these

phenomena in detail [23], highlighting many of the
parameters that have been successfully measured in
membrane protein MD simulations. To our knowl-
edge, only three studies have examined Ab in an
explicit bilayer environment [24–26], but none of
these have reported the behavior of the lipid mem-
brane in which the peptide was embedded and,
instead, have focused primarily on the properties of
the peptide.
In the present study, we aimed to expand previous
work by examining the properties of lipid molecules
surrounding the membrane-perturbing Ab40 peptide.
Although it is known that Ab can interact with the
plasma membrane and assemble in this environment
[6], a fundamental understanding of the molecular
basis for this phenomenon is missing. Central ques-
tions still remain, especially regarding the intrinsic
characteristics (i.e. both structural and chemical) of Ab
that allow it to disrupt the surrounding lipids. Detailed
studies with atomic resolution, such as the simulations
reported in the present study, are crucial to under-
standing this phenomenon. A greater knowledge of the
most basic interactions between Ab and a model mem-
brane can lead to a more complete understanding of
the membrane-aided assembly of Ab and the resulting
damage to cell membranes.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3061
Results
Description of simulation systems

The preparation of the ten Ab-DPPC simulation sys-
tems listed in Table 1 has been described in detail else-
where [25] and only a summary of their essential
characteristics is appropriate here. The coordinates
and topology for the DPPC bilayer were obtained
from a previous study by Tieleman and Berendsen
[27], and are available at the author’s website (http://
moose.bio.ucalgary.ca/index.php?page=Structures_and_
Topologies). The goals of these simulations were to
examine the effects of different variables (i.e. ionic
strength, temperature, and Ab positioning) on the
dynamics of the peptide and the behavior of the
surrounding lipids. Systems belonging to simulation set
A were designed primarily to understand the effects of
an increased salt concentration. Simulation set B
examined the effects of both the protonation state of
the peptide and temperature on the behavior of the
system. Finally, the systems in simulation set C also
examined the effects of increased salt, but contrasted
with simulation set A in that the Ab peptide was
placed more deeply in the membrane.
Two sets of negative control systems of pure DPPC
bilayers were prepared by a similar method. These sys-
tems were designed to examine whether the additional
solvation or increased ionic strength had any back-
ground effect on lipid dynamics. Two systems were
prepared from this structure: (a) the original bilayer
with the original water-to-lipid solvation ratio (‘Origi-
nal Solvation’; OS1) and (b) this bilayer in the pres-
ence of 100 mm NaCl (OS2). In addition, three

systems were prepared by placing an additional slab of
water to one side of the bilayer to approximate the
increased water-to-lipid solvation ratio and the system
size present in the peptide–bilayer systems. Similar to
the OS simulation set, these systems contained either
no salt (‘New Solvation’; NS1 and NS3) or 100 mm
NaCl (NS2). The solvation ratios of NS1, NS2, and
NS3 closely match those of the simulated Ab systems,
although not exactly. Instead, they were designed to
strike a balance between the dimensions of the system
and the number of water molecules aiming to examine
whether or not the asymmetry of the system and
increased solvation would affect the dynamics of the
lipids. System details are summarized in Table 2.
In each simulation, coordinates were saved every
2 ps, generating 50 000 data points per simulation.
Analyses were conducted using tools within the gro-
macs software package, version 3.3 [28] (for deuterium
order parameters) and code developed in-house [29]
(for lipid tilt, effective chain length, area per lipid
headgroups, and bilayer thickness). Averaging over
time was conducted, when appropriate, to generate a
time-dependent progression of these measurements.
Positive control systems were prepared with melittin
the presence of DPPC. The structure of melittin was
taken from the crystal structure, Protein Data Bank
entry 2MLT [30]. Two orientations were prepared: one
with melittin embedded in the DPPC bilayer, as in pre-
vious studies [17,18] [‘Embedded’ (E1) and with
100 mm NaCl (E2)], and the other with melittin paral-

lel to the bilayer interface, as reported previously [16]
[‘Parallel’ (P1) and with 100 mm NaCl (P2)]. These
systems were prepared in the same manner as the
Ab-DPPC systems, giving starting configurations com-
parable to those presented in the original studies.
Details of these systems are presented in Table 3.
The initial asymmetric orientation of Ab relative to
the DPPC bilayer creates an interesting situation when
analyzing the properties of the surrounding lipid
bilayer. Over time, the peptide interacts differently
with each leaflet. Such a situation resembles that of
melittin, whose interactions with lipids have been
Table 1. Simulation system details.
System
Ionic
strength
(m
M)
Net
charge
on Ab
DPPC
lipids
Temperature
(K)
Water
molecules
Solvation
ratio
A1 0

a
)3 122 323 6912 56.6 : 1
A2 100 6873 56.3 : 1
A3 0 )2 6922 56.7 : 1
A4 100 6878 56.4 : 1
B1 0 +1 122 300 6949 57.0 : 1
B2 0 323 6948 57.0 : 1
B3 0 +7 300 6951 57.0 : 1
B4 0 323 6948 57.0 : 1
C1 0 )3 120 323 6928 57.7 : 1
C2 100 6888 57.4 : 1
a
An ionic strength of 0 mM implies counterions sufficient only to
neutralize the charge of the system.
Table 2. DPPC control simulation details.
System
Ionic
strength
(m
M)
DPPC
lipids
Temperature
(K)
Water
molecules
Solvation
ratio
OS1 0
a

3655 28.6 : 1
OS2 100 128 323 3641 28.4 : 1
NS1 0 6907 54.0 : 1
NS2 100 6881 53.8 : 1
NS3 0 128 300 6907 54.0 : 1
a
An ionic strength of 0 mM implies that no ions were added to
these systems.
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3062 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
described experimentally [31–34] and computationally
[16–18].
Validity of melittin-DPPC controls relative to
previous work
Studies by Bachar and Becker [18] and Berne
`
che et al.
[16] provide meaningful reference points to the simula-
tions of the present study. We constructed systems that
had initial configurations similar to those produced by
the original investigators, but we simulated the systems
in a different manner in certain respects. We applied
new equilibration schemes to pack the lipids around
the peptides, using different force field parameters
(GROMOS ⁄ Berger instead of CHARMM). We also
conducted simulations that were far longer than in the
original reports (i.e. 100 ns instead of 300–500 ps). The
goal of this series of simulations was to produce data
not only to validate our simulation set-up, but also to
serve as a basis for comparing the effects of Ab40 on a

DPPC bilayer in light of the observations made with
respect to melittin.
Simulation E1 was inspired by the work of Bachar
and Becker [18], and simulation E2 arose from our
desire to examine the effects of increased ionic strength
on the peptide–membrane systems. Even in our longer
100 ns simulations, the positioning and orientation of
melittin at the end of the simulations were similar to
that reported by Bachar and Becker [18], although, in
our simulations, melittin became embedded more dee-
ply in the bilayer and more disordered at its termini.
The disordering at the termini was predicted by Bachar
and Becker [18], although it was not observed in the
timeframe that they simulated.
With respect to the lipid properties, the most mean-
ingful comparison between the present study and pre-
vious simulations arises with respect to lipid order.
Deuterium order parameters describe the orientation
of the lipid acyl chains, on average, relative to the
bilayer normal. These parameters are calculated by the
equation:
ÀS
CD
¼
3 cos
2
h À 1
2

ð1Þ

In Eqn (1), h represents the angle between the C-D
bond and the bilayer normal, and the angle brackets
denote that the values are averaged over all equivalent
atoms, and over time.
We observed that the lipids nearest melittin experi-
ence a greater degree of disorder, whereas more distant
lipids become more ordered relative to control simula-
tions in the absence of melittin. This disordering effect
is comparable to the results obtained in the original
studies [16,18]. In addition, the top leaflet of the
bilayer, which interacts with melittin most strongly,
was observed to be more disordered relative to the
bottom leaflet, which experienced a greater degree of
chain elongation and lipid packing. Bachar and Becker
[18] divided the lipids in their bilayer into ‘tiers’ based
on the distance between the protein and lipid molecule
center of mass. The average value of ) S
CD
was pre-
sented for the ‘plateau region’ of the acyl chain, which
extends from carbons 4 to 8 of the acyl chain (denoted
<)S
CD
>
[4,8]
). The values reported are 0.157 ± 0.009,
0.215 ± 0.006, and 0.215 ± 0.006 for the first, second,
and third tiers, respectively. We find very similar
values of 0.144 ± 0.010, 0.194 ± 0.007, and
0.219 ± 0.005 for these same subsets of lipids. We

attribute the small differences in these values to the
use of different force fields, application of different
equilibration schemes, and the length of our simula-
tions, which is several orders of magnitude longer than
that of the original study.
Similar conclusions can be made between our simula-
tion P1 (i.e. starting with melittin parallel to the inter-
face of the bilayer at the beginning of the simulation)
and the study by Berne
`
che et al. [16]. With respect to
the behavior of the lipids, we make the same observa-
tion that those of the top leaflet (which also interact
most strongly with melittin) become very disordered rel-
ative to the lipids of the lower leaflet, overall, although
our values for the deuterium order parameters are
higher. In the original study by Berne
`
che et al. [16], the
average order parameter was 0.149 in the top leaflet
and 0.188 (i.e. a difference of 21%) in the bottom leaf-
let. The corresponding values for these parameters from
our simulations are 0.157 and 0.220 (29% difference),
respectively. We attribute these differences to many of
the same factors as described above with respect to the
study by Bachar and Becker [18], and also the fact that
the simulations conducted by Berne
`
che et al. [16] uti-
lized DMPC as the membrane lipid instead of DPPC,

so that some differences should be expected.
Table 3. Melittin simulation details.
System
Ionic
strength
(m
M)
DPPC
lipids
Temperature
(K)
Water
molecules
Solvation
ratio
E1 0
a
119 323 7071 59.4 : 1
E2 100 6993 58.8 : 1
P1 0 118 323 7047 59.7 : 1
P2 100 7005 59.4 : 1
a
An ionic strength of 0 mM implies counterions sufficient only to
neutralize the charge of the system.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3063
Deuterium order parameters of the Ab40-DPPC
systems
In our studies with Ab40, we examine a peptide that is
primarily asymmetric with respect to its interactions

with the lipid membrane. As such, we analyzed the
deuterium order parameters of each leaflet separately,
including the whole acyl chain and the ‘plateau region’
described above. Taking the approach applied by
Bachar and Becker [18], we analyzed the lipids in ‘tiers’
at increasing distance from Ab40. The first and second
tiers contained 20 lipids each, and the third tier con-
tained the remaining lipids in the leaflet, between 18
and 24. The results of these calculations are presented
in Table 4. Note that the tiered analysis does not apply
to the DPPC-only controls; the values presented repre-
sent an average of the plateau region for each leaflet
of the bilayer.
From these data, it can be seen that, overall, the
)S
CD
values in the top leaflet of the Ab40-DPPC sys-
tems are lower than those in the bottom leaflet. One
explanation for this phenomenon was proposed by
Tieleman et al. [35], wherein the lipids that interact
strongly with the protein become increasingly tilted rel-
ative to the bilayer normal, causing the angle between
the C-D bond and the normal to decrease. This occur-
rence was reported in simulations of melittin [18], and
occurs in the present study as well in the case of both
melittin and Ab40. The lipids of the top leaflet tend to
adopt an angle such that they become tilted, with their
headgroups pointing towards Ab, and the lipids of the
lower leaflet elongate to become more ordered, filling
the void in the center of the bilayer (Fig. 1). The

results of simulations C1 and C2 reflect the fact that
the peptides in these simulations interacted more or
less symmetrically with both leaflets over time. The
peptide became deeply inserted in the bilayer in a
transmembrane orientation, with disordered N- and
C-termini protruding through the lipid headgroups
of both leaflets.
We also note that the lipids in the first tier tend to
be more disordered than those of the second and third
tiers. In fact, in most cases, the values of <)S
CD
>
[4,8]
increase as the distance between the peptide and the
lipids increases. The presence of the Ab40 peptide
causes substantial disorder on the lipids with which it
most closely interacts, simultaneously resulting in an
increase in order of the lipids that are further away.
This behavior is dependent upon the conformation of
the peptide. In cases where Ab40 lost much of its ini-
tial a-helicity, the nearby lipids become more disor-
dered and the more distant lipids increase in order. In
cases where the peptide unfolds to a lesser extent (e.g.
simulation B4), the distant lipids approach a value of
<)S
CD
>
[4,8]
that is comparable to that of the relevant
control (NS1), based on the average order parameter

of Tier 3 in the two leaflets. We thus conclude from
these data that Ab interacts with the membrane in a
Table 4. Average values of deuterium order parameters. Data are the mean (± SD).
Simulation
Top leaflet plateau region Bottom leaflet plateau region Whole acyl chain
Tier 1 Tier 2 Tier 3 Tier 1 Tier 2 Tier 3 Top Bottom
A1 0.163 (0.007) 0.223 (0.008) 0.233 (0.007) 0.180 (0.005) 0.211 (0.009) 0.216 (0.006) 0.177 0.217
A2 0.209 (0.008) 0.274 (0.006) 0.285 (0.011) 0.254 (0.010) 0.235 (0.006) 0.263 (0.010) 0.223 0.260
A3 0.147 (0.008) 0.219 (0.005) 0.231 (0.004) 0.192 (0.006) 0.191 (0.005) 0.203 (0.009) 0.173 0.231
A4 0.188 (0.009) 0.250 (0.004) 0.273 (0.006) 0.197 (0.015) 0.247 (0.006) 0.250 (0.003) 0.207 0.238
B1 0.205 (0.019) 0.243 (0.016) 0.318 (0.017) 0.189 (0.019) 0.271 (0.020) 0.268 (0.018) 0.231 0.263
B2 0.134 (0.008) 0.213 (0.006) 0.227 (0.004) 0.165 (0.004) 0.193 (0.006) 0.203 (0.006) 0.167 0.213
B3 0.233 (0.007) 0.249 (0.013) 0.307 (0.014) 0.212 (0.008) 0.246 (0.014) 0.286 (0.010) 0.240 0.280
B4 0.160 (0.010) 0.220 (0.006) 0.239 (0.008) 0.220 (0.006) 0.191 (0.004) 0.197 (0.007) 0.176 0.213
C1 0.161 (0.010) 0.240 (0.005) 0.288 (0.007) 0.200 (0.008) 0.221 (0.005) 0.247 (0.006) 0.204 0.196
C2 0.183 (0.012) 0.284 (0.007) 0.321 (0.012) 0.232 (0.010) 0.269 (0.005) 0.275 (0.009) 0.240 0.245
OS1 0.216 (0.004) 0.215 (0.004) 0.185 0.185
OS2 0.232 (0.004) 0.231 (0.005) 0.201 0.201
NS1 0.217 (0.003) 0.217 (0.004) 0.188 0.188
NS2 0.242 (0.004) 0.243 (0.003) 0.213 0.214
NS3 0.252 (0.007) 0.253 (0.009) 0.229 0.229
E1 0.166 (0.007) 0.218 (0.004) 0.235 (0.005) 0.206 (0.010) 0.193 (0.004) 0.211 (0.005) 0.178 0.206
E2 0.217 (0.005) 0.252 (0.005) 0.245 (0.005) 0.225 (0.004) 0.253 (0.008) 0.232 (0.007) 0.210 0.245
P1 0.177 (0.013) 0.200 (0.005) 0.203 (0.004) 0.176 (0.012) 0.205 (0.008) 0.198 (0.005) 0.157 0.220
P2 0.202 (0.009) 0.228 (0.004) 0.247 (0.006) 0.205 (0.007) 0.229 (0.006) 0.228 (0.004) 0.187 0.252
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3064 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
manner similar to melittin with respect to its effects on
the disordering of the surrounding lipids.
In addition, the )S

CD
values for the top leaflet lipids
of the peptide–membrane systems are primarily lower
than the respective controls (NS1 or NS2), whereas the
bottom leaflet lipids are more ordered than the con-
trols. This behavior is a result of the top leaflet inter-
acting strongly with the unfolded and charged portions
of the peptides in each simulation, and is especially
true in the case of the lipids closest to the peptide (Tier
1). The values of )S
CD
for the controls are in good
agreement with previous experimental and simulation
studies [36,37].
The contraction of the lipid headgroups, and con-
comitant disordering of the acyl chains of the lipids in
closest contact with Ab, results in no substantial
changes in the overall density of the lipid bilayer.
There is a slight increase in density among the lipids
nearest Ab (most likely a result of the strong interac-
tion between Ab and the lipid headgroups; see below),
but regions of slightly lower density exist to compen-
sate for this more tightly-packed region. The bottom
leaflet, which becomes more ordered over time,
increases in density slightly. The top leaflet appears to
be slightly less dense than the bottom leaflet as well as
the control. Factoring in the presence of the protein
and averaging between the two leaflets gives an over-
all result that the bulk density of the lipids in the
peptide–membrane systems is not substantially differ-

ent from that of the control (DPPC-only) systems (see
Fig. S1).
Bilayer thickness
It has been reported previously that monomeric Ab40
can intercalate into the hydrophobic core of reconsti-
tuted synaptic plasma membranes, resulting in a
decrease in the thickness of the membrane [7]. To
quantitatively assess this descriptor of membrane dis-
ruption, we measured the thickness of our simulated
bilayers in terms of the P–P distance between the top
and bottom leaflets of the bilayer, using gridmat-md
[29]. The results obtained are shown in Fig. 2. More
detailed results are provided in the Supporting infor-
mation (Figs S2–S6). The time averages over the last
Fig. 1. As shown in a snapshot from the end of trajectory A1, the
Ab40 peptide causes DPPC lipids of the top leaflet of the bilayer to
become more disordered, with their acyl chains becoming more
parallel to the bilayer surface. Lipids in the bottom leaflet become
more ordered, extending their acyl chains to fill in the growing void
in the center of the bilayer. Representative lipids (sticks) near the
peptide (ribbon) are shown.
Fig. 2. Bilayer thickness around the embedded peptides, taken
from the average thickness over the last 25 ns of simulation. Pep-
tide conformations are from the final frame of each simulation,
which is representative of the final 50 ns of simulation time. For
perspective, the embedded region of the peptide is colored gray,
whereas the region exposed to the water–bilayer interface is
shown in black. The legend shows bilayer thickness (nm), mapped
to the corresponding colors.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab

FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3065
25 ns of each trajectory are shown. The final confor-
mation of the peptide is also shown, placed at its aver-
age location over this time period. The final
conformation of the peptide is representative of the
last 50 ns of the trajectories because most of the prom-
inent secondary structure changes occurred during the
first half of each simulation [25].
The most striking observation overall is the amount
by which the Ab40 peptide depresses the bilayer in its
immediate vicinity, in the order of 1.0 nm. There are
hydrogen bonds and favorable electrostatic interac-
tions between the zwitterionic headgroups of the
DPPC lipids and the backbone and charged residues
of the peptide. The result of these interactions is that
the lipid headgroups tilt substantially around the pep-
tide, causing the acyl chains of the lipids to spread
outward, more parallel to the surface of the bilayer
(Fig. 1; see below). Control simulations above the
phase transition temperature (i.e. those without
embedded peptides, at 323 K) show good agreement
with the experimentally-determined thickness of
3.7 nm [38].
It is also observed that melittin can lead to a similar
magnitude of bilayer thinning, in the order of
0.5–1.0 nm. This thinning only occurs in regions where
the peptide became more disordered over time. For
simulations E1 and E2, these disordered segments were
the N- and C-termini of the peptide, whereas it was
the N-terminus in P1, and the middle of the peptide

became slightly disordered in P2.
Area per lipid headgroup
Experimental work has concluded that the average
area per lipid headgroup for fully hydrated DPPC at
50 °C is in the range 62–64 A
˚
2
[39,40]. Previous simu-
lations of DPPC examining the effects of increased
ionic strength have demonstrated that the area per
lipid headgroup decreases with an increasing salt con-
centration, from 62.7 A
˚
2
in the absence of NaCl to
60.5 A
˚
2
in the presence of 100 mm NaCl [41]. The
results from our control systems, averaged over the
last 50 ns of simulation (Table 5), compare well with
these findings. There is very little difference between
DPPC systems at the original solvation ratio, and
those in an asymmetric box with an increased amount
of water (NS1, NS2, and NS3).
Determining the area per lipid headgroup in the
presence of an irregularly-shaped protein presents a
unique challenge, and we utilize the gridmat-md
methodology, wherein each headgroup is assigned to a
polygon within the grid of the lateral bilayer surface

[29]. As shown in Table 5, a trend becomes clear. The
area per lipid headgroup for lipids in the top leaflet is
decreased substantially from the control simulations,
whereas the area per lipid headgroup for lipids in the
Table 5. Area per lipid headgroup (mean ± SD) in A
˚
2
(% difference from controls) over the last 50 ns of each trajectory.
Simulation
Residue initially at
bilayer–water interface
Simulated
temperature (K) Top leaflet Bottom leaflet
A1 K28 323 52.7 ± 1.4 ()17%) 60.0 ± 0.8 ()5.5%)
A2 49.0 ± 1.7 ()17%) 54.5 ± 1.1 ()8.1%)
A3 54.4 ± 1.0 ()14%) 58.5 ± 1.0 ()7.9%)
A4 49.0 ± 0.8 ()17%) 57.4 ± 1.0 ()3.2%)
B1 K28 300 46.8 ± 0.6 ()20%) 55.2 ± 0.5 ()6.0%)
B2 323 55.9 ± 1.3 ()12%) 61.6 ± 0.8 ()3.0%)
B3 300 48.0 ± 0.9 ()18%) 54.1 ± 0.8 ()7.8%)
B4 323 52.9 ± 1.4 ()17%) 60.4 ± 0.8 ()4.9%)
C1 V24 323 53.6 ± 1.4 ()16%) 61.7 ± 1.6 ()2.8%)
C2 49.7 ± 1.5 ()16%) 55.9 ± 1.4 ()5.7%)
OS1 NA 323 63.3 ± 0.7 63.3 ± 0.7
OS2 323 60.5 ± 1.0 60.5 ± 1.0
NS1 323 63.5 ± 1.1 63.5 ± 1.1
NS2 323 59.3 ± 0.7 59.3 ± 0.7
NS3 300 58.7 ± 0.7 58.7 ± 0.7
E1 W19 323 58.7 ± 1.1 ()7.6%) 59.5 ± 1.1 ()6.3%)
E2 56.0 ± 0.8 ()5.6%) 55.4 ± 1.3 ()6.6%)

P1 NA 62.5 ± 1.8 ()1.6%) 60.7 ± 0.9 ()4.4%)
P2 57.1 ± 1.1 ()3.7%) 55.7 ± 0.8 ()6.0%)
In the case of the OS ⁄ NS series, no peptide was present. For P1 and P2, the entire peptide was initially located at the membrane–water
interface. NA, not applicable.
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3066 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
bottom leaflet is only slightly less than that of the con-
trols. The reason for this result is related to the obser-
vations regarding deuterium order parameters. The
lipids of the top leaflet interact strongly with the back-
bone and charged amino acid side chains of the disor-
dered N-terminal segment of Ab via hydrogen bonds
and electrostatic interactions. The result is that the lip-
ids tilt substantially, reducing the vertical thickness of
the top leaflet. This behavior requires the lipids of the
lower leaflet to pack more tightly and extend their acyl
chains to maintain the integrity of the membrane. The
substantial tilt (and resulting disorder) of the top leaf-
let lipids and the slight increase in packing (and thus
order) in the bottom leaflet lipids is reflected in the
area per lipid headgroup.
The interaction between the N-terminal segment of
Ab and the DPPC headgroups develops over time.
After equilibration, the measured area per lipid head-
group in each system is close to the accepted experi-
mental value (62–64 A
˚
2
). Upon contact between the
N-terminal region of Ab and the membrane–water

interface (within 10 ns of simulated time), the area
per lipid headgroup begins to rapidly decrease as the
lipids associate with this disordered segment of the
peptide (Fig. 3; see also Figs S7–S9). Unfolding of
Ab occurs over the first 50 ns of each simulation,
after which the peptide conformation is largely
unchanged [25]. The area per lipid headgroup for the
control systems (simulation sets OS and NS) remains
steady over time at values appropriate for a fully
hydrated DPPC bilayer under the given conditions
(Fig. 4).
The lipids closest to Ab40 experience the greatest
decrease in area per lipid headgroup. From Fig. 5, it
can be seen that lipids closest to the peptide have the
smallest lateral area, whereas lipids further away tend
to occupy areas close to the bulk value of DPPC. In
Fig. 5, lipids of the top leaflet were ordered according
to their proximity to the center of mass of the Ab pep-
tide. Thus, the closer lipids have the smaller residue
Fig. 3. Area per lipid headgroup as a function of time for simulation set A. After making contact with the DPPC headgroups (within 10 ns in
all cases), the N-terminal segment of Ab attracts the lipids of the top leaflet, depressing their lateral area. The area per lipid headgroup in
the bottom leaflet is decreased as a result of the increased order and packing in this leaflet, which is a consequence of the disordering of
the top leaflet.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3067
designation. In the case of simulation A1, the area per
lipid headgroup is largely constant at the outset of the
simulation, fluctuating around a value of 62 A
˚
2

. Over
time, the lipids close to the peptide are drawn to it by
the interactions described above, whereas more distant
lipids maintain a more canonical value for their lateral
area. This trend is apparent in all other simulations of
Ab (see Figs S10–S15), except for A2. In simulation
A2, the peptide unfolded to the greatest extent of any
of the simulations, thus contacting the greatest number
of lipids. The lipids closest to the peptide center of
mass have a depressed value for their lateral area, as
Fig. 4. Area per lipid headgroup as a function of time for control DPPC simulations.
Fig. 5. Area per lipid headgroup as a function of distance from the protein; simulations A1 and A2 are shown at each of three time points
(0, 50, and 100 ns). Lipid residues are numbered such that those closest to the peptide have the lowest numbering, increasing as the lipids
are further away from the peptide. For clarity, running averages of the data are shown, using a window of ten data points.
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3068 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
do some lipids that are more distant from this point.
The explanation for this behavior is that, as the disor-
dered N-terminal segment elongates through the lipid
headgroups, it interacts with a greater number of lipids
than in any other simulation. In Fig. 5, lipids num-
bered from 41 to 49 are close to the peptide center of
mass and are attracted by contact with the polar back-
bone, lipid residues 50–60 make van der Waals con-
tacts with the amino acid side chains of Ab, and lipid
residues 60–70 interact strongly with the highly
charged and disordered N-terminal segment of Ab
(Fig. 6).
In all cases, the area per lipid headgroup in the bot-
tom leaflet was largely insensitive to proximity to the

peptide, even in the simulations wherein the C-termi-
nus of Ab interacted with the lipids of the lower leaflet
(A3, C1, and C2). This observation indicates that the
ability of Ab to condense nearby lipids lies primarily
in its highly-charged, unstructured N-terminal seg-
ment.
Simulations of melittin showed similar behavior.
Simulations E1 and E2 showed a slight decrease in
area per lipid headgroup in the vicinity of the pep-
tide (see Fig. S13). Because melittin largely maintains
its secondary structure over time, the effects of the
peptide on this parameter are less pronounced than in
the case of Ab. In simulations P1 and P2, wherein the
entire peptide was in contact with the DPPC head-
groups, the nearest lipids experienced a reduction in
their lateral area, which we attribute to hydrogen-
bonding between charged headgroup phosphates and
the backbone of the small section of the peptide that
became disordered over time (Fig. 2; see also
Fig. S15).
Lipid tilt and effective chain length
The attraction between the lipids and unfolded regions
of the Ab peptide described above gives rise to the
striking behavior of the lipid acyl chains. As noted
above, the acyl chains of lipids near the peptide tilt
substantially, increasing their disorder as the peptide
draws them close to itself. To quantify this observa-
tion, two related parameters were measured: acyl chain
tilt angle and effective chain length. We defined the
acyl chain tilt angle as the angle formed between the

bilayer normal and the vector defined by the first
methylene carbon and the terminal methyl carbon on
the acyl chain. A description of effective chain length
has been proposed by Petrache et al. [42] (therein
termed the ‘average chain length’; L
C
*). This descrip-
tor is simply defined as the distance along the bilayer
normal between the first methylene carbon and the ter-
minal methyl carbon. These two parameters (i.e. the
tilt angle and the effective chain length) should be
related under most circumstances, such that, as the tilt
angle increases (and the acyl chain becomes more
parallel to the bilayer surface), the effective chain
length should decrease.
Tilt angle and effective chain length have been ana-
lyzed for the systems simulated in the present study as
a function of distance from the peptide. There was no
substantial difference in the results for the sn-1 and
sn-2 chains; hence, for the purpose of clarity, the data
presented here are in direct reference to the sn-1 chain.
We find that the lipids in the top leaflet in closest con-
tact with the peptide (typically those interacting with
the disordered N-terminal segment) increase their tilt
angle over time, simultaneously decreasing their effec-
tive chain length (Fig. 7). In other words, the strong
attraction between the peptide backbone and charged
residues draws the headgroups of nearby lipids away
from other surrounding lipids, pulling the entire lipid
more parallel to the surface of the bilayer. Regions of

the most substantially tilted lipids correspond to those
with the smallest area per lipid headgroup and the
greatest amount of disorder. In the bottom leaflet, the
Fig. 6. Illustration of the contacts between the Ab40 peptide in
simulation A2 and the lipids of the top leaflet. The peptide is shown
as a black ribbon, and each lipid is represented by the phosphorus
of its headgroup, shown as spheres. The phosphorus atoms are
colored according to the lateral area of the corresponding lipid,
increasing as the colors change from blue to red. The small green
sphere represents the peptide center of mass, demonstrating that
not all of the lipids closest to this point experience the greatest
degree of association with the peptide in this simulation.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3069
lipid chains elongate, as demonstrated by a small
increase in effective chain length over time, as well as
an overall reduction in the tilt angle (Fig. 7).
Lipid tilting was minimal in simulations involving
melittin, which interacts more weakly with the lipid
headgroups as a result of its smaller size and greater
retention of secondary structure. These results demon-
strate that more extensive lipid tilting is induced by the
dynamic behavior of Ab. Because Ab unfolds to a
much greater extent than melittin, thus interacting with
more lipids, it is able to cause greater disruption of
canonical lipid dynamics and orientation. Control sim-
ulations of pure DPPC showed an effective chain
length of approximately 1.25 nm in both leaflets,
which is in agreement with the value proposed by
Petrache et al. [42].

Discussion
According to the ‘amyloid hypothesis’, Ab is central to
the development and progression of Alzheimer’s disease
[2], but relatively little is known about how this small
peptide interacts with a lipid membrane in the context
of neurodegeneration. Although a number of experi-
ments have examined the properties of lipids in the
presence of Ab [7,11,12], little detailed structural data
exist to indicate how Ab induces these observed phe-
nomena, providing the motivation for the present
study. To gain a clear picture of how Ab disrupts a
membrane environment, the peptide must be examined
in detail with respect to its structural and chemical fea-
tures and how they impact the surrounding lipid
matrix. A previous study by Ambroggio et al. [12] sug-
gested that Ab42 interacts strongly with a lipid environ-
ment, becoming part of it and disrupting interactions
between the lipids, leading to structural deformation.
Although the focus of their study was Ab42, whereas
Ab40 was the subject of investigation in the present
study, we consider that there are common features of
both alloforms contributing to membrane perturbation.
The most substantial observations of the present
study are that the Ab40 peptide causes nearby DPPC
Fig. 7. Effective chain length (top panel) and acyl chain tilt angle (bottom panel), as a function of distance from the protein, for simulation
A1. The results are indicative of all the other simulations involving Ab, and thus are representative of the general conclusions discussed in
the text. There is no substantial difference in the characteristics of the sn-1 and sn-2 chains; thus, only the results for the sn-1 chain are
presented for clarity.
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3070 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS

lipids to become more tilted and disordered relative to
controls, that the peptide is capable of reducing the
area per lipid headgroup of the lipids with which it
most directly interacts, and that it is capable of reduc-
ing the thickness of the membrane in its immediate
vicinity. Taken as a whole, these data suggest interest-
ing roles for the region of the peptide that is present in
the extracellular environment, and that which remains
embedded in the bilayer.
Lipid tilt and effective chain length
The factor that gives rise to much of the behavior dis-
cussed in the present study is the tilting of lipids that
are closest to the peptide present in each simulation.
The Ab peptide is capable of drawing lipid headgroups
to itself through electrostatic and hydrogen-bonding
interactions, weakening the interactions between these
lipids and others that are more distant from Ab.
Nearby acyl chains tilt substantially over time (Fig. 7),
leading to a slight thinning of the hydrophobic core of
the bilayer, which is manifested in a reduction in effec-
tive chain length of these same lipids. Lipids that dis-
played the greatest degree of tilting also correspond to
those with the smallest area per lipid headgroup and
the greatest amount of disorder. We attribute these
observations to the ability of the Ab40 peptide, espe-
cially through its N-terminal disordered region, to bind
lipid headgroups very closely to itself and draw them
closely to each other. Further details of this pheno-
menon are provided below, where the effects of lipid
tilting on each of the other parameters measured in the

present study are described.
Area per lipid headgroup
We previously described the unfolding of the Ab40
peptide in these simulations [25], and noted that, in
each simulation, the extracellular region of the peptide
(typically residues 1–28) interacted strongly with the
water–bilayer interface region and became disordered
in almost all simulations. We attributed this behavior
to strong hydrogen-bonding and electrostatic interac-
tions between the peptide and the zwitterionic phos-
phatidylcholine headgroups. This behavior creates an
interesting contrast with melittin, another membrane-
disrupting peptide. When embedded, melittin still has
a small portion of its structure exposed to the extracel-
lular environment, but it is substantially shorter than
the extracellular segment of Ab40, and thus retains a
more organized secondary structure than Ab40 in the
simulations conducted in the present study. Table 5
shows that the effect of melittin on the area per lipid
headgroup of the individual bilayer leaflets is minimal,
even when the whole peptide is positioned approxi-
mately parallel to the water–bilayer interface. There is
some reduction in the area per lipid headgroup by
melittin, but the difference in area per lipid headgroup
between the leaflets is almost indistinguishable. How-
ever, in the case of Ab40, there is often a prominent
difference in the area per lipid headgroup between the
leaflets, with the leaflet interacting with the extracellu-
lar (N-terminal) region of Ab40 experiencing a sub-
stantially reduced area per lipid headgroup relative to

the intracellular leaflet. This is true even in the case of
simulations A3, C1, and C2, in which the peptide
adopted a transmembrane orientation, interacting with
the lipid headgroups in the bottom leaflet as well.
We propose that this behavior arises because of dif-
ferent interactions between the peptides and the lipid
headgroups. In the case of Ab40, the peptide is capa-
ble of attracting lipid headgroups very close to its
long, mostly disordered, N-terminal segment, tilting
the lipids and arranging them very closely to each
other. This finding is independent of the protonation
state and ionic strength of the surrounding aqueous
medium, suggesting that electrostatic interactions and
hydrogen-bonding are likely to be involved in drawing
lipid headgroups in, but that these interactions are
nonspecific. In other words, they are not sensitive to
the protonation state of any particular residue or
group of residues. They occur simply because there are
many charged and polar amino acids in the extracellu-
lar region of A b40, which interacts with the elements
of the water–bilayer interface. Although some lipids
become oriented such that their headgroups interact
with the C-terminal region of melittin, they tend to
remain more dispersed compared to the lipids in the
Ab40 simulations. That is, there are fewer lipids tightly
associated with melittin than there are in the case of
Ab40.
Disordering of nearby lipids
In our simulations, both Ab40 and melittin demon-
strated the ability to cause disorder in nearby lipids. In

the case of Ab40, we consider this disorder primarily
to be the result of two factors: (a) a reduction in the
area per lipid headgroup and (b) the unfolding of the
C-terminal, embedded, region of the peptide. As dis-
cussed above, Ab40 is capable of increasing the tilt of
the surrounding lipids, thus disordering their acyl
chains. The fact that melittin also causes a similar
amount of disorder on the surrounding lipids leads to
an interesting question. If, as we have proposed, melit-
tin does not interact as strongly as Ab with the lipid
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3071
headgroups, how can we account for the fact that both
of these peptides can disorder the surrounding lipids to
the same extent?
One possible answer to this question was proposed
by the authors of the original melittin simulations in
which the peptide was embedded in the membrane
[18]. They observed that the surrounding lipids tend
to pack around the peptide and tilt their acyl chains,
as the protein itself tilts. Another contribution to the
disordering effect is the restriction of motion along
the acyl chains once the lipids have packed around
the protein. Because this behavior arises as a result of
the tilt of the peptide and interactions between the
embedded region of the peptide and the acyl chains,
it presents an interesting insight into the interactions
between asymmetrically-embedded peptides and
nearby lipids. Similar to melittin, Ab40 is embedded
asymmetrically in the lipid bilayer and, over time, tilts

with respect to the bilayer normal. As such, we attri-
bute some of the disorder experienced by the nearby
lipids to the motion and tilting of this embedded
segment.
Thinning of the bilayer in the presence of Ab40
Because Ab40 is known to disrupt the integrity of the
lipid membrane [11,12], another parameter of interest
is the local thickness of the bilayer. We previously
reported the capacity of water to penetrate into the
bilayer when Ab40 is present [25], and a more thor-
ough examination of the bilayer thickness is now
appropriate in the context of lipid parameters.
The experimentally-determined thickness of a fluid-
phase DPPC bilayer (in terms of the P–P distance) is
3.7 nm [38]. We achieve good agreement with this
value in all of our control simulations conducted at
323 K (Fig. 2). In the presence of Ab40 (at 323 K),
however, the bilayer may become depressed between
1.0 and 1.5 nm, as determined by averaging the bilayer
dimensions over the last 25 ns of each simulation. As
with the reduced area per lipid headgroup, the
decreased bilayer thickness is independent of
the peptide protonation state. Local thinning of the
bilayer, of comparable magnitude, is observed in all of
the Ab40-DPPC simulations.
Similar thinning of the bilayer also occurs in the
presence of melittin. Although the bilayer may deform
to decrease its thickness up to 1.0 nm (also determined
by the same averaging discussed above), the deformed
region is much smaller than that in the case of Ab40.

We attribute this observation to the fact that, unlike
Ab40, melittin retains much of its secondary structure
throughout the trajectory, thus remaining more
compact. The unfolding of Ab40 makes it more acces-
sible to a wider area, and thus more lipids, which leads
to a more pronounced depression in the bilayer.
Hydrophobic mismatch most likely plays a part in
the local deformation of the bilayer. Systematic analy-
sis of hydrophobic mismatch using KALP model pep-
tides by Kandasamy and Larson [22] illustrates that
short, helical peptides, with charges placed within the
lipid headgroups, can result in depression of bilayer
thickness or tilting of the peptide to accommodate the
size of the peptide. The systems reported in the present
study also include elements of hydrophobic mismatch
because neither Ab40 nor melittin completely span the
bilayer in their initial configurations. This orientation
positions the hydrophobic, embedded, regions of the
peptide approximately halfway through the bilayer.
Thus, it is not unreasonable to conclude that the
bilayer deforms to accommodate this orientation. Even
as the simulations progress, and the peptides alter their
orientation, becoming more deeply embedded in some
cases, the length of the short hydrophobic stretch
remains, and is less than the dimensions of the hydro-
phobic core of the bilayer. This mismatch, anchored at
the water–bilayer interface in many cases by charged
amino acids such as Lys16 and Lys28 in Ab40, likely
contributes to the deformation of the dimensions of
the surrounding bilayer.

The observations made from our simulations com-
pare well with experimental observations. An early
study by Mason et al. [7] indicated that Ab40 was
capable of penetrating into rat synaptic plasma mem-
branes, thus decreasing bilayer thickness. In addition,
Kayed et al. [11] report that, although oligomeric Ab
species are primarily responsible for membrane perme-
ability, monomeric and low molecular weight species
can penetrate into the bilayer interior and cause thin-
ning in the order of 0.5 nm. Although the pathogenic
agent of Alzheimer’s disease is widely believed to be
an oligomeric Ab species, it is also important to under-
stand the interactions of monomeric Ab with lipid
bilayers. The results reported in the present study
potentially shed light on the molecular interactions
that give rise to the experimentally-observed behaviors
described above, including a new hypothesis for a
functional role of the 16 N-terminal residues of Ab.
Although it has long been postulated that Ab is par-
tially embedded in the hydrophobic core of the bilayer
via its C-terminus, a functional role for its N-terminus
has not yet been established. Our simulations suggest
an important role for the N-terminus in associating
with the surrounding lipid matrix. The preponderance
of charged amino acid side chains and the exposure of
the disordered backbone interact favorably with the
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3072 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
strongly polar environment of the membrane–water
interface. It is not surprising that this region of the

peptide may contribute to neurotoxicity. The N-termi-
nal sequence is cleaved by a-secretase to generate the
non-amyloidogenic sAPPa and p3 fragments from
amyloid precursor protein, precluding the production
of Ab [43,44]. The p3 peptide and Ab differ by only
the absence or presence of these 16 N-terminal resi-
dues, and p3 is nontoxic, whereas Ab is neurotoxic.
The toxicity of Ab has been proposed to be exerted
through its interactions with membranes [45], suggest-
ing that this sequence of amino acids likely plays an
important role in Ab–membrane interactions. In the
present study, we report the molecular basis for the
perturbation of lipids, which is a result of both hydro-
phobic mismatch of the C-terminal region of Ab
within the lipid bilayer and the favorable interaction of
the N-terminal segment of Ab with the polar environ-
ment of the lipid headgroups. Understanding this sim-
ple system (i.e. monomeric Ab in a model membrane)
provides an excellent starting point for the study of
more complex systems of oligomeric Ab peptides in
membrane systems.
Experimental procedures
Equilibration
All simulations were performed using the gromacs soft-
ware package, version 3.3 [28]. All systems were equili-
brated under an isothermal–isochoric (NVT) ensemble for
100 ps. Position restraints were placed on all peptide heavy
atoms (if present) in all directions, and on the phosphorus
atoms of the lipid headgroups in the vertical direction. All
position restraints utilized a spring constant, k

pr
,of
1000 kJÆmol
)1
Ænm
)2
. The Berendsen thermostat [46] was
used to regulate temperature, with a relaxation time (s
T
)of
0.1 ps. Each group (protein, lipids, solvent ⁄ ions) was cou-
pled to a separate temperature bath. The parameters devel-
oped by Berger et al. [47] were applied to the DPPC lipids,
and the gromos96 53a6 parameter set was used to describe
the rest of the system (protein, solvent, ions). Lennard–
Jones interactions were cut-off at 1.4 nm, and short-range,
nonbonded interactions were calculated with a twin-range
cut-off scheme (0.9 ⁄ 1.4 nm), with the neighbor list updated
every five simulation steps. Long-range electrostatic interac-
tions were calculated using the particle mesh Ewald method
[48] with fourth-order spline interpolation and a Fourier
grid spacing of 0.12 nm. This treatment of electrostatics has
been shown to provide an accurate representation of lipid
properties [49], and is also commonly used in simulations
of proteins. The linear constraint solver method [50] was
used to constrain all bond lengths, allowing a 2 fs integra-
tion step.
Following NVT equilibration, isothermal–isobaric (NPT)
equilibration was performed for 500 ps, applying a pressure
of 10 MPa in the transverse direction and 0.1 MPa in the

vertical direction. The pressure of the system was regulated
anisotropically using the Berendsen barostat [46] with a
relaxation time (s
P
) of 2.0 ps. These conditions were
employed to accelerate packing of the lipids around the
peptide (if present) and applied to the pure DPPC systems
for consistency. The same position restraints and simulation
parameters applied in the NVT step were also used during
NPT equilibration.
Production MD
Following the 600 ps of equilibration, production MD was
conducted for 100 ns, using an NPT ensemble. A pressure
of 0.1 MPa was applied in all directions with all other
parameters being the same as in the NPT equilibration. All
position restraints were removed prior to the production
phase. Simulations were conducted using Virginia Tech’s
SystemX supercomputer (2.3 GHz PowerPC 970FX proces-
sors). Coordinates were saved every 2 ps for analysis. All
analyses were performed using tools within the gromacs
software package or code developed in-house.
Acknowledgements
The authors thank W. J. Allen for contributing the
analysis scripts used in the present study, the adminis-
trators of Advanced Research Computing at Virginia
Tech (SystemX) for computing time and technical sup-
port, and the anonymous reviewers, whose comments
substantially improved the quality of this paper.
References
1 Alzheimer A (1991) Arteriosclerotic brain atrophy. Int J

Geriatr Psychiatry 6, 129–130.
2 Haass C & Selkoe DJ (2007) Soluble protein oligo-
mers in neurodegeneration: lessons from the Alzhei-
mer’s amyloid b-peptide. Nat Rev Mol Cell Biol 8,
101–112.
3 Saido TC & Iwata N (2006) Metabolism of amyloid b
peptide and pathogenesis of Alzheimer’s disease:
towards presymptomatic diagnosis, prevention and ther-
apy. Neurosci Res 54, 235–253.
4 Coles M, Bicknell W, Watson AA, Fairlie DP & Craik
DJ (1998) Solution structure of amyloid b-peptide(1-40)
in a water-micelle environment. Is the membrane-span-
ning domain where we think it is? Biochemistry 37,
11064–11077.
5 Tischer E & Cordell B (1996) b-amyloid precursor pro-
tein: location of transmembrane domain and specificity
of c-secretase cleavage. J Biol Chem 271, 21914–21919.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3073
6 Yip CM & McLaurin J (2001) Amyloid-b peptide
assembly: a critical step in fibrillogenesis and membrane
disruption. Biophys J 80, 1359–1371.
7 Mason RP, Jacob RF, Walter MF, Mason PE, Avdulov
NA, Chochina SV, Igbavboa U & Wood WG (1999)
Distribution and fluidizing action of soluble and aggre-
gated amyloid b-peptide in rat synaptic plasma mem-
branes. J Biol Chem 274, 18801–18807.
8 Avdulov NA, Chochina SV, Igbavboa U, Warden CS,
Vassiliev AV & Wood WG (1997) Lipid binding to
amyloid b-peptide aggregates: preferential binding of

cholesterol as compared with phosphatidylcholine and
fatty acids. J Neurochem 69, 1746–1752.
9Mu
¨
ller WE, Kirsch C & Eckert GP (2001) Membrane-
disordering effects of b-amyloid peptides. Biochem Soc
Trans 29, 617–623.
10 Mattson MP, Cheng B, Davis D, Bryant K, Lieberburg
I & Rydel RE (1992) b-Amyloid peptides destabilize
calcium homeostasis and render human cortical neurons
vulnerable to excitotoxicity. J Neurosci 12, 376–389.
11 Kayed R, Sokolov Y, Edmonds B, McIntire TM,
Milton SC, Hall JE & Glabe CG (2004) Permeabiliza-
tion of lipid bilayers is a common conformation-depen-
dent activity of soluble amyloid oligomers in protein
misfolding diseases. J Biol Chem 279, 46363–46366.
12 Ambroggio EE, Kim DH, Separovic F, Barrow CJ,
Barnham KJ, Bagatolli LA & Fidelio GD (2005)
Surface behavior and lipid interaction of Alzheimer
b-amyloid peptide 1-42: a membrane-disrupting peptide.
Biophys J 68, 2706–2713.
13 Terwilliger TC & Eisenberg D (1982) The structure of
melittin. II. Interpretation of the structure. J Biol Chem
257, 6016–6022.
14 Vogel H & Ja
¨
hnig F (1986) The structure of melittin in
membranes. Biophys J 50, 573–582.
15 Dempsey CE (1990) The actions of melittin on mem-
branes. Biochim Biophys Acta 1031, 143–161.

16 Berne
`
che S, Nina M & Roux B (1998) Molecular
dynamics simulation of melittin in a dim-
yristoylphosphatidylcholine bilayer membrane. Biophys
J 75, 1603–1618.
17 Bachar M & Becker OM (1999) Melittin at a mem-
brane ⁄ water interface: effects on water orientation and
water penetration. J Chem Phys 111, 8672–8685.
18 Bachar M & Becker OM (2000) Protein-induced mem-
brane disorder: a molecular dynamics study of melittin
in a dipalmitoylphosphatidylcholine bilayer. Biophys J
78, 1359–1375.
19 Bond PJ & Sansom MSP (2007) Bilayer deformation by
the Kv channel voltage sensor domain revealed by self-
assembly simulations. Proc Natl Acad Sci USA 104,
2631–2636.
20 Khalid S & Sansom MSP (2006) Molecular dynamics
simulations of a bacterial autotransporter: NalP from
Neisseria meningitidis. Mol Membr Biol 23, 499–508.
21 Deol SS, Bond PJ, Domene C & Sansom MSP (2004)
Lipid-protein interactions of integral membrane pro-
teins: a comparative simulation study. Biophys J 87,
3737–3749.
22 Kandasamy SK & Larson RG (2006) Molecular
dynamics simulations of model Trans-membrane pep-
tides in lipid bilayers: a systematic investigation of
hydrophobic mismatch. Biophys J 90, 2326–2343.
23 Sperotto MM, May S & Baumgaertner A (2006) Mod-
elling of proteins in membranes. Chem Phys Lipids 141,

2–29.
24 Xu Y, Shen J, Luo X, Zhu W, Chen K, Ma J & Jiang
H (2005) Conformational transition of the amyloid
b-peptide. Proc Natl Acad Sci USA 102, 5403–5407.
25 Lemkul JA & Bevan DR (2008) A comparative molecu-
lar dynamics analysis of the amyloid b-peptide in a lipid
bilayer. Arch Biochem Biophys 470, 54–63.
26 Davis CH & Berkowitz ML (2009) Interaction between
amyloid-b (1-42) peptide and phospholipid bilayers: a
molecular dynamics study. Biophys J 96, 785–797.
27 Tieleman DP & Berendsen HJC (1996) Molecular
dynamics simulations of a fully hydrated dipal-
mitoylphosphatidylcholine bilayer with different macro-
scopic boundary conditions and parameters. J Chem
Phys 105, 4871–4880.
28 van der Spoel D, Lindahl E, Hess B, Groenhof G,
Mark AE & Berendsen HJC (2005) GROMACS: fast,
flexible, and free. J Comput Chem 26, 1701–1718.
29 Allen WJ, Lemkul JA & Bevan DR (2009) GridMAT-
MD: a grid-based membrane analysis tool for use with
molecular dynamics. J Comput Chem doi: 10.1002/
jcc.21172 (in press).
30 Terwilliger TC & Eisenberg D (1982) The structure of
melittin. I. Structure determination and partial refine-
ment. J Biol Chem 257, 6010–6015.
31 Bradrick TD, Philippetis A & Georghiou S (1995)
Stopped-flow fluorometric study of the interaction of
melittin with phospholipid bilayers: importance of the
physical state of the bilayer and the acyl chain length.
Biophys J 69, 1999–2010.

32 Bradshaw JP, Dempsey CE & Watts A (1994) A com-
bined x-ray and neutron diffraction study of selectively
deuterated melittin in phospholipid bilayers: effect of
pH. Mol Membr Biol 11, 79–86.
33 Monette M & Lafleur M (1995) Modulation of melittin-
induced lysis by surface charge density of membranes.
Biophys J 68, 187–195.
34 Ohki S, Marcus E, Sukumaran DK & Arnold K (1994)
Interaction of melittin with lipid membranes. Biochim
Biophys Acta 1112, 1–6.
35 Tieleman DP, Forrest LR, Sansom MSP & Berendsen
HJC (1998) Lipid properties and the orientation of aro-
matic residues in OmpF, influenza M2, and alamethicin
systems: molecular dynamics simulations. Biochemistry
37, 17554–17561.
Membrane perturbation by Alzheimer’s Ab J. A. Lemkul and D. R. Bevan
3074 FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS
36 Douliez J-P, Le
´
onard A & Dufourc EJ (1995) Restate-
ment of order parameters in biomembranes: calculation
of C-C bond order parameters from C-D quadrupolar
splittings. Biophys J 68, 1727–1739.
37 Tieleman DP, Marrink SJ & Berendsen HJC (1997) A
computer perspective of membranes: molecular dynam-
ics studies of lipid bilayer systems. Biochim Biophys
Acta 1331, 235–270.
38 Lewis BA & Engelman DM (1983) Lipid bilayer thick-
ness varies linearly with acyl chain length in fluid phos-
phatidylcholine vesicles. J Mol Biol 166, 211–217.

39 Nagle JF (1993) Area ⁄ lipid of bilayers from NMR.
Biophys J 64, 1476–1481.
40 Nagle JF & Tristam-Nagle S (2000) Structure of lipid
bilayers. Biochim Biophys Acta 1469, 159–195.
41 Pandit SA, Bostick D & Berkowitz ML (2003) Molecu-
lar dynamics simulation of a dipal-
mitoylphosphatidylcholine bilayer with NaCl. Biophys J
84, 3743–3750.
42 Petrache HI, Tu K & Nagle JF (1999) Analysis of simu-
lated NMR order parameters for lipid bilayer structure
determination. Biophys J 76, 2479–2487.
43 Selkoe DJ (1999) Translating cell biology into
therapeutic advances in Alzheimer’s disease. Nature
399, A23–A31.
44 Thinakaran G & Koo EH (2008) Amyloid precursor
protein trafficking, processing, and function. J Biol
Chem 283, 29615–29619.
45 Kayed R, Head E, Thompson JL, McIntire TM, Milton
SC, Cotman CW & Glabe CG (2003) Common struc-
ture of soluble amyloid oligomers implies common
mechanism of pathogenesis. Science 300, 486–489.
46 Berendsen HJC, Postma JPM, van Gunsteren WF,
DiNola A & Haak JR (1984) Molecular dynamics with
coupling to an external bath. J Chem Phys 81, 3684–
3690.
47 Berger O, Edholm O & Ja
¨
hnig F (1997) Molecular
dynamics simulations of a fluid bilayer of dipal-
mitoylphosphatidylcholine at full hydration, constant

pressure, and constant temperature. Biophys J 72, 2002–
2013.
48 Essmann U, Perera L, Berkowitz ML, Darden T,
Lee H & Pedersen LG (1995) A smooth particle mesh
Ewald Method. J Chem Phys 103, 8577–8593.
49 Patra M, Karttunen M, Hyvo
¨
nen MT, Falck E,
Lindqvist P & Vattulainen I (2003) Molecular dynamics
simulations of lipid bilayers: major artifacts due to
truncating electrostatic interactions. Biophys J 84,
3636–3645.
50 Hess B, Bekker H, Berendsen HJC & Fraaije JGEM
(1997) LINCS: a linear constraint solver for molecular
simulations. J Comput Chem 18, 1463–1472.
Supporting information
The following supplementary material is available:
Fig. S1. Density plots.
Fig. S2. Simulation set A.
Fig. S3. Simulation set B, with images rendered as
described in Fig. S1.
Fig. S4. Simulation set C, with images rendered as
described in Fig. S2.
Fig. S5. Simulation sets E and P, with images of the
melittin peptide, rendered as described in Fig. S2.
Fig. S6. Simulation sets O and N, showing the progres-
sion of bilayer thickness for pure DPPC membrane
systems.
Fig. S7. Area per lipid headgroup as a function of time
for simulation set B.

Fig. S8. Area per lipid headgroup as a function of time
for simulation set C.
Fig. S9. Area per lipid headgroup as a function of time
for simulation sets E and P.
Fig. S10. Area per lipid headgroup as a function of
distance from the protein: simulations A3 and A4.
Fig. S11. Area per lipid headgroup as a function of
distance from the protein: simulations B1 and B2.
Fig. S12. Area per lipid headgroup as a function of
distance from the protein: simulations B3 and B4.
Fig. S13. Area per lipid headgroup as a function of
distance from the protein: simulations C1 and C2.
Fig. S14. Area per lipid headgroup as a function of
distance from the protein: simulations E1 and E2.
Fig. S15. Area per lipid headgroup as a function of
distance from the protein: simulations P1 and P2.
This supplementary material can be found in the
online version of this article.
Please note: Wiley-Blackwell is not responsible for
the content or functionality of any supplementary
materials supplied by the authors. Any queries (other
than missing material) should be directed to the corre-
sponding author for the article.
J. A. Lemkul and D. R. Bevan Membrane perturbation by Alzheimer’s Ab
FEBS Journal 276 (2009) 3060–3075 ª 2009 The Authors Journal compilation ª 2009 FEBS 3075