Content uploaded by Peter John Bond
Author content
All content in this area was uploaded by Peter John Bond on Jun 10, 2020
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=imbc20
Molecular Membrane Biology
ISSN: 0968-7688 (Print) 1464-5203 (Online) Journal homepage: https://www.tandfonline.com/loi/imbc20
The simulation approach to bacterial outer
membrane proteins (Review)
Peter J. Bond & Mark S. P. Sansom
To cite this article: Peter J. Bond & Mark S. P. Sansom (2004) The simulation approach to
bacterial outer membrane proteins (Review), Molecular Membrane Biology, 21:3, 151-161, DOI:
10.1080/0968760410001699169
To link to this article: https://doi.org/10.1080/0968760410001699169
Published online: 09 Jul 2009.
Submit your article to this journal
Article views: 209
View related articles
Citing articles: 43 View citing articles
The simulation approach to bacterial outer membrane proteins (Review)
Peter J. Bond and Mark S. P. Sansom*
Laboratory of Molecular Biophysics, Department of
Biochemistry, The University of Oxford, South Parks Road,
Oxford OX1 3QU, UK
Summary
The outer membrane of Gram-negative bacteria serves as a
protective barrier against the external environment but is
rendered selectively permeable to nutrients and waste by
proteins called porins. Other outer membrane proteins
(OMPs) provide the membrane with a variety of other functions
including active transport, catalysis, pathogenesis and signal
transduction. A relatively small number of crystal or NMR
structures of these proteins are known, and it is therefore
essential that the maximum possible information be extracted.
In this respect, computational techniques enable extrapolation
from time- and space-averaged static structures to dynamic,
physiological events. Electrostatics approaches have been
used to investigate the structures of porins. The stochastic
simulation of ion trajectories through these channels has been
possible with Brownian dynamics, which treats the membrane
and solvent approximately, enabling the prediction of conduc-
tion properties. Molecular dynamics has also been applied,
enabling fully atomistic descriptions of ‘virtual outer mem-
branes’. This has provided atomic resolution descriptions of
solute permeation through porins. It has also yielded insights
into the dynamics of gating in active transporters and ion
channels, as well as providing clues to catalytic mechanisms in
outer membrane enzymes. Additionally, simulations are begin-
ning to reveal the common features of interactions between
membrane proteins and lipids, with biological implications for
OMP folding, stability and mechanism. Future prospects in-
clude the simulation of longer, larger and more complex outer
membrane systems, with more accurate descriptions of inter-
atomic forces.
Keywords: Outer membrane protein, porin, lipid-protein interac-
tions, Brownian dynamics simulation, molecular dynamics simula-
tion.
Introduction
Outer membrane proteins (OMPs) are found in many
prokaryotes and in certain organelles of eukaryotic cells.
Computational methods have been used to predict that ca.
2
/3% of the genes in Gram-negative bacteria encode
integral OMPs, and a significant proportion of these are
expressed ubiquitously (e.g. ca. 0.5% of the genome,
equivalent to 20 OMPs, are abundant in E. coli ; Molloy et
al. 2000, Wimley 2003). To date, only about 20 unique OMP
crystal or NMR structures have been solved, all from
bacterial sources (see http://blanco.biomol.uci.edu/Membra-
ne_Proteins_xtal.html). In contrast to the a-helical nature of
all other known (inner) membrane protein structures, the
basic architecture of the bacterial OMP is the b-barrel
domain. These are closed barrels, consisting of transmem-
brane anti-parallel b-strands that are strongly tilted with
respect to the barrel axis; each strand is connected by short
turns on the inner side of the membrane and by long, mobile
loops on the extracellular side. Matching the bilayer environ-
ment, the outer surface of the barrel is strongly hydrophobic,
whilst at the membrane-solvent interface, amphipathic aro-
matic groups (i.e. Trp, Tyr) are generally observed (Schulz
2000). OMPs also exist in mitochondria, chloroplasts and
peroxisomes; many of these are probably b-barrels (e.g. the
mitochondrial voltage-dependent anion channel), but it is
now suggested that integral bb-helical proteins may also
exist in these outer membranes (Wimley 2003). No simula-
tion work has been performed on these OMPs because of a
lack of structural data, and they are therefore not further
discussed in this review.
OMPs are found in the outer membranes (OMs) of Gram-
negative bacteria, as well as in the cell envelopes of certain
Gram-positive bacteria and possibly even archaea (Nikaido
2003). The OM itself is highly asymmetric. Its inner leaflet,
which faces the periplasmic space, is composed of phos-
pholipid, similar in composition to the inner membrane. By
contrast, the outer leaflet contains large lipopolysaccharide
(LPS) molecules, which contain multiple saturated fatty acid
tails and heterogeneous, charged polysaccharides that are
cross-linked by divalent cations. The structure of LPS varies
greatly between species and is even modified within a single
cell in response to changes in environmental conditions to
protect, for example, against host defence proteins. The OM
protects the cell against toxic agents: the combination of a
highly charged sugar region and tightly ordered, gel-like
hydrocarbon chains results in low permeability. Neverthe-
less, to aid in the exchange of nutrients and waste, the
membrane is rendered selectively permeable to solutes
smaller than ca. 600 Da by pore-forming OMPs called porins.
Thus, the sieve-like OM sharply contrasts with the tightly
coupled, energy-transducing inner membrane. Along with
these classical, ‘non-specific’ porins and the similar but
solute-specific porins, a number of other OMPs with varying
functions exist, many of whose structures are known (Figure
1). These non-porin OMPs range in function from enzymes
and active transporters, to structural linkers and proteins
involved in defensive or pathogenic recognition events
(Koebnik et al. 2000, Koronakis et al. 2000, Vandeputte-
Rutten et al. 2001a, Prince et al. 2002, Chimento et al.
2003). Reflecting this diversity in function, the structures of
OMPs vary in the number and stagger of the b-strands, the
shape and contents of the transmembrane pore, the struc-
ture of the extracellular loops, and their oligomerization state
(Schulz 2000).
Because of the lack of structural data available for
membrane proteins, it is important that we derive the
maximum possible information from the structures available
*To whom correspondence should be addressed.
e-mail: mark.sansom@biop.ox.ac.uk
Molecular Membrane Biology, May /June 2004, 21, 151/161
Molecular Membrane Biology
ISSN 0968-7688 print/ISSN 1464-5203 online #2004 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/0968760410001699169
Figure 2. Typical MD simulation set-up for a membrane protein embedded in a bilayer (Bond et al. 2002). The protein, OmpA, is shown in blue
cartoons format. Aromatic residues at the lipid
/water interface are implicated in anchoring proteins in the outer membrane and are shown in
green space-filling format. Lys and Arg residues are shown in yellow space-filling format, and are able to ‘snorkel’ up from the bilayer region into
the solvent. The membrane is composed of dymyristoylphosphatidylcholine (DMPC) molecules, which are represented in bonds format, coloured
by atom type. Sodium and chloride ions in the bulk solvent are displayed as red and purple spheres; the ca. 5000 water molecules are omitted for
clarity. The figure was made with vmd (Humphrey et al. 1996).
Figure 1. Structures of OMPs from representative structural classes for which MD simulations have been performed (see text for details).
The outer membrane is indicated schematically, with the extracellular and periplasmic sides at the top and bottom, respectively. Except where
stated, the proteins are displayed in cartoons format, coloured by secondary structure. OmpF (pdb code 2OMF), a member of the
homotrimeric, 16 b-stranded general porin family. One of the three monomers is coloured by secondary structure. FhuA (pdb code 1FCP), a
member of the 22 b-stranded TonB-dependent receptor family involved in ferrichrome-iron transport. The ca. 150 residue, amino-terminal,
internal cork domain is shown in green cartoon format. The ferrichrome-iron ligand is shown in ice-blue space-filling format. OmpA (pdb code
1BXW), a small, eight-stranded b-barrel implicated in structural and ion channel functions. Some of the b-strands are rendered transparent,
revealing in blue bonds format the salt bridge residues implicated in gating. OMPLA (pdb code 1FW3), a 12-b-stranded outer membrane
phospholipase which becomes active in its homodimeric, calcium-bound form (represented here). One of the two monomers is coloured by
secondary structure. The other is coloured in green, with its catalytic triad residues displayed in blue bonds format. The ‘catalytic site’ calcium
ions are shown as red spheres, and the active site-bound substrate analogues, located at the monomer/monomer interface, are shown in
bonds format, coloured by atom type. OmpT (pdb code 1I78), a 10-b-stranded outer membrane protease which belongs to the ‘omptin’ family.
Its Asp/His catalytic dyad is located on the extracellular side of the extended b-barrel, displayed in green space-filling format. The figure was
made with vmd (Humphrey et al . 1996).
152 P. J. Bond and M. S. P. Sansom
to us. Since the first crystal structure of a porin was obtained,
a multitude of computational methods has been applied to
further our understanding of OMPs, at various levels of
molecular detail. Indeed, simulations of most of the known
structural classes of OMPs have now been performed
(Figure 1). We will begin this review by describing how
computational methods have been used to analyse and
predict ion conductance properties of porins, before discuss-
ing how more ambitious simulations have been integrated
with structural data to gain insight into solute, lipid and
protein dynamics, providing information on the functional
implications of the conformational dynamics of this important
class of membrane proteins.
The computational methods discussed differ in their level
of granularity*
/i.e. the level of detail used to describe the
membrane proteins and their environment. Coarse-grained
methods based on Poisson-Boltzmann theory focus primarily
on electrostatic interactions. The protein, solvent and mem-
brane are treated as continuum dielectric regions. A static
analysis of such a system description, treating only the
charges on the protein explicitly, enables qualitative visuali-
zation and quantitative calculation of the electrostatic protein
surface. Explicit description of the stochastic trajectories of
individual ions, whose pathways are affected by the mean
force of the respective implicit regions of the system and
charges on the protein, is possible with Brownian dynamics
simulations. More fine-grained methods employ the molecu-
lar dynamics simulation technique. This enables a fully
atomistic representation of the entire system, including
protein, membrane, water and ions, and results in a
trajectory of all atoms undergoing thermal fluctuations at
equilibrium (Figure 2). Modifications of the algorithms also
enable the simulation of non-equilibrium processes. These
more fine-grained approaches are desirable from the point of
view that they should provide the highest level of accuracy
and do not neglect solvation effects and potentially important
correlations between atoms. Moreover, the complete con-
formational mobility offered theoretically enables the obser-
vation of biologically important protein and lipid
conformational changes. By contrast, the coarse-grained
techniques are less computationally demanding than mole-
cular dynamics owing to the reduced number of particles and
the simpler description of forces on these particles.
Electrostatics and ion conduction in porins
The crystal structures of several bacterial porins provided
clues as to mechanisms of solute permeation (Schirmer
1998, Koebnik et al . 2000). Computational studies enable
one to test and refine the proposed mechanisms. Classical
porins are homotrimers, each monomer consisting of a 16-
stranded b-barrel (Figure 1). The long, mobile loops face the
extracellular medium, except for loop L2, which folds over the
top of the barrel and makes stabilizing interactions with an
adjacent monomer, and the exceptionally long loop L3, which
folds into the barrel and forms the minimum cross-section of
the transmembrane pore (7 by 11 A
˚in OmpF) at the ‘eyelet’
or constriction zone, across which there is a segregation of
basic and acidic residues (Figure 3). Based on examination
of the structures, it was suggested that the variable size and
electric field across the eyelet of each porin formed the
primary determinant of selectivity. Computational studies
have been used to analyse the electrostatics of interactions
of ions with porins, thereby confirming and extending this
hypothesis.
An early qualitative approach enabled the visualization of
the surface electrostatic potential for the first OMP whose
crystal structure was obtained, namely the porin from
Rhodobacter capsulatus, allowing a relation of the potential
to its selectivity (Weiss et al . 1991). Subsequently, a more
quantitative approach has enabled the inclusion in the
calculation of the protein, solvent, and membrane as
continuum regions of defined permittivity. The electrostatic
potential is calculated by solving (numerically) the Poisson-
Boltzmann equation, given the charge density (on the
protein) and dielectric constant of different regions (Davis
and McCammon 1990). This method was used to investigate
two homologous porins from E. coli: the cation-selective
OmpF and the anion selective PhoE (Karshikoff et al . 1994).
An advantage of this technique is that it enables the
calculation of pK
A
values of ionizable groups. Thus, in both
porins, some of the basic residues at the constriction zone
displayed unusual titration behaviour that may stabilize the
strong electrostatic field inside channel and could also
explain experimentally observed pH-dependent changes in
effective channel size. Moreover, the electrostatic potential
for the two channels revealed a screw-like transverse electric
field. The electrostatic field, strongest at the constriction
zone, would facilitate the dehydration of polar solutes and
may aid in the transport of molecules with strong dipole
moments. Additionally, the selectivities of the two porins
were predicted to be due primarily to differences in electro-
static potential at the extracellular mouths, similar to the
proposal from earlier analyses of R. capsulatus porin (Weiss
et al. 1991). Further continuum electrostatic potential calcu-
lations were subsequently performed on other porins, reveal-
ing that the biologically relevant difference in selectivity at low
ionic strength between OmpK36 and the homologous OmpF
is due to electrostatic features (Dutzler et al . 1999), and that
the characteristic potential and titration behaviour for the
extremely narrow charge constriction in classical porins from
Figure 3. The features of the channel within OmpF porin. (a) The protein is shown in transparent red cartoons format. Some of the b-strands are
omitted, revealing the pore. Loop L3 (green ribbons format) folds back into the barrel lumen, and contains two acidic residues at the eyelet (red
bonds format). On the barrel wall opposite, four basic residues (blue bonds format) complete the charge constellation in the constriction zone,
which determines the major conductance properties of the porin. (b) The inner molecular surface is shown with a similar orientation to that in (a),
coloured by electrostatic potential on a scale from red (negative) through white to blue (positive). Residues within the cut-away molecular surface
are shown in green bonds format. Note the segregation of positive and negative charge on opposite sides of the channel, which produces a
particularly strong transverse electrostatic field at the constriction zone; this is thought to be important for dehydration of polar solutes, and
exclusion of hydrophobic molecules. The electrostatic potential was calculated by solving the Poisson-Boltzmann equation using Grasp (Nicholls
et al . 1991) with default parameters, and the membrane was not considered. The figures were made with vmd (Humphrey et al . 1996).
Outer membrane protein simulations 153
the a-, b-, and g-proteobacteria, represented by Omp32,
explain their selectivity and unusual electrophysiological
channel properties (Zachariae et al . 2002).
An extension of the continuum representation of electro-
statics enables explicit simulation of ion transport through
channels. With the Brownian dynamics (BD) method, ion
trajectories are generated by numerically integrating sto-
chastic equations of motion over discrete timesteps, using a
potential function, which is a sum of the interaction of an ion
with the protein charges and the dielectric boundaries
(Chung and Kuyucak 2002). This method is suitable for large
open channels and results in statistically significant simula-
tions of ion flux. One such study simulated thousands of
independent ion trajectories through various porins (Schir-
mer and Phale 1999). Qualitatively, anions and cations were
observed to follow different paths, according to the respec-
tive charge segregation at the constriction of each porin.
Quantitatively, the authors were able to predict accurately
the relative conductance and ion selectivity of each porin and
also their dependence on ionic strength (related to ionic
screening effects). An extension of this study successfully
predicted permeability characteristics for a number of OmpF
mutants (Phale et al. 2001). Taken together, these two
studies seem to justify the assumptions made for the
purpose of deriving conductance properties from large,
open-channel structures. Nevertheless, certain difficulties
remain, mainly as a consequence of the description of the
steady-state ion conduction process via an equilibrium
electrostatics theory. An attempt to solve these problems
was made by combining the BD method with a Grand
Canonical Monte Carlo (GCMC) algorithm, which allows
Boltzmann-weighted creation and destruction of particles,
enabling the simulation of a fluctuating non-equilibrium state
(Im et al. 2000). The method (GCMC/BD), which also
includes a microscopic representation of the transmembrane
potential, allowed a qualitative prediction of the cation
specificity of OmpF, and also reproduced the experimentally
observed asymmetric current
/voltage relationship, with
different ion fluxes being seen at opposite applied trans-
membrane potentials. Further refinement of the potential
function governing the multi-ion trajectories should lead to
even more accurate prediction of macroscopic properties
under non-equilibrium conditions. In particular, the inclusion
within the GCMC/BD method of a general reaction field,
arising from implicit salt in the exterior bulk solvent and
polarization at the system dielectric boundaries, resulted in
greatly improved predictions of cation selectivity and channel
conductance for OmpF (Im and Roux 2001, 2002a).
Simulations applied to solute behaviour in porins
Although BD methods enable the simulation of statistically
meaningful ion trajectories with timescales comparable to
real permeation processes, there is a neglect of microscopic
detail and in particular of protein mobility, as BD methods
(generally) treat the protein as a static object. Such approx-
imations inherent in this approach may have important
consequences for an accurate and complete understanding
of solute transport. Although more computationally demand-
ing, molecular dynamics (MD) simulations circumvent some
of these limitations. In particular, it is possible to represent
the protein, lipid, and solvent in atomic detail and with
unrestrained conformational mobility. In MD simulations,
interactions between explicit atoms are described by an
empirical potential function, alongside numerical integration
of the classical equations of motion over discrete timesteps
(Karplus and McCammon 2002b). In a typical simulation
using current methods, a complete OMP structure is docked
into a bilayer (see, e.g., Faraldo-Go´ mez et al. 2002)
consisting of a few hundred phospholipids (an approximation
to the complex OM lipid composition), before placing this
system into a box of salt solution of dimensions /10 nm
3
.
The result is a virtual patch of OM, containing between /10
4
and /10
5
atoms (Figure 2). Subsequent MD simulations
result in a ‘trajectory’ of the system (i.e. the atomic
coordinates of the system vs. time) extending over a time-
scale of tens of nanoseconds, from which structural and
dynamical quantities can be calculated. A snapshot from a
representative simulation illustrates the complexity of such a
system (Figure 3; Bond et al . 2002). In addition to OMPs,
similar MD simulations have been extended to a wide variety
of membrane proteins including ion channels, water-trans-
port proteins, proton pumps and ABC transporters, as well as
to pure lipid bilayers and detergent micelles (Merz and Roux
1996, Tieleman et al. 1997, Tobias et al. 1997, Forrest and
Sansom 2000, Domene et al. 2003a).
Several MD studies have investigated equilibrium solute
behaviour in OmpF. An early 1-ns MD simulation of this porin
(Tieleman and Berendsen 1998), fully solvated at a low salt
concentration (/0.1 M) and embedded in a simple bilayer,
enabled a detailed analysis of water and ions within the wide
aqueous channel. The mobility of water molecules in the pore
was reduced in comparison with bulk solvent, a result which
has been consistently observed in simulations of other
membrane channels, and which has consequences for the
estimation of macroscopic properties from simple geometric
(Smart et al. 1996) and continuum electrostatics models
(e.g. Im and Roux 2002c). Moreover, the large variation in
water diffusion and ordering across the pore is consistent
with previous electrostatics analysis. A more extended
simulation study of OmpF in a bilayer has since been
performed, with the system bathed in a solution of 1M KCl
(equivalent to /200 of each ion; Im and Roux 2002c).
Because of the limited timescales presently accessible to the
MD method, it was important to generate a starting ion
configuration as close as possible to equilibrium. Therefore,
the initial positions of ions were determined according to the
electrostatic potential calculated from the Poisson equation,
using a Metropolis Monte Carlo simulation. The resulting 5-
ns trajectory resulted in a statistically meaningful average ion
density along the pore. A screw-like separation of cations
and ions along the channel was found, consistent with the
strong electric field, and more potassium ions were located in
the pore in accord with the observed cation selectivity of
OmpF. These are conclusions that have also been derived
from BD simulations, as discussed above. Indeed, a recent
study compared ion behaviour in the channel of OmpF using
MD and BD, along with Poisson-Nernst-Planck (PNP)
electrodiffusion theory, which considers the solvent and
154 P. J. Bond and M. S. P. Sansom
ions as a continuum (Im and Roux 2002a). All three methods
appeared to reproduce the equilibrium ion distribution with
similar accuracy, including the screw-like pathway for anions
and cations. Nevertheless, certain features of the ion
conduction mechanism are inaccessible from the more
approximate methods, owing especially to the continuum
dielectric description of water in the BD and PNP ap-
proaches. For example, an interesting insight into OmpF
permeation was gained from MD, when it was observed that
the total solvation number of ions was preserved along the
length of the pore, because water molecules and protein
residues contributed in complement to maintain a constant
solvation shell (Im and Roux 2002b). This feature is
important for the high-throughput of ion channels (Roux et
al. 2000). Moreover, the cation selectivity was potentially
explained by the fact that, whilst isolated potassium ions
could permeate the channel, chloride could only pass the
constriction zone when paired with potassium.
A number of atomistic simulations have been performed
with the aim of observing directional (i.e. non-equilibrium) ion
or solute transport in porin channels. Several of these studies
have attempted to correlate macroscopic conduction proper-
ties with microscopic ion
/protein interactions. For example,
an early simulation of OmpF was carried out, with a constant
force across the pore applied to a sodium ion as an
approximation to a transmembrane potential (Suenaga et
al. 1998). This was used to identify Asp113 on loop L3 at the
constriction zone as important in binding cations during the
permeation process, particularly at low salt concentration. As
an extension of this study, single-channel conductance
measurements for different alkali metal ions through OmpF
were combined with free energy calculations based on the
original ion-bound Asp113 configuration (Danelon et al .
2003). It was concluded that the ion binding affinity to the
central aspartate residue increased with atomic radius, and
that this was correlated with greater conduction rates
because of increased local ion concentrations within the
pore. However, an explicit bilayer was omitted and so the
outer surface of the protein was fixed. This lack of detail,
combined with the short simulation times and consequent
lack of statistically significant sampling, means that exten-
sions to these studies are probably required for confirmation
of the results. Furthermore, in a recent MD study of the
strongly anion selective porin Omp32 from Delftia acidovor-
ans (Zachariae et al. 2003), no explicit bilayer was included
and, consequently, a major part of the b-barrel had to be
constrained. However, multiple simulations were performed
under different conditions with a combined length of ca. 50
ns, and these included a reconstruction of the potential of
mean force (PMF) for anion translocation around the con-
striction zone, constructed from multiple umbrella sampling
simulations. Chloride ion transport was observed to move in
a stepwise manner along a ‘basic ladder’ in the channel. The
constriction zone at the centre consists of three arginine and
no compensating acidic residues, in contrast with other
porins. The transfer mechanism from one residue to the
next, both at equilibrium and with an electric field applied,
proceeded via thermal fluctuations to overcome small energy
barriers and was aided by motion of the flexible basic side-
chains. The PMF energy minimum was observed at the
constriction zone, where chloride ions remained bound to
Arg75 and Arg38 for prolonged periods of time, consistent
with a previous continuum electrostatics study which showed
this region to be an electrostatic potential well (Zachariae et
al. 2002). At this site, anions interacted not only with
arginines but also with polar groups from other residues;
indeed, the protein solvated the ions in a complimentary
manner with water molecules to maintain constantly a full
hydration shell throughout channel conduction, as previously
observed in the MD simulations of OmpF (Im and Roux
2002b). Moreover, the strong binding of chloride at the
constriction prevented the entry of further anions into the
channel; this could explain the curious observation that the
conductance of Omp32 decreases with increasing salt
concentration, in contrast with many other porins. The study
of Omp32 highlights the strengths of the MD technique: a
detailed description of the ion conduction mechanism was
possible, which could not be achieved via BD simulations
because of the exceptionally small constriction size and the
evident requirement for side-chain flexibility in the mechan-
ism (Zachariae et al. 2003).
Simulations have also been carried out to analyse the
transport of non-ionic solutes through OmpF (Robertson and
Tieleman 2002). Non-equilibrium MD was recently per-
formed, by attaching a ‘virtual spring’ to the dipolar molecules
alanine and methylglucose and pulling the spring along the
pore axis. In contrast with the previous MD and BD ion
trajectories described above, the solutes did not follow
screw-like pathways during transport; the rate of transport
may have been too rapid to observe this phenomenon.
Nevertheless, the molecules aligned with the transverse
electric field at the eyelet and desolvation as the pore
narrowed resulted in compensatory hydrogen bonds being
made with the protein wall, in a similar manner to the MD ion
trajectories.
Computational studies have also been applied to maltose
transport by the solute-specific porin maltoporin. The protein
is similar in structure to classical porins, except that its
constriction zone is composed of three loops rather than one,
and the pore contains residues that confer solute specificity,
namely a helical line of aromatic residues (the ‘greasy slide’)
and an opposing group of polar residues (the ‘ionic track’),
which facilitate sugar transport (Koebnik et al . 2000). The
conjugate peak refinement method explores possible reac-
tion pathways and transition states (Dutzler et al . 2002), and
revealed a fast relay of hydrogen bonds between sugar
residues and ‘ionic track’ side-chains along the protein
channel wall, helping to explain both the specificity and rate
of sugar transport.
Simulations applied to conformational change: channel
gating
Whilst it is possible to capture some of the fundamental
elements of ion transport in wide aqueous channels using
coarse-grained methods such as BD (Im and Roux 2002a),
certain functions of OMPs involve protein conformational
changes more amenable to investigation via full atomistic MD
simulations. In particular, the permeability of the bacterial
Outer membrane protein simulations 155
OM may be regulated by a number of gating mechanisms
specific to each OMP channel or transporter. The first such
mechanism to be investigated using simulations was the
voltage gating of porins in planar lipid bilayers, whereby
membrane potentials of ca. 100 mV or more induce closure
of the channels (Koebnik et al . 2000). Although the biological
significance of this process has been questioned (on account
of the low electrical potential across the OM), it is feasible
that the presence of highly charged LPS or other local
environmental conditions may modulate the channel closure
characteristics, or that the gating serves as a ‘fail-safe’ upon
mis-incorporation of porins into the inner membrane (Nikaido
2003). Initially, it was generally assumed that loop L3 played
a key role in gating, owing to its constrictive nature and
the high electrostatic potential at the centre of the pore
(Figure 3). The lack of a porin crystal structure in the closed
state lead early on to a number of MD studies to investigate
the conformational flexibility of the loop, particularly at the
eyelet (Bjo
¨rkste´ n et al . 1994, Soares et al . 1995, Watanabe
et al. 1997). Although the results suggested that the loop
could move significantly and hence diminish the pore size,
these simulations were all performed in the absence of ions,
solvent or bilayer, and, consequently, it was necessary for
large regions of the protein to be constrained, leading to
some reservations concerning the physical realism of the
results (Watanabe et al. 1997). Subsequently, the more
detailed MD studies described (Tieleman and Berendsen
1998, Im and Roux 2002b) have shown that, at least under
equilibrium conditions, the crystal structure conformation of
loop L3 of bilayer-embedded OmpF is very stable in both
low-and high-molarity salt solutions. This result has been
supported by several cross-linking studies, which show that
covalent tethering L3 to the barrel wall does not prevent
voltage gating (e.g. Phale et al . 1997). Along with the
observation of voltage gating activity in large, aqueous b-
barrels lacking any sort of constriction (Bainbridge et al.
1998), these results call into question the role of L3 in voltage
gating of OmpF and related porins.
Simulation studies have also been applied to investigate
gating in OMP transporters without large, aqueous channels,
such as FhuA. This is a large, monomeric, 22-stranded b-
barrel with long, extracellular-facing loops, as well as a novel
amino-terminal, globular ‘cork’ domain of ca. 150 amino
acids, which occludes the b-barrel (Ferguson et al . 1998,
Locher et al. 1998; Figure 1). FhuA is a member of the
homologous TonB-dependent receptor family of proteins,
which tightly bind iron-chelating siderophores or vitamin B
12
;
the crystal structures of a number of these OMPs with
ligands bound and unbound are known, including FepA
(Buchanan et al. 1999), FecA (Ferguson et al. 2002), and
BtuB (Chimento et al . 2003). They are all ‘scavenger’
transporters, which couple the import of ligand against a
concentration gradient to the proton motive force across the
inner membrane, via interaction with the periplasm-spanning
TonB and with the assistance of the inner membrane
proteins ExbB and ExbD (Faraldo-Go´ mez and Sansom
2003). A number of allosteric conformational changes are
likely to be involved in this multi-step mechanism, and this
has been partially unravelled with the aid of 10-ns simula-
tions of membrane-embedded FhuA in the ligand-free and
bound states (Faraldo-Go´ mez et al . 2003).
The first step involves conformational changes in some of
the loops of the cork domain induced by ferrichrome binding,
which are allosterically propagated to regions that interact
with TonB and thereby signal that ligand is loaded (Faraldo-
Go´ mez and Sansom 2003, Chimento et al . 2003). Along with
loops from the b-barrel, the cork loops form a binding pocket
involving multiple protein
/ligand interactions, and some
long-lived hydrogen bonds with iron-coordinating groups
are persistently maintained during the MD simulations
(Faraldo-Go´ mez et al . 2003). The next stage of transport
may be initiated by movement of the extracellular loops, as
indicated by the change in conformation of loop L8 so as to
close the binding pocket during the simulation of the side-
rophore-bound state (Faraldo-Go´ mez et al . 2003). This
suggests an ‘air-lock’ gating hypothesis, whereby ligand-
binding induces a closure of the extracellular ‘hatch’,
enabling directional transport upon the formation of some
kind of channel. This mechanism is supported by in vivo
cross-linking studies (Scott et al . 2002), and by the FecA
crystal structure (Ferguson et al . 2002), which revealed
closing of the extracellular hatch in the ligand-bound state via
conformational changes in loops L8 and also L7. Finally, the
electrochemical energy of the inner membrane is somehow
transduced via TonB into further conformational changes,
causing the transport of ligand from the extracellular hatch to
the periplasm; this may proceed by the formation of a
channel within the b-barrel, or by partial or complete removal
of the plug (Faraldo-Go´ mez and Sansom 2003). In this
context, a potential channel-forming, water-filled region has
been identified at part of the interface between the cork
domain and barrel wall in FhuA (Ferguson et al . 1998).
Simulation analysis suggests that this cavity remains too
small to accommodate the siderophore, due largely to the
maintenance of multiple hydrogen bonds between plug and
barrel, and is actually smaller towards the periplasmic end in
the ligand-bound state (Faraldo-Go´ mez et al . 2003). This
therefore suggests that the plug domain would need to
undergo a significant motion or change in conformation to
allow passage of siderophore. Interestingly, the interface
between the barrel and plug domains was observed to be
extensively solvated during simulation. Moreover, the asso-
ciated water molecules were longer-lived and less permeable
for the ligand-bound FhuA on a 10-ns timescale. These
tightly bound waters may thus reduce the activation energy
of dissociation of the plug-barrel interface, by providing
alternative hydrogen-bonding donors and acceptors.
Thus, the interplay between static structural information
and simulations of protein and water dynamics are leading to
the elucidation of the mechanisms of gating in active
transport. Simulations have also been applied to the gating
process in an atypical ion channel, OmpA from E. coli . This
small, monomeric protein is composed of an eight-stranded
amino-terminal b-barrel domain (OmpA
NT
), located in the
OM, and a globular carboxy-terminal domain of unknown
structure, which lies in the periplasm (Figure 1). OmpA or
similar proteins are expressed at high levels in almost all
Gram-negative bacteria, and maintain the structural integrity
of the cell envelope by linking the OM to the periplasm, as
156 P. J. Bond and M. S. P. Sansom
well as serving recognition roles for bacterial conjugation and
pathogenesis (Koebnik et al. 2000). In contrast with porins,
the crystal structures of OmpA
NT
reveal that the extracellular
loops all point away from the b-barrel, whilst the barrel
interior itself does not contain a continuous channel, but
rather several aqueous cavities interrupted by charged and
polar sidechains (Pautsch and Schulz 1998, 2000). However,
many independent studies have shown that OmpA
NT
forms
small ion channels (conductance ca. 60 pS in 1M KCl) in
planar lipid bilayers (e.g. Arora et al . 2000). Additionally, the
complete protein is able to form higher conductance ion
channels (conductance ca. 300 pS in 1M KCl). It has been
suggested that formation of the higher conductance channels
may involve conversion to a porin-like conformer with
additional b-strands contributed by the C-terminal domain
(Nikaido 2003). Irrespective of the mechanism of formation of
the higher conductance pores, it would seem that some kind
of conformational change is necessary to gate between the
closed OmpA
NT
crystal structure and a more mobile form that
would allow transient 60 pS pore formation in this domain.
The physiological importance of this phenomenon is unclear
because of the high permeability already conferred to the OM
by porins. Nevertheless, for some bacterial cells, OmpA
homologues are the major OM channels, as is the case for
OprF in Pseudomonas aeruginosa; the consequent lower
permeability of the cell envelope confers resistance to a wide
range of antibiotics (Nikaido 2003).
To explore the possible gating mechanisms of OmpA
NT
,
multi-nanosecond MD simulations of this protein were
performed in membrane mimetic environments (Bond et al.
2002). The side-chains making up the aqueous cavities
within the b-barrel were observed to be quite mobile in
response to thermal fluctuations, leading to significant water
diffusion. In fact, only one region was observed to be
prohibitive to complete water permeation events during
simulation, at an Arg-Glu salt bridge within the centre of
the pore. Using molecular modelling, the Arg side-chain was
rotamerized, coordinating it with a nearby alternative, un-
paired Glu side-chain. This was hypothesized to be a
potential gating mechanism that would lead to an open state;
accordingly, simulation of this putative open-state conforma-
tion led to complete water permeation events on the
nanosecond timescale, and its empirically estimated con-
ductance was in good agreement with experimental data.
This kind of electrostatic-switch gating mechanism is not
without precedent, as illustrated by pore formation in
annexins (Benz and Hofmann 1997) and chloride channels
(Dutzler et al. 2003). Furthermore, the solution NMR
structure of OmpA
NT
in detergent micelles has provided
additional information on the protein dynamics; in particular,
a gradient of flexibility was identified along the axis of the b-
barrel, which the authors speculate may have a role in pore
formation (Arora et al. 2001). Comparative simulations of
OmpA
NT
in a detergent micelle and a lipid bilayer resulted in
the observation of a similar gradient of flexibility (Bond and
Sansom 2003); however, this gradient was more prominent
in the more mobile micelle environment. The increased
mobility led to spontaneous conformational changes in the
salt bridge region originally identified by molecular modelling
as a potential gate (Bond et al . 2002), resulting in water
permeation through the pore. Further NMR relaxation ex-
periments seem to support the possibility that pore formation
may occur via conformational dynamics of side-chains in the
region of the proposed gate (Tamm et al . 2003).
Simulations applied to conformational change: OMP
catalysis
MD simulations have also been used to aid the investigation
of catalytic mechanisms of OM enzymes. OMPLA is an outer
membrane phospholipase that degrades phospholipids only
in perturbed cell membranes, and it is implicated in toxin
release. It consists of a 12-stranded b-barrel, the interior of
which contains water-filled cavities formed by a hydrogen-
bonding network; the formation of a channel is prevented by
the loops, turns, and termini that fold over the centre of the
barrel (Figure 1). A number of structural (Snijder et al . 1999,
2001) and biochemical (Snijder and Dijkstra 2000) investiga-
tions of OMPLA’s enzyme activity have revealed that it is
regulated by calcium-dependent, reversible dimerization.
The enzyme is completely inactive in the monomeric form,
but active in the dimeric state and when two calcium ions are
bound per monomer, one at the ‘catalytic site’ (where it aids
catalysis by stabilizing negatively charged intermediates)
and one at a lower affinity site between loops L3 and L4. The
b-barrel of OMPLA consists of a convex and a flat side, and
the dimerization interface occurs at the flat side, consisting
primarily of hydrophobic interactions along with a few critical
hydrogen bonds. The active site of each monomer consists
of a serine hydrolase catalytic triad, located on the outside of
each barrel near to the dimerization interface and on the
outer membrane leaflet side.
Surprisingly, the crystal structures revealed very little
difference between the monomeric and dimeric forms of
OMPLA. A series of 5-ns MD simulations of different forms of
bilayer-embedded OMPLA have been used to compare the
dynamics of the monomer and dimer, to look for clues that
may explain the process of conversion between the respec-
tive inactive and active catalytic states, and also to investi-
gate the role of calcium and water in the catalytic mechanism
(Baaden et al. 2003). Of particular interest was the fact that
dimerization and substrate-analogue binding progressively
decreased dynamic fluctuations in the structure of OMPLA,
especially around the calcium-binding sites. In particular, the
presence of calcium at the active site seemed to stabilize the
arrangement of the catalytic triad, by positioning water
molecules that form an intricate hydrogen bonding network
with the triad sidechains; this observation is also suggestive
of a role of one of the long-lived water molecules in the
catalytic mechanism. By contrast, simulation of the calcium-
free OMPLA monomer revealed no stable water-mediated
hydrogen bonding network, resulting in multiple conforma-
tional states of the active site residues such that a config-
uration likely to be productive in catalysis was unstable and
present for only about a third of the trajectory. Finally, whilst
the substrate-binding clefts remained stable during simula-
tion of the OMPLA dimer when covalently bound to acyl tail-
containing substrate analogues at each active site, simula-
tions lacking the inhibitor actually resulted in the collapse of
Outer membrane protein simulations 157
these clefts, thus reducing the likelihood of phospholipid
entry. This may explain the experimental observation that
dimeric OMPLA exists in equilibrium between active and
inactive conformations; bilayer perturbation is presumably
required to open the blocked state observed during simula-
tion, thus preventing unregulated lipolysis of the cell mem-
brane.
Lipid
/protein interactions
MD simulations also offer the prospect of analysis of the
interactions between membrane proteins and the surround-
ing lipid molecules. This is of some importance, as only
limited data on lipid
/protein interactions may be obtained by
examination of interactions in those crystal structures where
some lipid molecules are present (Fyfe et al. 2001, Lee
2003). Early simulations provided a preliminary glimpse at
lipid
/protein interactions (Tieleman et al . 1999), but more
detailed analysis was hampered by the relatively short
simulation times (/1 ns) relative to the known multi-
nanosecond dynamics of lipid exchange in bilayers. As
simulations of membrane proteins out to /20 ns and
beyond are entirely feasible, it is possible to obtain a more
reliable picture of lipid
/protein interactions.
Particular attention has been paid to the interactions
between amphipathic aromatic (i.e. tryptophan and tyrosine)
side-chain ‘belts’ on the surface of the protein and interfacial
region of the bilayer (Yau et al . 1998). These aromatic belts
are thought to anchor the protein within the mobile and
flexible membrane. OmpA possesses a clear aromatic belt at
each end of the molecule: five such residues in the upper
(extracellular) and six in the lower (periplasmic) respective
belts. There is also an effective third aromatic belt provided
by four Tyr residues of the extracellular loops, above the
transmembrane b-barrel region. Note that Trp and Tyr
residues are able to form hydrogen bonds with interfacial
water molecules and polar head groups of lipids (see Figure
4(a)), whilst also forming hydrophobic interactions with the
bilayer core. Density profile analysis of multi-nanosecond MD
simulations of OmpA in a DMPC bilayer reveal that the
aromatic belts clearly overlap with the locations of the lipid
headgroups (Domene et al. 2003b). The mobility of the
extracellular loops leads to conformational changes resulting
in a narrower, more ‘uniform’ aromatic belt that matches the
bilayer interface. Moreover, all the tyrosine rings become
more perpendicular to the membrane plane, optimizing
interfacial interactions; this may result in tighter lipid
/protein
packing. In the bacterial outer membrane, the extracellular
loops would be expected to form multiple interactions with
the rather more complex lipopolysaccharide (LPS) head-
groups of the outer leaflet of this membrane. This is
important because the resultant tight linkage between
OmpA and the membrane is essential for maintaining
bacterial cell integrity.
In addition to amphipathic aromatic residues, it is proposed
that basic side-chains (i.e. lysine and arginine) also play an
important role in lipid
/protein interactions (Killian and von
Heijne 2000; see Figure 4(b)). The outer membrane pro-
tease OmpT provides an example of the characterization of
such interactions via simulation. On the basis of comparing
the crystal structures of FhuA with bound LPS and of OmpT it
Figure 4. Protein /lipid interactions observed during simulations of
OmpA in a DMPC bilayer. The protein is shown in cartoons format.
Selected side-chains and lipids are shown in light grey bonds format,
with nitrogen and oxygen atoms coloured black. Their molecular
surfaces (probe radius of 1A) are also shown in transparent grey
format. (a) A tyrosine residue in the lower aromatic belt of OmpA is
interacting with a lipid molecule. Its hydroxyl group is hydrogen
bonded with the DMPC glycerol backbone and headgroup via
bridging water molecules (shown in space-filling format), whilst the
aromatic ring is making van der Waal’s contacts with the acyl lipid
chain. (b) The amine group of a lysine residue on the outside of
OmpA above the extracellular membrane surface is shown forming
an ionic bond with a DMPC phosphate group.
158 P. J. Bond and M. S. P. Sansom
was suggested (Vandeputte-Rutten et al . 2001b) that a LPS
binding site for OmpT consisting of a cluster of predominantly
basic side-chains. In the FhuA structure, the primary
contacts from the protein are via a lysine and arginine cluster
to the diphosphate moiety of the Lipid A portion of LPS.
Analysis of multiple 10-ns simulations of OmpT in a DMPC
bilayer (M. Baaden and M. S. P. Sansom, unpublished data)
have focused on the principal contacts between the phos-
phate of DMPC and the OmpT side-chains of the proposed
LPS binding site. The results of this analysis revealed long-
lasting hydrogen bonds from the DMPC molecule to all of the
residues in the proposed LPS binding site. A more detailed
analysis of the lipid
/protein interactions in the simulations
suggested that the interactions are quite dynamic, with the
phosphate of a DMPC molecule being passed back and forth
between the side-chains of Arg138, Arg175 and Lys226. It is
conceivable that the diphosphate of Lipid A may form more
stable interactions, possibly with more than one basic side-
chain simultaneously, thus supporting the proposed prefer-
ential binding of LPS at this site.
Reflections and future directions
Coarse-grained methods, used for the analysis of electro-
statics interactions between protein surfaces and ions, have
proven to be useful in describing some of the structural
determinants of macroscopic properties such as selectivity
and conductance in the wide aqueous channels found in
porins. However, with the advances in structural biology that
have led to the achievement of a multitude of crystal
(Koebnik et al. 2000, Wimley 2003), and more recently
NMR (Arora and Tamm 2001, Tamm et al . 2003, Ferna´ ndez
and Wu
¨thrich 2003), structures of membrane proteins, it has
emerged that OMPs are not always static, hollow barrels and
hence require more advanced computational simulations to
relate their mobility and interactions with their environment to
their biological function. Most such biomolecular simulations
have used classical MD methods, which have improved
considerably since their first application to proteins over 25
years ago (Karplus and McCammon 2002a). Advances in
algorithms, forcefields and computing power have experi-
enced a rapid evolution, from preliminary explorations of the
(in vacuo) conformational flexibility of porins, to the investi-
gation of solute behaviour in confined pores, and presently to
the role of protein, water and lipid dynamics in active
transport, gating and catalysis.
One of the main limitations to the application of MD to
biological systems is the relatively high computational power
required. This restraint manifests itself in two (overlapping)
ways. First, accepting that the current time regime for MD
simulations is of the order of tens of nanoseconds, larger-
scale motions of protein chains are insufficiently sampled
during simulations. Indeed, a comparative simulation study of
a number of membrane proteins shows that for all but the
most simple structures, 10 ns is insufficient to sample fully
their respective conformational sub-spaces, especially for
the more mobile, extra-membraneous domains (Faraldo-
Go´ mez et al ., unpublished data). On the other hand, recent
simulations of OmpA in the crystal unit cell environment
(P. J. Bond and M. S. P. Sansom, unpublished data) suggest
that extending simulations from by an order of magnitude
(i.e. from a /10-ns to a /100-ns timescale) yields similar
trends in the sampling of protein conformational dynamics for
different structural domains. This lends some confidence to
extrapolating to even longer timescales. Validation may be
aided by recent advances in NMR relaxation techniques to
enable detailed comparison between experiment and MD of
motions over different time regimes (Tamm 2003).
Second, many biological processes of interest occur on a
timescale as yet only indirectly accessible by simulation.
Examples of such phenomena include ligand binding to
proteins and large-scale domain conformational changes (on
a/1-msto/1-ms timescale), and protein folding ( /1ms
and above). Nevertheless, preliminary studies are beginning
to shed light on these and related aspects of the long
timescale conformational dynamics of proteins (Simmerling
et al. 2002, Chowdhury et al. 2003). It is expected that, with
continued improvements in methodology and computing
power, increases in both simulation length and system
complexity will be attained, enabling direct observation of
these more complex phenomena in silico . This will also result
in the ability to carry out high-throughput simulations with the
aim of drawing common features of the dynamics of OMPs,
and eventually the simulation of large-scale ‘virtual outer
membranes’ will become possible (Arinaminpathy et al .
2003). Additionally, advances in simulation timescales will
enable further comparison and integration with structural and
folding experiments. This should also have a direct impact on
the accuracy of forcefield parameters.
Beyond these sampling issues, certain improvements in
the simulation methodology for OMPs will be necessary. For
example, most simulations of OMPs to date have employed
relatively simple phospholipid bilayers, yet bacterial OMs
contain LPS in their outer leaflet. LPS is thought to have
effects on the folding and insertion of membrane proteins
(Nikaido 2003) and may also influence the conformational
dynamics of OMPs within the membrane. In this context, the
first co-crystallized structure of LPS bound to a membrane
protein was obtained (Ferguson et al . 2000), whilst pre-
liminary simulations of LPS bilayers have been performed
(Lins and Straatsma 2001, Schroll and Straatsma 2002).
Along with refinements in biological detail, standard mole-
cular mechanics forcefields should also see improvements in
the near future. For example, application of methods that
include some treatment of electronic polarizability of atoms
(Halgren and Damm 2001) may result in a more accurate
description of ions and their interaction with OMPs during
transport, and of the influence of the polar membrane
interface on OMP structure and dynamics.
Acknowledgements
Research in MSPS’s laboratory is supported by grants from the
Wellcome Trust, the BBSRC and the EPSRC. PJB is a Wellcome
Trust research student. Our thanks to all of our colleagues for their
interest in this work, and especially to Marc Baaden and Jose´
Faraldo-Go´ mez.
Outer membrane protein simulations 159
References
Arinaminpathy, Y., Biggin, P. C., Bond, P. J., Domene, C., Pang, A.
and Sansom, M. S. P., 2003, Large scale biomolecular simula-
tions: current status and future prospects. In Proceedings of the
UK e-Science All Hands Meeting 2003 . Simon J. Cox, ed.
(Nottingham), pp. 901
/907.
Arora, A. and Tamm, L. K., 2001, Biophysical approaches to
membrane protein structure determination. Curr. Opin. Struct.
Biol. ,11, 540
/547.
Arora, A., Rinehart, D., Szabo, G. and Tamm, L. K., 2000, Refolded
outer membrane protein A of Escherichia coli forms ion channels
with two conductance states in planar lipid bilayers. J. Biol.
Chem.,275, 1594
/1600.
Arora, A., Abildgaard, F., Bushweller, J. H. and Tamm, L. K., 2001,
Structure of outer membrane protein A transmembrane domain
by NMR spectroscopy. Nat. Struct. Biol. ,8, 334
/338.
Baaden, M., Meier, C. and Sansom, M. S. P., 2003, A molecular
dynamics investigation of mono and dimeric states of the outer
membrane enzyme OMPLA. J. Mol. Biol. ,331, 177
/189.
Bainbridge, G., Gokce, I. and Lakey, J. H., 1998, Voltage gating is a
fundamental feature of porin and toxin beta-barrel membrane
channels. FEBS Lett. ,431, 305
/308.
Benz, J. and Hofmann, A., 1997, Annexins: from structure to
function. Biol. Chem.,378, 177
/183.
Bjo
¨rkste´ n, J., Soares, C. M., Nilsson, O. and Tapia, O., 1994, On the
stability and plastic properties of the interior L3 loop in R.capsu-
latus porin. A molecular dynamics study. Prot. Eng. ,7, 487
/493.
Bond, P. and Sansom, M. S. P., 2003, Membrane protein dynamics
vs. environment: simulations of OmpA in a micelle and in a
bilayer. J. Mol. Biol. ,329, 1035
/1053.
Bond, P., Faraldo-Gome´ z, J. and Sansom, M. S. P., 2002, OmpA*
/
A pore or not a pore? Simulation and modelling studies. Biophys.
J. ,83, 763
/775.
Buchanan, S. K., Smith, B. S., Venkatramani, L., Xia, D., Essar, L.,
Palnitkar, M., Chakraborty, R., van der helm, D. and Deisenhofer,
J., 1999, Crystal structure of the outer membrane active
transporter FepA from Escherichia coli .Nat. Struct. Biol. ,6,
56
/63.
Chimento, D. P., Mohanty, A. K., Kadner, R. J. and Wiener, M. C.,
2003, Substrate-induced transmembrane signaling in the cobala-
min transporter BtuB. Nat. Struct. Biol. ,10, 394
/401.
Chowdhury, S., Lee, M. C., Xiong, G. and Duan, Y., 2003, Ab initio
folding simulation of the trp-cage mini-protein approaches NMR
resolution. J. Mol. Biol. ,327, 711
/717.
Chung, S. H. and Kuyucak, S., 2002, Ion channels: recent progress
and prospects. Eur. Biophys. J. ,31, 283
/293.
Danelon, C., Suenaga, A., Winterhalter, M. and Yamato, I., 2003,
Molecular origin of the cation selectivity in OmpF porin: single
channel conductances vs. free energy calculation. Biophys.
Chem.,104, 591
/603.
Davis, M. E. and McCammon, J. A., 1990, Electrostatics in
biomolecular structure and dynamics. Chem. Rev. ,90, 509
/521.
Domene, C., Bond, P. and Sansom, M. S. P., 2003a, Membrane
protein simulation: ion channels and bacterial outer membrane
proteins. Adv. Prot. Chem. ,66, 159
/193.
Domene, C., Bond, P. J., Deol, S. S. and Sansom, M. S. P., 2003b,
Lipid-protein interactions and the membrane/water interfacial
region. J. Amer. Chem. Soc.,125, 14966
/14967.
Dutzler, R., Rummel, G., Albertı
´, S., Herna´ ndez-Alle´ s, S., Phale, P.
S., Rosenbusch, J. P., Benedı
´, V. J. and Schirmer, T., 1999,
Crystal structure and functional characterization of OmpK36, the
osmoporin of Klebsiella pneumoniae. Structure,7, 425
/434.
Dutzler, R., Schirmer, T., Karplus, M. and Fischer, S., 2002,
Translocation mechanism of long sugar chains across the
maltoporin membrane channel. Structure,10, 1273
/1284.
Dutzler, R., Campbell, E. B. and MacKinnon, R., 2003, Gating the
selectivity filter in ClC chloride channels. Science ,300, 108
/112.
Faraldo-Go´ mez, J. D. and Sansom, M. S. P., 2003, Acquisition of
siderophores in gram-negative bacteria. Nat. Rev. Mol. Cell. Biol. ,
4, 105
/115.
Faraldo-Go´ mez, J., Smith, G. R. and Sansom, M. S. P., 2002, Setup
and optimisation of membrane protein simulations. Eur. Biophys.
J. ,31, 217
/227.
Faraldo-Go´ mez, J., Smith, G. R. and Sansom, M. S. P., 2003,
Molecular dynamics simulations of the bacterial outer membrane
protein FhuA: a comparative study of the ferrichrome-free and
bound states. Biophys. J. ,85,1
/15.
Ferguson, A. D., Hofmann, E., Coulton, J. W., Diederichs, K. and
Welte, W., 1998, Siderophore-mediated iron transport: crystal
structure of FhuA with bound lipopolysaccharide. Science ,282,
2215
/2220.
Ferguson, A. D., Welte, W., Hofmann, E., Lindner, B., Holst, O.,
Coulton, J. W. and Diederichs, K., 2000, A conserved structural
motif for lipopolysaccharide recognition by procaryotic and
eucaryotic proteins. Struct. Fold. Des. ,8, 585
/592.
Ferguson, A. D., Chakraborty, R., Smith, B. S., Esser, L., van der
Helm, D. and Deisenhofer, J., 2002, Structural basis of gating by
the outer membrane transporter FecA. Science ,295, 1715
/
1719.
Ferna´ ndez, C. and Wu
¨thrich, K., 2003, NMR solution structure
determination of membrane proteins reconstituted in detergent
micelles. FEBS Lett. ,555, 144
/150.
Forrest, L. R. and Sansom, M. S. P., 2000, Membrane simulations:
bigger and better? Curr. Opin. Struct. Biol. ,10, 174
/181.
Fyfe, P. K., McAuley, K. E., Roszak, A. W., Isaacs, N. W., Codgell,
R. J. and Jones, M. R., 2001, Probing the interface between
membrane proteins and membrane lipids by X-ray crystallogra-
phy. Trends Biochem. Sci. ,26, 106
/112.
Halgren, T. A. and Damm, W., 2001, Polarizable force fields. Curr.
Opin. Struct. Biol. ,11, 236
/242.
Humphrey, W., Dalke, A. and Schulten, K., 1996, VMD *
/Visual
Molecular Dynamics. J. Molec. Graph. ,14,33
/38.
Im, W. and Roux, B., 2001, Brownian dynamics simulations of ion
channels: a general treatment of electrostatic reaction fields for
molecular pores of arbitrary geometry. J. Chem. Phys. ,115,
4850
/4861.
Im, W. and Roux, B., 2002a, Ion permeation and selectivity of OmpF
porin: a theoretical study based on molecular dynamics, Brownian
dynamics, and continuum electrodiffusion theory. J. Mol. Biol. ,
322, 851
/869.
Im, W. and Roux, B., 2002b, Ions and counterions in a biological
channel: a molecular dynamics simulation of OmpF porin from
Escherichia coli in an explicit membrane with 1 M KCl aqueous
salt solution. J. Mol. Biol. ,319, 1177
/1197.
Im, W. and Roux, B., 2002c, Ions and counterions in a biological
channel: a molecular dynamics simulation of OmpF porin from
Escherichia coli in an explicit membrane with 1 M KCl aqueous
salt solution. J. Mol. Biol. ,319, 1177
/1197.
Im, W., Seefeld, S. and Roux, B., 2000, Grand canonical Monte
Carlo-Brownian dynamics algorithm for simulating ion channels.
Biophys. J. ,79, 788
/801.
Karplus, M. and McCammon, J. A., 2002a, Molecular dynamics
simulations of biomolecules. Nat. Struct. Biol. ,9, 646
/652.
Karplus, M. J. and McCammon, J. A., 2002b, Molecular dynamics
simulations of biomolecules. Nat. Struct. Biol. ,9, 646
/652.
Karshikoff, A., Spassov, V., Cowan, S. W., Ladenstein, R. and
Schirmer, T., 1994, Electrostatic properties of two porin channels
from Escherichia coli .J. Mol. Biol. ,240, 372
/384.
Killian, J. A. and von Heijne, G., 2000, How proteins adapt to a
membrane-water interface. Trends Biochem. Sci. ,25, 429
/434.
Koebnik, R., Locher, K. P. and Van Gelder, P., 2000, Structure and
function of bacterial outer membrane proteins: barrels in a
nutshell. Mol. Microbiol. ,37, 239
/253.
Koronakis, V., Sharff, A., Koronakis, E., Luisi, B. and Hughes, C.,
2000, Crystal structure of the bacterial membrane protein TolC
central to multidrug efflux and protein export. Nature ,405, 914
/
919.
Lee, A. G., 2003, Lipid-protein interactions in biological membranes:
a structural perspective. Biochim. Biophys. Acta ,1612,1
/40.
Lins, R. D. and Straatsma, T. P., 2001, Computer simulation of the
rough lipopolysaccharide membrane of Pseudomonas aerugi-
nosa .Biophys. J. ,81, 1037
/1046.
160 P. J. Bond and M. S. P. Sansom
Locher, K. P., Rees, B., Koebnik, R., Mitschler, A., Moulinier, L.,
Rosenbusch, J. and Moras, D., 1998, Transmembrane signalling
across the ligand-gated FhuA receptor; crystal structures of free
and ferrichrome-bound states reveal allosteric changes. Cell ,95,
771
/778.
Merz, K. M. and Roux, B., 1996, Biological Membranes: A Molecular
Perspective from Computation and Experiment (Birkha
¨user,
Boston).
Molloy, M. P., Herbert, B. R., Slade, M. B., Rabilloud, T., Nouwens,
A. S., Williams, K. L. and Gooley, A. A., 2000, Proteomic analysis
of the Escherichia coli outer membrane. Eur. J. Biochem. ,267,
2871
/2881.
Nicholls, A., Sharp, K. A. and Honig, B., 1991, Protein folding and
association: insights from the interfacial thermodynamic proper-
ties of hydrocarbons. Proteins: Struct. Func. Genet.,11, 281
/
296.
Nikaido, H., 2003, Molecular basis of bacterial outer membrane
permeability revisited. Microbiol. Molec. Biol. Rev. ,67, 593
/656.
Pautsch, A. and Schulz, G. E., 1998, Structure of the outer
membrane protein A transmembrane domain. Nat. Struct. Biol. ,
5, 1013
/1017.
Pautsch, A. and Schulz, G. E., 2000, High-resolution structure of the
OmpA membrane domain. J. Mol. Biol. ,298, 273
/282.
Phale, P. S., Schirmer, T., Prilipov, A., Lou, J.-L., Hardmeyer, A. and
Rosenbusch, J. P., 1997, Voltage gating of Escherichia coli porin
channels: role of the constriction loop. Proc. Natl Acad. Sci. USA ,
94, 6741
/6745.
Phale, P. S., Philippsen, A., Widmer, C., Phale, V. P., Rosenbusch,
J. P. and Schirmer, T., 2001, Role of charged residues at the
OmpF porin channel constriction probed by mutagenesis and
simulation. Biochemistry ,40, 6319
/6325.
Prince, S. M., Achtman, M. and Derrick, J. P., 2002, Crystal structure
of the OpcA integral membrane adhesin from Neisseria meningi-
tidis.Proc. Natl Acad. Sci. USA ,99, 3417
/3421.
Robertson, K. M. and Tieleman, D. P., 2002, Orientation and
interactions of dipolar molecules during transport through OmpF
porin. FEBS Lett.,528,53
/57.
Roux, B., Berne` che, S. and Im, W., 2000, Ion channels, permeation
and electrostatics: insight into the function of KcsA. Biochemistry ,
39, 13295
/13306.
Schirmer, T., 1998, General and specific porins from bacterial outer
membranes. J. Struct. Biol. ,121, 101
/109.
Schirmer, T. and Phale, P. S., 1999, Brownian dynamics simulation
of ion flow through porin channels. J. Mol. Biol. ,294, 1159
/1167.
Schroll, R. M. and Straatsma, T. P., 2002, Molecular structure of the
outer bacterial membrane of Pseudomonas aeruginosa via
classical simulation. Biopolymers ,65, 395
/407.
Schulz, G. E., 2000, b-Barrel membrane proteins. Curr. Opin. Struct.
Biol. ,10, 443
/447.
Scott, D. C., Newton, S. M. C. and Klebba, P. E., 2002, Surface loop
motion in FepA. J. Bacteriol. ,184, 4906
/4911.
Simmerling, C., Strockbine, B. and Roitberg, A. E., 2002, All-atom
structure prediction and folding simulations of a stable protein. J.
Am. Chem. Soc. ,124, 11258
/11259.
Smart, O. S., Neduvelil, J. G., Wang, X., Wallace, B. A. and Sansom,
M. S. P., 1996, Hole: a program for the analysis of the pore
dimensions of ion channel structural models. J. Mol. Graph. ,14,
354
/360.
Snijder, H. J. and Dijkstra, B. W., 2000, Bacterial phospholipase A:
structure and function of an integral membrane phospholipase.
Biochim. Biophys. Acta ,1488,91
/101.
Snijder, H. J., Ubarretxena-Belandia, I., Blaauw, M., Kalk, K. H.,
Verheij, H. M., Egmond, M. R., Dekker, N. and Dijkstra, B. W.,
1999, Structural evidence for dimerization-regulated activation of
an integral membrane phospholipase. Nature ,401, 717
/721.
Snijder, H. J., Kingma, R. L., Kalk, K. H., Dekker, N., Egmond, M. R.
and Dijkstra, B. W., 2001, Structural investigations of calcium
binding and its role in activity and activation of outer membrane
phospholipase A from Escherichia coli .J. Mol. Biol. ,309, 477
/
489.
Soares, C. M., Bjo
¨rkste´ n, J. and Tapia, O., 1995, L3 loop-mediated
mechanisms of pore closing in porin: a molecular dynamics
perturbation approach. Prot. Eng. ,8,5
/12.
Suenaga, A., Komeiji, Y., Uebayasi, M., Meguro, T., Saito, M. and
Yamato, I., 1998, Compuation observation of an ion permeation
through a channel protein. Biosci. Reports ,18,39
/48.
Tamm, L. K., Abildgaard, F., Arora, A., Blad, H. and Bushweller, J.
H., 2003, Structure, dynamics and function of the outer mem-
brane protein A (OmpA) and influenza hemagglutinin fusion
domain in detergent micelles by solution NMR. FEBS Lett. ,
555, 139
/143.
Tieleman, D. P. and Berendsen, H. J. C., 1998, A molecular
dynamics study of the pores formed by Escherichia coli OmpF
porin in a fully hydrated palmitoyloleoylphosphatidylcholine bi-
layer. Biophys. J. ,74, 2786
/2801.
Tieleman, D. P., Marrink, S. J. and Berendsen, H. J. C., 1997, A
computer perspective of membranes: molecular dynamics studies
of lipid bilayer systems. Biochim. Biophys. Acta ,1331, 235
/270.
Tieleman, D. P., Forrest, L. R., Berendsen, H. J. C. and Sansom, M.
S. P., 1999, Lipid properties and the orientation of aromatic
residues in OmpF, influenza M2 and alamethicin systems:
molecular dynamics simulations. Biochem. ,37, 17554
/17561.
Tobias, D. J., Tu, K. C. and Klein, M. L., 1997, Atomic-scale
molecular dynamics simulations of lipid membranes. Curr. Opin.
Coll. Interface Sci. ,2,15
/26.
Vandeputte-Rutten, L., Kramer, R. A., Kroon, J., Dekker, N.,
Egmond, M. R. and Gros, P., 2001a, Crystal structure of the
outer membrane protease OmpT from Escherichia coli suggests
a novel catalytic site. EMBO J. ,20, 5033
/5039.
Vandeputte-Rutten, L., Kramer, R. A., Kroon, J., Dekker, N.,
Egmond, M. R. and Gros, P., 2001b, Crystal structure of the
outer membrane protease OmpT from Escherichia coli suggests a
novel catalytic site. EMBO J. ,20, 5033
/5039.
Watanabe, M., Rosenbusch, J., Schirmer, T. and Karplus, M., 1997,
Computer simulations of the OmpF porin from the outer mem-
brane of Escherichia coli .Biophys. J. ,72, 2094
/2102.
Weiss, M. S., Abele, U., Weckesser, J., Welte, W., Schiltz, E. and
Schulz, G. E., 1991, Molecular architecture and electrostatic
properties of a bacterial porin. Science ,254, 1627
/1630.
Wimley, W. C., 2003, The versatile b-barrel membrane protein. Curr.
Opin. Struct. Biol. ,13, 404
/411.
Yau, W. M., Wimley, W. C., Gawrisch, K. and White, S. H., 1998,
The preference of tryptophan for membrane interfaces. Biochem. ,
37, 14713
/14718.
Zachariae, U., Koumanov, A., Engelhardt, H. and Karshikoff, A.,
2002, Electrostatic properties of the anion selective porin Omp32
from Delftia acidovorans and of the arginine cluster of bacterial
porins. Prot. Sci. ,11, 1309
/1319.
Zachariae, U., Helms, V. and Engelhardt, H., 2003, Multistep
mechanism of chloride translocation in a strongly anion-selective
porin channel. Biophys. J. ,85, 954
/962.
Received 22 January 2004; and in revised form 16 March 2004.
Outer membrane protein simulations 161
