Biophysical Chemistry

Biophysical Chemistry
Membranes and Proteins

Biophysical Chemistry
Membranes and Proteins
Edited by
Richard H. Templer and Robin Leatherbarrow
Imperial College of Science and Technology, London, UK
ROYAL SOCIETY OF CHEMISTRY

The Proceedings of the first annual international conference, Biophysical Chemistry 200 1,
held at Imperial College, London, UK on 19-21 September 2001
Special Publication No. 283
ISBN 0-85404-851-0
A catalogue record for this book is available from the British Library
0 The Royal Society of Chemistry 2002
All rights reserved.
Apart from any fair dealing for the purpose of research or private study, or criticism or
review as permitted under the terms of the U K Copyright, Designs and Patents Act, 1988,
this publication may not be reproduced, stored or transmitted, in any form or by any means,
without the prior permission in writing of The Royal Society of Chemistry, or in the case of
reprographic reproduction only in accordance with the terms of the licences issued by the
Copyright Licensing Agency in the UK, or in accordance with the terms of the licences
issued by the appropriate Reproduction Rights Organization outside the UK. Enquiries
concerning reproduction outside the terms stated here should be sent to The Royal Society of
Chemistry at the address printed on this page.
Published by The Royal Society of Chemistry,

Thomas Graham House, Science Park,Milton Road,
Cambridge CB4 OWF, UK
Registered Charity No. 207890
For further information see our web site at www.rsc.org
Printed by George Over Ltd., London and Rugby

Preface
On the 19th of September 2001 the first annual meeting organised by the
Biophysical Chemistry Group of the Royal Society of Chemistry was held in
London. Called Biophysical Chemistry 200 1 the conference focussed on four areas:
modelling of biological systems; membrane structures and interactions; methods for
probing biornolecules and channels and receptors. The aim of the conference was to
mix together scientists from the physical and life sciences who were bound only by
their shared interest in understanding the molecular details of biological processes.
The areas we chose as a focus for this sociological experiment were designed to
have the maximum potential for cross-fertilisation of scientific thought.
Despite the tragic events that occurred in the week preceding the conference over
170 scientists from around the world were able to attend what was a lively and
invigorating three days of scientific discussion. This book consists of a selection of
the contributions that were made to the conference. The authors were asked to write
articles, which although they would be subjected to refereeing were expected to be
opinionated and that the authors’ opinions would not be altered. The articles should
be read in the same spirit. Indeed the opinions and views expressed during the
conference and we hope replicated to some extent in these papers is in some part a
reflection of the different ways in which our scientific training prepares us to view
the world around us. We hope that this diversity in viewpoint will be evident from
the collection of papers we present here and that you will find them as stimulating
as we did.
Richard Ternpler and Robin Leatherbarrow, Imperial College, London, May 2002
Contents
I Probing Biological Molecules: Theory and Experiment

Flow Oriented Linear Dichroism to Probe Protein Orientation in Membrane
Environments 3
A. Rodger, J. Rajendra, R. Mortimer, T. Andrews, J.D. Hirst, A.T.B. Gilbert,
R. Marrington, T.R. Dafforn, D.J. Halsall, M. Ardhammar, B. Nordkn,
C.A. Woolhead, C. Robinson, T.J.T. Pinheiro, J. Kazlauskaite, M. Seymour,
N. Perez and M.J. Hannon
Quantitative Protein Circular Dichroism Calculations 20

N.A. Besley and J.D. Hirst
Probing Cellular Structure and Function by Atomic Force Microscopy 31

M.A. Horton, P.P. Lehenkari and G.T. Charras
Physical Characterization of Wild Type and mnn9 Mutant Cells of

Saccharomyces cerevisiae by Atomic Force Microscopy (AFM) 50
A. Mindez-Vilas, I. Corbacho, M.L. Gonza'lez-Martinand M. J. Nuevo
Probing Supramolecular Organisation at Immune Synapses 58

F.E. McCann, K. Suhling, L.M. Carlin, K. Eleme, K. Yanagi, P.M. W. French,
D. Phillips and D.M. Davis
Probing the Structure of Viral Ion Channel Proteins: A Computational

Approach 72
W.B. Fischer and M.S.P. Sansom
The Impact of H202 on the Structure of Catalases by Molecular

Modelling Methods 78
S.G. Kalko, J.LI. Gelpiand M. Orozco
Entropy in the Alignment and Dimerization of Class C G-Protein Coupled

Receptors 85
M.K. Dean, C. Higgs, R.E. Smith, P.D. Scott, R.P. Bywater, T.J. Howe and
C.A. Reynolds
Electrostatic Stability of Wild Type and Mutant Transthyretin Oligomers 94

S. Skoulakis and J.M. Goodfellow
Simulations of Human Lysozyme: Conformations Triggering Amyloidosis in
156T Mutant 103
G. Moraitakis and J.M. Goodfellow
vii
viii Contents
Collective Excitation Dynamics in Molecular Aggregates: Exciton Relaxation,

Self-Trapping and Polaron Formation 118
M. Dahlbom, W. Beenken, V. Sundstrom and T.Pullerits
Surprising Electro-magnetic Properties of Close Packed Organized Organic

Layers - Magnetization of Chiral Monolayers of Polypeptide I36
1. Cameli, K Shakalova, R. Naaman and Z. Vager
Barrier Crossing by a Flexible Long Chain Molecule - The Kink Mechanism 147
K.L, Sebastian
I1 Proteins, Lipids and Their Interactions

Lipid Interaction with Cytidylyltransferase Regulates Membrane Synthesis 163
S. Jackowski and I. Baburina
Models and Measurements on the Monolayer Bending Energy of Inverse

Lyotropic Mesophases 177
A.M. Squires, J.M. Seddon and R.H. Templer
Hemolytic and Antibacterial Activities of LK Peptides of Various Topologies:

A Monolayer and PM-IRRAS Approach 191
S. Castana, B. Desbat, H. Wrdblewski and J. Dufourcq
A Novel Approach for Probing Protein-Lipid Interactions of MscL,

a Membrane-Tension-Gated Channel 199
P.C. Moe and P. Blount
Folding of The a-Helical Membrane Proteins DsbB and NhaA 208

D.E. Otzen
FhuA, an Escherichia coli Transporter and Phage Receptor 215

P.Boulanger, L. PlanCon, M.Bonhivers and L. Letellier
Morpholgical Aspects of in cubo Protein Crystallisation 22 1

C. Sennoga, B. Hankamer, A. Heron, J.M. Seddon, J. Barber and
R.H. Templer
Mobility of Proteins and Lipids in the Photosynthetic Membranes of

Cy anobacteria 237
C. W. Mullineaux and M. Sarcina
Partitioning and Thermodynamics of Chlordiazepoxide in n-OctanoVBuffer

and Liposome System 243
C. Rodrigues, P. Gameiro, S. Reis, J.L. F. C. Lima and B. De Castro
Contents ix
Distribution of Vitamin E in Model Membranes 248

P.J. Quinn
Theory on Opening-up of Liposomal Membranes by Adsorption of Talin 254

Y. Suezaki
Differential Scanning Calorimetry and X-Ray Diffraction Studies of

Glycolipid Membranes 267
0. Ces, J.M . Seddon, R.H. Templer, D.A. Mannock and R.N. McElhaney
Subject Index 277

I Probing Biological Molecules: Theory and Experiment
FLOW ORIENTED LINEAR DICHROISM TO PROBE PROTEIN ORIENTATION IN
MEMBRANE ENVIRONMENTS
Alison Rodger:* Jascindra Rajendra; Rhoderick Morther; Terrence Andrews; Jonathan

D. Hirst," Andrew T.B. Gilbert,' Rachel Marrington: Timothy R. Dafforn; David J.
Halsal1,dMalin Ardhammar: Bengt NordCn: Cheryl A. Woolhead: Colin Robinson:
Teresa J.T.Pinheiro; Jurate Kazlauskaite; Mark Seymour,gNiuvis Perez: Michael J.
Hannona
a Department of Chemistry, University of Warwick, Coventry, CV4 7AL, UK

b Crystal Precision Ltd, Little Farm,Faraday Road, Rugby, Warwickshire, CV22 SNB,
UK
c School of Chemistry, University of Nottingham, University Park, Nottingham, NG7
2RD, UK
d Cambridge Institute of Medical Research, Wellcome Trust/MRC Building, Hills Rd,
Cambridge, CB2 2XY, UK
e Physical Chemistry, Chalmers University of Technology, S-4 1296, Gothenburg,
Sweden
f Biological Sciences, University of Warwick, Coventry, CV4 7AL, UK
g Syngenta, Jealott's Hill, Bracknell, Berkshire, RG42 6EY, UK
h Center of Molecular Immunology, 216 St. and 15, Atabey, Playa, PO Box 16040,
Habana 11600 Cuba
* email: a.rodger@warwick.ac.uk Fax (+44) 24 76524112
1 INTRODUCTION
Processes occurring on or in membranes are essential in most biological systems, and the
study of these processes has been engendering an increasing interest for a long time, as has
the creation of artificial lipid membrane systems. Studies of, for example, membrane
transport and membrane protein function call for a thorough knowledge of molecular
interactions within the membrane, between the lipids themselves and between lipids and
other species (proteins, drugs, and ions). To this end, the locations and orientations of
molecules bound to the membrane can give important information. However, to date no
simple experimental method has been established to achieve this for membrane bound
proteins. In this work we report the first flow linear dichroism (LD)192*3study of proteins
bound to liposomes. Flow LD of molecules bound to the bilayer of shear-deformed
liposomes is one of the few direct methods potentially available for the study of the
orientation of membrane guest molecules, provided that the molecules of interest have
significant absorption in the visible and near-UV regions.
Linear dichroism is the difference in absorption of light polarised parallel to an
orientation direction and light polarised perpendicular to that direction. The LD signal is
4 Biophysical Chemistry: Membranes and Proteins
related to the oscillator strength of a transition (its absorbance intensity) and the
polarisation of the transition relative to the orientation axis. It is thus the ideal technique to
use to probe the orientation of an analyte and it is widely used, for example, for
determining the orientation of drugs bound to flow oriented DNA. The most effective
method for achieving flow orientation has proved to be Couette flow where two
concentric cylinders with a small annular gap (usually 500 pm) are aligned and one of
them rotates. The light is incident radially on the cell so the stationary cylinder needs to
have two windows and the rotating cylinder needs to be transparent to the intended
radiation.
Liposomes can be considered models of cell membranes, and be used for studying
transport and signal mechanisms of membrane proteins in situ. They are also used for drug
delivery and as transfecting agents in gene therapy. Ardhammar, Mikati, Lincoln and
Norden have shown that aromatic moieties or the aromatic ‘arm’ of ruthenium
dipyridophenazine can be oriented in liposomes and their orientation detected when the
liposomes are subjected to shear flow. In this work we probed the orientation achievable
by flow with different size model membranes and then the application of LD to determine
the orientation of proteins bound on or in liposomes. The proteins studied include
gramicidin, cytochrome c, pre-PsbW (a thylakoid membrane protein precursor) and a
monoclonal antibody.
2 LINEAR DICHROISM OF ANALYTES IN SHEAR DISTORTED LIPOSOMES
As noted above there are two literature precedents for flow orienting liposomes to assess
the orientation of solutes in their bilayer. Ardhammar et al. found that pyrene and
anthracene in soybean liposomes (produced by extrusion) had a negative LD for their long
axes, in accord with their expectations that the molecules would be oriented with their long
axis parallel to the lipid hydrocarbon chains and thus on average perpendicular to the
elongation and orientation axis of the liposome. With perylene, however, both the long axis
and the short axis had negative LD signals, which the authors took to mean that the
diagonal of the molecule was oriented parallel to the lipid chains. For the later ruthenium
work the authors modelled a deformed liposome as a cylinder with hemi-spherical caps,
assumed that the lipids redistributed evenly in the deformed liposome, and determined that
the reduced LD, LD’ = LD/isotropic absorbance, to be:
3s
LD‘ = -(I - 3~0s’a,) (1)
4
where a, is the angle the transition moment of interest makes with the normal to the
cylinder surface (i.e. to the lipid long axis), S is the orientation factor that denotes the
fraction of the liposome that is oriented as a cylinder perfectly parallel to the flow
direction.
This equation can be formally derived as follows. Consider an ellipsoidal liposome
under the flow conditions to be approximated by a cylindrical bilayer, whose lipids are
perpendicular to the long axis of the cylinder, capped by two hemispheres. Analytes
oriented within the hemispherical caps, assuming an average uniform distribution of
analytes, will have no net LD, and can therefore be ignored.
The LD experiment defines one axis system (x, y, z ) ,where z is the orientation axis
(the long axis of the cylinder) and the x/y plane is perpendicular to this (Figure 1). In order
to determine the LD, it is usefir1 to use another axis system {X, Y,Z) defined by the bilayer
I Probing Biological Molecules: Theory and Experiment 5
and an analyte. Z is the long axis of the cylinder (so z = 2)and X is the normal to the
cylinder surface that passes through the origin of the analyte.
Liposome Shear deformed Model of liposome

liposome
Figure 1 Schematic illustration of a shear deformed liposome.
The analyte orientation will not be affected by the shear flow (the forces are too
small), so on average any analyte transition moment, pi, will be uniformly distributed
about the Xaxis. Let a,be defrned as above. Let fi be the angle between the projection of
pi onto the Y/Z plane and the Z axis. Thus in the {X, Y, Z> coordinate system
pi = ,u(cosa,,sinfl, sina,,cospi sina,), (2)
where ,u is the magnitude of the transition moment and p i s the angle the YZ projection of p
makes with the Z axis.
The LD% by defmition:
LD= (A, - A y )
(3)
= W,’- P Y 2 )
where f i is the z component of the transition dipole in the {x, y, z } coordinate system and,
in this case, also in the {X, U,Z} coordinate system. may be written as the dot product of
the transition moment vector and the vector for the y axis in the {X,Y, z ) coordinate
system. Thus
Py = P-Y
= pi (cosa,, sin pi sin a,,cosfli sin a i ) H-(sin
z y , cos y ,O), (4)
= pi (cosa, sin y + sin pi sin a,cosy)
where y can take any value fiom 0 to 27~.Since the isotropic absorbance is $13, we can
write
--
LD’ -(cos~,sina,)’-(cosa,siny+sinS,sina,cosy)’
3s
= (cos’ p, sin’ a,) - (cos2a, sin’ y + sin’ p, sin2a,cos’ y + 2 cosa, sin y sin p, sin a,cosy)
(5)
Both fi and can take any value from 0 to 2%so upon averaging over them:
LD‘ 2 sin’ a,- 2 cos’ a,- sin’ a,
-=
3s 4
- 2 - 2cos’ a,- 2cos’ a,- 1+cos2 a,
-
4
-- 1 - 3 ~ 0 s ’a,
4
as given by Ardhammar et aL4Two special cases of interest are when a transition moment
is parallel to the cylinder normal (the lipids), for which aj = 0 and LD/3S = -1/2, and when
it is perpendicular to that, e.g. on the surface of the cylinder, for which ai = 90" and LD/3S
= +1/4.
3 MATERIALS AND METHODS
The solvents were all analytical grade obtained from BDH laboratory supplies and were
dried where necessary. Other chemicals including soybean lipid (L-a-Lecithin, Type IV-S:
from soybean approx. 40% (TLC)), Gramicidin D from Bacillus Brevis and cytochrome c
(Horse heart muscle) were obtained from Sigma Chemical Company and were used as
received. Lyso- 1-myristoyl-sn-glycero-3-phosphocholine (LMPC) and 1-palmitoyl-2-
oleoyl-phosphatidylcholine were obtained from Avanti Polar Lipids. Re-PsbW was made
according to the methods in reference '. The humanised monoclonal antibody was prepared
as described in reference '.
The methods were used to create the membrane mimicking environments for these
experiments depended on the sample to be studied and the size of membrane structure
required. The smallest structures were produced by sonication as the final stage; the largest
by vortexing the lipid preparations (this results in large multi-lamellar liposomes as shown
by microscopy; unilamellar liposomes were produced from multi-lamellar liposomes by
putting them through a LiposoFast Basic Extruder (with a polycarbonate membrane) from
12 and 25 times following reference3). In general lipids and analytes were in molar ratio
- -
approximately 50:l for pyrene and 5 mg/mL lipid to 1 mg/mL protein for the proteins,
though in some cases fbrther dilution was required to avoid excessive absorbance. If the
analyte was not water soluble, it was generally simplest to mix lipid and analyte in
chloroform. The solvent was then evaporated and kept under vacuum for 24 hours. To the
dry lipid mixture was added 5 mM phosphate buffer (PH 7) to a final concentration of lipid
- 5 mg/mL. Then the sample was sonicated or vortexed or extruded. If the analytes were
water soluble they were added with the buffer or to the liposome solution after this had
been prepared. Although Gramicidin is not very water soluble, in this case it was
introduced to the membrane by treating as if it were water soluble and extruding the
sample which seemed to facilitate its solubilisation by the lipids. Alternatively, Gramicidin
was added to a liposome preparation from an ethanol solution. The LD experiments were
performed on newly prepared samples.
UV-visible absorbance spectra were recorded using a Cary 1E spectrophotometer.
Flow linear dichroism spectra were recorded using a Jasco 5-715 circular dichroism
spectropolarimeter, which was adapted for flow LD measurements. The photomultiplier
tube was moved into the sample compartment next to the LD cell to reduce artefacts due to
scattered light using a housing fabricated for the instrument by European Chirality
Services. The cell used was developed for this work and is described below. The rotation
speed used in the experiments was -1000 rpm. In each case the speed chosen was the
maximum value that avoided significant bubble formation in the cell. An LD baseline was
measured on the same sample but without any rotation.
With light scattering samples, this method gave a slightly better baseline correction than a
water baseline. For non-scattering samples the same baseline was recorded for the LD
sample in a non-spinning cell and for a water or other non-orientatable solution in a
spinning cell.
The geometries for the determination of protein structural motif transition moments
were constructed using the CHARMM ~ r o g r a mTwo . ~ a-helices were considered: the ideal
geometry of Pauling and Corey with ((p=48", 4 ~ ~ 5 7and " ) the mean values observed in
experimentally determined structures of ((p=-63", 4~=-41'). For the p-strand a single
geometry with (cp=-135", 4~=135")was chosen. Two different geometries of gramicidin
were also considered? For the a-helices and p-strand, constrained minimisations of the
blocked peptides Amn-Ala200 cbx were performed in the absence of solvent effects. A
length of 200 amino acids was chosen to minimise end effects, and thus mimic an infinite
helix or strand.
The polarisation calculations were performed using the Matrix Method lo in which
the protein is considered to be a set of n chromophores, each with a characteristic set of
excitation bands. These effectively form a basis for the transitions for the entire protein,
and are allowed to couple through diagonalisation of the Frenkel Hamiltonian. This yields
a set of energies (eigenvalues) and transition couplings (eigenvectors) for the protein
transitions. In our calculations we consider only the n+n* and n+x* transitions for each
peptide link in the protein backbone. To date, this approach is the most accurate for the
calculation of protein circular dichroism from fmt principles." The polarisation tensor can
be calculated using a sum over states
where ,utb is the transition dipole moment from state a to state b in the h direction. This
can be obtained fkom the eigenvectors. This approach allows the contribution to the
polarisation from each individual transition to be considered and, in particular, allows the
polarisation due to the n+x* bands to be isolated.
4 DESIGN AND CONSTRUCTION OF LD CELL
Preliminary experiments on the lipids with a 500 pm annular gap LD cell were
unsatisfactory, at least in part due to excessive bubble formation reducing the signal to
noise ratio. For this and other work it was therefore decided to design a new LD cell with a
significantly smaller annular gap of 50 pm. Although this reduces the pathlength of the
sample, it was hoped that the increase in average orientation would compensate for this. It
was also decided to construct the optics out of CaFz to extend the measurement range into
the infia red region if required. This led to problems relating to the effect of heat shock on
CaF2 (it shatters). The final cell consisted of the following components as illustrated in
Figure 2:
Main Body: Stainless steel construction.

Rotating Bearings: Stainless steevwater and dust resistant.
Stainless steel crystal supporting spindles.
Stainless steel window housing.
1
Top Shalt
I
I Bearing
Figure 2 50 pm annular gap LD cell with CaF2 optics, demountable components, inner
rotating cylinder.
To enable flexibility in applications and later design modifications it was decided to
have the optical components as demountable units. An inner rotating cylinder design was
adopted with the motor housed under the unit. Practical considerations of sample loading,
cell cleaning etc. are in conflict with the optimal smooth running of the cell and a robust
design. In the end a compromise was adopted whereby the whole cell can be disassembled
when required but can be loaded and cleaned for normal operation with a syringe and
compressed air. The cell top, which needs to be in place for smooth running, can readily be
removed to aid cleaning, but the small annular gap means that if anything precipitates out
of solution or sticks to the cylinder walls disassembly is the only option
Figure 3 Surface roughness of the polished CaF2 crystal determined using a Wyko Optical
surface measuring instrument (left, Ra=3.21 nm, Rq= 4.1 1 nm, -3.57 nm, Re49.77
nm)and a Nan0 surf stylus-measuring instrument (right, surface characteristics: average
height:10.35 nm, mean deviation 1.92 nm).
To design and fabricate an instrument that rotates in a sealed environment where

water and solvents are in contact with the inner cell means that the materials should be
immune to water and solvent fatigue. Investigative studies of several types of bearing that
could perform within the working environment of the instrument were undertaken. Nylon
and glass bearings designed to run in a liquid environment were found to have a high noise
level and variable rotation. PTFE bearings which were of a solid construction were found
to break down in a short working time span. Stainless steel bearings were found not to be
100% water and dust resistant and also to have a high friction level, which put strain on the
motor drive interface. Though not ideal, the latter were incorporated into the design. The
finished polished CaF2 core crystal was supported between two precision ground spindles
180" apart. The crystal after being cut/ground/lapped and polished was scanned for surface
damage using a Wyko Optical surface measuring instrument and a Nan0 surf stylus-
measuring instrument (Figure 3). The machining of the crystal created no problems as the
hardness factor of CaF2 is between 5 and 6 on the Mohs scale. However, it was found
advisable to have a coolant in constant contact with the crystal during the polishing
process. After assembly the maximum alignment error was 6 pm. The drive for the
instrument is provided by a 6 V DC motor interfaced with a variable ampsholts control
box.
The new LD cell (couette LD cell (CaF2, 50 pm)) was initially tested with calf thymus
DNA (280 pM solutions in water) and the signal intensity compared with that of our
existing quartz cell with 500 p M annular gap. When rotating at approximately the same
speed (as determined by a ps kinetics run detecting variation in absorbance as the cells
rotate) of 1000 rpm, the LD signal for the two cells was almost identical (data not shown).
This suggests that the average orientation of the DNA is 10 times greater with the 50 prn
pathlength. The improved mechanical efficiency of the new cell may also be playing a role.
The smaller pathlength cell is of course attractive as it uses less sample (in our cases: less
than 400 pL versus 1800 yL in the longer pathlength cell). Both the cell and also the
sample (due to reduced path length) have a lower absorbance signal for the same LD signal
and thus a wider wavelength range. The lower intrinsic cell absorbance may be in part a
feature of the CaF2 versus quartz optics, though it may simply be due to better polishing.
5 PYRENE
Before attempting to incorporate proteins into lipid structures, we considered different

lipids and different preparation methods with the model compound pyrene that had
previously been inserted into lipid bilayers using an extrusion method by Ardhammar et al.
to give an LD spectrum! As noted above, if an analyte is incorporated with its long axis
parallel to the lipids then the long axis transitions will have an LDBS of -1/2 and the short
axis transitions a value of +1/4. Since the 337 nm band has approximately 1/3 of the
absorbance signal of the 274 nm and 242 nm bands, this LD spectrum should thus have the
magnitude pattern -6:+3:-2 for 240 nm:275 nm:340 nm. If the pyrene is on the surface
then the LD spectrum magnitude pattern would be: +3:+3:+1. If the short axis is parallel to
the lipids then one would expect: +3:-6:+1. If the pyrene lies parallel to the lipids but with
no distinction between its long and short axis, then we expect: -3:-3:-1.
We incorporated pyrene into lipid structures using the sonication method and a
range of lipids. The results for the different lipids are similar but not identical. With LMPC
our data are approximately: -6:-1:-3; with POPC: -12:-3:-6 (data shown in Figure 4a);
and with soybean: -2:-1:-2. These ratios are not consistent with each other nor with any
of the limiting cases given above suggesting that either the pyrene adopts a non-symmetry
defrned orientation as previously concluded by Ardhammar et al. for perylene, or there is
an orientation distribution with a varying preference for the pyrene long axis being parallel
to the lipids.
~ 0.02
c
a
0
5 0.015
E
2
i
5 0.01
250 300 350 400 250 300 350 400

Wavelengthlnm Wavelengthlnm
Figure 4 LD spectra of pyrene in (a) sonicated POPC micelles and (b) vortexed soybean
liposomes.
Soybean lipid structures prepared by vortexing (Figure 4b) gave LD ratios of

approximately -2:+2:-1 (though it is hard to estimate exactly where the baseline is at the
lower wavelengths). Using the extruder (data not shown) -2:+1:-1 was obtained in accord
with Ardhammar et aZ. who found a ratio of +1:-1 for the two longer wavelength bands
(and could not measure the lower wavelength one). These data suggest that the pyrene
adopts a distribution of positions with some surface bound (-1/4 for the vortexed sample)
and some with the plane of the pyrene parallel to the lipids and a preference for the long
axis being parallel to the lipids (-1/2). The proportion in each position depends on the lipid
environment.
The smaller structures are attractive as the baseline scatter was able to be removed
by simply subtracting the baseline LD recorded using a non-spinning sample. This was not
possible with the larger liposomes. Despite this, soybean lipid, which forms multi- or uni-
lamellar liposomes, was adopted for later experiments due to its significantly higher signal
to noise ratio. Although the multi-lamellar soybean experiment gave a slightly better
pyrene LD spectrum (presumably as the larger multi-lamellar liposomes were shear
distorted to a greater extent so oriented better), we later found that the extrusion process
facilitated protein incorporation into the liposomes in some cases. It also produces a more
reproducible liposome structure. This method was therefore used for all the protein
samples.
6 PROTEIN MOTIF TRANSITION POLARISATIONSAND EXPECTED LD
To analyse any protein data we obtained it is necessary to know the transition polarisations
of the protein, which means the transition polarisations of its secondary structural motifs.
A somewhat simplistic model for determining the backbone z+z* transition polarisations
of a-helical peptides is to assume that we may represent the peptide by a right-handed
helix of transition moments oriented along each C=O bond? The lowest energy excited
state (-208 nm from experiment) will result from the head-to-tail couplings of
neighbouring transition moments. The net transition moment of this state is parallel to the
helix axis. The head-to-head, tail-to-tail coupling will give the highest energy component
(at -190 nm) whose polarization is perpendicular to the helix axis. These deductions are
consistent with the work of Brahms et al. who oriented a film of poly-gamma-ethyl-l-
glutamate on glass plates and were able to resolve three peptide LD bands: a weak mainly
perpendicular polarised one around 225 nm; one parallel polarised band centered around
210 nm; and a strong n+x* transition polarized perpendicular to helix axis at around 190
nm. 12
P-sheets are in many ways simpler than a-helices since the n+x* transition does
not have the 208 nm component and its polarisation can be assumed to be along the C=O
bond axis, thus (assuming there is more 'sheet' than 'turn' in the structure) the net
polarisation is perpendicular to the backbone run. No simplistic modelling can give the
transition polarisation for the n -+ n* transitions and so it was necessary to calculate these
for a-helices and p-sheets as well as gramicidin.
In the case of the a-helices and p-sheet, as discussed above, we have 201 peptide
bonds, so consideration of the n+# and n+?P transitions will yield a total of 402
transitions for the protein. Similarly the 32 peptide bonds of the gramicidin structures give
rise to 64 transitions. Not all of these will contribute significantly, and in the results given
below we consider only n+# bands with significant oscillator strengths. In all cases the
axis of the helix (for the p-sheet this means the long axis of the structure illustrated in
Figure 5 ) is taken as z'. The results are summarised in Table 1.
For the n+?P band in an infinite a-helix, three main transitions are predicted, one
polarised parallel to the helix, and two perpendicular. For finite helices some splitting may
occur. In both a-helices and the p-strand considered here (Figure 5 ) the splitting is very
small, justifying our choice of 200 amino acids for the length of our peptide. For the
(q~-48",q~-57") helix, two transitions contribute to the parallel transition, both at 220.83
nm. The third and fiflh transitions in Table l a contribute to one perpendicular transition,
while the fourth constitutes the other. The three bands are all perpendicular to each other to
within +, 2". The total strengths of these bands should be the same, however, some
discrepancy can be seen here due to the neglect of other transitions with small oscillator
strengths (<0.4). The ( q = 6 3 O , y=-41") helix shows similar results, although the parallel
band (transitions one and two in Table lb) is a lot stronger.
Figure 5 Schematic illustration of a P-strand showing the characteristic twist. The z-axis
is along the long axis of the twisted sheet.
For the p-strand, the parallel band at 219.38 nm is not split. The intensity of the
perpendicular bands, however, is scattered over the five remaining transitions in Table lc.
With angles of up to 37" fiom perpendicular to each other, it is difficult to separate these
into the two transitions. The perpendicular bands of the 0-strand occur at a lower energy
than the parallel band. This contrasts with the results for the a-helices.
For the first gramicidin structure (MAG, Figure 6, the more likely one in our
experiments 13) the three transitions in Table Id are quite clear, and show the parallel
transition lying between the two perpendicular ones with the perpendicular polarisation
having greater intensity. Again there is some discrepancy between the strengths of the
perpendicular bands, as discussed above. The geometry of the lMIC structure (Figure 6) is
not as regular, and the direction of the axis is not as well defined. The transitions at 220.55
nm and 220.40 nm are only 4" from being parallel to each other, and form the 'parallel'
band. The remaining two transitions in Table l e make up the perpendicular bands,
although these are 26" from perpendicular to each other, indicating a strong mixing.
Table 1: Wavelengths in nm; short axes and long axis components; and oscillator strengths
of the n((* transitions of a (a) (=(48(, (=(57( 200 alanine residue (-helix; (b) (=(63(, (=(41(
200 alanine residue (-helix, Z is the helix axis; (c) (=(135(, (=135( 200 alanine residue (-
strand; (d) Gramicidin (1MAG); and (e) Gramicidin (1MIC).
Wavelenath/nm
220.76
220.76
220.72
220.72
220.72
219.38
(d) Wavelength/nm Padau Oscillator strength

220.57 -0.354 0.028 0.000 0.406
220.40 0.000 0.000 0.314 0.317
219.41 0.012 0.241 0.000 0.187
0.141
0.269
Figure 6 Two gramicidin structures used for the transition moment polarisation
calculations. lMAG (left), lMIC (right).
Thus for an a-helix, the net effect is a -221 nm band polarised parallel to the helix
axis and a larger 220 nm band polarised perpendicular to the helix axis. The z+7t* band
(calculation data not shown) has a band polarised parallel to the helix axis at 202 nm and
another perpendicular to it at 189 nm. The ratio of calculated transition intensities (for 190
nm:202 nm:220 nm:221 nm) is approximately 230:80:3:1.From these values and Equation
(6) it follows that the LD expected for a helical peptide whose long axis lies parallel
(Figure 7, bottom) to the lipid molecules of a liposome is very approximately:
+115:-80:+1.5:-1 (from low wavelength to higher wavelengths) with the last two
components and the first two components showing significant cancellation. By way of
contrast, if an a-helix lies on the bilayer surface (Figure 7, top) so that the short axis
polarised transitions are half polarised parallel and half perpendicular to the lipids, we
expect: -230:+80:-3 :+1 ,again with cancellation occurring.
Figure 7 Schematic illustration of the proposed thylakoid membrane insertion mechanism

for pre-PsbW indicating a fairly randomly oriented surface bound protein and a-helices in
different orientations on a lipid bilayer.
The net n+x* transition intensity is short axis polarised perpendicular to the axis
of the twisted P-strand of Figure 5 . Thus if a twisted P-strand is oriented so that the peptide
backbone runs parallel to the lipids (Figure 8, middle), then for both the n+x* and x+n*
transitions LD/3S = +1/4; if, however, the peptide backbone runs perpendicular to the
lipids (Figure 8, left and right), then LDr/3S= -1/4 since half the transitions will be parallel
to the lipids and half perpendicular.
II II I I II II
Figure 8 Schematic illustration of P-sheets (from left to right) on the surface of, oriented
parallel to the lipids and perpendicular to the lipids of the bilayer.
7 PROTEIN CD AND LD SPECTRA
The proteins we chose to study were: gramicidin, cytochrome c, a thylakoid membrane

protein precursor pre-PsbW, and a monoclonal antibody. Gramicidin is a well known
membrane protein with an unusual helix dimer spanning the bilayer;14 Cytochrome c is a
membrane associated protein with a mixture of helical and other motifs and also a strong
Soret band ab~orbance;'~ Pre-PsbW has two transmembrane a-helices in the thylakoid
membrane prior to being processed to having the one (Figure : 7 and the monoclonal
)
antibody has no a-helical character in aqueous solution as shown by CD spectroscopy.
Thus the proteins have a range of structural motifs. Our hope was that we would see
evidence of the orientation on the liposomes of these proteins.
Gramicidin is not water soluble so the samples were prepared in two different
ways. In the frst the protein was extruded with the liposomes and gradually incorporated.
In the second method it was dissolved in ethanol and added to an aqueous solution of
liposomes. The latter method produced much better spectra which were consistent with
those obtained by the other method but showed more details.
The gramicidin LD absorbance spectrum (Figure 9) has an unusually large
absorbance band in the aromatic region due to the high concentration of tryptophan
residues in this protein. The corresponding CD signal is a structured band (also of
unusually large intensity) with alternating positive and negative components, presumably
from opposite signed contributions from the vibronic components of the overlapping La
and L b (starts at longer wavelength) bands whose polarisations are approximately
perpendicular to one another and across the diagonals of the tryptophan." The aromatic LD
also has alternating signed bands. Assuming these are due to the components of the two
bands having opposite signs, then we must conclude that the tryptophans are on average
oriented so that one of these two bands is more parallel and one more perpendicular to the
lipids.
1.5
al
2 0.5
a
t
6
m 4
U
5 2
a
0 0
-2
0.01
g 0.000
c
8 0.006
m
c
2 0.004
0,
9 0.002
Q
4 0
-0.002
Wavelength I n m
Figure 9 Absorbance, CL)(averaged over 64 scans), and LD spectra of gramicidin (0.5 mg

in 2.5 mg/mL lipid in phosphate buffer) in soybean liposomes prepared by adding ethanol
solubilised gramicidin to an aqueous extruded liposome solution. The sample has held at
4" for 2 hours prior to analysis. Absorbance and CD baselines were collected with the
same sample preparation method but without gramicidin.
Following reference l 3 the CD confms that the head to tail dimer of Figure 6a is
the dominant species in our experiments. As this hydrogen bonding pattern resembles that
of a P-sheet rather than that of an a-helix, there is no 208 nm n+n* transition apparent in
the spectra. The tryptophan LD from 270 -300 nm requires the tryptophans not to lie
parallel to the liposome surf'ace. The LD spectrum below 250 nm has positive maxima at
227 nm and -200 ntn and a negative maxima at 2 13 nm and below 180 nm (from data for a
more dilute sample which is not shown). Assuming the longer wavelength bands are the
result of cancellation of the n+n* components (see Tables Id and le), the longest
wavelength component would be perpendicular to the helix axis and the shorter
wavelength component parallel to it. The 200 nm maximum is then a short axis polarised
n+x* transition. This requires the gramicidin helices to lie parallel to the lipids,
presumably inserted into the membrane.
Figure 10 LD spectrum of cytochrome c (1 mg in 1 mL of 2.5 mg/mL lipid in phosphate

buffer and serial dilutions 1:2 and 1 :4, with spectra decreasing in magnitude) in extruded
soybean liposomes.
The CD spectrum of cytochrome c (not shown) indicated that the protein was in its
usual folded conformation in the presence of the liposomes. The 1 mg/mL cytochrome c
-
LD spectrum (Figure 10) shows the presence of a positive tryptophan signal at 280 nm
and the in-plane polarised heme chromophore Soret band with a bisignate signal at 425 -
nm. The plane of the heme group is not parallel to the surface (otherwise both components
would be positive), however, the positive band is not half the magnitude of the negative
component, indicating the heme is not perpendicular to the surface either. With fbrther
calculations or experiments to determine the polarisations of the Soret band, coupled with
information about the tryptophan groups, it would be possible to determine the orientation
factor S and the orientations of these chromophores in the membrane. A positive
maximum, presumably due to the n+n* transition occurs at -240 nm, a negative
maximum at -220 nm followed by a large positive signal below 200 nm. These data all
indicate that the average a-helix orientation is more parallel than perpendicular to the
lipids.
Pre-PsbW is a thylakoid membrane protein that can exist in tris-buffer soluble
multimeric form that is folded into a significantly a-helical structure.’ There is still
significant debate about how it inserts into thylakoid membranes. The helix of the mature
protein and a hydrophobic region in the presequence are all a-helical in SDS micelles.’
Our attempts to orient the SDS samples gave signals of the order of lod5absorption units
and terrible signal to noise ratios. When added to soybean lipids, pre-PsbW gave a CD
spectrum indicative of either no secondary structure or some P-sheet structure. However,
due to uncertainties in concentrations it was not possible to fit these data.16 The LD
spectrum of this system is given in Figure 11. It shows a very large positive signal at 204
nm (no signal above 220 nm) and an even larger negative signal below 190 nm that
overlaps with the 204 nm band. It is thus likely that the 204 nm band is the 7c+z* band
(shifted to longer wavelength due to cancellation), which indicates that the protein is
unfolded on the surface of the liposome, perhaps retaining some p-sheet like hydrogen
bonding. Assuming similar behaviour on the thylakoid membrane prior to insertion, this
supports the mechanism of pre-PsbW insertion schematically illustrated in Figure 7.
200 210 220 230 240 250

Wavelength I nm
Figure 1 1 LD spectrum of pre-PsbW (1 mg in 2.5 mg/mL lipid in phosphate buffer) in

soybean liposomes.
The CD spectrum of the monoclonal antibody (data not shown) fitted using CDsstr
l6 indicates about 35% P-sheet, 12% turns, and 10% poly-proline type I1 helix. The LD
spectrum shows a broad positive band whose maximum is at 230 nm. Below 215 nm a -
large negative signal occurs. Assuming these two bands are the components of the n+n*
transition, we have the short axis polarised band being positive and the long axis polarised
being negative suggesting the strand axis is parallel to the lipids. Such a picture is
consistent with the schematic pictures of antibodies binding to bilayers.
0.004
I
$ 0
0
2 -0.002
i
9 -0.004
200 220 240 260 280 300 320

Wavelength I nm
Figure 12 LD spectrum of a monoclonal antibody (1 mg in 2.5 m g h L lipid in phosphate

buffer) in soybean liposomes.
8 CONCLUSIONS
The work reported here is the first time flow linear dichroism has been used to probe the
orientation of proteins in lipid bilayers. In order to accomplish this, we have designed and
constructed a new flow LD cell with 50 pm annular gap and calcium fluoride optics that
has enabled us to measure flow oriented spectra of micellar and liposomal systems down to
185 nm in some cases.
Although the wavelength range is potentially wider with the micellar systems the
orientation is less so the LD signal to noise was generally worse than with the liposomes.
The orientation of multi-lamellar liposomes was higher but the background light scattering
was worse than with unilamellar liposomes. So in each case a choice has to be made that is
optimal for a given system.
A set of model protein systems, encompassing helical and sheet proteins, as well as
an unfolded protein have been placed in membrane mimicking environments and their
secondary structure motif orientations probed using linear dichroism. It has been possible
to deduce qualitatively the orientations of the proteins with respect to the normal to the
lipid bilayers using transition moments for different secondary structure motifs calculated
by the Matrix Method.
The technique of flow LD of proteins has potential for monitoring the kinetics of
insertion of proteins into membrane mimicking environments as well as determining
steady state orientations. In order to use the data more quantitatively to determine absolute
orientations of proteins in membranes, a more effective way of removing the background
scatter is needed since adding the proteins to the liposomes increases the scatter as does
flowing the samples. Work is in progress on this problem.
ACKNOWLEDGEMENTS
Professor Malcolm MacMey and Sirilak Wannaborworn, Cambridge University, are

thanked for use of their shear flow microscope system to visualise the liposomes. Support
of the Engineering and Physical Sciences Research Council is gratefblly acknowledged
(GR/M91105(AR),GR/R40869/01(AR),GR/N66971(JDH)
REFERENCES:
1. B. NordCn, M. Kubista, T. Kurucsev, T. Q.Rev. Biophysics 1992,25,51-170

2. A. Rodger, B. NordCn Circular dichroism and linear dichroism, &ford University
Press, Oxford, 1997
3. A. Rodger, A. Methods in Enzymology, 1993,226,232-258
4. M. Ardhammar, N. Mikati, B. NordenJ. Amer. Chem. SOC.1998,120,9957-9958; M.
Ardhammar,P. Lincoln, B. Norden J. Phys. Chem B. 2001,105,11363-1 1368
5 . C. Woolhead, A. Mant, S.J. Kim, C. Robinson, A. Rodger J. Biol. Chem.,2001,276,
14607-14613
6. A.A.M.Morales, G. Nunez-Gandolff, N.P. Perez, B.C. Velk, I. Caballero-Torres, J.
Duconge, E. Fernandez, F.Z. Crespo, A. Veloso, N. Iznaga-Escobar, Nuclear Medicine
and Biology, 1999,26,717-723
7. B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S. Swaminathan, M.
Karplus, J. Comp. Chem. 1983,4,187-217
8 . L. Pauling, R.B. Corey, Proc. Nut. Acad. Sci. USA 1951,37,235-240 & D. J. Barlow,
J. M. Thornton J. Mol. Bid. 1988,201,601-619
9. R. R. Ketchem, K. C. Lee, S. HUO,T. A. Cross, J. Biomol. N M , 1996,8, 1-14; Y.
Chen, A. Tucker, B. A. Wallace, J. Mol. Biol. 1996,264,757-769
10. P. M. Bayley, E. B. Nielsen, J. A. Schellman, J. Phys. Chem., 1969, 73,228-243; W. J.
GOUX,T. M. Hooker Jr., J. Am. Chem. SOC.1980,102,7080-7087
1 1 . N. A. Besley, J. D. Hirst, J. Am. Chem. SOC.1999,122,9636-9644
12. J. Brahms, J. Pilet, H. Damany, V. Chandrasekharan, P. N.A.S. 1968,60, 1 130-1 137
13. T.P. Galbraith, B.A. Wallace, Faraday Discuss. 1998, I 1 I , 159-1 64
14. D. Voet, J.G. Voet, Biochemistry 2nd edn, John Wiley and Sons 1995
15. B. Albinsson, M. Kubista, B. Norden, E.W. Thulstrup J. Phys. Chem. 1989,93,
6646-6654
16. W.C. Johnson, Proteins: Structure, Function and Genetics 1999,35,307-3 12
QUANTITATIVE PROTEIN CIRCULAR DICHROISM CALCULATIONS
N. A. Besley and J. D. Hirst*
School of Chemistry, University of Nottingham, University Park,

Nottingham NG7 2RD, UK
1 INTRODUCTION
Despite inherent complexities and challenges, the determination of protein structure is an

active area of research. Much of the motivation for this interest arises from the potential
benefit to areas such as drug design. Furthermore, the recent fruition of genome
sequencing efforts have heightened the demand for protein structure determination.
Membrane proteins are an important class of proteins and constitute a significant
proportion of pharmaceutical targets. The structures of these proteins are very difficult to
determine through NMR or X-ray crystallography. Consequently, alternative techniques
need to be explored.
The measurement of protein electronic circular dichroism (CD) spectra in the far-UV
has become routine, and provides a probe of protein secondary structure.''2 The
characteristic CD spectrum of an a-helix consists of a positive band at 190 nm and two
negative bands at 208 nm and 220 nm?94The CD spectra of P-sheet proteins can be
divided into two classes.' The first class has a maximum at 195 nm and a minimum in the
region 2 10-220 nm.Alternatively, other P-sheet proteins have a CD spectrum similar to
that of a random coil with a minimum at about 200 nm.
In recent years, a number of developments have catalysed renewed interest in protein
CD. Synchrotron radiation CD measurements can determine protein CD down to
wavelengths as low as 160 nm,6 compared to -180 nm for traditional CD measurements.
This additional information allows secondary structural components to be resolved in
greater detail. In addition, time-resolved far-UV protein CD measurements on the
microsecond7and nanosecond' time scales coupled with the increasing length of molecular
dynamics simulations9means that detailed modelling and understanding of the early events
in protein folding is tantalisingly close. However, to exploit this convergence of
experiment and theory requires quantitative calculations of protein CD spectra. Herein, we
will describe our recent work toward the development of quantitative protein CD
calculations and highlight some current applications of these methods.
2 THEORY
Circular dichroism is the differential absorption of left and right circularly polarised light,
and can be expressed as the differencebetween the left and right extinction coeficients.
The rotational strength represents the integrated intensity beneath a band in the CD
spectrum. Rotational strengths can be evaluated through the Rosenfeld equation.l o For an
electronic transition A t 0, the rotational strength &A is given by
where Fcand fi,represent the electronic and magnetic transition dipole moments. In
contrast, the intensity of a band in traditional electronic absorption spectra is given by
Consequently, the intensity of bands in CD spectroscopy depends on both electronic and

magnetic dipole moments while in electronic absorption spectroscopy only the electronic
dipole moment determines the intensity. Physically, CD involves both linear and rotational
displacement of charge.
In seminal theoretical work on the CD of biopolymers, M~ffitt"-'~ applied exciton
theory to the study of a-helices. The 190 nm amide m* transition was shown to split into
components, polarized parallel and perpendicular to the helix axis. Furthermore, the right-
handedness of a-helical polypeptides was predicted. These predictions were later
confirmed by e~periments.3.~ Current1 , the majority of calculations of the CD of large
systems adopt the matrix method"'% formalism. In this approach, the biopolymer is
assumed to comprise M non-interacting chromophoric groups. Electronic excitations can
occur only within a chromophoric group and not between groups. The wave function of the
whole molecule is expressed as a linear superposition of basis functionsaiP
i a
These basis functions are a product of wave functions describing the individual
chromophoric groups, which can be in an electronically excited state
@iu = ~,,...~i~...~,~...~~~ (5)
where 'pi, represents the wave function of the chromophore i, which has undergone an
electronic excitation a t 0. A Hamiltonian matrix is constructed by considering the
electronic Hamiltonian of the whole system. The diagonal elements of this matrix are the
energies of the electronic excited states of the chromophoric groups, while the off-diagonal
elements describe the interactions of a chromophore with the rest of the molecule. The off-
diagonal elements typically have the form
U d J b = II'PiO'Pi~~ij'Pio'Pjbdzidzj
i j
If only electrostatic interactions are considered then
where pio,and pjobarethe respective permanent or transition densities. These densities are
usually represented by a set of point charges." Then eqn. (7) becomes
If only nn* and nn* transitions are considered, then in the simplest case of a diamide the
Hamiltonian matrix takes the form
This matrix is then diagonalised by a unitary transformation. The transition energies of the
interacting system are given by the diagonal elements. The electronic and magnetic
moments are obtained by transforming the local moments by the same unitary
transformation. Subsequently, the CD can be evaluated through eqn. (2).
Despite over 40 years since the publication of Moffitt's work, quantitative calculations
of protein CD from first principles have not been achieved. If the basic assumptions of the
matrix method are valid (we will return to this later) then the accuracy of the calculations
depends on the construction of the Hamiltonian matrix. In most matrix method calculations
only the protein backbone is considered. N-methylacetamide (NMA) is used as a model of
the protein backbone chromophore. Consequently, in order to construct the matrix we
require the energies of the excited states of NMA, in addition to accurate representations of
the permanent and transition densities associated with the electronic excitations. In recent
years there has been a flurry of investigations both experimental and theoretical toward
charactensing the electronic excited states of amides
3 ELECTRONIC EXCITED STATES OF AMIDES
3.1 Electronic Structure
A key feature of the electronic structure of amides is the tt electron system. The C, N and
0 atoms each contribute one-electron to the n electron system. The n bonding orbital, nb, is
the in-phase combination of three 2p atomic orbitals with no nodes perpendicular to the
plane defined by the amide group. The non-bonding n orbital, nnb, has a perpendicular
node at the C atom. The anti-bonding z* orbital has two perpendicular nodes. Another
important feature of amide electronic structure is the two lone pairs on the oxygen atom, n
and n', which lie in the plane defined by the amide group and are orientated along the C-0
bond and perpendicular to it, respectively.
3.2 Experimental Studies
A number of early experimental measured the electronic spectrum of a variety

of mono-amides in both gas-phase and solvent. In these studies the characteristic electronic
spectrum of amides in gas-phase was established.22The spectra are dominated by an
intense band, the V band (V for valence), that was assigned to the &&* transition. In
addition, a weak band, the W band (W for weak), corresponding to the nz* transition
appears as a shoulder on the high wavelength side of this band, Superimposed on these
bands are a number of sharp bands that are attributed to transitions to Rydberg states,
involving excitations to diffuse atomic-like Rydberg orbitals. A furcher broad band appears
in the spectrum at higher energy. This band is often termed the Q band and was originally
assigned to the Kbn* transition. Furthermore, these studies show that the position of the
&bk* band is sensitive to substitution of the amide hydrogens. More recently, the
electronic spectrum of fonnamide was measured.u The nlt* and 7c,,bn* bands were found at
5.82 eV and 7.36 eV respectively. It was also proposed that the Q band arises from a
superposition of transitions to several Rydberg states and the m,R* transition lies at higher
energy.
While there has been considerable effort in measuring the electronic absorption
spectra of amides, there has been relatively few studies that determine the orientation of
the transition dipole moment of the n&* band. This is of interest, since this angle is
critical for calculations of protein CD. In 1957 Peterson and Simpson'' determined a value
of -41' for myristamide, where the angle is characterised by the angle between it and the
carbonyl bond with the OCN angle taken to be positive. In later work, Clark" reported
values of -35' and -55' for propanamide and N-actetyl lycine respectively. For hydrated
crystals of glycine-glycine and angle of 46' was found.29
In solution the amide absorption spectra show large changes.22The nnbfl* band
undergoes a large red-shift of up to 0.5 eV and its intensity is reduced. In polar solvents the
nz* band is subject to a blue-shift. In addition, there is no evidence of Rydberg bands.
3.3 Theoretical Studies
Modern computers have made it possible to study electronic excited states of amides using
state-of-the-art ab initio methods. Multi-reference configuration interaction (MRCX)
calculations have been reported for a number of amides, including formamide26and
NMAO2'Amides have also been studied using complete-active-space self-consistent-field
with multi-configurational perturbation theory (CASSCF/CASPT2)28 and equation of
motion coupled cluster theory?' These studies illustrate the problems associated with the
study of amide excited states. The large number of low lying Rydberg states gives rise to
artificial Rydberg-valence mixing which results in an overestimation of the n&* transition
energy. Within CASSCFKASPTZ procedures this can be overcome by a second
calculation to determine the valence state properties in which the Rydberg states have been
'deleted'. Using this procedure the excitation ener ies of a number of amides were
determined within a few tenths of an electron volt?f In a recent study:' the thermally
broadened VUV absorption spectrum of formamide was determined using an
implementation of time-dependent density functional theory (TDDFT)in a plane wave
basis set formalism. Excellent agreement with experiment was found, and importantly the
Q band was shown to arise predominantly from n+3d transitions.
Gas phase Condensed phase
Pre- 1996 Current
Figure 1 Schematic mono-amide molecular orbital diagram
A number of studies have addressed the problem of modelling the effect of solvent on
amide electronic spectra. The description of solvent within an ab initio framework is a
difficult problem since a large number of solvent molecules and their numerous
configurations need to be considered. Generally, three approximate approaches are
adopted. In the continuum approach, the solvent is modelled by a continuum characterised
by a dielectric constant in which there is a cavity that contains the solute. This provides a
gross description of the solvent in which short-range solute-solvent interactions, such as
hydrogen bonding, are neglected. Alternatively, explicit solvent molecules can be included
within the a b initio treatment in a discrete model. The computational cost of this approach
rises rapidly and only a small number of solvent molecules can be considered. These two
approaches can be combined in a semi-continuum model.
The additional computational cost arising from the introduction of solvent has meant
most studies of solvated amide electronic spectra have concerned formamide. Sobolewski3'
reported CASSCFKASPT2 calculations on the formamide-H20 complex using a double-6
plus polarisation basis set. They calculated transition energies and oscillator strengths (in
parentheses) of 6.03 eV (0.001) and 7.52 eV (0.337) for the nn* and &bn* transitions
respectively. These showed a blue-shift for the nn* band and small red-shift for the &,bn*
band when compared to the corresponding values of 5.85 eV (0.001) and 7.67 eV (0.317)
for isolated formarnide. The formamide-H20 complex was also studied using multi-
configurational self-consistent field, a nlt* transition energy of 6.14 eV was found.32
Within a continuum approach, we have examined the electronic spectrum of formamide
and NMA.33 In this study, the continuum model was supplemented with a repulsive
potential34outside the cavity that represents the Pauli repulsion between the solute and
solvent. This repulsion has the effect of destabilising the larger Rydberg states relative to
the valence states. A by-product of this is a decrease in the Rydberg-valence mixing
problem experienced in gas-phase studies. Transition energies for the nnbZ* band in water
of 6.95eV (0.256)and 6.60eV (0.277)were found for formamide and NMA respective1
These are in good agreement with experimental values of 6.81 eVZ2and 6.67 eV.Ti
However, the computed nn* transition energies of approximately 5.50 eV did not show a
significant blue-shift. In a subsequent study,35the electronic spectrum of formamide was
investigated using a discrete solvent model. A motivation for this study was to confm the
destabilisation of the Rydberg states by the solvent. The introduction of a small number of
water molecules was to destabilise of the Rydberg states by up to 0.5 eV. When combined
with a continuum all the features of the electronic spectrum of foxmamide in solution were
reproduced, including the blue-shift in the nlc* transition energy. A similar approach was
applied to the study of NMA.36Following these studies the amide molecular orbital picture
has been revised, as depicted in Figure 1.
3.4 Diarnide Electronic Structure
Calculations of electronic structure of diamides represent the first step towards protein
electronic structure. Hirst and Persson3’ presented CASSCF/CASPT2 calculations of the
planar form of diketopiperazine (DKP). The high symmetry of DKP precludes optical
activity. However, the coupling between the two n&c* transitions can be examined. These
transitions were found to lie at 5.98 eV (0.580)and 7.09 eV (0.076), representing a
splitting of 1.1 1 eV with the intensity appearing in the band of lower energy. Another
feature of the study of diamides are charge transfer transitions. These are characterised by
a large displacement of electron density from one amide group to the other. In DKP these
are forbidden. In another CASSCFKASPT2 study?* a linear dipeptide was studied and the
electronic spectrum computed for a number of configurations. This work identified charge
transfer bands that were computed to lie at -7.1-7.5eV, matching experimentally observed
bands. In a recent experimental and theoretical infra-red, W and CD spectra were
reported for the bridged cyclic diamide diazabicyclo[2.2.2]octane-3,6-dione I.
Figure 2 Diazabicyclo[2.2.2]octane-3,6,dione, I
The bridged cyclic diamide is an attractive molecule in which to study the interactions
between the amide groups because it is conformational rigid and its structure is known by
X-ray crystallography.40 Consequently,ambiguity regarding its conformation in solution is
removed. Furthermore, its CZ symmetry provides computational advantages. Table 1
shows the electronic and CD spectra of I in gas-phase and solution.
Similarly to mono-amides, the spectra are dominated by an intense x&* transition
which is calculated to lie at 6.28 eV in gas-phase and 6.07 eV in solution. The symmetry of
I means that it is not possible to classify the electronic transitions rigorously as local or
charge transfer within a delocalised molecular orbital picture, It is possible to determine
transitions present in the diamide that have no analogues in mono-amides. These
transitions can be associated with charge transfer transitions found in dipeptides of lower
symmetry. However, some caution must be taken since these studies compute vertical
transitions and this may be a poorer approximation for these bands. Furthermore, I is chiral
and may be studied using CD spectroscopy. The computed CD spectra are also shown. The
results showed that the CD spectra arise predominately fiom transitions that can be
considered local in nature. Both nz* and &bz* transitions give rise to two bands in the CD
spectrum of opposite sign. However, the splitting of the nz* bands is small and only one
band is evident, while the larger splitting of the znbz*transitions leads to two resolved
bands. Overall, the agreement with experiment can be regarded as qualitative with results
in solution closer to experiment. Larger basis sets and non-vertical transition energies are
likely to lead to improved agreement with experiment.
Gas Phase Solution Experiment
State AE(eV) R(DBM) AE(eV) R(DBM) AE(eV) R(DBM)

2'A(nz*) 5.48 -0.04 5.70 -0.03
3'A(%br~*) 6.28 -0.3 1 6.67 -0.27 5.69 -0.17
1'B(nn*) 5.47 +O. 13 5.77 +o. 10 5.30 +0.11

2'B(K&rC*) 6.6 1 +0.23 6.48 +0.29 6.23 +0.11
Table 1 Experimental and Calculated Electronic and CD Spectra of I
4 CALCULATIONSOF PROTEIN CD
While ab initio calculations of amides become prohibitive beyond diamides, they can
prove important in the calculation of protein CD. The matrix method requires energies of
the excited states and parameters that describe the permanent and transition electron
densities. These are problems well suited to ab initio methods.
Recently, we have used ab initio studies of NMA as a basis for reparametrisation of
the amide chromophore. This has resulted in better agreement with experiment!t43 This
improved agreement arises for three main reasons. Firstly, ab initio methods should
provide a more accurate description of electron densities than the semi-empirical methods
used previously. Secondly, using parameters based on condensed phase calculations
provides a more suitable description of the amide chromophore since the environment of a
protein is usually considered to be condensed phase. A third source of improvement comes
fiom a more accurate representation of the electron densities in eqn. (8). Traditionally, a
small number of point charges, typically 6- 12, located around the atom centres are fitted to
reproduce the computed dipole and quadrupole moments. We have adopted an alternative
approach in which a larger number of charges, 20 or 32 depending on symmetry, are fitted
to reproduce the electrostatic potential. To assess the accuracy of the different parameter
sets the computed CD spectra for a set of 29 proteins are compared with experiment. This
set includes (PDB codes shown in parentheses) cytocbrome c (3cyt), hemoglobin (1hco),
myoglobin (1mbn), bacteriorhodopsin (2brd), alcohol dehydrogenase (5adh), glutathione
reductase (3grs), lactate dehydrogenase (61dh), lysozyme (71yz), papain (Spap), rhodanese
(lrhd), subtilisin (1sbt), thermolysin (4tln), triose phosphate isomerase (1 tim), flavodoxin
(2fx2), carbonic anhydrase (lca2), concanavalin A (3cna), h-immunoglobulin (lrei),
ribonuclease A (3rn3), ribonuclease S (2ms), erabutoxin (3ebx), plastocyanin ( lplc), porin
(Jpor), prealbumin (2pab), a-chymotrypsinogen A (2cga), a-chymotrypsin I1 (5cha),
elastase (3est), superoxide dismutase (2sod), trypsin inhibitor (4pti) and trypsin (3ptn).
This set of proteins is representative of all types of protein CD. Table 2 shows the
Spearman rank correlation coefficients are computed at three important wavelengths of

190 nm, 208 nm and 220 nm for a number of different parameter sets. Figures in bold
represent a significant correlation at the 1% level.
Table 2 Comparison with experiment (data set comprising 29 proteins)
The correlation between experiment and theory is poor at all three wavelengths for the
parameter sets based on semi-empirical calculations. Gas phase MRCI parameters,
comprising 6 charges fitted to reproduce the molecular dipole and quadrupole, show an
improved correlation at 220 nm. The CASSCF parameters derived from condensed phase
studies are the first to show moderate correlation at 190 nm and 208 nm. This is initially
surprising since the CASSCF wave function is inferior to the MRCI wave function,
containing no description of dynamic correlation. However, this improved agreement
arises from an improved orientation of the &bn* transition dipole moment. Physically this
may be because there is a significant difference between this angle in gas-phase and
solution. More likely is that the removal of Rydberg-valence mixing in the condensed
phase calculations results in a better description of the &I$* state. Using this wave
function and improving the description of the charge densities leads to a significant
improvement in the correlation at all three wavelengths. Spearman rank correlation
coefficients of 0.84, 0.73 and 0.90 at 190 nm, 208 nm and 220 nm, respectively, are
achieved. The slopes of the regression lines were 0.6, 1.22, 1.02, respectively. In particular,
the agreement at 220 nm is encouraging, since the intensity at 220 nm is associated with
the helical content of the proteins. Detailed analysis of the different class of proteins shows
that much of the error arises fiom the class I1 of P-sheet proteins. For these proteins the
computed CD are not even qualitatively correct, predicting bands of the wrong sign. One
possible reason is that fluctuations in backbone conformation may result in large changes
in the CD.
Other authors have also reported improved calculations of protein CD. Using an
improved semi-empirical based parametrisation Woody and S r e e n ~ m apresented
~~ matrix
method calculations of protein CD that showed good correlation with experiment at the
three wavelengths. This was achieved by fixing the transition dipole moment of the &bn*
transition to that measured by Clark24and improving the location of the point charges.
Following an alternative procedure, Bode and Applequist4’ used the dipole interaction
method to study the CD of a number of proteins. However, since this model does not
include the nn* transition, modelling of the band at 220 nm is necessarily incomplete.
A logical progression for improving these parameters would be to use condensed
phase MRCI calculations with the electrostatic potential fitting of charges. This may result
in improved correlations but is unlikely to remedy the problem of the class I1 p-sheet
proteins. However, there are more significant sources of error. In our calculations we have
only considered the protein backbone. Electronic excitations on side-chain chromophores
can contribute to protein CD in the far-UV?"' Including these excitations within the
matrix method framework is straightforward. Presently, high quality parameter sets for
side-chains are not available. In principle these can be generated in an analogous way to
NMA. However, the large number of side-chain chromophores makes this a considerable
undertaking. A principal assumption in the matrix method is that electronic excitations
cannot occur between backbone chromophores. The inclusion of charge transfer
excitations within the matrix method formalism is possible. These avenues are currently
being explored in our laboratory.
5 APPLICATIONS
The correlation of 0.9 at 220 nm suggests that the intensity at this wavelength can be
studied with some confidence. The dependence of intensity with helix length has been
expressed as4*
where [e&,,is the ellipticity at 220 nm, [el;, the ellpiticity at 220 nm for an infinite helix,
r the helix length and k is a constant. From computed values of [el, for Pauling-Corey
helices both [el;, and k can be determined through a least squares fit. Values of [eT;;,= -
37000 degcm2dmol-' and k-2.8 were reported.43 In addition, the dependence of the
intensity at 220 nm with helix conformation was examined for 124 helical fragments
extracted fiom the X-ray structures of the proteins in the data set.
In a recent the influence of hydro en bond length on the 220 nm intensity was
investigated. This was motivated by a report' % of unprecedented helicities for peptides of
the form AcHe-(AlaLys),Ala2-NH2 Calculation of [€I& for
, model helical fragments
generated by constrained minimisation in a molecular mechanics force field showed that
short hydrogen bond lengths resulted in unusually high intensities. Since the experimental
measurements used a watedethylene glycol solvent which is known to promote hydrogen
bonding, this may rationalise the experimental observations.
6 CONCLUSIONS
The development of time-resolved CD spectroscopy provides an exciting new probe of the

evolution of protein secondary structure. To realise the full potential of this technique
requires theoretical simulations. In this paper, we have described recent progress towards
quantitative calculations of protein CD form first principles. Through reparametrisation of
the electronic excited states of NMA, near quantitative calculations of protein CD have
been achieved within the matrix method formalism. In the near future, we anticipate that
further refinement of these calculations coupled with molecular dynamic simulations will
provide new insights into the structure and dynamics of proteins.
7 ACKNOWLEDGMENTS
Some of the work described here was carried out at The Scripps Research Institute under
support from NSF, and ACS Petroleum Research Fund. Our current work is supported by
EPSRC and BBSRC.
References
1 K. Nakanashi, N. Berova and R. W. Woody, Circular Dichroism Principles and

Applications, VCH, New York, 1994.
2 Circular Dichroism and the ConformationalAnalysis of Biomolecules, ed. G. D. Fasman,
Plenum Press, New York, 1996.
3 K. Rosenheck and P. Doty, Proc. Natl. Acad. Sci. USA,1961,47, 1775.
4 G. Holzwarth and P. Doty, Proc. Natl. Acad. Sci. USA, 1965,87,218.
5 P. Manavalan and W. C. Johnson Jr., Nature, 1983,305,83 1.
6 B. A. Wallace, Nature Str. Biol., 2000,7,708.
7 E. Chen, P. Wittung-Stafshede and D. S.Kliger, J. Am. Chem. Soc., 1999,121,381 1.
8 C.-F. Zang, J. W. Lewis, R. Cerpa, I. D. Kuntz and D. S. Kliger, J. Phys. Chem., 1993,
97,5499.
9 Y.Duan and P. A. Kollman, Science, 1998,282,740.
10 L. Rosenfeld, 2.Phys., 1928,52, 161.
11 W. Moffitt and J. T. Yang, Proc. Natl. Acad. Sci. USA, 1956,42,596.
12 W. Mofitt, J. Chem. Phys., 1956,25,467.
13 W. Moffitt, D. D. Fitts and J. G. Kirkwood, Proc. Natl. Acad. Sci. USA, 1957,43,723.
14 R. W. Woody, J. Chem. Phys., 1968,49,4797.
15 P. M. Bayley, E. B. Nielsen and 3. A. Schellman,J. Phys. Chem., 1969,73,228.
16 W. J. Goux and T. M. Hooker Jr, J. Am. Chem. Soc., 1980,102,7080.
17 I. Tinoco, A h . Chem. Phys., 1962,4, 113.
18. H. D. Hunt and W.T. Simpson, J. Am. Chem. Soc., 1953,75,4540.
19. D. L. Peterson and W. T. Simpson, J. Am. Chem. SOC.1957,79,2375.
20. K. Kaya and S. Nagakura, Theor. Chim. Acta, 1967,7,124.
21. E. B. Nielsen and J. A. Schellman,J. Phys. Chem., 1967,71,2297.
22. H . Basch, H. B. Robin and N. A. Kuebler, J. Chem. Phys., 1968,49,5007.
23. J . M. Gingell, N. J. Mason, H. Zhao, I. C. Walker and M. R. F. Siggel, Chem. Phys.,
1997,220,19 1.
24. L. B. Clark, J. Am. Chem. Soc., 1995,117,7974.
25. V. Pajcini and S.A. Asher, J. Am. Chern. SOC., 1999,121,10942.
26. J. D. Hirst,D. M. Hirst and C. L. Brooks 111, J. Phys. Chem., 1996,100,13487.
27. J. D. Hirst, D. M. Hirst and C. L. Brooks 111, J. Phys. Chem., 1997,101,4821.
28. L. Serrano-Andr6s and M. P. Fillscher, J. Am. Chem. SOC.,1996,118, 12190.
29. P. Szalay and P.Fogarasi, Chem. Phys. Lett., 1997,270,406.
30. N . L.Doltsinis and M. Sprik, Chem. Phys. Lett., 2000,330,563.
31. A. L. Sobolewski, Photochem. Photobiol., 1995,89,89.
32. M . b u s s and S. P. Webb, J. Chem. Phys., 1997,107, 5771.
33. N. A. Besley and J. D. Hirst, J. Phys. Chem. A, 1998,52,10797.
34. A. Bemhardsson, R. Lindh, G. KarlstrUm and B. 0. ROOS, Chem. Phys. Lett., 1996,
151, 141.
35. N. A. Besley and J. D. Hirst, J. Am. Chem. SOC.,1999,121,8559.
36. N. A. Besley and J. D. Hkst, J. Mol. Stmchrre (THEOCHEW, 2000,506,161.
37. J. D. Hirst and B. Joakim-Persson,J. Phys. Chem. A, 1998,102,7519.
38. L. Serrano-Andrks and M. P. Fiilscher,J. Am. Chem. SOC.,1998,120,10912.
39. N. A. Besley, M.-J. Brienne and J. D. Hust, J. Phys. Chem. B, 2000,104, 12371.
40. M.-J. Brienne, J. Gabard, M. Leclercq, J.-M. Lehn, M. Cesario, C. Pascard, M. Chevk
and G. Dutruc-Rosset, Tetrahedron Letts., 1994,35,8157.
41. J. D. Hirst, J. Chem. Phys., 1998,109,782.
42. J. D. Hirst and N. A. Besley, J. Chem. Phys., 1999,111,2846.
43. N . A. Besley and J. D. Hirst, J. Am. Chem. SOC.,1999,121,9636.
44.R. W. Woody and N. Sreerama,J. Chem. Phys., 1999,111,2844.
45. K. A. Bode and J. Applequist, J. Am. Chem. Soc., 1998,120, 10938.
46. R. W. Woody, Biopolymers, 1978,17,1451.
47. A. Chakrabarthy, T. Kortemme, S. Padmanabhan and R. L. Baldwin, Biochemistry,
1993,32,5560.
48. Y.-H. Chen, J. T. Yang and H. M. Martinez, Biochemistry, 1972,11,4120.
49. Z . Dang and J. D. Hirst, Angew Chem. Intl. Ed., 2001,40,3619.
50. P. Wallimann, R. J. Kennedy and D. S. Kemp, Angew. Chem. Intl. Ed., 1999,38,1290.
PROBING CELLULAR STRUCTURE AND FUNCTION BY ATOMIC FORCE
MICROSCOPY
Michael A. Horton, Petri P. Lehenkari and Guillaume T. Charras
Department of Medicine, University College London, London WClA 655, U.K.

E-mail: m.horton@ucl.ac.uk
1 SUMMARY
Scanning probe microscopy techniques have been applied increasingly to cell biological
problems that make use of their precision in spatial and force control rather than their
ability to image at high resolution. We have developed a ‘life sciences platform’ that
integrates a commercial atomic force microscope (AFM) with an inverted optical
microscope equipped with bright field video recording, fluorescent imaging, and laser
scanning confocal microscopy. This has enabled us to image and analyse live cells in real
time by AFM including measurement of cell topography and material properties, force-
distance analysis and binding site mapping, together with simultaneous monitoring of cell
morphology and fluorescence by optical means. Further, we have used the AFM as a
precise micro-manipulator to apply forces and deliver chemical signals to cells and have
followed downstream phenotypic and signal transduction events with conventional
microscopy. We suggest that this latter feature, for example, will have wide application to
the analysis in real time of receptor-ligand interactions or enable the investigation of
mechanically operated receptors and channels in live cells. Details of the experimental set-
up are given with examples of the different analytical techniques.
2 INTRODUCTION
A large number of techniques have been developed to measure the chemical interactions
upon which physiological phenomena in living organisms are based. Many of the most
important molecular events occur at cell swrfaces and several methods, of which atomic
force microscopy (AFM) is an example (Table l), have been developed specifically to
interrogate surfaces. Interactions at cell membranes are inexorably linked to the
intracellular processes, such as activation of signal transduction cascades or modification
to the structure of the cellular cytoskeleton, that regulate, for example, tissue
differentiation and function or cell survival. These latter phenomena have been already
examined by traditional optical or electron-based microscopies and the development of
methods that integrate approaches, such as scanning near-field optical microscopy
(SNOM), scanning ion conductance microscopy (SICM), and multi-photon and
fluorescence resonant energy transfer (FRET) methods, are now being applied to cell
physiological processes. Cell surfaces also have been investigated by various micro-
manipulative techniques, such as the surface force apparatus', pipette laser
tweezers or with magnetically trapped All of these are based either on the use of a
physical, and often hctionalised, probe. A classical example of such an approach is the
use of laser tweezers in the study of the molecular basis for muscle contraction where
myosin molecules move on actin filaments7. Many of these complementary methods are
summarised in Table 1. In this review we will focus on AFM as a specific type of non-
optical scanned probe microscopy that is developing into a method suitable for cell
biology.
Table 1 Techniquesfor cellular imaging at high resolution
Techniaue
Light microscopy Standard wide-field methods, confocal (laser
scanning) microscopy, evanascent wave microscopy,
proximity techniques such as FRET
Electron and X-ray Transmission, scanning, environmental (unfixed,
microscopy FESEM); immunological contrast and specificity
(gold, enzyme methods)
Scanning probe Atomic Force Microscopy (AFM): including
microscopy topography, chemical imaging, molecular recognition
(antibody on tip) microscopy, material property
analysis, use as micro-manipulator
High resolution AFM: topography down to molecular

scale, measurement of intermolecular interaction
forces and molecular mechanics
Scanning near-field optical microscopy (SNOM):

optical and fluorescence imaging below the visible
light wave limit
Scanning ion conductance microscopy (SICM): ion

conductance feed-back for functional imaging and
mapping of channels in cells
The AFM was devised by Binnig and colleagues in 19868and frst applied to the life
sciences by Marti and co-workers'. AFM, in its present form, can be used to acquire
images of intact cells at high, sub-micrometre scale, resolution and for topographic maps to
be generated. Measurement of an increasing variety of cellular events under physiological
conditions have become possible and Table 2 presents a summary of some of these
applications to cell biology that have been reported in the literature. Intriguingly, the
ability to apply forces via the AFM probe has guided its application increasingly towards
use as a precise micro-manipulator rather than an imaging tool. Indeed, pure imaging
applications have been limited as image capture rate is disappointingly slow and limited in
comparison to optical and scanning electron
Table 2 The use of AFM in cell biology
Cellular property studied References
Cell morphology, membrane structure, 10-32

hctional changes, etc.
Cytoskeleton and cytoskeletal elements 12,14,19,33-36
Cell division 37-39
Cell migration and locomotion 40-44
Organelles and organelle movement 35,41,45,46
Secretory structures 3 5,47,48
Viral budding, infected cells 49-52
Cell-matrix and cell-cell binding forces 53-55
Receptor-ligand binding forces on cells 54
Micromanipulation, microdissection 12,56,57
Elastic properties of cells 14,5843
plasma membrane, nucleus, 34,36,43,47,60
cytoskeletal fibres etc
and after cytoskeletal disruption 36,63,64
vinculin-deficient cells 65
microscopy. However, and importantly, AFM can be performed on living samples with the
application of appropriate techniques and equipment design.
2.1 Principles of AFM
In AFM, a micro-fabricated cantilever (usually 100-200 nm long) with a very small tip
(with a few nm2 contact area) is raster scanned on the surface (for example, of a cell)
whose topography is to be measured (Figure 1). A piezo-electric ceramic driver is used to
raise or lower the cantilever in order to maintain a constant bending of the cantilever. A
laser beam is reflected via the top of the cantilever towards a photo-detector that detects
any movement or bending of the cantilever, thus enabling the position of the cantilever to
be calculated and adjusted. As a result, the AFM records images of surface topography
under a constant applied force in the low nN range (Figure 2).
Many different imaging modes have been developed (reviewed by Hansma et al.66).
Contact mode imaging consists of scanning the whole surface while maintaining the tip in
constant contact. In contact mode, it is also possible to measure the torque applied on the
cantilever as it is scanned sideways over the surface. Differences in torque e uate to
differences in the friction of the surface and variation in material properties6’.’. Non-
contact, or ‘tapping’, mode consists of rapidly oscillating (mechanically or magnetically)
the cantilever in the z direction whilst slowly lowering it towards the surface and scanning.
When the cantilever comes close to the surface, the amplitude of the oscillations is
dampened and the surface can be detected. This mode is much less damaging to samples”.
Finally, in force-distance measurements, the cantilever is slowly lowered towards the
surface and its deflection is constantly recorded. This yields a curve showing the
Biophysical Chemistry: Membranes mid Prorrim
Figure 1 Equipment setup of the bio-AFM

(LHS),A schematic illustration of the AFM operation. The sample being
analysed is a cell adhering to a glass coverslip and loaded with the calcium
sensitive dye Fho-3 (see Figure 5). The tip of the AFM cantilever
approaches and scans the surface of the cell (Figure Z), its position being
controlled and measured using a light lever and photodetector. The tip can
also be used to interact with the cell surface, acting as a micromanipulator
to interrogate cell surface receptors (see Figure 4) or to deliver a
mechanical force (Figure 5). Fluorescence is measured from below via an
integrated confocal microscope. (RHS), The hybrid AFM-confocal
microscope is illustrated. The AFM is fitted onto a specially designed
inverted microscope inteflace allowing the t@ to be aligned into a desired
position independent of the movement of the microscope stage and the
sample holder. This configuration allows simultaneous light, phase contrast
and video capture, and epijluoresecence confocal imaging. The equi ment
used was a Thermomicroscopes AFM with modijications’$and a
BioRad Radiance confocal microscope. (Adapted from (Lehenkari et. al.,
2000).
bending of the cantilever as a function of the distance traveled at a single point rather than
under scanning conditions (see Figure 3). Using the spring constant of the cantilever, it is
then possible to calculate the force needed to create that deflection, forming the basis for
measurement of material properties of cells and of ligand-receptor binding forces. If such
measurements are made across a surface, then a 3D map of binding events can be
developed (Figure 4).
A .%.. B
Figure 2 Topography on live cells

(A), An exemplary three-dimensional surface image of the osteoclast
cultured on a glass coverslip; this is a large cell, with dimensions exceeding
100x100x8p. (B), A live osteoblast scanned in contact mode using
constant force. As the tip of the cantilever is much stirer than the cell, it
deforms the cell membrane revealing intracellular cytoskeletal structures.
3 APPLICATIONS OF AFM
3.1 Topography measurements
AFM was initially designed for the study of surface topography, and its lateral resolution is
on the atomic level when it is used in contact mode on surfaces that are harder than the
probe'. By the same principle, AFM can also be used to create a height map at lower
resolution on soft samples, such as cells, this being derived by measuring the cantilever
position during scanning across the surface. From this map, it is possible to construct a
three-dimensional profile of a cell surface (Fig. 2A). As the cell surface is softer than the
tip, resolution is reduced due to tip indentation along the probed surface6'. Indeed, on
materials as soft as living cells, the resolution can be as low as on the order of 200 nm.In
comparison, the resolution of confocal laser scanning microscopy is limited by the
wavelength of the light used and is significantly lower than that of scanning electron
microscopy. However, as resolution is proportional to the deformation caused by the force
used for imaging, contact imaging with higher forces can reveal relatively 'hard'
intracellular structures in soft, unfixed cells (Fig. 2B). Operation of the AFM in non-
contact mode can enhance the resolution where it can reach the order of a few tens of
nanometers7’. Non-contact imaging is a very demanding mode to operate in liquids,
particularly on living cells. Some topographic images of cellular plasma and nuclear
membranes have, though, been published that reveal interesting sub-cellular details that
would not be resolved by standard optical imaging (for example, 32); it has though not
reached molecular resolution on native plasma membranes of eukaryotic cells. With
advances in equipment and software design, the ability to routinely image, at high
resolution, live, underivatised cellular material by AFM may yet be realised.
Pure imaging by AFM can be used to reveal details of membrane and sub-surface
biological processes in real time and under ambient conditions. Early examples included
vesicle trafficking, and cytoskeletal More recently the range of
physiological responses that have been analysed by AFM have begun to rival those
achieved by optical microscopy, and a number of these are given in Table 2. La1 et all9
examined neurite outgrowth fiom cultured PC 12 cells, Schneider and colleagues26930 used
the atomic force microscope to measure aldosterone-induced volume increase in live
endothelial cells, Quist et alM examined the role of cellular gap junctions by AFM, and
Spudich and Braunstain3’ monitored basophil degranulation. Similar1 the group of
Radmacher have examined physiological processes such as cell division3’: cardiomyocyte
beatb~g~~and the control of cellular material proper tie^^^.
3.2 Binding force measurements
Since its invention, AFM has proven its worth in the quantification of binding forces for
interactions, for example, between proteins and other macromolecules, receptors and
ligands, or cells and substrates (Table 3). Intermolecular receptor-ligand binding forces in
the range 15-250 pN have been reported for protein interactions, and up to 220 nN for total
cellular binding forces. Using the AFM,it is possible to evaluate the binding force between
a single receptor molecule and its ligand directly71AFM has also been used extensively in
‘molecular stretching’ experiments (see 72-74) with a wide range of linear biomolecules,
such as the giant muscle protein titin, DNA and polysaccharides, and high resolution
molecular imaging in purple membranes and artificial lipid bilayers (see 75-78).
In order to achieve such a measurement, the tip of the AFM is ‘fbnctionalised’ with a
specific molecule. Different techniques have been developed to achieve this. One relies on
molecules being randomly picked up from a surface prior to molecular pulling. Linkage of
molecules to the AFM tip utilise either passive chemisorption of the molecule in solution
onto the tip, often with polyethylene glycol (PEG) as a linker/adsorbents4, or covalent
coupling using various chemistries to directly link the molecule to the Force-
distance curves are then recorded in different locations until an adhesion event occurs,
being recognised by the shape of the force distance curve. The aspect and interpretation of
a typical force-distance measurement are presented in Figure 3.
Table 3 Inter-molecularinteractionforces measured by AFM
Ligand-Receptor Force (pN) Reference
Isolated molecules: disruptionforces

Meromyosin-act in 15-25 7
Avidin-biotin 160-200 3,8 1,82
Streptavidin-biotin 200-257 8 1,83
Cell adhesion proteoglycans 40-1 25 84,85
Antibody-antigen 49-244 86-88
P-selectin-PGSL-1 165 89
VE cadherin 35-55 90
Cholera toxin-ganglioside 90 91
Lectin-carbohydrate 35-65 92
Targeting proteins (aSNAP-NSF ; 93,94
VAMP 1-syntaxin)
Isolated molecules: adhesionforce
mapping
Antibody-lysozyme 95
Antibody-ICAM1 96
Enzymes on tips: substrate detection 97- 100

(ATP, glucose, acetyl cholinesterase)
Lectin-cell carbohydrate interaction

Con A-yeast 75-200 101
Lectin-epithelium, fibroblast 50 102,103
Receptors in cells:forces
RGD-integrin 35-120 54,104
Antibody-cell 40-127 54,105
Receptors in cells: mapping

RGD-integrin Lehenkari et al, in prep.
VIP receptor 106,107
Blood group A on red cells 108
Cell-to-cellhrfaceforces
Dictyostelium on tip/csA adhesin single 23 109
molecule interaction
Yeast on tip 110
Bacteria on tip 111,112
Trophoblast-uterine epithe1ium 1000-16000 53
Cell-uncoated surface 19000- 100000 55
Cell-coated surface 100000-220000 55
Figure 3 The jcorce-distance curve

A representation of a force-distance curve between a tip-bound ligand and a
surface-bound receptor molecule is illustrated. The AFM tip approaches the
surface (1) and contact is made. The cantilever bends (2) until it reaches the
speciJiedforce limit which is to be applied and is then withdrawn (3). The
tip then leaves contact with the surface and, as the ligand on the tip remains
bound to surface receptor molecule, both are extended. With further
application of retractionforce, the molecule-ligand complex dissociates (4).
The cantilever then returns into its resting position, (I), and is ready for
another measurement. The maximum diflerence between the approach and
retraction curves, and the curve shape in the region of (4), gives
information on the interaction ‘binding’force between the tip and surface
and their physical properties.
To enable these binding force measurements to be performed, the ligand can either be
adsorbed to atomically smooth, cleaved mica or be present as a natural component of the
cell membrane. Measurements on living cells are inherently difficult: the target molecule is
easily lost fiom the cell membrane, can be at too low a density to make a successful
interaction likely, or the tip may become contaminated by cellular debris during analysis.
Recently, Lehenkari and HortonS4have reported the first measurements of binding forces
of an Arg-Gly-Asp (RGD) peptide to its cognate receptor, integrh, on living cells (Figure
4).
A C
4.6
j -7.9
-11.1
Figure 4 Integrin receptors on osteoclasts analysed by AFM

(A, B) Measuring interactionforces between ligands and cell surface receptors
by AFM. Interactionforces were evaluated between F11 antibody molecules on
the AFM cantilever tip and surface-expressed a& integrin dimers on jkeshly
isolated rat osteoclasts. (A), Force-distance analysis revealed a typical single
interaction (the yump' in the AFM retraction curve seen is marked). (B),
Individual release forces were plotted as a histogram and accumulate around
certain values; these can be further analysed by multi peak Gaussian curve
jtting, as described in j4. The results (B) showed multiples of a particular
bindingforce between F l I antibody and osteoclast a& of 127 f 16 pN (mean
SO). (A, B adapted j-om Lehenkari and Horton, I 999s4,with permission}.
(C)Mapping of integrin receptor distribution with RGD peptide in living cells.
A height map (Cl, z-range 0 - 8 p ) was created using a 2Opm 'glass ball'
cantilever tQ producing a low resolution topogra hic image of a fieshiy
8
isolated rat osteoclast (as in Lehenkari et al., 200 ). The interactionforces
between a GRGDSPpeptide, coated onto the AFM tip (Lehenkari and Horton,
199@4),and cwp3 integrin receptors were analysed by taking repeated force-
distance measurements across the entire surface of the cell. The distribution of
the RGD-integrin receptor bindingforces on the cell was thus evaluated (C2,
force range 0-5nN). A negative control GRGESP peptide failed to show
binding (not shown). The topography and force map were then merged into an
image of the binding forces displayed on a pseudo-3D height image of a cell
(C3). The RGD binding sites were located at the edge of the cell (arrow) and
on the top of the osteoclast (asterisk}.
3.2.I Integrin-RGD recognition - an example of intermolecular force measurement

in cells by AFM.
Members of the integrin family of cell adhesion receptors play a key role in cell-matrix
re~ognition"~. They are found in virtually all cell types as transmembrane receptors that
consist of two non-covalently linked subunits that require divalent cations for optimal
function. In cells adhering to extracellular matrix substrates, integrins are found
concentrated at the underside of cells in focal adhesion plaques associated with
intracellular linker proteins such as vinculin, the F-actin cytoskeleton and signalling
molecules. Many integrins recognise the Arg-Gly-Asp (RGD) peptide consensus sequence.
Ligands containing this sequence include several cell surface and extracellular matrix
proteins. The affinity of the receptor for a given ligand depends not only on the integrin
type but aIso on conformational changes in the receptor and hence receptor activation
status. These changes are triggered by alterations in the extracellular microenvironment of
the receptor and by interactions between the cytoplasmic part of the receptor and
intracellular signalling molecules and the cytoskeleton. The properties of integrins suggest
that, to evaluate their affinity for ligands reliably, this must be performed on the surface of
intact cells54.It is likely that this is true for many other systems of cell surface membrane
receptors.
RGD binding forces, measured by AFM in live bone cells, osteoclasts and osteoblasts,
is dependent not only on the type of integrin expressed by a cell and its activation status
but also on the spatial conformation of the sequence in which the RGD sequence is
present54This may be because a more favorable orientation between the ligand and the
receptor results in shorter interaction ranges for the molecules and thus higher interaction
forces between RGD and the binding site. Surprisingly the dynamic range of RGD binding
forces is small compared to affinity measurements, based upon inhibition of receptor-
ligand interaction, such as I& estimates. Here, the efficacy of a linear RGD peptide and a
potent snake venom inhibitor differs by approximately a factor of lo5,whereas the binding
forces measured by AFM vary by about 2-f0ld~~. The relationship between these two
values remains to be clarified and it is probable that hndarnental differences exist between
the parameters measured.
AFM measures binding force and binding probability separately, whereas 'affinity'
measures a combination of these86and, thus, AFM can provide M h e r infomation on the
nature of molecular interactions. The recognition events observed in a force-distance curve
do not represent unitary constants, as unbinding forces depend upon the rate of dissociation
of the two species37893' 14. Ligand-receptor dissociation involves the rupture of multiple
bonds and transitions between intermediates of varying stability and energie~"~.By
measuring unbinding forces across a range of loading rates, information on the non-
covalent bond strengths that mediate intermolecular interactions, the 'energy landscape',
can be obtained'16. Meanwhile at a practical level, the pull-out speeds and cantilever
properties must be carefully taken into account when 'absolute' force values are cited.
3.3 Measuring the material properties of a cell
The intracellular cytoskeleton gives the cell its physical integrity and adapts to the
environment and the activity of the cell. However, the exact role of each element of the
cytoskeleton is still unclear. By measuring the material properties of the cell - such as
stiffness, plasticity, visco-elasticity - and determining the effect of induced cytoskeletal
changes on these, one can gain insight into the particular role of each cytoskeletal element.
If the slopes of force-distance curves recorded during indention with an AFM probe are
analysed, the elastic modulus, that is stifhess, of the cell can be determined at a particular
location by use of the theory of indentati~n”~ (see Table 2). Thus, AFM can be used to
determine the cellular elasticity of tissues (reviewed by Radmacher6’)and this information
can be accumulated during raster scanning across a cell to generate 3D maps of elastic
properties and height information.
Common materials such as steel or bone have stifiesses of 200 GPa and 10 Gpa,
respectively, whereas living cells are more compliant with stiffnesses between 1- 150 kPa
(see Table 2 for references). It appears that the different elements and regions of a cell
have very different stiffnesses. The effect of cytoskeleton-disrupting drugs on the cell
stiflbess has been studied e ~ t e n s i v e l y ~ ~ * ~ 2). In particular, it was found that
(Table
disrupting the F-actin network with cytochalasin reduced the cell stiffness, whereas
chemical cross-linking increased cell stiffness. Together, these studies show that F-actin
filaments participate in the maintenance of cellular elasticity. The F-actin cytoskeleton is
linked to cell surface integrin receptors by vinculin in focal adhesion plaques. Recently,
Goldmann et a165 reported that the elastic modulus of vinculin-deficient cells was
decreased compared to wild type; when vinculin expression was returned by gene transfer,
a near wild-type elastic modulus value was attained. This points to the role of vinculin in
stabilising focal adhesions and transferring mechanical stresses to the cytoskeletal network.
’*.
Cellular cytoskeletal mechanics has been extensively reviewed’ Other material
properties such as the visco-elasticity and the plasticity of cells have also been measured.
Visco-elastici can be measured by indenting the cell and following its recovery over a
period of tirn$. Plasticity measurements consist of measuring the permanent deformations
inflicted on the cell through indention. These fundamental measurements are essential if
the response of cells to mechanical perturbation is to be understood at any level other than
the purely phenomenological. Many types of cells and tissues, for example those of the
skeleton or cardiovascular systems, are responsive to states of mechanical strain. Probing
the molecular and physiological details of their control with be important for a full
understanding of the pathologies that affect these tissues (see section 3.4.1).
3.4 AFM as a micro-manipulator
Several papers have reported experiments that fall under neither of the previous categories.
These use the AFM as a micro-manipulator or a micro-detector. Domke et a14*used the
AFM to map the mechanical pulse of cultured cardiomyocytes. Thie et alS3examined the
adhesive forces between a trophoblast and uterine epithelium using whole cells to
fhctionalise the tips instead of isolated molecules. The adhesion forces recorded between
cells were around 3nN, which is an order of magnitude higher than the molecule-ligand
adhesion forces. Recently, Sagvolden et a t 5 developed a new use of AFM to quantify the
adhesion forces of cells to a substrate. This is of particular interest to the fields of tissue
engineering and implant biomaterial design, where knowledge of cellular adhesive
interactions will be fimdamental to the development of biocompatible orthopaedic and
vascular grafts.
3.4.I Mechanical stimulation of live osteoblasts.
Virtually all cell types have been reported to sense mechanical stimuli. Amongst these,
bone cells are particularly interesting as they govern the adaptation of bone structure in
response to mechanical usage, We have used AFM as an indentor to mechanically
stimulate bone cells'1g (Figure 5). Although there are many other techniques of applying
mechanical strain to single cells, AFM provides an unique combination of positional
precision, control of the force applied, innocuity to cells, and the ability to estimate cellular
elasticity and strain. The latter, combined with monitoring of intracellular calcium
reactions by linked optical methods (Figure l), enables the threshold strain needed to elicit
intracellular calcium responses in cells to be determined. Examination of the strain
distribution in cells allows one to speculate on possible mechanisms of strain detection.
Cells can sense mechanical stimuli in a number of ways: strain can be detected
through stretch-activated ion channels, integrin trans-membrane receptors that link the
extracellular matrix to the cell cytoskeleton or receptor type tyrosine kinases'20''22.The
exact mechanism involved and the downstream events that may be influenced by the type
of mechanical stimulation are yet to be fully evaluated. Cells respond to mechanical stimuli
by a rise in intracellular calcium concentration. Both the amplitude and the frequency of
calcium transients modulate a number of downstream events such as gene expre~sion'~~.
lntracellular Calcium Intensity
c 1201 IB
20 40 60 80 100 120
Time (in s)
Figure 5 Intracellular calcium responses in osteoblasts to strain applied via an AFM

equipped with a $$ass ball' tip cantilever
Rat osteoblasts were loaded with Fho-3 prior to indentation as shown in
Figure 1. (A), The cell about to be indented is indicated with an arrow. (B),
The osteoblast after indentation has an increased intracellular calcium
concentration. (C), Time course of the calcium intensity within the indented
cell is shown. The times that correspond to thejuorescence images (A and
B) are identijed. TD indicates the time when the AFMcantilever enters in
contact with the cell and LO indicates the time when the AFMcantilever is
lifted out of contact of the cell surface. (Adapted from Charras et al,
2001 "'with permission).
Calcium can either originate fiom the extra-cellular environment or fiom intracellular
stores and is intricately controlled (for a review, see 124). Furthermore, having sensed a
mechanical stimulus, the cell can transmit this information to neighbouring to
sensitise a larger area of tissue and give rise to a coherent tissue response.
The precision of prediction of cellular strain using AFM has been estimated and a
good correlation between the predicted values for the radius of indentation and the
experimentally measured ones was found"'. Although part of the overall error is imputable
to errors in optical measurement, it still remains necessary to find a more realistic model
than the current linear elastic isotropic models in order to describe living cells (for a
review, see Radmacher61). Several limitations can be pointed out. Firstly, the cell is a
complex structure, and it is best described as a fluid bilayer that is non-uniformly tethered
to an intricate network of interconnected struts (the intracellular microtubular network) and
cables (the F-actin cytoskeleton and intermediate filaments)'". Secondly, living cells
exhibit viscoelastic and plastic properties63and can actively adapt their cytoskeleton after
long term exposure to mechanical strain12*.
The strain distributions determined throughout the cell thickness using finite element
modelling have enabled a number of hypotheses regarding the way in which cells may
sense applied strains to be formulated"'. Maximal and minimal radial strains were situated
on the cell surface, suggesting a detection mediated by components in the cell membrane,
such as mechano-sensitive cation channels that open in response to membrane strain and
let extracellular cations into the cell (I2' for a review). As there is a high strain gradient at
the interface between the indented and un-indented regions, the strain would vary greatly
in amplitude over a short distance, possibly stretching the membrane sufficiently to open
mechano-sensitive cation channels. The vertical strain distribution shows a large
compressive strain just under the area of indentation. This sub-membranous area is a prime
area for connection between the cell membrane and the cellular cytoskeleton. Hence, the
signal may be detected by a direct deformation of cellular cytoskeleton, which may in turn
activate mechanisms that increase intracellular calcium levels, for example
phosphoinositides, ryanodine receptor and chloride channel mediated response^'^^.
AFM, via its use both as an engineering tool and a micro-manipulator, shows geat
promise in helping our understanding of mechano-transduction pathways. A nagging
-
question, though, comes to mind do all cells react to similar applied mechanical strains?
Through its capacity to quantify biological phenomena in engineering terms, AFM may
bring biology into the era of solution engineering and enable such a question to be
answered.
4 CONCLUSIONS: OUTSTANDING RESEARCH QUESTIONS
The use of AFM in biology in the fbture might, intriguingly, be based not so much on its
atomic scale imaging capabilities, but rather on the opportunity it offers to apply and
measure forces between biological entities, such as molecules, cells or biomaterials. As a
consequence, AFM is being used increasingly as a precise micro-manipulator rather than
an imaging tool. Interesting work using this capability has focused not only on quantiQing
inter-molecular, but also htra-molecular, bond strengths. For example, unfolding forces for
large linear molecules have been measured and m ~ d e l l e d ~ As
~ ~ equipment
'~. improves and
the range of different commercially available systems widens, it is predicted that
measurements of molecular interaction forces by AFM will become a routine procedure for
many cell biologists.
Of the many applications that we have described, chemical force or molecular

recognition microscopy has the potential to become used industrially in the evaluation
process of candidate pharmaceuticals as an adjunct to standard pharmacological tools.
Likewise, affinity mapping of receptor distribution is particularly useful for examining the
location of receptors to which antibodies may not exist or form the large group of ‘orphan
receptors’ in the human genome, as well as for evaluating the functionality of receptors
present within the cell membrane; with development, an AFM-based method could replace
the current gold standard technique of receptor autoradiography.
Adhesion measurement of whole cells, as described by Sagvolden et a15’ or Thie et
a153,may enable new materials to be characterised for implantation in patients. Thus, new
orthopaedic implants could be evaluated to select materials for their specific purpose so
that, for example, they would promote adhesion of osteoblasts over other cell types and
promote ideal integration into the body.
Cellular bio-mechanics has been hindered by the lack of a precise tool to enable
hypotheses to be experimentally verified. AFM may help us understand how cells react to
strain, by which mechanisms they adapt to life in strained environments, or how
mechanical and signalling athways interact at the cellular level, as has been hypothesised
(for example, percolationRg and ten~egrity’~~). Furthermore, in conjunction with fmite
element modelling, AFM may help to answer some of the more intriguing questions posed
by biology. For example, how do erythrocytes manage to pass through capillaries whose
diameter is smaller than their own?
Currently, there is also a lack of suitable methods to analyse the 3D structure of
membrane glycoproteins at high resolution in their native context and configuration. The
pioneering work of Muller and colleagues77378 used proteins of bacterial purple membranes,
which are naturally present as 2D crystals, such as bacteriorhodopsin and Ompf, or
eukaryotic membranes naturally enriched in specific rotei ins^^'^^. This makes equivalent
methods for molecules present in the membranes of eukaryotic cells particularly attractive,
especially if high resolution, ‘soft’ imaging techniques can be developed. Here, membrane
glycoproteins are typically present at much lower densities and below levels that would be
expected to form crystalloid features. By refming such experiments on eukaryotic cells,
one may be able to gain a definitive insight into the structure and fhction of, for example,
ion channels or receptors in their native membrane macro-molecular complexes.
Finally, in order to realise its full potential in cell biology, a number of design and
technical problems still need to be optimised or solved and made available in a commercial
setting. The AFM has to be interfaced to an optical microscope to be able to choose the cell
to be examined, maintained in a near physiological condition which must be easily
changeable during imaging. As certain cell types are very tall and certain substrates very
uneven, the z-range of the AFM has to be extended beyond the range that is generally
available commercially. Phase-contrast and fluorescence imaging of cells examined using
AFM during or post-experimentation must be possible and of high optical quality and
resolution. For chemical force AFM,a robust, easy and reliable way of hctionalising tips
needs to be devised and these need to be optimised for cellular imaging. In order to reliably
examine biological phenomena in real time, an increased scanning speed applying less
force would be desirablei3’. In order to realise the full potential of integration with the
whole range of biological examination techniques, the AFM apparatus must leave easy
access to the sample to enable simultaneous AFM and micro-manipulation, micro-injection
or electrophysiology. Finally, serious attention must be paid to ease of use, both in terms of
the hard- and soft-ware, for AFM to gain as wide acceptance in the biological community
as the confocal microscope.
Acknowledgements
This work was supported by a Programme Grant to M.A.H. from The Wellcome Trust.
P.P.L. was in receipt of an EMBO fellowship and G.T.C. was fbnded by a Johnson and
Johnson COSAT award.
References
1 D.E. Leckband, J.N. Israelachvili, F.J. Schmitt and W. Knoll, Science, 1992, 255,
1419.
2 E. Evans, D. Berk, A. Leung and N. Mohandas, Biophys. J., 1991 59,849.
3 R. Merkel, P. Nassoy, A. Leung, K. Ritchie and E. Evans, Nature, 1999,397,50.
4 K. Svoboda ,C.F. Schmidt, B.J. Schnapp and S.M. Block, Nature, 1993,365,721.
5 E.L. Florin, A. Pralle, J.K. Horber and E.H. Stelzer, J. Struct Biol., 1997,119,202.
6 M. Glogauer, P. Arora, G. Yao, I. Sokholov, J. Ferrier and C.A. McCulloch, J. Cell
Sci., 1997,110, 11.
7 H. Nakajima, Y. Kunioka, K. Nakano, K. Shimizu, M. Set0 and T. Ando, Biochem.
Biophys. Res. Commun., 1997,234, 178.
8 G. Binnig, C.F. Quate and C. Gerber, Phys. Rev. Lett., 1986,56,930.
9 0. Marti, V. Elings, M. Haugan, C.E. Bracker, J. Schneir, B. Drake, S.A. Gould, J.
Gurley, L. Hellemans, K. Shaw, et. al., J Microsc., 1988,152,803.
10 J.K. Horber, W. Haberle, F. Ohnesorge, G. Binnig, H.G. Liebich, C.P. Czerny, H.
Mahnel and A. Mayr, Scanning Microsc., 1992,6,919.
11 J.H. Hoh, G.E. Sosinsky, J.P. Revel and P.K. Hansma, Biophys J., 1993,65,149.
12 V. Parpura, P.G. Haydon and E. Henderson, J. Cell Sci., 1993,104,427.
13 M. Beckmann, H.A. Kolb and F. Lang, J. Membr. Biol., 1994,140,197.
14 J.H. Hoh and C.A. Schoenenberger,J Cell Sci., 1994,107,1105.
15 C. Le Grimellec, E. Lesniewska, C. Cachia, J.P. Schreiber, F. de Fornel and J.P.
Goudonnet, Biophys. J., 1994,67,36.
16 H. Oberleithner, E. Brinckmann, A. Schwab and G. Krohne, Proc. Natl. Acad. Sci.,
USA,1994,91,9784.
17 K.A. Barbee, Biochem. Cell Biol., 1995,73,501.
18 R. Lal, S.A. John, D.W. Laird and M.F. Arnsdorf, Am. J. Physiol., 1995,268, C968.
19 R. Lal, B. Drake, D. Blumberg, D.R. Saner, P.K.Hansma and S.C. Feinstein, Am. J.
Physiol., 1995,269, C275.
20 C. Le Grimellec, E. Lesniewska, M.C. Giocondi, C. Cachia, J.P. Schreiber and J.P.
Goudonnet, Scanning Microsc., 1995,9,401.
21 D.J. Muller, F.A. Schabert, G. Buldt and A. Engel, Biophys. J.,1995,68, 1681.
22 A. Schwab, K. Gabriel, F. Finsterwalder, G. Folprecht, R. Greger, A. Kramer and H.
Oberleithner, Pflugers Arch., 1995,430,802.
23 F. Braet, R. De Zanger, W. Kallem, A. Raap, H. Tanke and E. Wisse, Scanning
Microsc. Suppl., 1996,10,225.
24 T. Danker, M. Mazzanti, R. Tonini, A. Rakowska and H. Oberleithner, Cell Bid. Int.,
1997, 21, 747.
25 U. Ehrenhofer, A. Rakowska, S.W. Schneider, A. Schwab and H. Oberleithner, CeZZ
Biol. Int., 1997,21,737.
26 S.W. Schneider, Y. Yano, B.E. Sumpio, B.P. Jena, J.P. Geibel, M. Gekle and H.
Oberleithner, Cell Biol. Int., 1997,21,759.
27 R.A. Garcia, D.E. Laney, S.M. Parsons and H.G. Hansma, J. Neurosci. Res., 1998,52,
350.
28 C . Le Grimellec, E. Lesniewska, M.C. Giocondi, E. Finot, V. Vie and J.P. Goudonnet,
Biophys. J., 1998,75,695.
29 A. Rakowska, T. Danker, S.W. Schneider and H. Oberleithner, J. Membr. Biol., 1998,
163, 129.
30 S.W. Schneider, P. Pagel, J. Storck, Y. Yano, B.E. Sumpio, J.P. Geibel and H.
Oberleithner, Kidney Blood Press Res., 1998,21,256.
31 H. Schillers, T. Danker, H.J. Schnittler, F. Lang and H. Oberleithner, Cell Physiol.
Biochem., 2000,10,99.
32 V. Vie, M.C. Giocondi, E. Lesniewska, E. Finot, J.P. Goudonnet and C. Le irimellec,
Ultramicroscopy,2000,82,279.
33 E. Henderson, P.G. Haydon and D.S. Sakaguchi, Science, 1992,257,1944.
34 S.G. Shroff, D.R.Saner and R. Lal, Am. J. Physiol., 1995,269, C286.
35 A. Spudich and D. Braunstein, Proc, Natl. Acad. Sci., USA, 1995,92,6976.
36 U.G. Hofmann, C. Rotsch, W.J. Parak and M.J. Radmacher, Struct. Biol., 997, 119,
84.
37 Y.G. Kuznetsov, A.J. Malkin and A. McPherson, J. Struct. Biol., 1997, 120, 180.
38 J.A. Dvorak and E. Nagao, f i p . Cell Res., 1998,242,69.
39 R. Matzke, K. Jacobson and M. Radmacher, Nut. Cell Biol., 2001,3,607.
40 H. Oberleithner, G. Giebisch and J. Geibel, PJlugers Arch., 1993,425,506.
41 M. Fritz, M. Radmacher and H.E. Gaub, Biophys. J., 1994,66, 1328.
42 J. Domke, W.J. Parak, M. George, H.E. Gaub and M. Radmacher, Eur. Biophys. J.,
1999,28, 179.
43 C. Rotsch, K. Jacobson and M. Radmacher, Proc. Natl. Acad. Sci., USA, 1999, 96,
921.
44 S.W. Schneider, P. Pagel, C. Rotsch, T. Danker, H. Oberleithner, M. Radmacher and
A. Schwab, PJIugersArch., 2000,439,297.
45 B.P. Jena, S.W. Schneider, J.P. Geibel, P. Webster, H. Oberleithner and K.C.
Sritharan, Proc. Natl. Acad. Sci., USA, 1997,94, 13317.
46 A.P. Quist, S.K. Wee, H. Lin and R. Lal, J. Cell Biol., 2000,148, 1063.
47 V. Parpura and J.M. Fernandez, Biophys. J., 1996,71,2356.
48 V. Parpura, R.T. Doyle, T.A. Basarsky, E. Henderson and P.G. Haydon, Neuroimage,
1995,2,3.
49 W. Haberle, J.K. Horber, F. Ohnesorge, D.P. Smith and G. Binnig, Ultramicroscopy,
1992, 42-44, 1 161.
50 F.M. Ohnesorge, J.K. Horber, W. Haberle, C.P. Czerny, D.P. Smith and G. Binnig
Biophys. J., 1997,73,2183.
51 J.A. Dvorak, S. Kobayashi, K. Abe, T. Fujiwara, T. Takeuchi and E.J. Nagao,
Electron. Microsc., (Tokyo), 2000,49,429.
52 E. Nagao, 0. Kaneko and J.A. Dvorak, J. Struct. Biol., 2000,130,34.
53 M. Thie, R. Rospel, W. Dettmann, M. Benoit, M. Ludwig, H.E. Gaub and H.W.
Denker, Hum. Reprod. 1998,13,3211.
54 P.P. Lehenkari and M.A. Horton, Biochem. Biophys. Res. Commun., 1999,259,645.
55 G. Sagvolden, I. Giaever, E.O. Pettersen and J. Feder, Proc. Natl. Acad. Sci., USA,
1999,96,471.
56 J.H. Hoh, R. Lal, S.A. John, J.P. Revel and M.F. Arnsdorf, Science, 1991,253, 1405.
57 M.G. Langer, A. Koitschev, H. Haase, U. Rexhausen, J.K. Horber and J.P.
Ruppersberg, Ultramicroscopy, 2000,82,269.
58 M. Sato, K. Nagayama, N. Kataoka, M. Sasaki and K. Hane, J Biomech., 2000,33,

127.
59 A.L. Weisenhorn, M. Khorsandi, S. Kasas, V. Gotzos and H.J. Butt, Nanotechnology,
1993,4, 106.
60 M. Radmacher, M. Fritz, C..M. Kacher, J.P. Cleveland and P.K. Hansma, Biophys. J.,
1996,70,556.
61 M. Radmacher, IEEE Eng. Med. Biol. Mag., 1997,16,47.
62 E. A-Hassan, W.F. Heinz, M.D. Antonik, N.P. D'Costa, S. Nageswaran, C.A.
Schoenenberger and J.H. Hoh, Biophys. J., 1998,74, 1564.
63 H.W. Wu, T. Kuhn and V.T. Moy, Scanning, 1998,20,389.
64 C. Rotsch and M. Radmacher, Biophys. J.,2000,78,520.
65 W.H. Goldmann, R. Galneder, M. Ludwig, W. Xu, E.D. Adamson, N. Wang and R.M.
Ezzell, Exp. Cell Res., 1998,239,235.
66 H.G. Hansma, K.J. Kim,D.E. Laney, R.A. Garcia, M. Argaman, M.J. Allen and S.M.
Parsons, J. Struct. Biol., 1997,119,99.
67 G. Shang, X. Qiu, C. Wang and C. Bai, Appl. Phys. A., 1998,66, S333.
68 M. Benoit, T. Holstein and H.E. Gaub, Eur, Biophys. J., 1999,26,283.
69 P.P. Lehenkari, G.T. Charras, A. Nykben and M.A. Horton, UZtramicroscopy, 2000,
82,289.
70 C.A. Putman, K.O. van der Werf, B.G. de Grooth, N.F. van Hulst and J. Greve,
Biophys. J., 1994,67, 1749.
71 A. Chilkoti, T. Boland, B.D. Ratner and P.S. Stayton, Biophys. J., 1995,69,2125.
72 T.E. Fisher, P.E. Marszalek and J.M. Fernandez, Nut. Struct. Biol., 2000,7,719.
73 J. Zlatanova, S.M. Lindsay and S.H. Leuba, Prog. Biophys. Mol. Biol., 2000,74,37.
74 M. Carrion-Vazquez, A.F. Oberhauser, T.E. Fisher, P.E. Marszalek, H. Li and J.M.
Fernandez, Prog. Biophys. Mol. Biol., 2000,74,63.
75 T. Danker and H. Oberleithner, Pflugers Arch. 2000,439,671.
76 H. Schillers, T. Danker, M. Madeja and H.J. Oberleithner, Membr. Biol., 2001, 180,
205.
77 A. Engel and D.J. Muller, Nut. Struct. Biol., 2000,7,715.
78 D.J. Muller, J.B. Heymann, F. Oesterhelt, C. Moller, H. Gaub, G. Buldt, A. Engel,
Biochirn. Biophys. Acta., 2000,1460,27.
79 O.H. Willemsen, M.M. Snel, A. Cambi, J. Greve, B.G. De Grooth and C.G. Figdor,
Biophys. J., 2000,79,3267.
80 G.J. Schutz, M. Sonnleitner, P. Hinterdorfer and H. Schindler, Molec. Memb. Biol.,
2000,17, 17.
81 V.T. Moy, E.L. Florin and H.E. Gaub, Science, 1994,266,257.
82 E.L. Florin, V.T. Moy and H.E. Gaub, Science, 1994,264,415.
83 S.S. Wong, E. Joselevich, A.T. Woolley, C.L. Cheung and C.M. Lieber, Nature, 1998,
394,52.
84 U. Dammer, 0. Popescu, P. Wagner, D. Anselmetti, H.J. Guntherodt and G.N.
Misevic, Science, 1995,267, 1 173.
85 G.N. Misevic, Microsc. Res. Tech., 1999,44,304.
86 P. Hinterdorfer, W. Baumgartner, H.J. Gruber, K. Schilcher and H. Schindler, Proc.
Natl. Acad. Sci., USA, 1996,93,3477.
87 S. Allen, X. Chen, J. Davies, M.C. Davies, A.C. Dawkes, J.C. Edwards, C.J. Roberts,
J. Sefton, S.J. Tendler and P.M. Williams, Biochemistry, 1997,36, 7457.
88 U. Dammer, M. Hegner, D. Anselmetti, P. Wagner, M. Dreier, W. Huber and H.J.
Guntherodt, Biophys. J., 1996,70,2437.
89 J. Fritz, A.G. Katopodis, F. Kolbinger and D. Anselmetti, Proc. Natl. Acad. Sci. USA,
1998,95,12283.
90 W. Baumgartner, P. Hinterdorfer, W. Ness, A. Raab, D. Vestweber, H. Schindler and
D. Drenckhahn, Proc. Natl. Acad. Sci., USA, 2000,97,4005.
91 P.F. Luckham and K.Smith, Faraduy Discuss, 1998, (1 11), 307.
92 W. Dettmann, M. Grandbois, S. Andre, M. Benoit, A.K. Wehle, H. Kaltner, H.J.
Gabius and H.E. Gaub, Arch. Biochem. Biophys., 2000 383, 157.
93 K.C. Sritharan, A.S. Quinn, D.J. Taatjes and B.P. Jena, Cell Biol. Int., 1998,22,649.
94 D.J. Ellis, T. Berge, J.M. Edwardson and R.M. Henderson, Microsc. Res. Tech., 1999,
44,368.
95 A. Raab, W. Han, D. Badt, S.J. Smith-Gill, S.M. Lindsay, H. Schindler and P.
Hinterdorfer, Nat. Biotechnol., 1999, 17,901.
96 O.H. Willemsen, M.M. Snel, K.O. van der Werf, B.G. de Grooth, J. Greve, P.
Hinterdorfer, H.J. Gruber, H. Schindler, Y. van Kooyk and C.G. Figdor, Biophys.
J., 1998,75,2220.
97 S.W. Schneider, M.E. Egan, B.P. Jena, W.B. Guggino, H. Oberleithner and J.P.
Geibel, Proc. Natl. Acad. Sci., USA, 1999,96, 12180.
98 2. Yingge, Z. Delu, B. Chunli and W. Chen, Life Sci., 1999,65, PL253.
99 R. de Souza Pereira, FEBS Lett., 2000,475,43.
100 M. Fiorini, R. McKendry, M.A. Cooper, T. Rayment and C. Abell, Siophys J., 2001,
80,2471.
101 M. Gad, A. Itoh and A. Ikai, Cell Biol. Int., 1997,21,697.
102 T. Osada, S. Takezawa, A. Itoh, H. Arakawa, M. Ichikawa and A. Ikai, Chem. Senses,
1999,24, 1.
103 A. Chen and V.T. Moy, Biophys. J.,2000,78,2814.
104 N.B. Holland, C.A. Siedlecki and R.E. Marchant, J. Biomed. Mater. Res., 1999, 45,
167.
105 G. Kada, L. Blayney, L.H. Jeyakumar, F. Kienberger, V.P. Pastushenko, S. Fleischer,
H. Schindler, F.A. Lai and P. Hinterdorfer, Ultramicroscopy, 2001,86,129.
106 P. Lundberg, A. Lie, A. Bjurholm, P.P. Lehenkari, M.A. Horton, U.H. Lerner and M.
Ransjo, Bone, 2000,27,803.
107 P. Lundberg, I. Lundgren, H. Mukohyama, P.P. Lehenkari, M.A. Horton and U.H.
Lerner, Endocrinology, 2001, 142,339.
108 M. Grandbois, W. Dettmann, M. Benoit and H.E. Gaub, J. Histochem. Cytochem.,
2000,48,719.
109 M. Benoit, D. Gabriel, G. Gerisch and H.E. Gaub, Nat. Cell Biol., 2000,2,313.
110 W.R. Bowen, R.W. Lovitt, C.J. Wright, J. Colloid. Interface Sci., 2001,237,54.
1 1 1 A. Razatos, Y.L. Ong, M.M. Sharma and G. Georgiou, Proc. Natl. Acad. Sci., USA,
1998,95,11059.
1 12 S.K. Lower, M.F. Jr Hochella, T.J. Beveridge, Science, 2001,292, 1360.
113 C.M. Isacke and M.A. Horton, The Adhesion Molecule FactsBook, 2nd.Ed. Academic
Press, London, UK, 2000.
114 T. Strunz, K. Oroszlan, R. Schafer and H.J. Guntherodt, Proc. Natl. Acad. Sci., USA,
1999,96,11277.
115 B. Isralewitz, M. Gao and K. Schulten, Curr. Opin. Struct. Biol., 2001,11,224.
116 E. Evans, Annu. Rev. Biophys. Biomol. Struct., 2001,30, 105.
117 K.L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge, UK,
1985.
I I8 D.E. Ingber, Sci. Am., 1998,278,48.
119 G.T. Charras, P.P. Lehenkari and M.A. Horton, Ultramicroscopy, 2001,86,85.
120 A.J. Banes, M.Tsuzaki, J. Yamamoto, T. Fischer, B. Brigman, T. Brown and L.

Miller, Biochem. Cell Biol., 1995,73,349.
121 F. Sachs and C.E. Morris, Rev. Physiol. Biochem. Pharmacol., 1998,132,l.
122 J. Sadoshima, S. Izumo, Heart Vessels, 1997, Suppll2, 194.
123 M.J. Berridge, Neuron., 1998,21, 13.
124 M.J. Berridge, P. Lipp and M.D.Bootman,. Nat. Rev.Mol. Cell Biol., 2000,1, 1 1.
125 S.L. Xia and J. Ferrier, Biochem. Biophys. Res. Commun., 1992,186, 1212.
126 N.R. Jorgensen, 2. Henriksen, C. Brot, E.F. Eriksen, O.H. Sorensen, R.Civitelli and
T.H. Steinberg, J. Bone Miner.Res., 2000,15, 1024.
127 D.E. Ingber,. J. CelZSci., 1993,104,613.
128 P.R. Girard and R.M.Nerem, Front Med. Biol. Eng., 1993,5,31.
129 G. Forgacs, J. CeZl Sci., 1995,108,2131.
130 J. Tamayo, A.D. Humphris, R.J.Owen and M.J.Miles, Biophys. J., 2001,81,526.
PHYSICAL CHARACTERIZATION OF WILD TYPE AND mnn9 MUTANT CELLS
OF Saccharomyces cerevisiae BY ATOMIC FORCE MICROSCOPY (AFM)
A.Mendez-Vilas', I.Corbacho2,M.L.Gonzalez-Martin' and M.J.Nuevo'
'Department of Physics. University of Extremadura. Avda. Elvas s/n. 06071 Badajoz,

SPAIN. E-Mail: maria@unex.es
'Department of Microbiology. University of Extremadura. Avda. Elvas s/n. 06071
Badajoz, SPAIN.
1 INTRODUCTION
The study of glycosylation processes in eukaryotic cells is a very relevant field of Cell and
Molecular Biology. Glycosylation is the main covalent modification which proteins
secreted by eukaryotic membranes undergo. Glycosylation mechanisms are well preserved
in eukaryotic cells and they all follow the same fundamental scheme. The percentage of the
glucidic part in a glycoprotein can range from 2 to 90%. The carbohydrate contribution to
the proteic function is variable. Saccharomyces cerevisiae was used as a model, for
different reasons: (a) it is unicellular and can be easily grown in cheap culture media; (b) it
is a well known species; (c) it is easy to obtain glycosylation mutants; (d) it is surrounded
by a thick cell wall of about 200 nm, which represents 1530% of the overall weight. This
cell wall is composed of mannoproteins (40%), glucans and chitin. Mannoproteins are
glycoproteins whose glucidic portion is almost exclusively mannose. Glucan and chitin are
structural, and the mannoproteins are the filling. In this work, among the three types of
links in glycoproteins, the N-glycosidic links were those analyzed. A review of the
significance of N-glycosylation can be found on Ref 1.
In order to study the N-glycosidic remains structure, several kinds of mutants in
glycosylation processes were isolated, including the mnn mutants.*'. The difference
between the distinct kinds of mutants is the glycosylation process phase related to the
muted gen. For example, the mnnl mutant lacks 6 (1,3) linked terminal mannoses. The
mnn6 mutant presents a minor inclusion of mannosylphosphate groups, and the mnn9,
which is the most dramatically affected, lacks most of the external chain, and presents a
lower growth rate.
The morphology and physico-chemical properties (adhesion, mechanical properties,
etc.) of these cells, is a near-total-darkness field. The influence of the defects of the N-
glycosylation processes on the physical properties of the cell wall is of great interest, and
Atomic Force Microscopy offers new possibilities. AFM permits an extremely high lateral
resolution, without previous manipulation of the sample, unlike other imaging techniques
such as Scanning Electron Microscopy (SEM). Moreover, different physical properties can
be also measured by AFM, such as surface roughness, friction, adhesion force and
mechanical properties, with the same lateral resolution as topography.
AFM scans the surface of a sample with a sharp silicon or Si3N4 tip (a few pm long
and less than 0.01 pm in diameter) located at the free end of a cantilever (about 100 pm
1 Probing Biological Molecules: Theory and Experimerzr 51
long). Forces between the tip and the sample surface cause the cantilever to bend or
deflect. A detector measures the deflection of the cantilever and produces a map of the
surface topography of the sample scanned under the tip. AFM can be used to study
insulators or conductors in two general modes (contact or non-contact regimes), depending
on the way the tip interacts with the surface of the sample. In the AFM contact mode,
additional studies can be performed by lateral force microscopy (LFM). This technique
measures lateral deflections (twisting) of the cantilever that arise from forces on the
cantilever parallel to the plane of the sample surface5. It is useful for imaging variations in
surface fiction that can arise from inhomogeneities in the surface and also for
obtaining edge-enhanced images of any surface'. The LFM and AFM images are, in
general, collected simultaneously, so that changes in the composition of the sample surface
can be observed.
The effect of different kinds of mutations on the morphology and physical surface
properties of cells have been previously studied by some authors by Atomic Force
Microscopy. Ishijima et a t studied the ultrastructure of the cell wall of the cps8 actin
mutant cell in Schizosaccharomycespombe. This mutant was used to determine the role of
the actin cytoskeleton in cell wall formation. in addition to the studies about the
ultrastructure of cells, AFM can also be used to measure surface forces such as adhesion,
as well as mechanical surface properties. Differences in adhesion to biomaterials by E.co2i
K12 mutant (cells with truncated LPS molecules) and its parental strain D21 (which
synthesize the complete carbohydrate chain) were studied by Y.L.Ong et a2".
In order to analyze the influence of this mutation on the morphology of the cells
and their surface structure, contact and non-contact images were obtained on both wild and
mnn9 mutant cells of S.cerevisiae. Moreover, the surface ultrastructure of both surfaces
was characterized by measuring their so-called root mean square (rms) surface roughness.
2 METHOD AND RESULTS
2.1 Sample Preparation
Both wild type and mnn9 strains were grown in YEPD (Yeast Extract - Peptone -
Dextrose) solid medium, at 30 "C for 48 hours. A small portion of cells were then collected
using a toothpick and extended over a microscope slide with a drop of water and allowed
to dry at room temperature. The strains were kindly supplied by LM Hernhdez and I.
Oliver0 from the Department of Microbiology of the University of Extremadura.
2.2 Atomic Force Microscopy
All the experiments were carried out with an Autoprobe CP atomic force microscope (Park
Scientific Instruments, Geneva, Switzerland), equipped with a scanner of maximum ranges
of 100 pm and 7 pm in xy and z directions, respectively. This instrument has an optical
microscope for easy location the region of interest, by monitoring the sample on a TV
screen. The images were acquired using silicon nitride cantilevers with a nominal force
constant of 0.4 Nlm and a typical probe curvature radio of lOnm, supplied by the
manufacturer. The scanner speed ranged between 1 and 5 pm/s and the images were
acquired at 512x512 pixels. The maximum applied force was = 10 nN. The non-contact
images were taken at the cantilever resonance frequency of 1 12 kHz.
52 Biophysical Chemistty: Membranes and Proteins
2.3 Wild Type Cells
In order to study the cells morphology, contact (repulsive imaging regimen) and non-
contact (attractive regimen) images were taken. Fig. 1 (a) shows a contact mode image of a
set of wild type cells. As can be derived from these images, we have piles of cells with an
average diameter of around 4 pm. Fig.l(b) also shows a non-contact image, taken at a
slightly lower frequency than the resonance frequency of the cantilever (referred to as
Tapping Mode). As can be seen equivalent information is obtained from both imaging
modes, which indicates that in contact mode (the most aggressive imaging mode) the tip is
not damaging the cell surface. Both types of image show a very smooth cell surface,
without any noticeable characteristics.
Figure 1 Contact and non-contact images of wild type cells.
Friction images were also made in order to gain more insight into the physical
properties of these cell surfaces, in particular regarding surface composition homogeneity.
Friction can also be used in many cases for enhancing topographical features. Fig.2(a)-(b)
show a forward and backward friction image, respectively, of a wild type cell. No contrast
between the two images is appreciated, which seems to indicate the high composition
homogeneity of the cell surface, even at sub-micrometer scales. This coincides with what
was expected from the bibliography.
Figure 2 (a)Fonvard and @)backwardfriction images of wild type cells.
An example of the utility of friction image for enhancing topographical features can
be seen in Fig. 3 and 4, where (a) is a contact topographical image and (b)-(c) are the
corresponding forward and backward friction images, for two different samples,
respectively. The circular structures that appear in these images seem to be the chitin rings
formed in the mother cell after cell division (budding). As can be observed, these surface
structures are much better resolved in friction images than in topographical images. The
diameters of these rings are highly homogeneous and are about 1 pm, with a width of
about 200 run, which is in excellent agreement with previous results from other
microscopy techniques. In addition, it can be observed that adjacent chitin rings are present
in all the images, as would be expected from a haploid strain such as ours.
Figure 3 (a)Topographical contact image: (b)Fomlard friction image; Ic) Backward

friction image for wild type cells.
Figure 4 (a)Topographical contact image; (b)Fonuardfviction image; (c) Backward

friction image for wild type cells.
Another physical property which can be measured by AFM is surface roughness.

Although there are various parameters for characterizing surface roughness", the most
used is the rms roughness, defined as the standard deviation of height data over a chosen
area of the surface. It is important to note that this magnitude depends on the scale on
which it is measured. Fig.5 shows the rms surface roughness of wild type cells, measured
on several scales of the surface. As can be observed, nanometric roughness values are
obtained, Such a high resolution is possible due to the high vertical resolution of the AFM
photodetector (up to 1 A), This result indicates that AFM is a much more suitable
technique for roughness estimation and ultrastructure observation than any other
microscopy technique requiring a surface coverage for measurements, as for example in
SEM analysis.
Figure 5 Rms surface roughness scaling with the length of the scanned area for wild
type cells.
2.4 mnn9 Mutant Cells
Fig.6 shows a contact-mode AFM image of a set of mutant cells. As can be clearly seen,
there is evidence that the tip is scraping off the surface. It was impossible to obtain high
quality images in contact mode, even at low force imaging (<lo nN), which seems to
indicate the existence of regions with different (softer) surface properties from the wild
type one. This observation is in agreement with previous results, and shows that the mnn9
cell wall is more osmotically fragile than wild type cells'*.
Figure 6 Topographical contact image of mutant cells at an imaging force of IOnh?
In order to obtain a more detailed view of the surface structure of these mutant
cells, non-contact (Tapping mode) images were also made of them. Fig.7 shows some non-
contact images of the surface of mnn9 mutant cells. From the images the existence can be
observed of smooth regions which are very similar in appearance to the wild type surface,
as well as highly rough regions which cannot be appreciated in the wild type. Fig.8 shows
topographical line profiles recorded over a smooth area (a), and over a rough area (b). It is
clear from our results that although the mutation does not perceptibly alter the cell
I Probing Btologicai Molerrdes: Theory and Experiment 55
dimensions and overall morphology, a great impact on the cell wall surface properties is
observed. In orderto quantify the observed change in surface morphology, surface
Figure 7 Topographical non-contact images of mutants cells.
Figure 8 Topographical line profiles recorded of (a) smooth regions and (b) rougher
regions of the mutant cell surface.
roughness measurements were also taken of the mutant cells and the results summarized in
Fig.9. For comparison, the corresponding roughness values of the wild type are also
shown. Surface roughness of the mutants cells is observed to be greater than that of the
wild type cells, on all the measured scales, confirming what the images presented visually.
From the comparison of the surface roughness in the two cells types, differences of about
10-15 nm are obtained. These differences obtained in surface roughness by AFM for both
samples could not be detected by other microscopy techniques requiring a surface coverage
that could be of the same magnitude as the differences observed.
00 05 10 :5 20 25 30 45 40
L (mi
Figure 9 Surface roughness measured values of mutant and wild type cells.
On the other hand, it is interesting to point out the non-existence of chitin rings in
the AFM images of the mutant cells. It is possible that extensive damage of the cell wall of
the mutant cells caused the chitin rings to disappear. Apart from the high roughness
regions observed, highly irregular structures of a diameter of about 0.5 pm are observed on
the surface, as can be seen in Fig.10. The origin of these observed structures is not
altogether clear to us, but they could be deformed bud scars or even the initial stages of
new budding processes, since the growing rate for the mutant cells lower than for the wild
ones,
Figure 10 Rough regions with highly irregular structures observed on the surface of
mutant cells.
3 CONCLUSION
Atomic Force Microscopy is revealed as a powerful technique for studying the influence of
N-glycosylation process defects on the physical cell wall properties of S. cerevisiae cells.
Non-contact images show highly rough and irregular structures on the mutant cells, this
observation being confirmed by the surface roughness measurements over wild type and
mnn9 cells. The fact that AFM is capable of probing biological surfaces without previous
treatment and its unprecedented lateral and vertical resolution, make this instrument a
unique tool for nanometric studies of membranes and cell surfaces.
Acknowledgements
This work was supported by Project IPROOA083 from Junta de Extremadura (Consejeria
de Educacibn, Ciencia y Tecnologia).
References
1 S.Munro. FEBS Letters, 2001,498,223.

2 C.E. Ballou. Methods Enzymol., 1990, 185,440.
3 L. Ballou, R.E. Cohen, C.E. Ballou. J. Biol. Chem., 1980,255,5986.
4 P . Mafias, I. Oliveros, M. Avalos y L.M. Hernhdez. GZycobioZogy, 1997,4,487
5 S.N. Magonov, M.-H. Whangbo, in Surface analysis with STM and AFM, VCH,
Weinheim, 1996,323.
6 R.M. Overney, E. Meyer, J. Frommer, D. Brodbeck, R. Luthi, L. Howald, H.-J.
Guntherodt, M. Fujihira, H. Takano, Y. Gotoh, Nature, 1992,359, 133.
7 J. Colchero, H. Bielefeldt, A. Ruf, M. Hipp, 0. Marti, and J. Mlynek, Phys. Stat. Sol.
(a), 1992, 131, 73.
8 H.-Y. Nie, M.J.Walzak, N.S. Mchtyre, A.M. EL-Sherik, Appl. Surj". Sci.,1999, 144-
145,633.
9 S.A. Ishijima, M. Konomi, T. Takagi, M. Sato, J. Ishiguro, M. Osumi, FEMS
Microbiol. Lett., 1999,180(1), 3 1.
10 Y-L. Ong, A. Razatos, G. Georgiou, M.M. Sharma, Langmuir, 1999, 15,2719.
1 1 J.D. Kiely, D.A. Bonnell, J. Vac. Sci. Techno1B, 1997, 15(4), 1483.
12 P.K. Tsai, J. Frevert, C.E. Ballou. J. Biol. Chem., 1984,259(6), 3805.
PROBING SUPRAMOLECULAR ORGANISATION AT IMMUNE SYNAPSES.
Fiona E. McCann', Klaus Suhling', Leo M. Carlin', Konstantina Eleme', Kumiko Yanagi',
Paul M.W. French2,David Phillips3, Daniel M.Davis'*4.
Department of Biological Sciences, Sir Alexander Fleming Building, * Department of

Physics, Department of Chemistry, Imperial College of Science, Technology and Medicine,
London, SW7 2AZ,UK. E-Mail: d.davis@ic.ac.uk
1 INTRODUCTION: MOLECULAR RECOGNITION OF DISEASE BY NATURAL

KILLER CELLS
Natural Killer (NK) cells are large granular lymphocytes that have the ability to lyse certain
target cells that have undergone viral infection or tumour transformation.1 One way that NK
cells can discriminate infected cells from healthy cells is by detecting a loss or downregulation
of self proteins, in particular major histocompatibility complex (MHC) class I molecules on
the target cell surface.* Viral infection and tumourigenesis are frequently associated with
alterations in expression of MHC class I molecules as a means to avoid detection by cytotoxic
T cells.3 NK cell recognition of self protein is mediated by groups of receptors that bind
specific MHC class I molecules that transduce an inhibitory signal, thus preventing the natural
cytotoxicity incurred with NK cell docking. In particular, human NK cells detect MHC class I
protein via a variety of immunoglobulin-like and C-type lectin receptors.4-6 Killer
Immunoglobulin-like Receptors (KIR) with two immunoglobulin (Ig) domains, KIR2DL 1 and
KIR2DL2 recognise the MHC class I proteins, I-LA-Cw4 or -Cw6, and HLA-Cw3 or -Cw7,
respectively.7-11 However, our understanding of how NK cells detect diseased cells is further
complicated by the expression of stirnulatory as well as inhibitory NK receptors for MHC
class I protein.12~13In addition, several classes of NK receptors have recently been identified
that recognise non-MHC proteins on target cell surfaces.14-19 In short, regulation of NK cell
cytotoxicity occurs through a complex and little understood integration of signals from
stimulatory and inhibitory cell surface receptors. Here, we first review how fluorescence
imaging has elucidated much about the organisation of proteins at immune synapses. We then
discuss some molecular mechanisms that may influence this organisation of proteins at
immune synapses. Next we discuss the recent unexpected observation that cell surface
proteins transfer between cells at immune synapses. Finally, as an example of a new technique
that could be employed in the study of immune synapses, we discuss the use of fluorescence
lifetime imaging (FLIM) to probe the environment of cell surface proteins.
2 SUPRAMOLECULAR ORGANISATION AT IMMUNE SYNAPSES
It has recently emerged that immunological intercellular contacts, where information is

exchanged between effector and target cells, consists of specialised membrane domains
reminiscent of the neuronal synapse. This analogy was first suggested by Norcross,*o and the
term ‘the immunological synapse’ was proposed by Paul and Seder a decade later with
reference to T lymphocytes and antigen presenting cells.21 Although it is largely unknown
how this organisation of cell surface proteins is orchestrated, several mechanisms have been
proposed. These include lateral segregation of membrane proteins by lipid rafts, cytoskeletal
involvement and distribution of proteins according to the size of their extracellular domains.
One of the most striking recent developments in molecular immunology is the discovery
that T cell receptors and adhesion molecules cluster in unexpectedly large segregated
membrane domains at immune synapses.22-24In a model system using T cells interacting with
protein-rich two dimensional lipid bilayers, Grakoui et al 24 report on protein dynamics at the
immune synapse following TCR ligation. They propose that after contact between the T cell
and the protein-rich lipid bilayers, a bull’s-eye pattern of protein domains forms with a central
region containing integrins surrounded by a ring containing MHC protein. After a few minutes
the pattern inverts so that the central region is predominantly occupied by T cell receptor
(TCR) and MHC protein while the integrins, LFA-1 and ICAM-1 are segregated to the
periphery.24 These discrete domains have been termed supramolecular activating clusters
(SMAC), subdivided into central, i.e. cSMAC, and peripheral, i.e. pSMAC, regions.22
Subsequent stabilisation of this arrangement occurs which may serve to sustain the T cell
activation.
The transmembrane phosphatase CD45 can dephosphorylate Lck and therefore block T cell
activation during the early stages following TCR ligation. After about 3 min. of contact
between a T cell and APC, CD45 is excluded from the developing T cell immune synapse,
which is consistent with the notion that exclusion of tyrosine phosphatases from membrane
compartments is required for triggering the kinase cascade. Intriguingly, after an initial
exclusion of CD45 from the T cell immune synapse some CD45 relocates close to the cSMAC
in the vicinity of the TCR/MHC complex within about 10 min.25 These and other observations
2326-30 suggest an important role for immune synapses in controlling T cell signalling
(reviewed in references 31-33-34).
Redistribution of receptors and ligands at NK cell immune synapses were first imaged by
time-lapse laser scanning confocal microscopy (LSCM).35 Interestingly, the inhibitory NK
cell immune synapse has inverted arrangements of MHC, ICAM-1 and LFA-1 in comparison
to the mature activating T cell synapse.35 Strikingly, the NK inhibitory immune synapse
prevails in the presence of drugs that disrupt cytoskeleton or deplete ATP, which suggests that
synapse formation is independent of cytoskeleton and/or ATP mediated events.35
The different arrangements of proteins in a T cell activating and an NK cell inhibitory
immune synapse are schematically represented in Figure 1. In line with the terminology used
to first describe T cell immune synapses, discrete regions at inhibitory immune synapses have
been termed p- and c-supramolecular inhibitory clusters (SMIC).36 Recent high-resolution
fluorescence imaging by confocal microscopy of target cells transfected with enhanced green
fluorescent protein (GFP)-tagged MHC protein revealed that the patterning of p- and c-SMIC
is fluid, i.e. the patterning of membrane domains can alter during contact with the NK cell.36
In the case of NK cells where inhibitory and activating receptors co-exist on the cell surface,
segregation of proteins into discrete membrane domains may be critical in integrating
stimulatory and inhibitory signals. The location of activating receptors and ligands at NK
immune synapses is one major unknown in this area and more broadly, virtually nothing is
known of the biophysical processes governing formation of immune synapses. Possible
molecular mechanisms that may influence lateral segregation and redistribution of membrane
proteins and ultimately synapse formation include lipid rafts (2. I), cytoskeleton (2.2), and
distribution of proteins according to the size of their extracellular domains (2.3).
-2-4 ~ u n
-5-10Hm 4
Inhibitory NK cell synapses
Activating T cell synapses
TCFUCD3 'Ip MHC Class 11-peptide )? MHC Class I
@ CD2 0 LFA-3
I KrR
1 LFA-I 0 ICAM-1 1 CD4/CD8
Figure 1: Comparison of activating and inhibitory immune synapses. To date activating T cell
synapses and inhibitory NK cell synapses have been imaged and are schematically represented
above. Both inhibitory T cell synapses, e.g. with anergised T cells, or activated NK cell
synapses e.g. upon recognition of CMV- infected cells remain to be imaged.
2.1 Lipid Rafts
One important method by which proteins within plasma membranes can be separated is by
association with different lipids (reviewed in references 37-39). Early evidence for membrane
domains came from fluorescence lifetime measurements that identified two distinct
environments within the plasma membranes of living cells. These distinct populations
corresponded to different lipid phases with different degrees of order of their hydrocarbon
chains.40 The saturated tails of sphingolipids associated with cholesterol are proposed to exist
in a liquid-ordered phase separated from the liquid-disordered phase of the phospholipid
containing plasma membrane. These sphingolipid, cholesterol-rich lipid rafts drift amongst the
phospholipids, and association of proteins with the rafts segregates them from other non-
associated proteins distributed within the phospholipid rich plasma membrane.37t38 Rafts have
been shown to exclude most membrane proteins and are thought to constitute approximately
10% of the cell surface.33 Proteins known to associate with lipid rafts include
glycosylphosphatidylinositol (GP1)-linked, myristylated and acylated proteins. Addition of a
GPI anchor inserts the protein into the outer membrane leaflet while acylation normally
promotes association with the cytoplasmic membrane leaflet$1942 one exception being the
transmembrane transferrin receptor, which is excluded from lipid rafts despite being multiply
acylated.42 Examples of proteins constitutively associated with lipid rafts include the Src-
family kinases, and G proteins. Others such as the T and B cell receptors, along with some
downstream signalling molecules are recruited into rafts upon ligand binding, demonstrating
the dynamic nature of events occurring within these membrane microdomains.37-39t42
Several studies of lipid rafts have relied on their insolubility in the detergent Triton X-100
at 4"C, in which they form glycolipid-enriched complexes, and are thus often referred to as
detergent insoluble glycolipid enriched complexes (DIGs).43 Due to their high lipid content,
rafts 'float' to a low density during gradient centrifugation enabling the isolation and
identification of associated proteins. This method cannot be used to quantitatively report lipid
raft composition in the intact cell. Milder detergents such as octyglucoside will solubilise the
rafts, and thus a larger than expected fraction of membrane proteins are insoluble in Triton-X-
Finally, the lipid composition of the rafts has not been proven to be distinct from the
plasma membrane. Thus, the question can still be raised whether this methodology induces
artefacts in identification of the contents of lipid rafts and this highlights the importance of
employing detergent-independent methods to study lipid rafts in cell membranes at
physiological temperatures.
Bioimaging can be used to reveal lipid rafts in living cells using fluorescent labels of
proteins or lipids enriched in these domains. In resting cells, lipid rafts are not clearly
identifiable and only when they are induced to coalesce into larger structures following
crosslinking of raft components are they visible by microscopy.45 Fluorophore-conjugated
cholera toxin fl subunit, which binds GMl, is commonly used as a stain for cholesterol-rich
lipid domains.46~47 That clustering of cholera-toxin staining was induced by co-stimulation of
T cells suggests an immunological importance of lipid raft motility.47 Using fluorescence
resonance energy transfer (FRET), Varma and Mayor48 found cholesterol dependent
clustering of GPI-anchored proteins into domains sized less than 70 nm, but using a similar
methodology, another group found no evidence for such ~lustering.~g
Current evidence supports a model in which lipid rafts play a role in immune cell
activation.50~51In resting T cells, TCR are excluded from lipid rafts but ligand interaction may
result in oligomerisation of the TCR and subsequent recruitment to lipid rafts where they are
brought into contact with Src-family kinases.39 The first biochemical event that can be
detected following TCR ligation is Lck-mediated phosphorylation of tyrosine residues within
the immunoregulatorytyrosine activating motifs (ITAMs).~~ T cell signalling was abolished in
cells expressing only a mutant, non-palmitoylated isoform of Lck that is excluded from lipid
rafts.46 Strikingly, signalling was rescued when separate clusters of the mutant Lck and
gangliosides within rafts were brought together using bridging antibodies, demonstrating that
Lck was not inactivated by the mutation. This suggested that Lck must be physically close to
its signalling partners for effective signalling and this association normally required clustering
of the proteins within lipid rafts.46
Phosphorylation of ITAMs is immediately followed by recruitment of cytoplasmic ZAP-70

to lipid rafts where Lck constitutively resides (reviewed in reference 53). ZAP-70 is then in
close proximity to linker for activation of T cells (LAT), which is subsequently
phosphorylated triggering recruitment of further signalling proteins to lipid rafts. Despite
localisation within the plasma membrane, a mutant LAT that is not palmitoylated and is
excluded from cholesterol-rich lipid rafts was unable to be phosphorylated by ZAP-70 leading
to the disruption of functional T cell signalling.54 Clearly, segregation of membrane proteins
by association with lipid rafts has the potential to selectively promote or exclude specific
protein interactions. However, the relationship between lipid rafts and the supramolecular
organisation at immune synapses remains to be directly investigated. Importantly, Lou et a1 55
have demonstrated that lipid rafts, revealed by staining with cholera toxin, are not clustered at
inhibitory NK cell immune synapses, inferring that lipid rafts are at least not necessary to
create distinct protein domains at immune synapses.
2.2 The Role of the Cytoskeletonin Immune Synapse Formation.
In opposition to segregation of membrane proteins by association with lipids, several groups

have suggested that protein organisation at immune synapses is controlled by association with
intracellular cytoskeletal proteins.32.56 Evidence for active transport of receptors to the
immune synapse was first described using the large beads to observe the movement of the
cortical actin cytoskeleton and linked receptors.23 Upon TCR engagement, beads that are
either specifically attached to ICAM- 1, or non-specifically attached via streptavidin,
translocate to the T cell immune synapse. The movement of non-specifically attached beads
was abolished in the absence of ICAM-1, indicating the process to be activation dependent.
Evidence that myosin motor proteins controlled this movement was that the drug butanedione
monoxime (BDM) impaired bead movement across the surface of the T cell. It should be
noted however that BDM is not a specific inhibitor of myosin motor proteins and the
involvement of other processes cannot be ruled out.57
It is the T cell cytoskeleton that is thought to drive formation of the T cell immune synapse.
When B cells are used as APC, no polarisation of their cytoskeleton has been observed during
interaction with T cells, and inhibition of APC cytoskeleton rearrangement had no effect on T
cell activation. However in contrast, in immune synapses between dendritic cells (DC) and T
cells, Al-Alwan et a158 observed polarization of filamentous (f-) actin and fascin, an actin
bundling protein unique to dendritic cells. Pretreatment of DC with cytochalasin D, which
disrupts assembly of filamentous actin, reduced the conjugation of DC with T cells by 76%.
2.3 The Role of Protein Size in Immune Synapse Formation
The current dogma33amongst immunologists is that supramolecular organisation at immune

synapses is actively driven by ATP-dependent processes including myosin driven movement
of the effector cell’s cytoskeleton. However, our own experimental evidence suggests that the
actin cytoskeleton is not necessary, at least for formation of the inhibitory NK immune
synapse.35 Another mechanism that could drive the formation of segregated protein domains
is that they thermodynamically separate according to the sizes of their extracellular domains,
as schematically represented in Figure 1. Initial adhesion of T celYAPC or NK /target cell
occurs via binding of the large integrins LFA- 1 and ICAM- 1, which form a complex bridging
approximately 40 nm across the synapse. This interaction is thought to initiate binding of
smaller molecules e.g. TCR or KIR binding MHC protein, which would span approximately
15 nm. If it is energetically favourable for the membrane to bend rather than the proteins,
segregated protein domains are created (Figure 1). Separation of proteins in this manner has
been postulated to occur at a sub-micron scale, 56959 but it is possible that such microdomains
coalesce to create the supramolecular organisation of proteins seen by LSCM.
Molecular modelling suggests that such spontaneous organisation of immune synapses
could indeed occur, based on thermodynamicsgoverning the lateral mobility of proteins in the
membrane, membrane mechanics, and the sizes of the extracellular portions of the cell surface
proteins.60 Thus, if membrane biophysics contributes, at least in part, to the organisation at
immune synapses, membrane fluidity should be different in comparison to elsewhere at the
cell surface. Furthermore, membrane viscosity should vary across the immune synapse such
that domains in varying viscosity segregate in concordance with the domains rich in different
proteins. Such aspects of the biophysical chemistry at immune synapses are wholly uncharted.
3 INTERCELLULAR TRANSFER OF CELL SURFACE PROTEINS AT IMMUNE

SYNAPSES
Fluorescent probes along with use of LSCM have proved to be invaluable in visualising
segregation of proteins at immune synapses with high temporal and spatial resolution.
Unexpectedly, evidence is emerging that several cell surface proteins transfer between
contacting cells at immune synapses.
Several groups have shown the receptor dependent transfer of antigen presenting MHC
protein to T cells 61-65 by a number of experimental techniques. Other cell surface proteins
also transfer from APC to T cells, namely CDSO (B7-I), OX40L and CD54 (ICAM-1).66T
cells have even been observed to capture membrane fragments from APC, reliant upon TCR
signalling.64 The intercellular transfer of proteins may serve a number of immunoregulatory
functions. For example, it has been demonstrated that T cells that have ‘captured’ target cell
MHC protein onto their own cell surfaces can then stimulate neighbouring T cells to either
proliferate 61 or be killed by cytotoxic T cells with the same specificity, termed fratricide.63 T
cell fratricide might support the notion of lymphocyte ‘exhaustion’ where T cell responses are
reduced after prolonged exposure to antigen. Recently, B cells have also been shown to
acquire antigen from professional APC after immune synapse f~rmation.~’In B cells
intercellular transfer of antigen is suggested as a way that B cells can process and present
membrane bound antigen.67
Our own study shows that NK cells expressing the KIR2DLl receptor can acquire cognate
HLA-C ligands from target cells at inhibitory NK cell immune synapses. A vesicle or group of
vesicles of Green Fluorescent Protein (GFP)tagged HLA-C can clearly be seen moving from a
target cell into an NK cell in Figure 2. Surprisingly large amounts, e.g. up to 30%,of the MHC
protein can be ‘captured’ from target cells by NK cells.36 An immunological function for this
intercellular transfer is not known but we speculate that in NK cells intercellular transfer of
MHC may serve to affect the threshold at which inhibitory signals are initiated, possibly by
sequestering KIR or key signalling molecules away from the NK immune synapse.
Figure 2. Intercellular transfer of GFP tagged HLA-Cw6 from a 721.221 target cell to a
human blood NK cell. (a) Transmitted light image showing a target cell (721.221, EBV
transformed B cell line) expressing GFP-tagged HLA-Cw6 in contact with a KIR2DL1
expressing human blood NK cell (smaller cell). (b) A series of confocal images showing the
GFP fluorescence within the white box in (a) over 40 seconds. Each is a single 0.3 pm slice
taken from the same focal plane. The white arrowhead in the first frame indicates a vesicle, or
group of vesicles containing GFP-tagged HLA-Cw6 moving away from the HLA-Cw6
clustered at the NK inhibitory immune synapse and into the NK cell cytoplasm. The position
of the white arrowhead remains constant throughout to indicate the movement of the vesicle,
or group of vesicles. The zero time-point is arbitrarily defined. The scale bar in the lower right
corner of (a) represents 10 pm. This figure is representative of at least three independent
experiments (after reference 36).
The half-life of soluble KIR2DL1 binding to soluble HLA-C is around 0.3 s, more than an
order of magnitude faster than the binding of TCR to MHC ligands and at least two orders of
magnitude faster than mAb binding.68 Interestingly however, the half-life of KIR2DL l/HLA-
Cw6 has been measured to be similar to TCR/MHC binding in the presence of zinc.69The few
minutes that NK cells remain in contact with target cells seems unlikely to be long enough for
appreciable amounts of HLA-C to be captured without the induction of an active process by
KIR. This is consistent with the requirement for NK cell ATP. Thus, the half-life of KIRMHC
and TCR/MHC complexes may be similar in order to facilitate a common molecular
mechanism of capturing MHC protein from target cells. Possible molecular mechanisms that
may influence intercellular cell surface protein transfer include (3.1) protein cleavage or
ectodomain shedding, (3.2) exosomes, (3.3) protein and lipid mixing, (3.4) cytoskeleton-
mediated transport.
3.1 Protein Cleavage or Ectodomain Shedding
Extracellular domains of cell surface receptors can be cleaved by proteases. This method of
protein release has been documented in lymphocytes previously, for example in shedding TNF
receptor and L-selectin.70 However, this seems an unlikely mechanism for intercellular
transfer of proteins at immune synapses for several reasons. Firstly, in our own studies, the
MHC protein seen to transfer into NK cells was tagged with GFP on the cytoplasmic tail of
the protein. Thus, for GFP fluorescence to be observed on the NK cell, the whole GFP-MHC
construct is likely to have been transferred and certainly not just the extracellular portion.36
More directly, Western blotting the lysates of NK cells purified after incubation with target
cells revealed that the NK cell acquired a protein of the size corresponding to the whole of the
GFP-tagged HLA-C, indicating that the transferred protein was not first cleaved.36 Finally, the
transfer of OX40L in the presence of protease inhibitors 66 excludes the action of ectodomain
shedding enzymes in facilitating intercellular protein transfer at immune synapses.
3.2 Exosomes
Exosomes, the product of fusion of late endosomal or lysosomal vesicles with the plasma
membrane, are released by a number of lymphocytes, including cytotoxic T cells, B cells and
dendritic cells. These exosomes can contain several cell surface proteins in the Ig superfamily,
including CD54, CD86 (B7-2) and MHC class I and I1 proteins, reviewed by Denzer et al.71
Antigen-presenting MHC complexes carried in the vesicles are accessible, as purified
exosomes can stimulate T cells in vitru.7*973 This could explain why the transfer of certain
proteins is dependent on receptor recognition, as docking and internalisation of vesicles into
the effector cell may depend on receptor ligation andor signalling. However, constitutive
release of MHC protein in APC derived exosomes is too slow to account for the large amounts
of MHC protein transferred to NK cells, unless exosomal secretion is specifically triggered at
immune synapses.
3.3 Protein and Lipid Mixing at Immune Synapses
The plasma membrane itself is highly dynamic in structure; lipids and proteins can relocate to
various points at the cell surface. It could be that when two such membranes come into close
proximity at the immune synapse, and receptor ligation occurs, lipids and proteins could
transfer between the opposing lipid bilayers. In fact, there is evidence that lipids are
transferred between cells along with proteins at the T cell immune synapse.* When the
fluorescent lipid PKH26 was incorporated into target cell membranes it was shown to transfer
to T cells only if TCR signalling has occurred. However CFSE, a stable internal protein dye
used to label target cell cytosol did not transfer. Receptor specificity could play a part in a
process like this, whereby receptors internalise and drag their ligands along with membrane
fragments into the effector cell.
3.4 Cytoskeletal-mediatedTransport at Immune Synapses
Tubulin cytoskeleton is often involved in intracellular vesicular transport, and as can be seen
for example in Figure 2, the movement of HLA-C containing vesicles is directional,
reminiscent of vesicles travelling along the cytoskeleton, rather than moving by Brownian
motion. Thus, it is possible that the cytoskeleton is required to move HLA-C that has already
transferred to the effector cell away from the immune synapse. However, treatment of the cells
with cytochalasin D and BDM had no affect on the number cell conjugates in which
intercellular protein transfer could be seen,36 suggesting that polymerisation of the actin
cytoskeleton, and myosin motors are not necessary for intercellular protein transfer. Studies
using sodium azide to deplete ATP showed that ATP is required by the NK cell, and not the
target cell, for transfer to occur. Thus, some active transport system must be invoked to
facilitate intercellularprotein transfer.
4 PROBING THE PROTEIN ENVIRONMENT BY FLUORESCENCE LIFETIME

IMAGING
Immune synapse formation involves a complex rearrangement of proteins and lipids at

intercellular contacts creating distinct local environments of each protein type. Clearly, new
techniques are needed to probe the environment of protein domains within immune synapses.
Fiuorescence intensity imaging only reveals the location of proteins at immune synapses. But
fluorescence emission is a multi-parameter signal that can report on, for example, the local
viscosity,74 refractive index,75-77 pH,78 solvent polarity79~80or the orientation of the
fluorophore and its interaction with either the environment or other fluorophores.81182
Fluorescence lifetime imaging (FLIM)83-86 is a technique that, in addition to position and
intensity, also captures the fluorescence lifetime, i.e. the average time needed for the
fluorescence to decay after an excitation pulse. This allows the local distribution and physical
environment of the fluorescence probe or fluorophore-tagged proteins to be studied.87 FLIM
can probe the physical properties of each protein microenvironment and this information will
help elucidate the molecular mechanisms of immune synapse formation. However, due to the
number of parameters on which the fluorescence lifetime depends, the interpretation of data
can be complex, particularly in a biological or biomedical context where many of these
parameters may be unknown. Thus, in order to interpret FLIM of GFP in cells, we set out to
determine the parameter that affects the fluorescence lifetime of GFP tagged receptors.
4.1 The Fluorescence Lifetime of GFP Reports the Local Refractive Index
In the absence of fluorescence quenching processes, the fluorescence lifetimez can be

calculated from the absorption and emission spectra using the Strickler-Berg formula,8* taking
into account the fluorescence quantum yield 4 :
Formula 1
where n is the refractive index of the medium, Z is the intensity of the fluorescence emission, E
is the molar extinction coefficient and i j is the wave number (inverse of the wavelength L, i.e.
? =k-'). The relationship is calculated from the Einstein A and B coefficients for sharp atomic
transitions and is applicable to molecules where the absorption band is broad and the emission
Stokes shifted. The refractive index dependence is due to the medium in which the absorption
and emission processes occur77
Figure 3 shows three decays of GFP fluorescence upon excitation with a femtosecond laser.
The fluorescence intensity (vertical axis, logarithmic) decays with time (horizontal axis)
within nanoseconds after the excitation. The upper fluorescence decay is GFP in PBS buffer,
the middle decay GFP in 50% glyceroWO% buffer and the lower decay 90% glyceroYlO%
buffer. A gradual decrease of the fluorescence lifetime as the glycerol content increases is
clearly evident. These fluorescence decays were obtained by the time-correlated single photon
counting method,89 with excitation at 470 nm and the emission collected at 510 nm using a
I Probing Biological Molecules: Theo? and Experiment 67
monochromator. An emission polarizer oriented at the magic angle 54.7" was used to
eliminate rotational depolarization effects. A deconvolution of the decay curves with the
instrumental response function yields the fluorescence lifetime 2. It ranges from 2.80 ns in
PBS buffer (refractive index n51onrn= 1.337) to 2.20 ns in 90% glyceroY10% buffer (n510nm=
1.463).
A plot of the inverse lifetime t' versus the square of the refractive index of solvent
solutions with an increasing amount of glycerol, n2, is shown in Figure 4. It is a linear
relationship as predicted by the Strickler-Berg formula assuming E, I and @areconstant. A
straight line fit yields a gradient of 0.239H.010 ns-', which is in good agreement with the
value of 0.223M.020ns-' calculated directly from the absorption and emission spectra and the
quantum yield 4 using E =50000 1 mol-' cm-' and @ =0.60 (Formula l).90 These data are not
particular to using mixtures of glycerol and water as a solvent (Suhling et al., manuscript in
preparation).
We have shown that in homogeneous and isotropic media the GFP fluorescence decay
reports on the refractive index of the environment. Our aim now is to use KIM and the GFP
fluorescence decay directly, to report on the refractive index in different domains of immune
synapses. This raises the question what biologically relevant parameter the refractive index
reports on. We hypothesise that, for example, a cholesterol rich lipid microdomain 91 will
have a distinct refractive index.
I " " ' " " " " I
0 5 10
time I ns
Figure 3: Fluorescence decays of GFP in PBS buffer/glycerol mixtures. The upper curve
(continuous line) is the GFP fluorescence decay in PBS buffer, the middle curve (dashed line)
in 50% glyceroY50% PBS buffer and the lower curve (dotted line) in 90% glycerol/lO% PBS
buffer. As the amount of glycerol is increased, the refractive index increases and the
fluorescence decay is shortened. The excitation wavelength kX=470nm, and the fluorescence
emission wavelength was L,=5 10 nm collected at the magic angle.
0.46
0.44
0.42
.
u)
re 0.40
0.38
0.36
nz
Figure 4: The inverse fluorescence liferime T-' versus the square of the refractive index, n2. It
is a linear relationship as predicted by the Strickler-Bergformula (Formula I ) assuming E, I
and @are constant. The straight line is a jit to the data points and yields a gradient of
0.239~l.010ns-'.
5 SUMMARY
Understanding of the formation and function of immune synapses is a foundation from which
new pharmaceutically active compounds can be designed to augment immune responses to
viruses or suppress autoimmune responses. For example, candidate novel drugs could be
identified by screening for small molecules that interfere with specific patterns of protein
domains at immune synapses. To date, limited techniques have been used to probe
supramolecular organisation at cell surfaces. Fluorescence Lifetime Imaging reporting on the
physical environment of proteins as they redistribute into immune synapses is a highly
promising technique to apply to studying immune synapse formation.
Acknowledgements
We are grateful to Jan Siege1 for critical reading of this manuscript. Work in our laboratory is
supported by grants from the Medical Research Council, the Biotechnology and Biological
Sciences Research Council and The Royal Society.
References
1 G. Trinchieri, Adv Zmrnunol, 1989,47,187.

2 H. G. Ljunggren and K. Karre, Immunol Today, 1990,11,237.
3 L. L. Lanier, Annu Rev Immunol, 1998,16,359.
4 E. 0. Long and S . Rajagopalan, Semin Zmmunol, 2000,12,101.
5 E. 0. Long, D. F. Barber, D. N. Burshtyn, M. Faure, M. Peterson, S. Rajagopalan, V.

Renard, M. Sandusky, C. C. Stebbins, N. Wagtmann and C. Watzl, Immunol Rev,
2001,181,223.
6 D. H. Raulet, R. E. Vance and C. W. McMahon, Annu Rev Zmmunol, 2001,19,291.
7 M. Colonna, G. Borsellino, M. Falco, G. B. Ferrara and J. L. Strominger, Proc NatE
Acad Sci Il S A, 1993,90,12000.
8 M. Colonna, E. G. Brooks, M. Falco, G. B. Ferrara and J. L. Strominger, Science,
1993,260,1121.
9 J. C . Boyington, S. A. Motyka, P. Schuck, A. G. Brooks and P. D. Sun, Nature, 2000,
405,537.
10 Q. R. Fan, E. 0. Long and D. C. Wiley, Nat Immunol, 2001,2,452.
11 J. C . Boyington, A. G. Brooks and P. D. Sun, Immunol Rev, 2001,181,66.
12 0. Mandelboim, H. T. Reyburn, M. Vales-Gomez, L. Pazmany, M. Colonna, G.
Borsellino and J. L. Strominger, J E x p Med, 1996,184,913.
13 A. Moretta, S. Sivori, M. Vitale, D. Pende, L. Morelli, R. Augugliaro, C. Bottino and
L. Moretta, . IExp Med, 1995,182,875.
14 R. Biassoni, C. Cantoni, D. Pende, S. Sivori, S. Parolini, M. Vitale, C. Bottino and A.
Moretta, Zmmunol Rev, 2001,181,203.
15 D. Pende, S. Parolini, A. Pessino, S. Sivori, R. Augugliaro, L. Morelli, E. Marcenaro,
L. Accame, A. Malaspina, R. Biassoni, C. Bottino, L. Moretta and A. Moretta, J Exp
Med, 1999,190,1505.
16 M. Vitale, C. Bottino, S. Sivori, L. Sanseverino, R. Castriconi, E. Marcenaro, R.
Augugliaro, L. Moretta and A. Moretta, J Exp Med, 1998,187,2065.
17 S . Sivori, M. Vitale, L. Morelli, L. Sanseverino, R. Augugliaro, C. Bottino, L. Moretta
and A. Moretta, J Exp Med, 1997,186,1129.
18 A. Pessino, S. Sivori, C. Bottino, A. Malaspina, L. Morelli, L. Moretta, R. Biassoni
and A. Moretta, J Exp Med, 1998,188,953.
19 C. Cantoni, S. Verdiani, M. Falco, A. Pessino, M. Cilli, R. Conte, D. Pende, M. Ponte,
M. S. Mikaelsson, L. Moretta and R. Biassoni, Eur J Zmmunol, 1998,28,1980.
20 M. A. Norcross, Ann lmmunol (Paris), 1984,135D, 113.
21 W. E. Paul and R. A. Seder, Cell, 1994,76,241.
22 C. R. Monks, B. A. Freiberg, H. Kupfer, N. Sciaky and A. Kupfer, Nature, 1998, 395,
82.
23 C. WUlfing and M. M. Davis, Science, 1998,282,2266.
24 A. Grakoui, S. K. Bromley, C. Sumen, M. M. Davis, A. S. Shaw, P. M. Allen and M.
L. Dustin, Science, 1999,285,221.
25 K.G . Johnson, S. K. Bromley, M. L. Dustin and M. L. Thomas, Proc Natl Acad Sci U
SA, 2000,97,10138.
26 M. A. McCloskey and M. M. Poo, J Cell Biol, 1986,102,2185.
27 B. C. Schaefer, M. F. Ware, P. Marrack, G. R. Fanger, J. W. Kappler, G. L. Johnson
and C. R. Monks, Immunity, 1999,11,411.
28 C. Wiilfing, M. D. Sjaastad and M. M. Davis, Proc Natl Acad Sci U S A, 1998, 95,
6302.
29 C. Wiilfing, A. Bauch, G. R. Crabtree and M. M. Davis, Proc Natl Acad Sci U S A,
2000,97,10150.
30 M. F. Krummel, M.D. Sjaastad, C. WUlfing and M. M. Davis, Science, 2000, 289,

1349.
31 J. Delon and R. N. Germain, Curr Biol, 2000,10, R923.
32 M. L. Dustin and J. A. Cooper, Nut Zmmunol,2000,1,23.
33 S . K. Bromley, W. R. Burack, K. G. Johnson, K. Somersalo, T. N. Sims, C. Sumen, M.
Davis, A. S. Shaw, P. M. Allen and M. L. Dustin, Annu Rev Zmmunol, 2001,19,375
34 E. Donnadieu, P. Revy and A. Trautmann, Immunology, 2001,103,417.
35 D. M. Davis, I. Chiu, M. Fassett, G. B. Cohen, 0. Mandelboim and J. L. Strominger,
Proc Nut1 Acad Sci U S A, 1999,96,15062.
36 L. M. Carlin, K. Eleme, F. E. McCann and D. M. Davis, manuscript submitted.
37 D. A. Brown and E. London, J Biol Chem, 2000,275,17221.
38 K,Simons and E. Ikonen, Nature, 1997,387,569.
39 A. Cherukuri, M. Dykstra and S . K. Pierce, Immunity, 2001, 14,657.
40 R. D. Klausner, A. M. Kleinfeld, R. L. Hoover and M. J. Karnovsky, J Biol Chem,
1980,255,1286.
41 D. A. Brown and E. London, Annu Rev Cell Dev Biol, 1998,14,111.
42 P. W. Janes, S. C.Ley, A. I. Magee and P. S. Kabouridis, Semin Zmmunol, 2000, 12,
23.
43 D. A. Brown and J. K. Rose, Cell, 1992,68,533.
44 R. Schroeder, E. London and D. Brown, Proc Nut2 Acad Sci U S A, 1994,91,12130.
45 E. D. Sheets, D. Holowka and B. Baird, J Cell Biol, 1999,145,877.
46 P. W. Janes, S. C. Ley and A. I. Magee, J Cell Biol, 1999,147,447.
47 A. Viola, S. Schroeder, Y.Sakakibara and A. Lanzavecchia, Science, 1999,283,680.
48 R. Varma and S. Mayor, Nature, 1998,394,798.
49 A. K. Kenworthy, N. Petranova and M. Edidin, MoZ Biol Cell, 2000,11,1645.
50 R. Xavier, T. Brennan, Q. Li, C. McCormack and B. Seed, Immunity, 1998,8,723.
51 0. Leupin, R. Zaru, T. Laroche, S. Muller and S. Valitutti, Curr Biol, 2000,10,277.
52 A. C . Chan, D. M. Desai and A. Weiss, Annu Rev Zmmunol, 1994,12,555.
53 W . Zhang and L. E. Samelson, Semin Zmmunol, 2000,12,35.
54 W . Zhang, J. Sloan-Lancaster, J. Kitchen, R. P. Trible and L. E. Samelson, Cell, 1998,
92, 83.
55 2. Lou, D. Jevremovic, D. D. Billadeau and P. J. Leibson, J Exp Med, 2000,191,347.
56 A. P. Van der Menve, S. J. Davis, A. S. Shaw and M. L. Dustin, Semin Zmmunol,2000,
12,5.
57 N. Urwyler, P. Eggli and H. U. Keller, Cell Biol Znt, 2000,24,863.
58 M. M . Al-Alwan, G. Rowden, T. D. Lee and K.A. West, JZmmunol, 2001,166,1452.
59 M. K. Wild, A. Carnbiaggi, M. H. Brown, E. A. Davies, H. Ohno, T. Saito and P. A.
Van der Merwe, J Exp Med, 1999,190,3 1.
60 S. Y. Qi, J. T. Groves and A. K. Chakraborty, Proc Nut2 Acad Sci U S A, 2001, 98,
6548.
61 P. Y. Arnold, D. K. Davidian and M.D. Mannie, Eur J Zmrnunol, 1997,27,3198.
62 P. Y.Arnold and M. D. Mannie, Eur J Zmmunol,1999,29, 1363.
63 J. F. Huang, Y. Yang, H. Sepulveda, W. Shi, I. Hwang, P. A. Peterson, M. R. Jackson,
J. Sprent and 2.Cai, Science, 1999,286,952.
64 D. Hudrisier, J. Riond, H. Mazarguil, J. E. Gairin and E. Joly, JZmmunol, 2001, 166,
3645.
65 I. Hwang, J. Huang, H. Kishimoto and e. al., J Exp Med, 2000,191, 1137.

66 E. Baba, Y. Takahashi, J. Lichtenfeld, R. Tanaka, A. Yoshida, K. Sugamura, N.
Yamamoto and Y. Tanaka, J Zmmunol, 2001,167,875.
67 F. D. Batista, D. Iber and M. S. Neuberger, Nature, 2001,411,489.
68 M. Vales-Gomez, H. T.Reyburn, R. A. Erskine and J. Strominger, Proc Natl Acad Sci
U S A , 1998,95,14326.
69 M. Vales-Gomez, R. A. Erskine, M. P. Deacon, J. L. Strominger and H. T. Reyburn,
Proc Natl Acad Sci U S A, 2001,98, 1734.
70 J. J. Peschon, J. L. Slack, P. Reddy, K. L. Stocking, S. W. Sunnarborg, D. C. Lee, W.
E. Russell, B. J. Castner, R. S. Johnson, J. N. Fitzner, R. W. Boyce, N. Nelson, C. J.
Kozlosky, M. F. Wolfson, C. T. Rauch, D. P. Cerretti, R. J. Paxton, C. J. March and R.
A. Black, Science, 1998,282, 1281.
71 K. Denzer, M. J. Kleijmeer, H. F. Heijnen, W. Stoorvogel and H. J. Geuze, J Cell Sci,
2000,113,3365.
72 L. Zitvogel, A. Regnault, A. Lozier, J. Wolfers, C. Flament, D. Tenza, P. Ricciardi-
Castagnoli, G. Raposo and S. Amigorena, Nat Med, 1998,4,594.
73 G . Raposo, H. W. Nijman, W. Stoorvogel, R. Liejendekker, C. V. Harding, C. J.
Melief and H. J. Geuze, J Exp Med, 1996,183,1161.
74 W. Sibbet and J. R. Taylor, J. Lurnin, 1983,28,367.
75 S . Hirayama, Y. Iuchi, F. Tanaka and K. Shobatake, Chem Phys, 1990,144,401.
76 R. A. Lampert, S. R. Meech, J. Metcalfe, D. Phillips and A. P. Schaap, Chem Phys
Lett, 1983,94, 137.
77 S . Hirayama and D. Phillips, J Photochem, 1980,12,139.
78 M. Kneen, J. Farinas, Y. Li and A. S. Verkman, Biophys J, 1998,74,1591.
79 B. Sengupta, J. Guharay and P. K. Sengupta, Spectrochim Acta A Mol Biomol
Spectrosc, 2000,56A,1433.
80 T. Parasassi, G. Ravagnan, R. M. Rusch and E. Gratton, Photochem Photobiol, 1993,
57,403.
81 J. B. Birks, Photophysics of Aromatic Molecules, Wiley-Interscience, Chichester,
1970.
82 J. R. Lakowicz, Principles of Fluorescence Spectroscopy, 2nd edition, JSluwer
Academk/Plenum Publishers, New York, 1999.
83 A. D. Scully, A. J. MacRobert, S. Botchway, P. O'Neill, R. B. Ostler and D. Phillips,
J Fluoresc, 1996,6, 119.
84 A. D. Scully, A. J. MacRobert, S. Botchway, P. O'Neill, R. B. Ostler and D. Phillips,
Bioimaging, 1997,5,9.
85 K. Dowling, M, J. Dayel, M. J. Lever, P. M. W. French, J. D. Hares and A. K. L.
Dymoke-Bradshaw, Opt Lett, 1998,23,810.
86 K, Dowling, M. J. Dayel, S. C. W. Hyde, P. M. W. French, M. J. Lever, J. D. Hares
and A. K. L. Dymoke-Bradshaw, J Mod Optic, 1998,46,199.
87 P. I. Bastiaens and A. Squire, Trends Cell Biol, 1999,9,48.
88 S . J. Strickler and R. A. Berg, J Chem Physi, 1962,37,814.
89 D. V . O'Connor and D. Phillips, Time-correlated single photon counting, Academic
Press, London, 1984.
90 K. Suhling, D. M. Davis, P. PetrBSek, J. Siege1 and D. Phillips, Proc. S.P.Z.E., 2001,
4259,92.
91 C. Dietrich, L. A. Bagatolli, Z. N. Volovyk, N. L. Thompson, M. Levi, K. Jacobson
and E. Gratton, Biophys J, 2001,80, 1417.
PROBING THE STRUCTURE OF VIRAL ION CHANNEL PROTEINS:
A COMPUTATIONAL APPROACH.
W.B. Fischer'" and M.S.P. Sansom2
Department of Biochemistry, Oxford University, South Parks Road, Oxford OX1 3QU, UK
*Correspondence address: wolfgang.fischer @bioch.ox .ac.uk
Laboratory of Molecular Biophysics, University of Oxford, Oxford OX1 3QU, UK
1 INTRODUCTION
There are significant difficulties with the structural characterization of membrane proteins at a
molecular level. These difficulties often arise from (i) insufficient quantities of protein
available and (ii) high costs of particular chemicals (e.g. specifically labelled amino acids)
necessary to obtain spectroscopic data with high structural resolution (e.g. NMR, X-ray). It is
expected that these difficulties might be overcome in the future. In the meantime
computational methods may play an important role in the construction of reliable models for
investigations of the structure and function of this class of proteins. This role can be achieved
not only by enabling in silico based experiments which are almost impossible to obtain
experimentally but also by proposing models prior to their experimental verification.
Examples for this will be presented for the viral ion channel proteins NB from influenza B and
Vpu from HIV-1.
The genome of both enveloped viruses influenza B and HIV-1 encode for short
membrane proteins called NB (100 amino acids)' and Vpu (81 amino respectively. In
similarity with an equivalent short membrane protein from influenza A called M2, which has
been found to be involved in viral entry/exit pathway and is conducting protons, it is expected
that NB and Vpu have similar functional roles in the life cycle of the viruses. What all these
viral membrane proteins have in common is that they span the bilayer once and have to
assemble to form ion channels. For NB, ion channel activity has been confirmed for the full
length protein expressed in Escherichia coli and reconstituted into lipid bilayers: Also a
synthetic peptide analogous to the transmembrane (TM) segment of NB under similar
conditions as mentioned exhibits channel a~tivity.~Full length Vpu and the TM segment alone
both show ion conducting properties. Besides these results, 'in vivo' channel formation has
not yet been fully proven for these proteins. The cytoplasmic part of Vpu has been found to
inhibit the transport of CD4 and other receptor molecules of the infected cell to the cell
membrane.
The current knowledge of structural properties of these proteins is based on solution and
solid state NMR-, FTIR-, and CD-spectroscopy. NB's topology of the TM segment has been
established by CD-spectroscopy to be predominantly helical? Simulations have been done of
NB bundles of different size in a fully hydrated lipid bilayer system. For Vpu it is confirmed
that the TM segment consists of a helical motif?-* In addition, helical motifs have been
identified for the cytoplasmic part of the protein?." MD simulations have been performed on
pentameric bundles consisting of the TM segment of Vpu in a hydrophobic slab representing
the lipid bilayer" and in a slab of the bilayer mimetic octane.12
In this study we compare simulations on the TM segment of NB5 (Fig. 1) with those
performed on TM bundles of Vpu (Fig. 2). Both bundles each consisting of TM segments of
28 amino acids in length are embedded in a hydrated lipid bilayer. The analysis focuses on the
structural and functional role of particular amino acids.
2 METHOD
We generated bundles consisting of 4, 5 , and 6 TM segments of NB (NB, Lee, with C-38

mutated to Y):
IRGS" IIITICVSL130VILIVFGYIA4' KIFI
andVpu6-33 :
IVAIV" ALVVAIIIA12' VVWSIVIIEY3' RKI
using a SNMD protocol13 based on the program Xplori4. For NB no starting tilt angle was
used, whereas for Vpu we used an initial angle of 5" during the generation process. A detailed
description of generating the helical TM segments is given el~ewhere.'~"~ In brief, the SNMD
protocol comprises two stages. In Stage 1 the bundles were constructed with parallel, idealized
a-helices based on the positions of the Ca-atoms of the peptide unit. All other atoms of the
individual amino acids are superimposed. In Stage 2 potential energy functions with the
PARAM19 parameter set are introduced into the protocol. Partial charges on side-chain atoms
of polar side chains are gradually scaled up (from 0.05 to 0.4 times their full value) during a
temperature reduction from 500 to 300 K. We obtain 5 x 5 = 25 structures as a result of the
SNMD protocol. The most symmetric structure in respect to the pseudo five-fold symmetry
axis of the bundle was chosen for the simulations.
All bundles were placed within a lipid bilayer of 1-palmitoyl-2-oleoyl-sn-glycerol-3-

phosphatidyl-choline (POPC) and consequently solvated with water molecules (>30 water
molecules (SPC water model16) per lipid). Our simulations were done with a system of ca.
20,000 atoms.I7 Simulations were run for 1 ns using GROMACS 1.6 software
(http://rugmdO.chem.rug.nl/-gmx/gmx.html) OR a 10 processor SGI Origin 2000 machine and
a twin cutoff of 1.O and 1.7 A. For data analysis Gromacs software was used. The structures
were visualized with the program MOLSCRIPT.
A B
Fig. 1: View from the N to the C terminal end of the pentameric (A) and hexameric
(B) bundle of NB. The serine side chains are shown in black and the phenylalanines
in light grey. Lipid bilayer and the water molecules are omitted for clarity.
3 RESULTS AND DISCUSSION
All our models are built with the intention that the hydrophilic residues are facing the pore of
the water filled channel. This idea is based on experimental findin s of the nicotinic
acetylcholine receptor where serines and threonines are lining the Both NB (Fig. 1)
and Vpu (Fig. 2) contain serine residues, as well as threonines in NB only, which all point into
the pore. And in both cases this automatically locates the aromatic residues phenylalanine (in
NB) and tryptophans (in Vpu) towards the helix / lipid interface. Such a position of the
aromatic residues in agreement with findings in other channels such as porins2' and
g r a r n i ~ i d i n .In
~ ~the
- ~case
~ of NB, other hydrophilic residues like arginines (R-18)and lysines
(K-41) are located at the helix / lipid interface (data not shown). For Vpu we find a sequence
of amino acids (glu-29, tyr-30, arg-31) able to form hydrogen bonds from which the arginines
are pointing into the pore (Fig. 2 B and C).
The models adopt a stable structure shown by the leveling off of the root mean square
deviation (RMSD), which calculates the positional deviation of each of the subsequent
structures from the starting structure (data not shown). Leveling off is achieved after ca. 200
ps and reaches values below 3 nm for each of the bundles. For NB all bundles remain fairly
helical. Only single amino acids, especially in the bundle consisting of 4 segments (NB-4) and
bundle of 5 segments (NB-5), do not remain in a helical configuration.' Val-27 with Q, = -
66.9" -9.8' and Y = -28.1' f -12.7', and Ser-28 with <p = -88.9' 2 -15.5" and Y = -41.8" f -
+_
10.9" adopt a- and Y-values which deviate from helical values (@ = -60', Y = -50' 2 5 ) in one
segment of NB-4, For NB-5, residues at the N terminal end in one segment show deviation
from 'normal': Arg-18, which faces the lipid / head group interface (@ = -120.3' i--12.0°, Y =
-61.7' k -31.5') and Gly-19 (@ = -106.5' k -33.0", Y = -52.2' k -15.7'). For NB-6 the values
for all residues do not deviate from the standard values for a helix. The average kink angle
seems to decrease with increasing number of segments (Tab. 1). The average helix tilt angle
for all segments adopt values around ca. 6' which indicates the helices tilt slightly compared
Fig. 2: View from the N to the

A B C terminal end of the
pentameric (A, C) and
hexameric (B, D) bundle of
Vpu. In A and B the serines are
high lighted in black, the
tryptophans in light grey. In C
and D black denotes arginines
(arg-3l), light grey tyrosine
(tyr-30), and grey glutamic acid
(glu-29).
C D
to their starting structure with no tilt implemented. NB-6 shows the lowest average crossing
angle of ca. 4". All bundles allow for a continuous column of water molecules through their
interior with approximately 50 (NB-4), 105 (NB-5), and 145 (NB-6) molecules.
Also for Vpu-4 strong deviations are found for a serine residue (Ser-24: @ = -28.6' k -
20.4', Y = -52.0" f -12.2'). In addition residues Glu-29, Tyr-30, Arg-31, and Lys-32 adopt @-
and Y-values in two segments, which are indicative of an unraveling of the bundle at the C
terminal end (for one segment: Glu-29: Q, = -53.7' k 9.5', Y = -30.6' k 10.0"; Tyr-30: Q, = -
60.9" k 11.2', Y = -45.7" k 11.7'; Arg-31: Q, = -140.2' k 20.5", Y = 94.4" & 13.5'; Lys-32: Q,
= -47.7' f 15.9', Y = -73.8' f 18.0'). These values tend to be less dramatic for Glu-29, Tyr-
30, and Arg-31 in Vpu-5 and Vpu-6. The average kink angles are slightly higher for the Vpu
bundle (Tab.1). For one segment in Vpu-4 we find a value of ca. 32'. It is interesting to note
that the average tilt angles for Vpu-4 and Vpu-5 increase during the simulation from an initial
5" to ca. 15' after 1 ns of simulation. Vpu-6 adopts an almost identical angle than NB-6 of
around 6". The average crossing angle decreases from ca. 20" for Vpu-4 to ca. '5 for Vpu-6.
We can only fill the bundles Vpu-5 and Vpu-6 with ca. 92 and ca. 110 waters, respectively.
Vpu-4 does not show any space for water molecules within the assembled helices. Also during
the simulation no water can access the pore.
As a result for both NB and Vpu the segments seem to straighten with increasing
number of segments forming the bundle. During the simulation both NB and Vpu tetrameric
Tab. 1: Kink, tilt, and crossing angles averaged over all the segments in each of the bundles.
Structural
Data
NB-4 1 NB-5 1 NB-6
Kink
Angle (") *
20.6 6.8 13.8 5.9 8.8 f: 4.4 16.9 19.9 f 7.8 12.7 7.7
f 10.0
Tilt
Angle (") 4.8 & 1.3 9.2 A 2.8 6.2 f 1.8 15.3 * 5.0 14.5 f 3.4 6.0 f 2.1
Crossing
Angle (") 6.5 f 2.6 6.6 f 4.5 3.9 k 2.0 20.2 f 7.5 16.7 f 2.2 4.6 f 2.3
and pentameric bundles increase their tilt angles by about 7" - 8". For both of NB's and Vpu's
hexameric bundles we obtain identical tilt angles of ca. 6" at the end of the simulations. The
major sequential differences in both bundles are (i) an increase of p-branched residues from
NB (14 residues) to Vpu (17 residues), and (ii) a decrease of hydrophilic residues from 3 (2
serines and one threonine) in NB to one (serine) in Vpu (1). Thus, these differences do not
affect the helix tilt angle of the hexamer. The hydrophobic parts of the tetra- and pentameric
bundles of Vpu with only one serine allow for the segments to slide around each other
adopting higher average crossing angles than their NB counterparts with two serines per
segment. It is striking that in both cases for the tetrameric bundle the serines deviate from a
regular helical conformation. For Vpu the C terminal end with Glu, Tyr, Arg (EYR-motif)
seems to be highly fragile in all bundles in remaining in a helical conformation during the
simulation.
4 CONCLUSION
Our simulations indicate that major structural differences between bundles of TM segments
are most pronounced if less than 6 segments form the bundle. The reason for this lies in the
amino acids used for the TM segments and as a consequence the resulting non-covalent
interactions. Hydrophilic residues seem to adopt and support to a lesser degree proper helical
conformation. Especially if located at the end of the helices this causes severe deviation from
helicity. However, this might reflect the macroscopic behavior in as much as it might form a
barrier for the formation of bundles with less than 5 or 6 segments, especially in the case of
Vpu. We can expect that the specific properties of an ion channel forming protein can be
effective if the bundle size is due to 5 or less segments. Bundles of 6 segments seem to be less
sensitive to the amino acid sequence and initial starting structure.
5 ACKNOWLEDGMENT
We acknowledge helpful and stimulating discussion with G. Smith, L. Forrest and P.

Tieleman. WBF thanks the EC for a research fellow ship (TMR).
References:
M.W. Shaw, R.A. Lamb, B. W. Erickson, D.J. Briedis, P.W. Choppin, Proc. Natl. Acad.
Sci. USA, 1982,79,68 17.
K. Strebel, T. Klimkait, M.A. Martin, Science, 1988, 241, 1221.
E.A. Cohen, E.F. Terwilliger, J.G. Sodroski, W.A. Haseltine, Nature, 1988,334,532.
N.A. Sunstrom, L.S. Prekumar, A. Prekumar, G. Ewart, G.B. Cox, P.W. Gage, J. Membr.
Biol., 1996, 150, 127.
W.B. Fischer, M. Pitkeathly, B.A. Wallace, L.R. Forrest, G.R. Smith, M.S.P. Sansom,
Biochemistry, 2000,39, 12708.
V. Wray, T. Federau, P. Henklein, S. Klabunde, 0. Kunert, D. Schomburg, U. Schubert,
In?. J. Peptide Protein Res., 1995, 45, 35.
V. Wray, R. Kinder, T. Federau, P. Henklein, B. Bechinger, U. Schubert, Biochemistry,
1999,38,5272.
A. Kukol, I.T. Arkin, Biophys. J., 1999,77, 1594.
T. Federau, U. Schubert, J. Flofldorf, P. Henklein, D. Schomburg, V. Wray, In?. J. Peptide
Protein Res., 1996,47, 297.
10 D. Willbold, S. Hoffmann, P. Rosch, Eur. J. Biochem., 1997,245, 581.
11 A.L. Grice, I.D. Kerr, M.S.P. Sansom, FEBS Lett., 1997,405,299.
12 P.B. Moore, Q. Zhong, T.Husslein, M.L. Klein, FEBS Lett., 1998,431, 143.
13 I.D. Kerr, R. Sankararamakrishnan, O.S. Smart, M.S.P. Sansom, Biophys. J., 1994, 67,
1501.
14 A.T. Briinger, X-PLOR Version 3.1. A System for X-ray Crystallography and NMR; Yale
University Press: New Haven, Ct., 1992.
15 F. Cordes, A. Kukol, L.R. Forrest, I.T. Arkin, M.S.P. Sansom, W.B. Fischer, Biochim.
Biophys. Acta, 200 1,1512,29 1.
16 H.J.C. Berendsen, J.R. Grigera, T.P. Straatsma, J. Phys. Chem., 1987,91,6269.
17 Fischer, W. B.; Forrest, L. R.; Smith, G. R.; Sansom, M.S. P. Biopolymers 2000,53,529
18 J.P. Changeux, J.I. Galzi, A. Devillers-ThiCry, D. Bertrand, Quart. Rev. Biophys., 1992,
25, 395.
19 H. Lester, Ann. Rev. Biophys. Biomol. Struct., 1992, 21, 267.
20 F. Hucho, V.I. Tsetlin, J. Machold, Eur. J. Biochem., 1996,239,539.
21 A. Hirsch, T. Wacker, J. Weckesser, K. Diederichs, W. Wolfram, Proteins: Struc. Func.
Genet., 1995,23,282.
22 A.M. O'Connell, R.E. Koeppe 11, O.S. Andersen, Science, 1990,250, 1256.
23 N.D. Lazo, W. Hu, T.A. Cross, J. Chem. SOC., Chem. Commun., 1992, 1529.
24 W. Hu, N.D. Lazo, T.A. Cross, Biochemistry, 1995,34, 14138.
25 C. Branden, J. Tooze, Introduction to Protein Structure. Garland, New York., 1991.
THE IMPACT OF H202 ON THE STRUCTURE OF CATALASES BY MOLECULAR
MODELLING METHODS
Susana G. Kalko, Josd Ll. Gelpi, and Modesto Orozco
Departament de Bioquimica i Biologia Molecular, Universitat de Barcelona. C/ Marti i

Franques 1. Barcelona 08028. Spain
1 INTRODUCTION
Catalases are responsible for the reduction of reactive oxygen species such as small
peroxides (including hydrogen peroxide), which are dismutated into water (or alcohol)
and oxygen. The speed of this reaction is very high, reaching in some cases the
diffusion limit. Malfunction of catalases may lead to severe effects, amon! them
increased susceptibility to thermal injury,' inflammation2and accelerated ageing.
Crystallographic studies show that the active site of catalase is deeply buried in the
interior of the protein, at the end of a long and narrow channel connecting the exterior
surface with the heme groupd4Other smaller channels were suggested as possible
product release pathways>5 but no clear experimental evidence has been provided to
c o n f m their existence in active catalases.
In a previous study' the substrate recognition mechanisms of peroxisomal catalase
fiom Saccharomyces cerevisiae were explained by using crystallographic and
theoretical data. Our calculations suggested that water was a competitive inhibitor of
the enzyme, blocking the access of hydrogen peroxide to the active site. Furthermore,
theoretical calculations gave support to the putative role of secondary channels6 like
those suggested by Fita and coworkers based on crystallographic data!*'
Catalases are designed by evolution to work in environments having a very large
concentration of H202, such as those existing at the peroxisomes. The impact of the
large concentration of hydrogen peroxide on the structure and reactive properties of
catalases is unclear due to difficulties in obtaining crystals in such environments.
However, this effect might be crucial for the physiological role of the enzyme. In this
paper we will use state ofthe art molecular dynamics simulations to gain insight into
the impact of a high concentration of H202 on catalase structure and ligand binding
properties.
2 METHODS AND RESULTS
2.1 Molecular Dynamics
Simulations in pure water and a hydrogen peroxide/water mixture were performed

starting fiom the crystallographic structure of the Saccharomyces cerevisiae
peroxisomal catalase solved at 2.8 A resolution: The active form is an homotetramer

but due to the size of the system a reduced model was used in our simulations (see ref. 6
for details). The optimised structure in water was used to generate a starting model for
the protein in a c.a. 27% mixture of hydrogen peroxide and water. AMBER-957 and
TIP3P8 force-fields were used to describe protein atoms and water, respectively.
Bonded parameters for hydrogen peroxide were determined using the PAPQMD
strate$ and HF/6-3 1G(d) calculations as reference. Charges were determined using
RESP and HF/6-3 lG(d) wave functions.
2.1.1 Analysis of the root mean square deviation (Rusrc))). The all atoms
RMSDs between the flexible part of the protein and the crystal structure were small in
both trajectories, around 1.0 A in pure water and 1.5 A in the mixee. The greater
RMSD for the simulation in the mixture agrees with the fact that the crystal
environment is closer to pure water than to the 27% H202/H20mixture. The all atoms
RMSD between the MD-averaged structure and the trajectory along a simulation can be
used to obtain a measure of the flexibility of the protein in both solvents. This analysis
was extended (see Table 1) by inspection of the fluctuations of four regions of special
interest: gl, the internal part of the main channel (A66-A73, A109-A126 and A148-
A156 residues); g2, the external part of that channel (Al57-Al89 residues); 83, a
channel connecting the heme with the interface between monomers (residues D59-
D63); and 84, a channel connecting the heme with the N A D P O binding position
(residues A142-A147).6 It becomes evident that g2 is the most sensitive region to the
solvent composition, while g4 has a very stable conformation, either in pure water or in
the H20dH20 mixture.
Table 1 RMSD between the MD-averaged structures and the corresponding pure
water and mixture trajectories. Standard deviations in parenthesis.
refirence gUA g2/A g3/A g4/A total/A
Pure 0.7(0.1) l.l(O.1) 1.2(0.3) OS(0.3) l.O(O.1)

water
Mixture l.O(O.1) 1.7(0.3) l.O(O.2) 0.5(0.1) 1.2(0.1)
The all atom RMSD between the average structure in pure water and the
trajectory of the H202/H20 mixture provides us with a measure of the influence of
solvent composition on the structure (see Figure 1). Moderate structural deviations are
found during this trajectory, the largest of them located at the g2 region. The g4 region
shows the smallest RMSD values, but a noticeable transition occurs around 1 ns due to
the "flipping" of the ASN143 side chain in the H202 solution. It is worth noting that this
residue was suggested to participate in a "back-door" mechanism of this enzyme? No
other major changes were found in the structure related to the change of solvent
indicating that the main structure of catalase is insensitive to changes in the
concentration of H202(within the range that occurs in the peroxisome).
n
s
I I I I
I
4
v
0
0 1000 8000 woo 4000 6000
&nulation t h e (pa)
Figure 1 RMSD plot between MD trajectory of mixture sirnulation and M D

average structure of pure water simulation in regions g l (black), g2
(red), g3 (green) and g4 (blue). RMSD for total mobile residues (data
not shown) are similar to gl values.
2.1.2 Analysis of secondary structure. The secondary structure of the protein is

highly conserved irrespective of the composition of the solvent (see Fig. 2). A detailed
analysis of dihedral angles in the g l and g2 regions of the main chain indicates a
slightly larger flexibility (around 20-30% larger) in the simulation in the mixture
solution compared with the simulation in pure water. This finding agrees with the
RMSD analysis noted above, suggesting these regions as the most sensitive to changes
in solvent composition.
Figure2 Secondary structure representation for M D averaged structure of the

main channel for mixture simulation (red) and pure water simulation
(blue). Relevant residues and heme group are explicitly represented with
balls and sticks.
2.1.3 Analysis of kzy distances in the mouth of the main channel. The principal
effect of the presence of H2Oz in terms of heme accessibility is a slight widening of the
main channel. Thus, residues VAL1 11 and PHE149 defme the narrowest point of the
main channel in water (shortest heavy atom-heavy atom distance around 6 A). On the
contrary, in the HzOz/H20 mixture the side chain of PHE149 is more flexible, showing a
Cflip-flop transition, which opens the channel at this point. Therefore, PRO124 and
PHE159 define then the narrowest point of the channel (shortest heavy atom-heavy
atom distance around 6.8 A). In summary, the accessibility to the heme site is higher in
the presence of a large concentration of peroxide in the solution. This fact might
increase the catalytic efficiency of the enzyme under conditions of high hydrogen
peroxide concentration.
2.1.4 Analysis of protein-protein hydrogen b o d in channels. VAL111 has a

key role in defining the channel accessibility, as demonstrated by site-directed
mutagenesis," which makes it interestingto study the pattern of hydrogen bonds (HB).
The occupancy degree of HB for this residue, together with those involving the essential
HIS70 and its close neighbour SER109 are shown in Table 2. The stability of these
important protein-protein hydrogen bonds is very large in the case of pure water
simulation, whereas in the mixture solvent the stability is reduced. This fact may be
related to a higher number of hydrogen bonds between protein olar atoms and the
hydrogen peroxide molecule, as suggested in previous calculations.?
Table 2 Occupancy degree of HB during the simulation.
Partners Pure water (99) Mixture PA)
OVALlll-
NH HIS70 96.2 77.4
NHvAL111-
0 PRO124 97.3 92.6
* 0 SER109-
N GLY126 99.4 86.2
* OG
SERlO9- 98.4 0
NDl HIS70
* HB in crystal structure
2.1.5 Analysis of preferential solvation. The total and relative number of
solvent molecules at less than 3 A distance of the protein atoms were calculated along
the trajectory and are shown in Table 3. The quotient between the solvation number
(SN) and the total number of solvent molecules was calculated for all residues -fist raw
of Table 3- whereas for residues classified as charged, polar, aromatic, exposed12 or
buried,12 the relative value was referred to the SN of all residues. The analysis show
that in all the cases hydrogen peroxide is more closely associated to the protein than is
water, suggesting that H202 is a better solvent than water for catalase. Interestingly,
there are not differences between the solvation of water or hydrogen peroxide in
exposed residues, but on the contrary, important differences are found between solvent-
protein interactions for buried residues, since these residues are more accessible to
hydrogen peroxide than to water. Finally, it is worth noting that water has
approximately the same level of contact to residues in both simulations, indicating that
the intrinsic ability of water to interact with different types of residues does not vary in
the presence of hydrogen peroxide.
Table 3 Average number of molecules in the first solvation shells (SN). Error limits
correspond to standard deviations fiom average values. Bold numbers are
described in the text.
WATER WATER PEROmDE

Residues in PURE in in
WATER MIXTURE: MLXTUrn
(Total:3754) (Total:2096) (Total:560)
ALL 854f 18 503 f 17 179 f 8

0.23 0.24 0.32
........................................,,........,,...,............,..,..,,..............................................
CHARGED 437f 10 263 f 11 98f6
0.51 0.52 0.54
POLAR 320 f 9 189 f 9 75 f 6

0.38 0.38 0.42
AROMATIC 139f7 84*7 32 f 5

0.16 0.17 o.ia
EXPOSED 778f 17 465* 16 167 f 9
0.91 0.92 0.92
BURIED 111 f 5 63f4 34k3

0.13 0.13 0.19
2.2 Classical Molecular Interaction Potential (CMIP)
CMIP ~alculations'~ were carried out using a neutral TIP3P80 as a probe, and the MD-
averaged structures in pure and mixture solution. The grid was positioned in the centre
of the main channel, and a spacing of 1 A was used. This type of calculation allowed us
to obtain a picture of the variation of the binding site accessibility in presence of large
concentration of hydrogen peroxide. CMIP plots in Fig. 3 clearly define the shape of the
main channel. Remarkably, in the structure derived fiom mixture simulation there is a
continuous spine of solvation, whereas in the pure water case the spine is discontinuous,
at the VAL1 1 1 position. A branching of the main channel is observable in the mixture
simulation, going fiom the heme site to the NADP(H) binding site, supporting previous
hypotheses on the role of the secondary channel for the release of reaction products?*'
Figure3 CMP plots corresponding to 0 Kcal/mol interaction energy between a

neutral TIP3P 0 and MD average structure of pure aqueous solution (le$)
and mixture solution (right) around the active site of cataiase.
3 CONCLUSIONS
The structure of catalase is very similar in the crystal, in pure aqueous solution, and in
the presence of c.a. 27% of hydrogen peroxide. The stability of the structure in very
different environments, including those with a very large concentration of H202 should
be important fiom a biological point of view, since it allows the enzyme to be active
under these conditions. The enzyme shows a significant widening of its channels in the
presence of peroxide. This suggests a solvent-induced conformational change, which
might increase solute accessibility to the active site under conditions of large oxidative
stress. In summary, the results presented here indicate that hydrogen peroxide does not
lead to dramatic effects on catalase structure, but its presence in the medium promotes
conformational changes of the enzyme for an easier arrival of the ligand to the active
site and the posterior clearance of the reaction products.
References
J.A. Leff, Inflammation, 1993,1?, 199.

B. Halliwell and 0.1. Aruoma, FEBS Lett.,1991,281,9.
J. Taub, J.F. Lau, C. Ma,J.H. Hahn,R. Hoque, J. Rothblatt and M. Chalfie, Nature,
1999,399,162.
M.J. Mat&,M. Zamocky, L.M.Nykyri, C.Henog, P.M. A M , C. Betzel, F. Koller
and I. Fita, J MoZ.Bioi., 1999,286, 135.
M.S.Servinc, M.J.Mat& J. Switala, I. Fita and P. Loewen, Protein Sci., 1999, 8,
490.
S. G. Kalko, J. L1. Gelpf, I. Fita and M. Orozco, J.Am.Chem.Soc.,2001,123,9665.
W.D. Cornell, P. Cieplak, C.I. Bayly, 1.R Gould, K. Merz, D.M.Ferguson, D.C.
Spellmeyer, T. Fox, J.W. Caldwell and PA. Kollman, J.Am.Chem.Soc.,1995, 117,
11946.
W.L. Jorgensen, J. Chandrasekhar, J. Madura, R.W. Impey, and M.L Klein,
JChemPhys, 1983,79,926.
C . Ale&, E. I. Canela, R. Franco and M.Orozco, JComp.Chem.,1991,12,664.
10 C.I. Bayly, P. Cieplak, W.DCornell and P.A. Kollman, J.Phys.Chem., 1993, 97,
10269.
11 M. Zamocky, C. Herzog, L.MNylcyri, and F. Koller, FEBS Let?.,1995,367,241.
12 S.J. Hubbard, S.F. Campbell and J. Thornton, J. Mol. Biol., 1991,220,507.
13 J.Ll. Gelpi, S.G. Kalko, X. Barril, J. Cirera, X. de la Cruz, F.J. Luque and M.
Orozco, Proteins, 2001,45,428.
ENTROPY IN THE ALIGNMENT AND DIMERIZATION OF CLASS C G-PROTEN
COUPLED RECEPTORS
Mark K Dean', Christopher Higgs', Richard E Smith', Paul D Scott2, Robert P Bywater3,
Trevor J Howe4 and Christopher A Reynolds'
'Department of Biological Sciences, Central Campus, University of Essex, Wivenhoe Park, Col-
Chester, Essex, C04 3SQ, United Kingdom
Department of Computer Science, University of Essex, Wivenhoe Park, Colchester, Essex,
C 0 4 3SQ, United Kingdom
Biostructure Department, Novo Nordisk MS, Novo Nordisk Park DK-2760 Miilsv, Denmark,
Novartis, Horsham Research Centre, Wimblehurst Road, Horsham, West Sussex, RH 12 5AE3,
United Kingdom
1 INTRODUCTION
Class C G-protein coupled receptors (GPCRs), which include the metabotropic glutamate,
calcium sensing and GABAB receptors, have played an important role in elucidating the
link between G-protein coupled receptor dimerization and function. Here we present a se-
quence-based study on the alignment and dimerizationof class C GPCRs.
The observation of disulfide-linked metabotropic glutamate receptor dimers gave rise
to one of the earlier reports of GPCR dimers'; this was soon followed by a report of disul-
fide-linked calcium-sensing receptors2. The most significant link between G-protein cou-
pled receptor (GPCR) dimerization and function however, came with the observation that
GABAB receptor subtypes R1 and R2 were not hctional unless co-expre~sed~~. It was
suggested that the GABABdimer was held together by coiled-coil interactions between the
C-termini6; these articles3 also introduced the idea of heterodimerization. Structural evi-
dence for class C receptor dmers has been provided by crystal structures of the N-terminal
domain of the metabotropic glutamate receptor, in both closed and open dimeric forms
(with and without agonist), implying that the agonist induced activation involves large
relative movement between the two parts of the N-terminal domain'; we suggest that the
h c t i o n of this movement is to bring the transmembrane domains together.
Within the class A GPCRs, Devi has shown that heterodimerization of the K and p
opioid receptors could give rise to novel pharmacology not seen in the K or p systems
alone' and these ideas have also arisen in other articles describing GPCR heterodirneriza-
'
tion'-' . However, the current interest in G-protein coupled receptor dimerization probably
arose fiom the work of Maggio et al. Maggio constructed chimeric muscarinic-adrenergic
receptors in which the N-tennius through to intracellular loop 3 was taken fiom the mus-
carinic M3 receptor and helix 6 through to the C-terminus was taken fkom the a 2 -
adrenergic receptor. The receptors were inactive: they did not bind ligand or activate the
G-protein. The alternative adrenergic-mwarinic chimeras were similarly inactive but
when co-expressed, the chimeras bound both adrenergic and muscarinic ligands and acti-
vated the G-protein. This hctional rescue on coexpression was interpreted as “cross-talk”
by the authod2. Equally important was the work of Hebert et al. who showed using CO-
immunoprecipitation that transmembrane helix 6 of the P2-adrenergic receptor inhibited
both dimerization and a~tivation’~, implying that the dimerization is driven by interactions
between the transmembrane domains. Our theoretical work on receptor dimerization has
provided an interpretation of Maggio’s results in terms of domain swapping, with helices 5
and 6 nominally forming the receptor dimer interfa~e’~’’~.
Studies using fluorescence resonance energy transfer (FRET) and bioluminescence
resonance energy transfer (BRET) have given further evidence for GPCR di~nerization’~~”,
but generally, the link between dimerization and h c t i o n is not understood. We note that
parallel dose-response curves for signallin and ligand-induced dimerization have been
observed in some case^'^?'^ but not in However, recent studies have shown that
heterodmerization can affect the G-protein coupling preferences of the receptors2*and this
provides evidence that the GPCR dimer is relevant to activation.
Approaches to predicting protein-protein interactions fiom sequence are based rimar-
ily on correlated mutation analysis2’ or the evolutionary trace (ET) method22-2! Both
methods are essentially data-mining approaches for determining highly correlated patterns
of change within a multiple sequence ali nment and both have been usehlly applied to
GPCRs and their associated G-proteins’”l.”. Here we apply a related entropy-based
method to the problem of G-protein coupled receptor dimerization in the class C receptors.
Initially, we have used the entropy analysis to assist in the alignment of class C sequences
with the class A opsin receptor sequences; this enabled the entropy analysis to be mapped
onto the rhodopsin X-ray crystal structure. As a control, we have also studied the protein
interface and substrate binding site in the ras-rasgap X-ray crystal structure.
2 METHODS
2.1 Determination of functionally important residues
The basic assumptions of the ET method, as applied to a multiple sequence alignment of a

protein family are:
0 the f d y retains its fold throughout the series of proteins
the location of the functional sites is conserved within the family

these hctional sites have lower mutation rates than the rest of the protein
0 the lower mutation rate is punctuated by mutation events that cause divergence.
Clearly these assumptions will not apply rigorously because certain structure-stabbing
residues may also have low mutation rates while certain binding sites may display conver-
gent evolution. Similarly, certain family members may display subtle variations in the fold.
Nevertheless, they do provide a working hypothesis. In practical implementations of the
ET method, conserved-in-class residues are identified for positions associated with succes-
sively more branched regions of the dendrogram (derived fiom the multiple sequence
alignment) and plotted onto a structure until the residues no longer cluster but rather are
scattered in a seemingly random manner over the protein surface.
These ET assumptions apply equally well to the entropy analysis used here, where the
entropy (in an information sense) at a given position j is given by equation (1)
where N is the number of sequences (24 for class C, 112 for rhodopsin), P is the number of
amino acid residues (21, including gaps), V is the variability and Mi is the number of
amino acids of type i at position j. Here the entropy was normalized so that the maximum
value was 100. To determine hctional sites, the entropy values were sorted into 20 bins
of increasing relative entropy and residues from successive bins were plotted until the resi-
dues no longer appeared to cluster, as in previous applications of the ET method. There are
two conceptual advantages of this method over the ET method. Firstly, a dendrogram is not
required and so discussions on the most appropriate tree (e.g. neighbour UP-
GMA33)are not relevant. Secondly, while a sequencing error prevents a residue from being
identified as conserved-in-class in the ET method, the error merely increases the entropy
but does not necessarily prevent the residue fiom being identified in a higher bin.
2.2 Alignment of class A and class C receptor sequences
The class A GPCRs can be readily aligned34935 and for this family, the X-ray structure of
rhodopsin provides a suitable structural However, the similarity between class
A receptors and those of class B and class C is too low to obtain a multiple sequence align-
ment using conventional methods. Elsewhere an alignment between the class A and class B
receptors has been determined by assuming that the positions of the functionally important
residues are conserved3* even if their identity is not conserved; the positions of the
hnctionally important residues were identified as having a conservation level of 70% or
more. The alignment then reduced to finding the alignment with the maximum coincidence
of hctionally important residues subject to the constraint of aligning the internal residues,
as determined using PERSCAN39. Here we have followed a similar procedure but have
used a relative entropy cut-off of 30% or less to identify the functionally important
residues. The GABAB sequences were aligned to the other class C sequences using essen-
tially the same method.
3.1 The rhodopsin-class C alignment
The entropy-based alignment of the 7 transmembrane helices of the rhodopsin family and
the class C M y are shown in Table 1. Several conserved motifs of the class A receptors
are picked out by the entropy analysis, including the GN on helix 1, the DRY on helix 3,
the CWXXP on TM6, the charged residues at the intracellular end of helix 6 and the
NPXXY motif of helix 7. In each case apart fiom helix 7,at least one of the corresponding
positions in the class C receptor family is identified by the entropy analysis. There is a rea-
sonable degree of overlap between the low entropy residues of the rhodopsin f d y and
the low entropy residues of the class C receptors, and these matches are denoted by an X in
Table 1.
The extent to which we Table 1. The alignment of class A opsin family

should expect the low entropy with class Cfamily GPCRs. The alignment of low
residue positions to align depends entropy residue positions within the transmem-
on the similarity between the brane (TM) regions 1-7, denoted in bold, e.g. A,
structwe h c t ion relationships, is indicated by “X”. Identities, high similarities
and this similarity essentially and medium similarities and predicted internal
arises fiom the common seven residues are denoted (‘1 ”, “:”, “. and by under-
transmembrane helical structure linin,
and a common G-protein coupling TM1 AVL3+&YSLLFLELL--YILXILR
mechanism. However, there are I::.: ... l . . l : 1.x : x:
AIAPVFFACLGILATLFVIVTEVRYNDT
notable differences that will be TM2 RTP_TNYELLN&AVm-LLEALT&PPEAL
reflected in low entropy residues :. x. .:xx .: : l X ... I
occurring in different regions of ASGRELSYILLTGIFLCYCITFLMIAK
the structure. Thus, for the class C TM3 ALCKLVSFLDVLNGTASIFSLTAIAIDRYLAICHP
receptors, the ligand binding site I:X I .. . I . : I X x xx .x. xx
is within the N-terminus’ but for TM4
the class A opsin receptors it re- x::: . x. : X I I . : . . I . .
sides with the helical ASQLVITFSLISVQLLGWIWLWE
Moreover, the role of the closed TM5
N-terminus structure is probably . I I 1 ... . l : l : IX. x.: x
S DLSLICSLGYSGLLMVTCTFKTRGVP
to favour dimerization’ but other TM6 K A E R l y ~ - v I I J G v g L L ~ P ~ I V g L L
regions, e.g. helix 6, may fulfil I . xx. xx I1xx:x:.x:..x. I I
this role within the class A recep- KFIGETMPTTCI IWLAFI PIFFGTAQSAEK
t o r ~ ’ ~While
. these variations on TM7 - FLIEWLAALN&LNNIIEAF*
the basic theme could underlie
some differences in the co-
location of low entropy residues,
the optimal match of low entropy residue positions does give rise to an alignment, as
shown in table 1. In addition to the alignment of low entropy residue positions, the few
identities, high similarities and medium similarities are denoted by “I”, “:” and “.” respec-
tively. Moreover, the optimal entropy alignment is consistent with the alignment of internal
residues as predicted by PERSCAN. These internal residues do not filly coincide with the
Figure 2. Conserved and low entropy residues (Fey) plotted onto the structure of ras.
(a) The interface with rasgap, (3) The binding site for GTPpY.
internal residues in the rhodopsin crystal structure but are used in preference to the crystal
structure data for two reasons. Firstly, no crystal structure is available for the transmem-
brane domain of the class C receptors and secondly, the rhodopsin crystal structure is for
the inactive structure and while the class C inactive structure may differ slightly, the PER-
SCAN results are equally applicable to both the active and inactive structures. Thus, the
alignment enables a model of the transmembrane domain of the class C receptor to be con-
structed and the entropy analysis to be applied to the aligned class C receptor sequences.
3.2 Functional sites predicted by entropy analysis: ras
The entropy analysis has been applied to a seed alignment, obtained from the P f m data-
base4' (accession number PF00071) of 61 sequences of the ras family. Residues @omthe
first 9/10 entropy bins, i.e. up to a relative entropy of 45/50%, were plotted onto the struc-
ture of the ras protein (a small GTPase, taken fiom the ras-rasgap complex4', pdb code
1WQl), as shown in figure 1. Figure l a shows that the residues in the vicinity of the inter-
face with rasgap, a protein that modifies the catalytic properties of ras, are clearly identi-
fied by the entropy analysis. In addition, the GDP binding site is also identified, as shown
in figure lb. Although figure l b also shows residues from the first 9/10 bins, most of the
residues in contact with GDP identified by the entropy were identified in the first 4 bins.
The observation that substrate binding sites are identified at a Iower entropy than protein
interfaces appears to be a common trend across many proteins42.The observation that en-
tropy analysis works well for the protein-protein interface between ras and rasgap implies
that it could also be used to identi.@putative protein interfaces in the class C receptors.
3.3 Functional sites predicted by entropy analysis: class C GPCRs
Figure 2a shows the low entropy

residues plotted onto the external
face of transmembrane helices 5 and
6. Studies on the class A rece tors,
P
particularly those of MaggioI2. and
H e t ~ r t but
' ~ also those of Ng44have
shown that the transmembrane heli-
ces play a crucial role in receptor
dimerization, and this has been su -
ported by computational studiesI5; 4'.
These results are in marked contrast
to the initial studies on class C re-
ceptor dimerization. Firstly, Romano
et al showed that the glutamate re-
ceptor dimer formed through disul-
fide bonds between the N-termini';
Figure 2. (a) Low entropy residuesfor the similar results were obtained for the
class C receptorsplotted onto the 3 0 struc- calcium sensing recepto?. These
ture of rhodopsin; helices 5 and 4form the results were M e r clarified by the
centre of theBgure, with helix 7 to the right. crystallization of the glutamate N-
(6) A sfor (a), but helices 2 and 3form the terminus dimer in both an open and
centre of the figure. a closed form? Secondly, the initial
studies on the GABAB heterodimer6, involving yeast two hybrid studies, suggested that it
is held together by coiled coil interactions between the C-termini. Subsequent studies have
indicated that the coiled coil interaction is an endoplasmic reticulum retention motif and
that removal of the coiled coil does not prevent dimer Moreover, modelling
studies have shown that the C-terminus of the GABABreceptors is sdficiently long for the
transmembrane domains to form a 5,6-dimer simultaneously with a coiled coil interac-
tion27,46. Here in figure 2a we have shown that the class C receptors display the appropriate
hctionality on helices 5 and 6, as predicted by entropy analysis, to dimerize through their
transmembrane domains. Naturally, the prediction of a functional site on helices 5 and 6
does not ident@ the interacting partners. Thus, while it seems reasonable to suggest that
these signifj, the regions involved in homodimerjzation, the same regions are also involved
in heteroditr~erization~~ but a more in depth analysis would be required to identfi which
heterodimer partners would be permitted by particular sequences.
Elsewhere we have performed an evolutionary trace analysis and reported similar re-
s u l t ~ However,
~~. the entropy analysis shows a stronger signal on helix 7, Visible in figure
2a. This region displays some interesting structural features in the Vicinity of the conserved
class A NPXXY motif (which is not present in class C receptors). Weinstein has shown
from NMR studies on peptides, fiom molecular modelling and from structural database
searches that NP does not readily occur in a-helical regions but that it does have a strong
tendency to form p - t ~ r n s ~ *The > ~ ~2D
',
. and 3D structural data on the other hand is un-
equivocal that this region of helix 7 is relatively strai t in the inactive receptor (at least in
comparison to Weinstein's proposed structure^)^^^^'^^ -52 though the X-ray data shows that
this key re ion has a 3lo-helical structure rather than an a-helical structure, as predicted by
Gouldson' f. (Distortions elsewhere in TM7 were earlier predicted by Pappin and
FindlaJ3.) These conflicting observations would be resolved if the activation process in-
volved conformational changes, and indeed the importance of residues in this region may
underlie the cluster of low entropy residues on helix 7 shown in figure 2a.
The entropy analysis identifies a bctional site on helices 2 and 3, as shown in Sgure
2b. A wide range of effects resulting kom mutating external residues on helices 2 and 3
have been recorded, as listed in reference25but generally the function, if any, of these ex-
ternal residues is unknown for both the class A and class C receptors. Prelirmnary evidence
however suggests that residues in this region may be involved in binding R A M P s ~ ~ > ~ ~ .
It is interesting to compare the functional site predicted by entropy analysis, the evolu-
tionary trace method and by correlated mutation analysis. Correlated mutation analysis has
been extremely useful for identwig specific hctional residues, particularly those in-
volved in selective binding, for both ligand binding and subtype specific dimeriza-
4,15,27,29,56,57
. However, in our hands the evolutionary trace method is superior for iden-
t w i g the spatial extent of the binding site2728.The entropy analysis generally gives simi-
lar results to the evolutionary trace method42.Here, entropy gives excellent results for Ras
(figure 1) but the uncertain nature of biological sequence data dictates that this may not
always be the case. Thus, while entropy analysis identifies a putative dimerization interface
on helices 5 and 6 of the class C receptors, the evolutionary trace method appears to give
superior results25. We note that the evolutionary trace and entropy d y s e s give slightly
different information, thus while the dimerization interface is not identified so clearly by
entropy analysis, residues on helix 7 are identified more readily.
4 CONCLUSIONS
We have presented several applications of entropy analysis for identifying fimctional resi-
dues fiom multiple sequence alignments. Entropy analysis is related to correlated mutation
analysis and the evolutionary trace method but has the advantages of not requiring a den-
drogram and not requiring residues to be 100% conserved-in-class. The entropy measures
conservation, so it can be used to create alignments between sequence families that share
low levels of similarity by aligning positions of low entropy. Here the assumption is that
while identity of amino acids is not conserved, the functional positions are conserved. Us-
ing this method we have aligned the class A and class C GPCRs. The method has been
applied to a seed alignment of the ras f8mily. The GTP bindmg site is identitied at rela-
tively low entropy and the rasgap binding site has been identified at medium entropy val-
ues. The method also identifies potential functionality amongst the external resides of the
class C GPCRs. A putative dimerization site is identfied on helices 5 and 6, but not as
clearly as with the evolutionary trace method. This dimerization site appears to extend onto
helix 7, and this extension may be related to requirements for flexibility in this region. Ad-
ditional potential functionality has been observed on helices 2 and 3. In the class B recep-
tors this site may be associated with RAMP (Receptor Activity Modif)ing Protein) binding
but this region has no known h c t i o n in class C receptors.
Acknowledgements. We wish to acknowledge Novo Nordisk (CH), Novartis (RES) and

the BBSRC (MKD, RES) for funding.
References
1 C. Romano, Yang,W. L., O'Malley, K. L. J Biol. Chem. 1996,271,28612.

2 M . Bai, Trivedi, S., Brown, E. M. J. Biol. Chem. 1998,273,23605.
3 H. Mohler, Fritschy, J. M. Trends Pharmacol. Sci. 1999,20, 87.
4 K. Kaupmann, Malitschek, B., Schuler, V., Heid, J., Froestl, W., Beck, P., Mosbacher,
J., Bischoff, S., Kulik, A., Shigemoto, R., Karschin, A., Bettler, B. Nature 1998,396,
683.
5 K . A. Jones, Borowsky, B., Tamm, J. A., Craig, D. A., Durkin, M. M., Dai, M., Yao,
W. J., Johnson, M., Gunwaldsen, C., H u g , L. Y., Tang, C., Shen, Q., Salon, J. A.,
Morse, K., Laz, T., Smith, K. E., Nagarathnam, D., Noble, S. A., Branchek, T. A., Ge-
rald, C. Nature 1998,396,674.
6 J. H. White; Wise, A.; Main, M. J.; Green, A.; Fraser, N. J.; Disney, G. H.; Barnes, A.
A.; Emson, P.; Foord, S. M.; Marshall, F. H. Nature 1998,396,679.
7 N. Kunishima, Simada, Y.; Tsuji, Y.; Sato, T.; Yamamoto, M.; Kumasaka, T.; Naka-
nishi, S.; Jingami, H.; Morikawa, K. Nature 2000,407,971.
8 B. A. Jordan, Devi, L. A. Nature 1999,399,697.
9 M. RocheviUe; Lange, D. C.; Kumar, U.; Patel, S. C.; Patel, R. C.; Patel, Y. C. Science
2000,288, 154.
10 M . Mellado; Rodriguez-Frade, J. M.; Vila-Coro, A. J.; Fernandez, S.; Martin, D. A.;
Jones, D. R.; Toran, J. L.; Martinez, A. EMBO J 2001,20,2497.
11 B. A. Jordan, Trapaidze, N.; Gomes, I.; Nivarthi, R.; Devi, L. A. Proc. Natl. Acad. Sci.
U.S.A 2001,98,343.
12 R. Maggio, Vogel, Z., Wess, J. Proc. Natl. Acad. Sci. US.A 1993, 90,3 103.
13 T. E. Hebert, Moffett, S., Morello, J. P., Loisel, T. P., Bichet, D. G., Barret, C., Bou-
vier, M. J Biol. Chem. 1996,271, 16384.
14 P. R. Gouldson, Snell, C. R., Reynolds, C. A. J. Med. Chern. 1997,40,3871.
15 P. R. Gouldson, Snell, C. R., Bywater, R. P., Higgs, C., Reynolds, C. A. Protein Eng
1998,11, 1181.
16 M. Rocheville, Lange, D. C., Kumar, U., Sasi, R., Patel, R. C., Patel, Y. C. J Biol.
Chem 2000,275,7862.
17 S. Angers, Salahpour, A., Joly, E., Hilairet, S., Chelsky, D., Dennis, M., Bouvier, M.
Proc. Natl. Acad. Sci. US.A 2000,97,3684.
18 S. R. George, Lee, S. P., Varghese, G., Zeman, P. R., Seeman, P., Ng, G. Y., O'Dowd,
B. F. J. Biol. Chem. 1998,273,30244.
19 S. Cvejic, Devi, L. A. J.Biol.Chem. 1997,272,26959.
20 S. AbdAlla, Lother, H., Quitterer, U. Nature 2000,407,94.
21 F. Pazos, Helmer-Citterich, M., Ausiello, G., Valencia, A. J. Mol. Biol. 1997,271,
511.
22 0 .Lichtarge, Yamamoto, K. R., Cohen, F. E. J. MoZ. Biol. 1997,274,325.
23 0. Lichtarge, Bourne, H. R., Cohen, F. E. Proc. NatZ. Acad. Sci. US.A 1996, 93,
7507.
24 0. Lichtarge, Bourne, H. R., Cohen, F. E. J Mol. Biol. 1996,257,342.
25 M . K. Dean, Higgs, C., Smith, R. E., Bywater, R. P., Snell, C. R., Scott, P. D., Upton,
G. J. C., Howe, T. J., Reynolds, C. A. J. Med. Chem., 2001,44,4595.
26 W . Kuipers, Oliveira, L., Vriend, G., Ijzerman, A. P. Receptors. Channels 1997,5,
159.
27 P. R. Gouldson, Higgs, C., Smith., R. E., Dean, M. K., Gkoutos, G. V., Reynolds, C. A.
Neuropsychopharmacology 2000,23, S60-S77.
28 G. V. Gkoutos, Higgs, C., Bywater, R. P., Gouldson, P. R., Reynolds, C. A. Int. J.
Quant. Chem. Biophys Quarterly 1999,7437 1.
29 P. R. Gouldson, Dean, M. K., Snell, C. R., Bywater, R. P., Gkoutos, G., Reynolds, C.
A. Protein Eng, 2001, 14,759.
30 M. E. Sowa, He, W., Slep, K. C., Kercher, M. A., Lichtarge, O., Wensel, T. G. Nat.
Struct. Biol. 2001,8,234.
31 M. E. Sowa, He, W., Wensel, T. G., Lichtarge, 0. froc. Natl. Acad. Sci. U S . A 2000,
97, 1483.
32 N. Saitou, Nei, M. Mol. Biol. Evol. 1987, 4,406.
33 S . Sneath Numerical Tmonomy 1973.
34 F. Horn, Weare, J., Beukers, M. W.,Horsch, S., Bairoch, A., Chen, W., Edvardsen, O.,
Campagne, F., Vriend, G. Nucleic Acids Res. 1998,26,275.
35 Vriend, G. GPCRDB: Information system for G protein-coupled receptors (GPCRs),
http://www.gpcr.org/7tm/. 2000.
36 K. Palczewski, Kumasaka, T., Hori, T., Behnke, C. A,, Motoshima, H., Fox, B. A., Le,
T., I, Teller, D. C., Okada, T., Stenkamp, R. E., Yamamoto, M., Miyano, M. Science
2000,289,739.
37 D. C. Teller, Okada, T., Behnke, C. A., Palczewski, K., Stenkamp, R. E. Biochemistry
2001,40,7761.
38 T. M. Frimurer, Bywater, R. P. Proteins 1999,35,375.
39 D. Donnelly, Overington, J. P., Blundell, T. L. Protein Eng 1994,7,645.
40 A. Bateman, Birney, E., Durbin, R., Eddy, S. R., Finn, R. D., Sonnhammer, E. L. Nu-
cleic Acids Res. 1999,27,260.
41 K. SchefEek, Ahmadian, M. R., Kabsch, W., Wiesmuller, L., Lautwein, A,, Schmitz,
F., Wittinghofer, A. Science 1997,277,333.
42 M. K. Dean, Smith, R. E., Scott, P. D., Upton, G. J. C., Reynolds, C. A. to be submit-
ted. 2001.
43 R. Maggio, Barbier, P., Fornai, F., Corsini, G. U. J.Biol.Chem. 1996,271, 31055.
44 G. Y. Ng, O'Dowd, B. F., Lee, S. P., Chung, H. T., Brann, M. R., Seeman, P., George,
S. R. Biochem. Biophys. Res. Commun. 1996,227,200.
45 Calver, A. R.,Robbm, M. J., Medhurst, A. D., Hirst, W. D., Pangalos, and M.N. Ef-
fects of GABAB receptor subunit chimeras on receptor localisation and ligand bind-
ing, European Neuroscience 2000,Brighton, UK. 2000.
46 Higgs, C and Reynolds, C. A. Modelling G-protein coupled receptors, in Theoretical
Biochemistry, ed. Eriksson, L., Elsevier, 2001, Amsterdam, pp341-376.
47 F. Chela, Escriche, M., Burgueno, J., Angulo, E., Casado, V., Soloviev, M. M., Ca-
nela, E. I., Mallol, J., Chan, w. y.,Lluis, C., McIlhinney, R. A,, Franco, R. J
BioLChem 2001,276,18345.
48 K. Konvicka, Guamieri, F., Ballesteros, J. A., Weinstein, H. Biophys J 1998,75,601.
49 D.Fu, Ballesteros, J. A., Weinstein, H., Chen, J., Javitch, J. A. Biochemistry 1996,35,
1 1278.
50 A. Krebs, Villa, C., Edwards, P. C., Schertler, G. F. JMol.BioZ. 1998,282,991.
51 J. M.Baldwin, Schertler, G. F., Unger, V. M. J.Mol.BioZ. 1997,272,144.
52 V.M. Unger, Hargrave, P. A., Baldwin, J. M., Schertler, G. F. Nature 1997,389,203.
53 J. B.Findlay, Pappin, D. J. Bi0chern.J. 1986,238,625.
54 N . Tilakaratne, Christopoulos, G., Zumpe,E. T., Foord, S. M., Sexton, P. M. J
Pharmacol. Exp. Ther. 2000,294,61.
55 E.T. Zumpe, Tilakaratne, N., Fraser, N. J., Christopoulos, G., Foord, S. M., Sexton, P.
M. Biochem.Biophys Res.Cornmun. 2000,267,368.
56 P.R.Gouldson, Bywater, R. P., Reynolds, C. A. BiochemSoc.Trans. 1997, 25,4343.
57 P. R. Gouldson, Bywater, R. P., Reynolds, C. A. BiochemSoc.Trans. 1997, 25,5298.
ELECTROSTATIC STABILITY OF W E D TYPE AND MUTANT TRANSTHYRETIN
OLIGOMERS.
S. Skoulakis and J.M. Goodfellow.
Department of Crystallography, Birkbeck College, University of London, Malet Street,

London WClE 7HX, UK
1 INTRODUCTION
Transthyretin (TTR), formerly know as prealbuim', is a protein that binds thyroid

hormones and transports retinal binding protein that carries retinol. Many structures of
TTR and of its mutants have been determined2. The physiologically active form of wild
type 'ITR is a homotetrameric plasma protein in which the monomer is composed by 127
amino acid residues (Figure 1). The residues in the monomeric unit are in two four-
stranded beta-sheets, which form a beta-sandwich structure. In general, all mutant
structures crystallise in the same space group as the wild type, and display only minor
changes compared with the wild type. 'TTR, both wild type and mutant, is one of at least 20
human proteins that under suitable conditions has been shown to form a m y l ~ i d s ~ ' ~ .
Deposits of wild-type TTR amyloids accumulate in the heart, in a disease called senile
systemic amyloidosis (SSA) with the age of onset at about 80. A related disease, familial
amyloidotic polyneuropathy (FAP), occurs at a much younger age, in some cases during
the second decade of life, and affected individuals are found to have a mutation of TTR.
So far 73 point mutations are associated with FAP and many of these destabilise the
molecule leading to the formation of fibrils.
One of the models that has been proposed, for the formation of amyloids, is the
conformational change hypothesis'. First, by lowering the pH, the tetrameric structure of
the protein is disrupted and disassociates into a monomeric structure, which is structurally
different from the monomer at normal pH. These altered monomers can associate and
eventually form amyloid fibres. The mutations shift the tetramer-monomer equilibrium in
favour of the structurally altered monomer, leading to a relative loss of stability of the
mutated proteins relative to the wild type.
Figure1 Diagram of the tetramer of wild type transthyretin (TTR)taken form crystal
structure wt2,plotted using Molmol (version 2. I k. 1.O)9.
The pH-dependent effects in proteins are mainly electrostatic in nature and originate
from changes in the protonation states of acidic and basic Some of the
titratable residues in a protein exhibit different behaviour, compared with that of the same
residue in isolation, due to their interactions with other residues in the protein, and due to
the altered interactions with the solvent.
In order to test the conformational change hypothesis we calculate the protonation state
of all titratable residues. We find that some residues undergo a large change in their
protonation state upon dimer and tetramer formation. In order to understand further the role
of these residues in the binding process, we calculate their contribution to the electrostatic
free energy of binding.
2 METHODS
Initial models are constructed using the GROMACS package". The atomic coordinates are
taken from the RCSB Protein Data Bank for the wild type transthyretin (ltta) and (lf41) at
1.781, and 1.581resolution respectively and the mutants M119T (lbze) at l.8A and V30M
(lttc) at 1.781 resolution. Only the coordinate files of the dimers are provided and the
tetramers are generated using the CCP4 suite of programs. As explained in Hornberg et al.
(2000)2,the placement of the first 9 and the last 2 residues is not correct in the above pdb
files due to the increased flexibility of these segments. We exclude these segments from all
the structures in our p& calculations and electrostatic calculations, and since these are
exposed to the solvent, we do not expect their pKa values to be significantly perturbed. The
positions of hydrogen atoms, that are added to the structures, are minimised using 50
cycles of steepest descent followed by conjugate gradient with a force tolerance of 100
kJ/mol/nm. During this minimisation procedure, all the atoms except those of hydrogen are
frozen.
The electrostatic potential can be found by solving the linearized form of the Poisson-
Boltzmann equation (PB)11,12n’3,as follows:
V [E(r)V @(r)]- k2&(r)Hr)=-4np(r) (1)
where @(r)is the electrostatic potential at position r, p(r) is the charge density and E(r) is
the dielectric constant as a function of position, and k2 is the modified Debye-Huckel
parameter, which accounts for counterion screening in the solvent. Since water is more
easily polarized by an electric field than the protein at least two values for E must be used.
In the following we use value 80 for the water and 20 for the protein, and repeat the
calculations with value 4 for the protein.
We can define the PIG, value of a residue i on the basis of protonation probability (q(i)),
of this residue, from the equation7:
pKa(i)=pH + (U2.303) ln[(q(i))/( 1-(q(i)))]. (2)
The pH value at which the protonation probability is 0.5 is named as pKln and is often
used to describe the titration behaviour of the residues. A large change in p K 1 ~
indicates a
large change in the titration curves, upon oligomer formation. It can be proved6*’,that the
change in the protonation curves upon oligomer formation is related to the pH dependent
electrostatic free energy contribution according to the following formula
where, A( Q(pH)) is the pH-dependent mean charge difference between two states of the
system and is given by the following equation
A { Q ) bind = ( Q) O ~ i- n x ( Q ) , (4)
where n is the number of monomers involved in the binding process, ( Q ) oli and ( Q ) mon
are the mean charges for the oligomer and monomer respectively and are given by the sum
of the mean charges of all the residues,
This implies that the residues with the largest shift in the pK112 values, are the most
important in the pH related binding process.
To understand further the contribution of specific residues, or networks of residues, to
the energetics of binding, we calculate the electrostatic free energy difference that results
from simultaneously mutating the residues or network of residues to uncharged ones. The
calculated free energy corresponds to the reversible work associated with the process of
1 Probing Biological Molecules: Theory and Experiment 97
inserting all charges, to each side chain of the network in the oligomeric form of the
protein, relative to the monomer. The total electrostatic free energy AGtot, can be
decomposed into three distinct term^'^,'^ (Figure 2).
AGtot = AGdslv + AGbrd + AGprot (6)
AGdslv is due to desolvating the network of residues in the dimer or the tetramer when all
the other charges are 0, the “bridging” term AGbrd is the electrostatic free energy of
interaction of the residues embedded in a uncharged protein, and AGprot is the difference
in the interaction, of the network with the charges of the protein, between the dimer or
tetramer and the monomer.
= AGdslv
= AGbrd
= AGprot
Figure 2 Decomposition of the AGtot for the dirner. The network of charges is shown in
light grey. The Poisson Boltzman equation is solved in every case shown. Only
the charges of the network are included in the calculations of AGdsl and,
AGbrd. All the charges are included to the structures’ snapshots noted by (#O).
3 RESULTS
Initially we undertook calculations to determine the titration curves of all the residues in
the monomer, dimer and tetramer of wild type and mutant conformations. Using MEAD16
and MontiI7, we solve the Poisson-Boltzman equation and we calculate the titration
behaviour of the residues. The titration curves of the residues in the wild type are similar to
those of the mutants, which is consistent with the close structural similarity of the
conformations. The results of the pKln calculations for the monomers of the non-
amyloidogenic mutant M 119T (mut-noa), the amyloidogenic mutant V30M (mut-a), and
the two structures for the wild type (wtl), (wt2), are shown in Table 1. The values are very
similar for all structures. The calculations are performed with dielectric constant E = 2018.
Table 1 pK1n valuesf o r ITR monomers.

Residues mut-noa wtl
~~
wt2 mut-a
Arg2 1 11.1 11.6 11.4 11.4
Arg34 11.4 11.6 11.4 11.7
Argl03 12.0 11.9 12.0 12.0
Argl04 11.4 11.6 11.4 11.4
Lysl5 9.6 9.5 9.5 9.6
Lys35 9.6 9.8 9.8 9.3
Lys48 10.2 10.1 9.3 9.9
Lys70 9.3 9.2 9.3 9.3
Lys76 9.9 9.8 9.6 9.6
Lys80 10.2 10.1 9.9 9.9
Tyr69 9.3 10.4 9.9 10.2
Tyr78 8.7 9.2 8.7 9.3
Tyr 105 9.3 9.2 8.7 9.3
Tyrll4 9.6 9.5 9.6 9.6
Tyrll6 9.3 9.2 9.6 9.3
CyslO 8.1 7.7 7.5 7.8
His3 1 5.1 5.0 5.1 5.1
His56 4.8 4.1 4.8 5.1
His88 3.9 4.1 3.9 3.9
His90 3.6 4.1 3.8 3.9
Glu42 2.4 3.2 3.0 2.7
Glu5 1 3.6 3.5 3.6 3.6
Glu54 2.1 1.7 1.8 1.8
Glu6 1 3.3 2.9 3.O 3.3
Glu62 3.3 3.5 3.3 3.6
Glu63 3.3 2.9 3.O 3.0
Glu66 3.6 3.8 3.6 3.9
Glu72 0.9 0.5 1.2 0.6
Glu89 2.1 2.0 2.1 2.1
Glu92 0.3 0.5 0.3 0.3
Asp 18 -0.8 -0.1 -0.8 -0.4
Asp38 3.0 2.6 2.7 2.7
Asp39 2.4 2.6 2.4 2.4
Asp74 0.9 0.2 0.3 0.9
As1199 1.8 2.0 L. 1 1.8
Most of the residues have reasonable p K 1 ~values. His88, His90, Glu42, Glu54, Glu72,
Glu92, Aspl8, Asp74, Asp99, have lower values than expected. Due to the long range of
the table interactions and the number of titratable residues, it is not easy, in general, to
attribute the anomalous p K 1 ~value to the proximity of specific residues. However, the
interactions between the residues can be analysed from visual inspection of the crystal
structures and from the matrix of interactions (as output from MEAD). We find that there
is a network of strongly interacting residues, consisting of Glu72, His88, Glu92, His90,
Tyrll6. The distances between the closest atoms are shown in Table 2. The closest atoms
are the ones that are strongly interacting as we confirm from the output of the matrix of
interactions. For the other residues, Glu42 is influenced by Arg34, Glu54 by Lysl5, Asp18
by Tyr78, and Asp99 by TyrlO5.
Table 2 Close contacts within the monomers.
Di stances(A)
Closest non-hydrogen atoms mut-noa wtl wt2 mut-a
Glu72(OEl)-His90(NE2) 4.12 3.84 4.55 4.29
His9O(NE2)-Glu92(OE2) 5.90 4.26 5.61 4.26
His90(ND l)-Glu92(OE1) 5.60 4.57 3.29 4.65
His90(NDl)-Glu92(OE2) 4.19 2.57 5.83 3.39
Glu92(OEl)-Tyrl16(OH) 3.44 3.74 5.27 7.60
His88(NE2)-Tyrl16(CE2) 3.84 3.70 6.17 3.90
His88(NE2)-Tyrl16(OH) 5.20 4.95 3.99 5.29
His88(NDl)-Tyr116(CE2) 3.81 3.92 5.29 3.93
Upon dimer formation, the p& of some the residues changes and the results for the largest
changes are shown in the Table 3. These changes cannot easily be explained by the
approach of the other monomer, for example, by formation of salt bridges. The only extra
hydrogen bond that is formed involving titratable residues is between OG1 (Thr96) and
OE2 (Glu89), in all structures studied.
Table 3 ApKln for dimer formation. &Kin values ((pKln(dimer)- pKm(monomer))for

the residues with the largest changes in values i.e. IApKmJ>0.64 8 kcaUmol at
300 K. Values for the residues of only one monomer are shown. Due to symmetry,
we find similar behaviour for the other monomer. The values are calculated with
dielectric constant E =20.These residues noted by (#) titrate at a p H less than -4
in the dimer.
mut-noa wtl wt2 mut-a

Lys35 -0.8 -1.0 -0.8 -0.9
Lys70 -1.9 -2.1 -2.4 -2.5
Lys76 -1.1 -1.2 -1.0 -1.6
Tyr 105 -1.1 -1.3 -1.6 -1.1
Tyrll4 -1.o -1.2 -1.1 -1.2
Tyrll6 -1.7 -1.6 0.0 -1.5
His31 -0.9 -0.9 -0.8 -0.7
His88 -5.3 -5.3 -5.7 -5.7
His90 -7.6 -7.4 -6.9 <-7.6'
Glu66 -0.4 -0.4 -0.6 -0.7
Glu72 -2.3 -2.8 -2.7 -2.6
Glu89 -2.5 -2.5 -2.8 -3.9
Glu92 <-5.3# <-5.3# <45# <-4.3#
Asp18 -0.8 -0.4 -0.5 -0.6
Asp39 -1.8 -2.0 -1.4 -2.0
Asp74 -0.9 -0.7 -0.6 -0.9
Asp99 -2.4 -2.1 -2.2 -1.8
Notable is the close proximity of Glu92 in one monomer to the Glu92 and the OH atom of
Tyrll6 table the other monomer. These belong to the network of strongly interacting
intramolecular residues, that become buried upon dimer formation, and interact
intermolecularly with the same network of the other monomer, (Table 4).
On dimer association to form the tetramer, a similar pattern of behaviour is observed.
The values of pK112, shown in Table 5, decrease but not to the same extent as upon dimer
formation. This is not surprising because the dimer-dimer interface is populated by
hydrophobic residues. This is in contrast to the monomer-monomer interface, which is
occupied mainly by charged and polar residues. Once more the interactions between
residues are the most important factor leading to the low p K l values.
~ The most affected
Table 4 New contacts upon dimer formation. Upon dimer formation the above titratable
groups come in close proximity resulting in low p K 1 values.
~
monomer a monomer b Distances(8i)

mut-noa wt 1 wt2 mut-a
Glu92(0) Tyrl16(OH) 2.98 3.OO 2.76 2.88
Glu92iOEl) Glu92(OE1) 4.53 4.14 4.11 2.63
Glu92(OE1) Glu92(OE2) 6.03 6.02 3.29 3.89
Glu92(OE2) Glu92(OE1) 5.81 5.07 4.46 4.81
Tyrl16(OH) Tyrl16(OH) 4.03 4.79 4.25 4.31
residues are the ones closest to the interface between the dimers, as expected, even though
the only titratable residues that come closer than 581 to the dimer, due to tetramer
formation, are Tyrll4 and Arg21 from the third and fourth monomer of the second dimer.
The appearance of the second dimer changes the strength of the interactions already
present in the first dimer, and this causes the shift of the pK112 values.
Table 5 ApKln for tetramer formation. ApKln values ((pKm(tetramer)- pKlQ(dimer))for

the residues with the largest changes i.e. lApK1nl>O.6 ~ 0kcal/mol
% at 300 K. The
values for the residues of only one monomer, (a), are shown. Due to symmetry
we find similar behaviour for the others. The dielectric constant used is E =20.
monomer a mut-noa wtl wt2 mut-a

Lys 15 -2.5 -2.1 -1.8 -2.3
Tyr78 -1.2 -1.2 -1 .o -1.2
Tyrll6 -2.4 -1.7 -2.9 -3.0
His56 -0.8 -1.0 -1.1 -1.0
His88 <-2.6 <-2.6 <-2.4 <-2.2
Glu54 -2.6 -2.5 -2.5 -2.7
Asp 18 <-2.4 <-2.8 <-3.4 <-3.0
Asp74 -0.7 -0.1 -0.5 -0.5
From the above calculations we find that A( Q ) bind is negative, implying that AGbind is
increasin as we lower the pH (equation 3). This is consistent with the experimental
results3A$ proving the instability of the oligomer structure as the pH is lowered. Moreover
the importance of the network of Glu72, His88, Glu92, His90, and Tyrll6, in the pH
dependent binding process, makes the investigation of its role in the binding process in
general worthwhile.
We calculate the contribution of this network (Ntl), in the electrostatic free energy of
dimer and tetramer formation. We also repeat the calculations for a subset (Nt2) of this
network composed by Glu92, His90, and Tyrl16. In both cases we find that the network
formation is unfavourable, mainly due to the desolvation term, which is large since the
network is located at the interface between the monomers.
Table 6 Eelectrostaticfree energy contribution,for tetramer, AAGtot(t-m), (dimer,

MGtot(d-m))formation, from the monomers,for subsets of residues, namely N t l :
Glu72, His88, Glu92, His90, and Tyrll6 and Nt2: Glu92, His90, and Tyrll6. All
energies are in kcallmoi.
MGtot(d-m) AGprot AGbrd AGdslv AGtot
&=4 &=20 &=4 E=20 &=4 E=20 &=4 &=20
wtl Ntl -11.7 -4.1 7.2 3.6 24.5 3.6 20.0 3.1
Nt2 -10.2 -2.0 5.6 1.8 22.2 4.1 17.6 3.9
wt2 Ntl -10.7 -4.6 7.3 1.9 20.7 3.0 17.3 0.3
Nt2 -8.8 -4.0 5.1 1.7 20.8 2.8 17.1 0.5
mut-noa Ntl -11.6 -3.8 7.0 3.3 27.4 4.4 22.8 3.9
Nt2 -7.9 -1.5 5.3 1.6 24.1 3.5 21.5 3.6
mut-a Ntl -10.8 -4.0 7.8 3.9 17.2 3.2 14.2 3.1
Nt2 -7.0 -1.4 6.0 3.9 13.6 3.2 12.6 5.7
MGtot(t-m) AGprot AGbrd AGdslv AGtot

E=4 &=20 &=4 E=20 &=4 &=20 E=4 E=20
wtl Ntl -13.2 -4.3 7.3 3.6 24.8 5.3 18.9 4.6
Nt2 -11.2 -2.3 5.7 1.8 22.5 4.1 17.0 3.6
wt2 Ntl -12.3 -4.1 7.7 3.4 21.3 5.5 16.7 4.8
Nt2 -10.3 -2.2 5.5 1.7 20.1 4.0 15.3 3.5
mut-noa Ntl -13.7 -4.3 7.0 3.3 26.5 4.3 19.8 3.3
Nt2 -8.9 -1.7 5.2 1.8 23.7 3.4 20.0 3.5
mut-a Ntl -12.9 -4.4 7.6 3.9 16.1 3.1 10.8 2.6
N t2 -6.8 -1.4 5.6 2.2 12.4 2.1 11.2 2.9
Our results (Table 6) are robust with respect to changes in the dielectric constant. When we
use E = 20, AGtot is less unfavourable, as expected from the smoothing role of a higher
value of E, but still consistent with general view of electrostatic strain in this network. The
results are very similar for the network in the dimer or tetramer, indicating that the
monomer-monomer interface is the one contributing the most to the electrostatic
interactions.
Table 7 Electrostatic free energy of binding.
&=4
- . E = 20
mut-noa wtl wt2 mut-a mut-noa wtl wt2 mut-a
d-m 53.5 47.9 35.0 45.0 9.6 9.4 8.2 9.5
AGelect t-d 33.7 48.9 32.1 30.2 10.2 12.0 8.9 10.2
t-m 140.7 144.7 102.1 120.2 29.4 30.8 25.3 29.2
102 Biophysical Chemistry:Membranes and Proteins
We also calculate the total electrostatic free energy of binding for all the structures, (Table
7). The table energy found is positive in all cases and with all dielectrics used. This means
that the electrostatic interactions oppose the oligomer formation. Even though we can
predict the trend of the free energies of binding the calculations are not sensitive enough to
discriminate the mutant from the wild type structure, which is not surprising considering
the high accuracy required for such discrimination. Our calculations show that the reduced
stability of the oligomer when we lower the pH is not due to the loss of favourable
interactions upon binding of H', but a 'bad' situation becoming 'worse' as the electrostatic
interactions are unfavourable even at low pH.
4 CONCLUSION
We predict the loss of stability of oligomeric 'ITR, when the pH is lowered. The
importance of the network of residues, Glu72, His88, Glu92, His90, and Tyrll6, is
indicated. We also find that the contribution of the above residues in the electrostatic free
energy of binding is unfavourable and the binding of H' increases this contribution,
leading to the formation of monomeric structures.
Acknowledgements
We would like to thank the Wellcome Trust for a training fellowship in mathematical
biology to SS and the BBSRC for some computer resources. The work was carried out
within the Bloomsbury Structural Biology Centre.
References
1 C.C.F. Blake, M.J. Geisow, S.J. Oatley, B. RCrat, andC. RCrat, J. Mol. Biol., 1978,
121,339.
2 A.Hornberg, T. Eneqvist, A. Olofsson, E. Lundgren, and A. E. Sauer-Eriksson, J.
Mol. Biol., 2000, 302,649.
3 J.W. Kelly, Structure, 1997, 5, 595.
4 A.M. Damas, M.S. Saraiva, Journal of Structural Biology, 2000, 130, 290.
5 J.W. Kelly, Current Opinion in Structural Biology, 1998, 8, 101.
6 A.S. Yang, B. Honig, J. Mol. Biol., 1993, 231,459.
7 G. M. Ullmann, E.W. Knapp, Eur.Biophys. J., 1999, 28,533.
8 M. Schaefer, M. Sommer, M.Karplus, J. Phys. Chem., 1996,101, 16663.
9 R. Koradi, M. Billeter, K. Wiithrich, J. Mol. Graphics, 1996, 14, 51.
10 H.J.C. Berendsen, D. van der Spoel, R. van Drunen, Comp. Phys. Comm., 1995,91,
43.
11 J. Warwicker, H.C. Watson, J. Mol. B i d , 1982, 157,671.
12 M.K. Gilson, B. Honig, Proteins Struct. Funct. Genet., 1988,4, 7.
13 B. Honig, A. Nicholls, Science, 1995, 268, 1144.
14 V. Lounnas, R.C. Wade, Biochemistry, 1997,30,5402.
15 Z.S. Hendsch, B. Tidor, Prot. Sci., 1994,3,211.
16 D. Bashford, K. Gerwert, J. MoZ. Biol., 1992, 224,473.
17 P.Beroza, D.R. Fredkin. M.Y.Ukamura, G. Feher, Proc. Natl. Acad. Sci. USA,
1990,87,5804.
18 J. Antosiewicz, J, J.A. McCammon, M.K. Gilson, Biochemistry, 1996, 35, 7819.
SIMULATIONS OF HUMAN LYSOZYME:CONFORMATIONS TRIGGERING
AMYLOIDOSIS IN I56T MUTANT.
G. Moraitakis' and J.M. Goodfellow'

1
School of Crystallography,Birkbeck College, University of London, Malet Street, London
WClE 7HX,UK
1 INTRODUCTION
Human Lysozyme is a 130-residue protein found in secretions (e.g. saliva, sweat and
mucus) and more generally in leukocytes and kidneys. It is an enzyme which hydrolyses
preferentially the p- 1,4 glucosidic linkages between N-acetylmuramic acid and N-
acetylglucosamine which occur in the mucopeptide cell wall structure of certain micro-
organisms.' The wild type human lysozyme has been crystallised and its structure
elucidated by Artymiuk and Blake2 at 1.581 resolution (Figure IA). Its native structure
consists of two domains: an a domain that has four a-helices (A-D) and one 310 helix, and
a p domain, which consists mainly of an antiparallel p-sheet and a long loop. The active
site is located in the cleft that is formed between these two domains. The protein contains 4
disulphide bonds of which two are located in the a-domain, one in the long loop of the p-
domain, and one that connects the two domains.
There are two known natural mutations of the human lysozyme: I56T and D67H.3
They both cause autosomal dominant hereditary non-neuropathic systemic amyloidosis.
This is a condition whereby there is tissue deposition in viscera and other body cavities, of
normally soluble autologous proteins as insoluble fibrils called umyloids. The core of these
fibrils structure consists of P-sheet with the strands perpendicular to the long axis of the
fibre! Amyloids can be formed from proteins of diverse sequence, fold and function and
are known to lead to serious medical conditions such as Alzheimer's disease and
spongiform encephalopathies.' Although the mechanism of lysozyme fibrilogenesis is not
very clear, experimental evidence6 suggests that it is related to changes in stability and
tendency of aggregation due to the mutations. More specifically, Booth et aL7 proposed
that, during folding, a partially folded transient population of the two human lysozyme
amyloidogenic proteins, which lacks global cooperativity, undergoes structural
transformation (a helix-to-sheet transition) and creates the first template, the "seed", for
further protein deposition and fibril formation. In a recent paper: Morozova-Roche et al.
show that the presence of these "seeds" for both wild and the two mutants of human
lysozyme facilitates the formation of fibrils.
We focus on the behaviour of the I56T variant. The crystallised structure of this
mutant7 at 1.8 81 resolution, is quite similar to that of the wild type (0.12 81 difference) and
'40
Figure 1 The crystal structures of the wild type (A) and the 156T ( B ) human lysozyme
and the mutation site, Ile.56, located in the interface of the a and Bdomains,
are shown as ball and sticks. The 4 disulphide bonds are indicated as ball and
sticks. The sulphur atoms of these bonds as well as the N and C termini atoms
are shown as spheres.
is shown in figure 1B. There are very small changes near the mutation site whereby the
introduction of a charged group (hydroxyl) and the removal of a methyl group lead to the
N-terminal approach more Thr56 and result to minor changes at the C-terminal of helix C.
In order to understand the behaviour of the partially folded structures that may trigger
amyloid formation, we explore the conformations adopted during the induced unfolding of
the wild type and mutants at high temperatures (500K) using molecular dynamics (MD)
simulations. The temperature 'denaturation' of proteins using MD is considered as one of
the most straightforward computational experiment^^.^ The great advantage of unfolding
simulations is that they allow the investigation of the conformational properties at every
point along the unfolding pathway." There are many examples of their use'' and these
have led to detailed insi ht of experimental results of protein unfolding'2313and the study
of the folding pathway."Lysozyme, from hen egg-white has been extensively investigated
by means of unfolding simulations with the aim to investigate several issues such as the
stability and folding of the protein. 15*16*17*18~19
The human and hen forms of lysozyme have
60% sequence similarity but very similar 3D structure.
In our study, we attempt to examine and compare the structure and properties of the
partially folded intermediates obtained from the induced unfolding of the wild type and the
I56T form.
2 METHODS
2.1 Model and SimulationDetails
The crystal structures of the wild type (1REX) and the mutant (1LOZ) lysozyme are the
starting point of the simulations. Hydrogen atoms are added and the final system includes
solvent molecules which are represented by the SPC216 model.2o All atoms are explicitly
represented. The solvated model is contained in a rectangular box (70x70~70A) using
periodic boundaries conditions? The building of the protein models and all their
simulations are carried out with the GROMACS suit of programs.22The GROMOS 96
force field is used to describe the atomic interaction^.^^ For the correct treatment of long-
range electrostatics we make use of the particle mesh Ewald summation alg~rithm.~'The
high frequency degrees of freedom from the covalent bonds of hydrogen atoms are
constrained using the LINCS algorithm2' and thus the time step is increased from 1 to 2 fs.
The system is coupled to an external temperature bath with a separate bath for the solvent
and the solute. The regulation of the pressure is achieved by means of a "pressure" bath.
Data on the trajectories are saved every 0.2~s.For these simulations, we use an in-house
multi-processor Origin 2000 with 4 processors. The total CPU time for all simulations was
about 48 days.
In order to minimise the initial system, we use a combination of the conjugate
gradients and steepest descent methods in which after every 50 steps of conjugate
gradients, one step of steepest descent is performed. The minimisation is terminated when
the overall force of the system is lOON or after 5000 steps. This protocol is first used to
minimise hydrogen atom and water molecules positions and then extended to the whole
system. To further optimise the arrangement of the solvent around the protein and alleviate
high energy regions (hot-spots), the water molecules only, are assigned initial velocities
from a Gaussian distribution generated from a random seed and then warmed-up from 50K
to 300K over lops of MD. This is followed by equilibration of the water molecules at
300K for 30ps. Then the whole system (including the previously constrained protein) is
assigned initial velocities from a Gaussian distribution generated by a random seed for
50K,warmed up to 300K for lops and finally equilibrated at this temperature for 1OOOps.
The looOps trajectory at 300K serves as a control simulation. The last conformation of the
system obtained from this trajectory is used as the starting structure for the unfolding
simulation in which the system is warmed up to 500K for 10 ps and is subsequently
maintained at this temperature. To enhance the sampling of the unfolding pathway, long
multiple trajectories are performed by assigning velocities generated fiom different seed
numbers. Three 5000~shigh temperature (500K) and three lO0Ops control (300K)
simulationsare performed for each lysozyme form.
2.2 Analysis of Molecular Dynamics
In the analysis, the root mean square deviation (RMSD), solvent accessible surface area
and secondary structure content plots are calculated using the analysis programs provided
by GROMACS.26The secondary structure analysis is carried out using the Kabsch and
Sander algorithm2' incorporated in their DSSP program. The inter-residue distances and
the clustering analysis is performed with software written in-house. The percentages of
secondary content per region across all trajectories are derived as follows: First the
percentage in each of the four secondary structure conformation (helix, strand, turn, coil)
contained in the specified regions is calculated for each structure sampled from the
multiple unfolding trajectories of each lysozyme form. Then, these percentages are
obtained across all multiple trajectories (Table 1).
The number of pairwise distances of all residues within 8A (closest atoms distance) is
calculated for each conformation in the trajectory and for the crystal structure. For the
latter, all these residue contacts are referred to as native contacts as opposed to the nun-
native contacts of pairs of residues that are situated within the 8A cut-off but are not
present in the native structure (crystal Structure).
The distance matrix used in the clustering, is constructed from the column vectors of a
properties’ matrix. Here the properties’ matrix of the conformations observed during the
trajectories contains four rows corresponding to four properties: (1) the number of native
residue contacts, (2) the number of non-native residue contacts, (3) the number of residues
in secondary structure elements (a-helix, B-sheet, &turn) and (4) the number of residues in
random coils. The reason that those properties have been chosen is because it has been
observed from our plots that they change with time during the unfolding unlike other
properties like Solvent Accessible Surface area (SASA), radius of Gyration (Rg)and
number of H-bonds which oscillate around their initial value. Thus, the latter cannot
discriminate easily between conformations obtained in different times during the
unfolding).
The Mahalanobis distance2*is used for the construction of the distance matrix. The
elements of this matrix are calculated by the following formula:
D; = (xi- XjlT C’(xi - xj) (1)
Here, C is the covariance matrix and x is a column vector of the properties’ matrix. The
use of Mahalanobis distance removes several of the limitations of Euclidian distances used
by others for classifications of MD sir nu la ti on^^^ as it automatically accounts for the
scaling of the coordinate axes and also it corrects for correlation between the different
features (e.g. residue contacts and secondary structure content).30
We use hierarchical clustering to group the conformations by transforming a set of data
points with a given measurement for dissimilarity (the distance matrix) into a sequence of
nested partitions (a dendrogram). We use agglomerative methods to built the dendrogram.
These involve starting with each data point as a single cluster and subsequently merge
these together. In particular we make use of the group average distance hierarchical
algorithm: At each step, we seek the shortest distance of a pair of clusters in the distance
matrix and merge them together. A new distance matrix is created in which the new
distances of all the clusters from the newly created matrix are based on the average
distance of their This process is continued until no distance in the distance
matrix occurs below a set cutoff. We have selected this cutoff to be the point in the process
at which a cluster formed contains more than 20% of the conformations. Our results using
this clustering algorithm have been verified against other statistical packages such as S-
Three 5ns trajectories at high temperature are performed for the wild type (WT 1, WT 2,
WT 3) and three for the mutant (I56TH 1, I56T 2, I56T 3) in order to increase the sampling
of the unfolding conformational space.?’ Clustering techniques are used to probe for the
most populated average conformations. Analysis of these conformations is carried out to
ascertain the structural features of any “seeds” that have the potential to lead to fibre
formation.
3.1 Analysis of the individual trajectories
At 300K we observe that the root mean square deviation (RMSD)based on the Ca atoms
plateaus around 0.2 nm from the crystal structure for all trajectories (Figure 2) and thus
agrees with previous results on the equilibration of a native protein structure at this
-El 0.7
w 0.6
P
0.5
3 0.4 (1.1
Equilibration
0.3
Figure 2 The plot of root mean spare deviation (RiMSD) (in nm) for Ca atoms as a
function of time (ps) with respect to the crystal conformation. Results are
shown for unfolding (500K),control and equilibration trajectories (300K).
temperature by Kazmirski et alF9 At 500K, we find that all simulations have similar rates
of increase in RMSD for the fust 1ooOps. However, after this point, the I56T mutant
trajectories exhibit a larger increase in RMSD compared to that of the wild type
trajectories. The values of the RMSD near 5000 ps seem to converge for all trajectories
with the exception of one that reaches about 1.lnm after 5ns.
The root mean square fluctuations of the atoms across all the simulations of the mutant
and the wild type form have been calculated for the C a atoms (Figure 3). We observe that
during unfolding, I56T tends to have slightly higher fluctuations in most of the regions.
This difference is most notable in the p domain and, in particular, in the loop region of this
domain between residues 68 and 75. The mutation site is however slightly more rigid in the
mutant during the simulations than in the wild type. The higher flexibility of parts of the p-
domain in I56T implies that it is less stable than that of the wild type, which is in agree-
0.9
0.8
0.7
0.6
0 0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130
Gatpha
Figure 3 Root mean square fluctuations (RMSF) of the C a atoms during the unfolding
and 300K simulations. The triangles represent the C a atom at residue 56. The
atoms of the residues involved in disulphide bonds are also shown (larger
circles). Regions of P-strand and a-helix are shown.
ment with the work of Booth et al.7 where the calculated B-factors for this domain are
increased. We also observe some increased fluctuations in the loop region just after helix B
in the mutant conformations.
Inter-residue distances change dramatically during the unfolding simulations with
residues originally distant in the native structure now being found in close proximity and
vice versa. Residue contacts can be classified as native contacts (those found in the native
structure) and non-native contacts which are formed during the simulation but not found in
the native structure. For the unfolding simulations at 500K,we have observed that the
trajectories have a similar rate of loss of native contacts and formation of non-native
contacts for the first 1000ps. Subsequently, the mutant simulations show a very slight
increase in loss of native contacts and a similar increase formation of non-native with all
trajectories converging by 5000 ps. For the control simulations at 300K,we observe that
the number of inter-residue contacts remains steady fairly constant. Thus, results from
these contact plots are consistent with the analysis of RMSD.
The major impact of unfolding can be shown from plots of inter-residue contact
distances, coloured according to the fraction of conformations sampled exhibiting each
contact. A similar pattern in native contact loss is observed for the wild type and the
mutant. Some differences are observed in the area of the beta domain and in particular in
the contact region of residue 56 and the C-terminal of helix B. Helix C shows very high
contacts retention for the mutant. In the non-native plots (figure 4). We observe that new
relatively strong contacts (50-60%)appear in the beta domain of the wild type. These are
near the native contacts suggesting that this domain is more resistant. A relatively stronger
contact (30-50%)forms between Thr56 and helix 310.This is more clearly shown in figure
5 where we plot the percentage of the total conformations sampled carrying each of the
contacts that residue 56 makes with the rest of the aminoacids in the polypeptide chain. We
observe that Thr56 has indeed particularly strong contacts with the residues 80-90 (helix
310) and residues 59-61 (part of the beta domain loop). In contrast Ile56 is in a lot less
conformations connected to this helix. However it shows some very loose contacts with the
C-terminal residues (1 10-129) as opposed to its mutant.
The solvent accessible surface area (SASA) does not change greatly with time in any
trajectory at 500K, for wild type or mutant. During the unfolding of a protein, it is
expected that more residues become accessible to solvent. Although a slight increase is
observed for all the trajectories between 0 and 1000 ps in the SASA of hydrophobic
residues, most of the time it is around the average value of 40 nm2. The total SASA
increases slightly in the first 500ps and then acquires a steady value. We noted also a small
“jump” near 4500 ps that is common for all the trajectories. These results for SASA are
consistent with the analysis of the radius of gyration against time, which similarly shows
no major change with time during the simulations.
The reason for the lack of overall change of the radius of gyration and SASA may be
the presence of four disulphide bonds. In the classical force field, which is used in MD
simulations, the disulphide bonds are kept intact irrespective of the increased temperature
conditions. In a separate set of simulations (data not shown) in which the disulphide bonds
of the two forms are reduced, it is observed that at high temperature both SASA and the
radius of gyration increase after 1.5 ns. At the same time, the percentage of secondary
structure decreases faster than in the normal simulations. One effect of the disulphide
bonding is that the residues involved in this are less flexible, as illustrated from our root
mean square fluctuation plots (Figure 3), where the Ca of these residues have the lowest
fluctuations. This is consistence with the crystallographically derived B-factors for the
structure.
1‘1
LO 1
81
61
II
21
Figure 4 Maps of non-native contacts of (A) the wild vpe and ( B ) the mutant across all
conformationssampled. Both axes represent residue number. The labels on the
diagonal correspond to regions of the molecule. The cut-off for a pair of
residues to be in contact is 861 between their Ca atoms. Colour shade is based
on percentage of conformations sampled have each specific contact and is
shown. The area of contact between residue 56 and helix 310 is shown in the
rectangle.
g 601 I
Figure 5 Distribution of contacts that Ile56 (A) and Thr.56 (B)make with the rest of the
residues across all the conformations sampled.
For all the trajectories, we find that the secondary structure profile changes
significantly during the simulation. Initially, most residues are in a specific secondary
structural conformation but the number of such residues drops gradually and reaches a
plateau after about 4ns. At the same time, random coil conformations become gradually
predominant. Analysis of the time course of secondary structure shows that (a) the original
P-strands are destroyed before the helices, (b) the helical elements are eventually replaced
by coils, bends and turns and then (c) transient beta strands and bridges appear in many
areas across the polypeptide chain.
The changes of the secondary structure content during the unfolding in separate
regions of the lysozyme are depicted in Table 1. As the regions selected are entirely
comprised of one of the four conformations in the native structure, we can have a rough
estimation of (a) how much of the original conformation is retained during the unfolding
and (b) what other conformations are acquired during the simulation. Our first observation
is that helical regions original secondary conformation is converted mostly to coils and
turns during the simulations. However, there is a wide difference in the percentage of each
helical region undergoing this conversion between the mutant and the wild type. Helix A
in the latter has helical components for 50% of the conformations sampled while only one
third of the structures in the mutant are helical in this region. In the same way, helix 310
and helix D are a lot more retained in the wild type than in the mutant. In contrast helix B
in the mutant tends to be helical in twice as many structures as the wild type. Similarly
helix C retains helical components mostly in the mutant conformations. In very few cases
there is partial conversion from helix to strand and even less from strand to helix.
Generally we observe that in most of the cases the wild type structures sampled tend to
have higher percentages of the original secondary structure conformation than the mutant
with the sole exception of helix B. Conversely, the turn and coil content is higher for the
mutant in all the regions.
The hydrogen bonding across the protein, during unfolding, has also been examined,
although in part it is related to the inter-residue contacts analysis described previously. The
total number of H-bonds is initially about 100 and then, after 500ps, it drops to around an
average of 75 (within the range of 60 to 90) for the rest of the trajectory. This applies to all
of the trajectories at 500K.The explanation for this observation may be related to the
changes in the secondary structure; the native structure is rich in helices and strands, both
of which involve many hydrogen bonds. During the unfolding, we have observed that these
elements are converted to coils (no hydrogen bonds) and turns (contain smaller hydrogen
bonding network).
3.2 Clustering of the trajectories’ conformations and analysis
More informative analysis of multiple trajectories can be obtained by accumulating all

snapshots from each trajectory and clustering into those with similar properties. This is
particularly useful since we cannot discriminate which trajectory is more “significant”than
the others.34All the conformations generated from the three trajectories of each lysozyme
form, are clustered (using an hierarchical clustering procedure) to generate average
structures ranked by the population of the cluster they represent. This clustering is based
on residue contacts and secondary structure content (as described in methods). The
resulting conformations are shown in Figure 6.
Table 1 Secondary structure content in lysozyme during the unfolding simulations.

Helix (9%)” Strand (%)’ Turn (%)‘ Coil(%)’
Regions (residues) WT 156T WT I56T WT 156T WT 156T
Helix A (5-14) 54 31 1 4 27 30 19 34
Helix B (25-36) 23 43 4 6 37 30 36 20
Strand 1 (43-46) 1 1 20 17 21 21 59 61
Strand 2 (51-54) 0 1 35 20 16 28 49 50
LOOP (60-78) 4 3 11 9 50 51 35 37
Helix 310 (81-85) 23 8 2 2 44 50 31 40
Helix C (90-100) 40 57 2 3 35 29 23 9
Helix D(110-115) 29 13 6 2 44 43 21 41
’The percentage is calculated from accumulative sums of secondary structure within given regions across all
conformations sampled during the trajectories
WC? 20.37% WCZ 14.63% WC3 10.71% WC4 10.20% WCS8.149C
WCBA709( WC75.87% WC84..BbK WCo4.17H WC103.41%

MCI 2 1 . 4 ~ MCZ 17.03% ~ c 1 03. 4 7 ~ m.4 a.im MCS 7 . m
Figure 6 The average structures of the ten most populated clusters are shown. The
mutation site (residue 67) is shown as ball and sticks. The N-and C-terminal
are shown as green and blue spheres respectively. The disulphide links are
also shown as spheres. The clustersfor the WT are denoted as WCl, WC2 etc.
while the ones from the mutant as MCI, MC2 etc. The models have been
obtained after least square fitting each of these conformations to the respective
initial models (crystal structure + hydrogens). Thus, the orientation and shape
of the structures are relative to the ones shown in Figure 1. The dotted lines
show the distance between Ile56 and helix 310 (orange) and the distance
between Ile56 and helix B (green).
Analysis of the wild type trajectories leads to 21 clusters. The most populated, WC1,
contains 20.4%of the total population, followed by two clusters with the order of 14% and
two more of 10%.Five more clusters ranging between 3-7% are found and the rest are
below 2%. For the I56T mutant, where 22 clusters are assigned using the same cutoff
criteria as the WT, the distribution is slightly different. The largest group contains 21.5%
and followed by another big cluster of about 17%of the total population. Two clusters of
about 10%exist and six more groups ranging from 4-7% are assigned. There is one group
of about 2% and two of about 1% and the rest have populations ranging between 0.54%
and 0.07%.A small proportion of the conformations of both lysozyme forms are grouped
in clusters (named WC8 and MC7) that both contain elements from the initial stages of the
unfolding trajectory.
The average structural features of these clusters are given in Table 2. The most
populated conformations for both the WT and the mutant have lost most of their secondary
structure content and the original native contacts. It appears though, that the WT clusters
are usually populated by conformations retaining more native contacts (i.e. a higher
fraction of the initial conformation) than its mutant counterpart. Additionally, the
populated mutant clusters contain conformations usually with greater non-native contacts.
In many of the clusters there are conformations that contain a high proportion of residues
in random coils. However, the WT clusters appear to contain a higher percentage of
secondary structure elements (helices and sheets) than the mutant.
The major impact of unfolding can be shown fiom plots of inter-residue contact
distances. These contact maps for the average structures of the two most populated clusters
and the crystal structures are plotted in a 2D colour map of all-against-all residue distances
(Figure 7). In the most populated cluster of the mutant (MCl), the beta domain (especially
in the region of the beta strands) exhibits a very different set of contacts compared with
those in crystal structure and with those in other clusters &om the mutant trajectories.
Specifically, in the MC1 cluster, the region around Ile56, located in the interface
between the a and p domain, loses contact with helix B but has stronger contacts with helix
310 compared to the wild type. Those two contacts are present in the crystal structures and
300K simulations of both lysozyme forms and they exist as well in the average structure of
the most populated clusters of the wild type. The distribution of these distances in the
conformational space sampled with the simulations is shown in Figure 8. We observe that
at 500K,for the Ile56-helix B contact, the mutant exhibits a distribution lying further away
Table 2 Average structural properties of the (A) Wild Type and (B) the 156T mutant
clusters given as the numbers of residues in a particular secondary structure
conformation.
A. WILD TYPE
Population Native Contacts Non-Native No. residues in No. residues in
Cluster (%I (% retained) Contacts helix or sheet or turna Coil"
WCl 20.37 140 (38) 240 31 52
wc2 14.63 182 (4% 175 55 42
wc3 13.71 191 (52) 173 51 48
wc4 10.20 117 (32) 267 39 56
wc5 9.74 164 (4) 237 50 45
WC6 6.79 116 (31) 238 32 61
wc7 5.97 107 (29) 247 33 54
WC8 4.66 270 (73) 94 78 28
wc9 4.17 148 (40) 265 34 54
WClO 3.41 108 (19) 283 30 49
Crystat 370 (100) 0 89 19
Population Native Contacts Non-Native No. residues in No. residues in

Cluster (%I (9%retained) Contacts helix or sheet or turn" Random Coil"
MCl 21.49 131 (36) 245 32 54
MC2 17.03 183 (50) 189 56 42
MC3 10.47 108 (29) 263 40 52
MC4 9.19 125 (34) 287 35 57
MC5 7.64 127 (34) 288 30 49
MC6 6.23 116 (31) 299 42 49
MC7 6.10 273 (74) 98 70 34
MC8 5.18 174 (47) 212 51 39
MC9 4.89 149 (40) 250 25 57
MClO 4.10 113 (30) 222 31 54
Crystat 357 (loo) 0 93 18
" The rest of the 130 residues are in f3-bendsand f3-bridges.
For comparison, the structural features of the respective crystal structures are shown.
121 121
101 101
81 81
61 61
41 41
21 21
1 I
I21 121
101 I01
81 81
61 61
41 41
21 21
1 I
Figure 7 Contact maps of (A) the WT crystal structure, (B) the mutant crystal structure,
(C) the WCI cluster and (0) the MCl cluster. Both axes represent residue
number. The labels on the diagonal correspond to regions of the molecule. The
cut-ofl for a pair of residues to be in contact is 8A between their Ca atoms.
The residue 56 contacts with helix B (rectangle) and helix 310 (dashed
rectangle) are shown.
(1.3 nm) from the distributions of the 300K simulation and the wild type 500K simulations
(around 0.9 nm). The latter however have a small population of conformations where this
distance ranges between 1 and 2.2 nm. On the contrary we observe that I56T mutant
conformations have in most of the time shorter distance between Thr56 and helix 310.A
comparison of the means and standard deviations of the plots are shown in Table 3. The
mean and the standard deviations of the distributions of the unfolding simulations have
been compared using standard statistical methods (t-test and F-test) and found that their
differenceis significant (i.e it does not stem from random data differences).
Table 3 Mean and standard deviation of the distance distribution (in nm) of Ile56-Helix
B and Ile56-Helix 310 contacts.
11e56 - HelixB (nm) Ile56 - Helix 310(nm)

Simulation WT I56T WT I56T
Crystal 0.90 0.91 0.99 1.01
Control 0.91 f 0.04 0.83 f 0.04 0.94 f 0.05 0.99 f 0.04
Unfolding 0.89 i 0.35 1.23 f 0.26 1.34 i 0.42 0.82 f 0.28
Distance (niii)
Figure 8 Distribution of the distances between (A) residue 56 (Ile) to helix 310 and (B)
between residue 56 (Ile) and Helix B in the sampled conformations of the
unfolding and control trajectories. The distances at the native structure are
shown as black (wild type) and red (mutant) rectangles just above the x-axis.
The tendency of Thr56 to approach helix 3 10 (residues 8 1-85)might be related to some

favourable hydrogen bonding occurring in the area. In fact from the contact maps of figure
7D and also from the distribution of residue contacts in figure 5, it is clear that Thr56 has a
high tendency to be in contact distance (less than 8A) with residues 80 and 82. These two
are serines which implies that hydrogen bonding may be involved between them and Thr56
thus giving an explanation for the persistence of the latter to migrate towards helix 3 10.
3.3 Comparison with experimental data and other simulations
Comparisons with experimental data can be made by consideration of the RMSD and
residue contacts against time. Thus, we observe that at 500K the RMSD of the mutant
increases at a higher rate than the RMSD of the wild t e after 1OOOps. These results are in
agreement with data from circular dichroism'and stopped-flow fluorescence
experiment^,^' which show that both natural mutant forms of lysozyme are less
thermostable than the wild type.
In high temperature (498K) unfolding MD simulations carried out in hen egg-white
lysozyme, the partially folded intermediates identified are showing increased non-native
structure. This is consistent with our findings for the most populated clusters of
conformations sampled during the high temperature simulations whereby non-native
contacts are increased across the polypeptide chain. Another common feature is that the
radius of gyration of the intermediates increases slightly (-10%) in comparison to the
300K simulations. These intermediate states show a slight change in non-polar solvent
I Probing Biological Molecules: Theory and Experiment I15
accessible surface area (SASA) in comparison to the crystal structure and again this is
consistent with our findings
During induced unfolding at 500K,both wild type and mutant lysozyme demonstrate
increasing loss of their original secondary structure. The results fi-om the analysis of
secondary structure content against time show that most of the original secondary structure
elements are lost within the frst 2.5 ns. An interesting feature is that the beta domain is
destabilised first and then the alpha domain follows. This is in agreement with previous
unfolding simulations of hen egg-white lysozyme at 5o0K.l8 Also data from refolding
experiments (usin stopped-flow amide hydrogen exchange and mass spectroscopy by
Miranker et a]?’ show that about 80% of the refolding molecules have their amide
hydrogen atoms in the a-domain protected before those in the P-domain. Pepys et al.’ also
showed that the beta domain is more unstable than the alpha domain.
The destabilised secondary structure in both wild type and mutant lysozyme forms
gives rise to random transient p-turns across the whole polypeptide chain. p- turns do not
generally constitute stable structural elements3*which explains their transient nature in our
simulations. However based on both experimental and also lattice
it is argued that p-turns may play a role in the initial stages of the formation
of helices and sheets during folding and thus they appear to direct folding pathways while
tending to adopt conformationsthat minimise the “local” conformationalfree energy of the
residues in the turn!2 In our results (Table 1) we observe that the sampled conformations
of the mutant, contain slightly higher percentage of turns than the WT in most of the
regions. Using Fourier Transform Infra-Red spectroscopy data’ propose that the partly
folded forms of the lysozyme variants associate through the unstable beta domain to form
the initial seed for the generation of amyloids. From Table 1 we observe that for this beta
domain (strand 1, strand 2 and loop) the partially unfolded intermediates of the mutant
have higher beta turn content, which, may be an indication of the relative instability in this
region in the mutant lysozyme compared to the wild type.
Booth and collaborators7suggest that the motion in the P-domain due to the mutation
may not be the direct cause for amyloidogenicity. Rather, changes in the interface region of
the two domains, transmitted fiom the disturbed P-sheet may be the actual cause. In
particular Ile56, found to have increased B-factors for both amyloidogenic mutants of
human lysozyme, could play an important role in the mutant’s amyloidogenic properties.
In our unfolding simulations, the fluctuation for this residue is higher in the variant and
thus in agreement with these experimental results.
4 CONCLUSIONS
Our results show that the unfolding of the mutant appears to occur faster than the wild type
and we observe that it takes longer for the wild type to lose its native structure, under our
artificial higher temperature conditions. In addition to that, in the conformational space
sampled, it seems that the wild type retains higher content of its original secondary
structure in most of the regions relatively to the mutant. Both wild type and mutant
simulations include conversions of the original secondary structure conformations to coils
and beta turns. These f3-turns, however, seem to be more predominant in the beta domain
of the mutant, a region which has been suggested to be involved in fibril~genesis?’~ Also
in the beta domain we observe higher atomic fluctuations in I56T than the wild type
implying that this region is more flexible and thus reinforcing the previous data. However,
Thr56 appears to be more rigid than Ile56 with the former tending towards helix 310.Turns,
are thought to occur in abundance prior to formation of helices and sheets during folding4’
and thus the overall higher content in turns in the sampled unfolding conformational area
of the mutant may be an indication that it unfolds faster than the wild type.
The clustering analysis reinforces the points mentioned above. The most populated
clusters of the wild type retain near half of their native contacts compared with one third
for I56T mutant. The lack of retention of native contacts is a useful measure of the degree
of unfolding. So, it appears that most of the conformations of the mutant are further away
from their native structure than those of the wild type and thus, the mutant protein is more
susceptible to changes under unfolding conditions. Interestingly, the most populated
structures of the mutant have their Thr56 near helix 310 unlike the wild type ones. Some
favourable hydrogen bonding with Ser80 and Ser82 may be related to this observation. The
restrain due to this hydrogen bonding to the motion of Thr56, a pivotal residue in the
interface of the two lysozyme domains, could result in destabilisation of the structure of
the molecule that has been suggested is related to amyloid~genesis.~”
In our unfolding simulations we attempt to identify features that distinguish the
partially unfolded conformationsof the WT and 156T human lysozyme. Although there is a
vast diversity in the conformational features sampled for both the forms, the different
distances of the Ile56 and two regions of the alpha domain seems to differentiate a number
of conformations of the mutant from the wild type. The significance of this finding cannot
directly be linked at this time to amyloidogenesis. However, it does demonstrate clearly
that one of the effects of the mutation at residue 56 is the resulting distortion of the
important region at the interface of the two domains.
Acknowledgements
This work has been carried out within the BBSRC sponsored Bloomsbury Centre for
Structural Biology. We thank the BBSRC for computer hardware and the EPSRC for a
studentship to GM.
References
1 D.M. Chipman and N.Sharon, Science, 1969,165,454.

2. P.J. Artymiuk and C.C.F. Blake, J. Mol. Biol., 1981, 152,737.
3. M.B.Pepys, P.N. Hawkins, D.R. Booth, D.M. Viguishin, G.A. Tennent, A.K. Soutar,
N. Totty, 0. Nguyen, C.C.F. Blake, C.J. Terry, T.G. Feest, A.M. Zalin and J.J. Hsuan,
Nature, 1993,206, 553.
4. M.B. Pepys, in Amyloidosis: The Oxford Textbook of Medicine, 3rd Edn., ed. D.J.
Weatherall, J.G.G. Ledingham and D.A. Warell, Oxford University Press, Oxford,
1996, vol. 2, p. 1512.
5 . J.W. Kelly, Curr. Opin. Struct. Biol., 1998,8, 101.
6. J. Funahashi, K. Takano, K. Ogasahara, Y. Yamagata and K. Yutani, J.
Biochem.,1996,120, 1216.
7 . D.R. Booth, M. Sunde, V. Bellotti, C.V. Robinson, W.L. Hutchinson, P.E. Fraser, P/N
Hawkins, C.M. Dobson, S.E.Radford, C.C.F. Blake, and M.B. Pepys, Nature, 1997,
385,787.
8. L.A. Morozova-Roche, J. Zurdo, A. Spencer, W. Noppe, V. Receveur, D.B. Archer,
M. Joniau and C.M. Dobson, Journal of Structural Biology, 2000,130,339.
9. C.L. Brooks III, Curr. Op. Struct. Bio1.,1998,8,222.
10. A. Li, V. Daggett, Pro=.-Natl.AcadSci. USA., 1994,91, 10430.
1 1 . V. Daggett, Cum. Op. Struct. Biol., 2000, 10, 160.
12. A. Li, and V. Daggett, J. Mol. Biol., 1998,275,677.
13. D.O.V. Alonso and V. Daggett, Biophysics, 2000,97,133.

14. M. Karpfusand A. Sali, Cum. Opin. Struct. Biol. 1995,5,58.
15. A.E. Mark and W.F. van Gunsteren, Biochemistry, 1992,31,7745.
16. P.H. Hunenberger, A.E. Mark and W.F.van Gunsteren, Nature: Struct. Funct. Genet.,
1995,21, 196.
17. M.A. Williams, J.M. Thornton and J.M. Goodfellow,Protein Eng., 1997, 10,895.
18. S.L. Kazmirski and V. Daggett, J. Mol. Biol., 1998,284,793.
19. B. Gilquin, C. Guilbert and D. Perahia, Proteins, 2000,41,58.
20. H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren and J. Hermans, in
Zntemolecular Forces. Dordrecht, Reidel. 198 1.
21. M.P. Allen and D.J. Tildesley, in Computer Simulations of Liquids, Clarendon,
Oxford, 1989.
22. H.J.C. Berendsen, D. van der Spoel and R. van Drunen, Comp. Phys. Comm., 1995,
91,43.
23. W.F. van Gunsteren, S.R. Billeter, A.A. Eising, P.H. HUnenberger, P. KrUger, A.E.
Mark, W.R.O. Scott I.G. Tironi in Biomolecular Simulation: The GROMOS96 manual
and user guide, HochschulverlagAG an der ETH Zurich, 1996.
24. T. Darden, D. York and L. Pedersen, J. Chem. Phys., 1993,98,10089.
25. B. Hess, H. Bekker, H.J.C. Berendsen and J.G.E.M. Fraaije, J. Comp. Chem.,1997,
18, 1463.
26. H.J.C. Berendsen, P.M. Postmaj, W.F.van Gusteren, A. Di Nola and J.R. Haak, J.
Chem .Phys. 1984,81,3684.
27. W. Kabsch and C.Sander, Biopolymers, 1983,22,2577.
28. P.C. Mahalanobis, Journal of the Asiatic Society of Benagal, 1983,26,541.
29. S.L.,Kazmirski, A. Li and V. Daggett, J. Mol. Bio1.,1999,290,283.
30. M. Otto, in Chemometrics: statistics and computer application in analytical
chemistry, Weinheim, Cambridge, 1998.
31. B.S. Everitt, in Cluster Analysis, London, E. Arnold New York Halsted, 1995.
32. H. Spilth,. in Cluster analysis algorithms for data reduction and class@cation of
objects, Chichester Ellis Horwood New York Chichester, 1980.
33. W.N.Venables and B.D.Ripley, in Modem Applied Statistics with S-PLUS.3rd Edn.,
Springer, 1999.
34. G. A. Worth,F. Nardi and R.C. Wade, J. of Phys. Chem. B, 1998,102,6260.
35. D. Canet, , M. Sunde, A.M. Last, A. Miranker, A. Spencer, C.V. Robinson and C.M.
Dobson, Biochemistry,1999,38,6419.
36. A. Miranker, S.E. Radford, M. Karplus and C.M.Dobson, Nature,1991,349,633.
37. A. Miranker, C.V. Robinson, S.E.Radford, R.T. Alpin and C.M. Dobson, Science,
1993,262,896.
38. D.J. Tobias, S.F. Sneddon and C.L.Brooks III, J. Mol. Biol., 1990,216,783.
39. S.S. Zimmerman and H.A. Scheraga, Proc. Natl Acad. Sci. USA.,1977,744126.
40. H.J. Dyson, J.R. Sayre, G. Merutka, H.C. Shin, R.A. Lerner and P.E. Wright, J. Mol.
Biol., 1992,226, 819.
41. J. Skolnick and A. Kolinski, J. Mol. Biol., 1991,221,499.
42. A. Yang, B. Hitz and B.Honig, J. Mol. Biol., 1996,259,873.
COLLECTIVE EXCITATION DYNAMICS IN MOLECULAR AGGREGATES:
EXCITON RELAXATION, SELF-TRAPPING AND POLARON FORMATION.
M. Dahlbom, W. Beenken, V. Sundstrom, T. Pullerits
Department of Chemical Physics, Lund University, Box 124, S-221 00 Lund, Sweden
1 INTRODUCTION
The excited state properties and dynamics in various types of condensed systems are of
great interest to the community of chemical physics. In particular, a diverse range of
molecular systems where collective excited states occur has in recent time been under
intense study due to their cooperative electronic behavior. Systems, such as conjugated
polymers, fluorescent proteins and biological light-harvesting systems have attracted a
great interest since ultra-short laser pulses became available and a large number of studies
have been performed both theoretically and experimentally, see review' and references
therein.
In these types of systems, the inter-molecular interactions can lead to delocalization of
excited states, denoted exciton states (excitons). Molecular excitons were introduced by
Frenke12 early in the last century and further studied by Davydov3. The properties and
ultra-fast dynamics of cooperative excited state behavior in molecular aggregates have
been studied extensively over the last decade. For example, a number of studies have
utilized different techniques in order to characterize the exciton delocalization in bacterial
photosynthetic It has been suggested that an optical excitation in the tightly
coupled peripheral antenna of purple bacteria spans a considerable size of the aggregate
and relaxes within a few hundreds of femtoseconds to an equilibrium size of just a few
chrom~phores~. Delocalized excitations can also be trapped at lattice sites due to the
displacement of the excited state potential and the localization of the wave function as a
result of the excitation-phonon coupling. Peierls6 and Frenkel' introduced the possibility of
exciton self-trapping as early as in the 1930's. Excitons that interact strongly with the
nuclear degrees of freedom are denoted polarons. Due to this coupling they distort their
surrounding to a considerable degree. Polaron formation has, for example, been suggested
'.
to occur in photosynthetic antenna aggregates at low temperatures Polaron with their
sensitivity to nuclear motions can also be seen as possible non-destructive spectroscopic
probes of protein andor membrane dynamics.
The aim of this paper is to study the collective excitations coupled to the molecular
vibrations in the real-space representation. We present a model that treats the molecular
vibrations classically via the Langevin equation whereas the electronic system is treated
quantum mechanically'. The excitation dynamics can henceforth be studied as a function
of the molecular vibrations. It will be shown that in the presented model, following an
optical excitation, the excitons are trapped at low energy sites (self-trapping) due to the
population enhancement of the displacement factor introduced via a mean field correction.
The degree of excitation migration and self-trapping depends on temperature, exciton-
phonon couplings (Huang-Rhys factors) and static distribution of site energies"*". Special
interest is directed towards the real-space motion of the excitation on the femtosecond
time-scale and polaron formation. In order to distinguish between the two processes, we
studied the wave function mobility and localization as a function of the vibrational
frequency, displacement, temperature and intermolecular interaction strength. Besides the
general considerations of real-space excitation dynamics in molecular aggregates on the
femtosecond time-scale there is also the question of large scale energy funnelling in
photosynthetic light-harvesting pigment-protein systems. In order to make the transfer
between the B850 aggregate and the LH1 antenna complex more efficient in purple
bacteria, the excitation in the B850 ring should be localized at the junction where the
distance to the LHl is the closest. Polivka et. a1.* has attributed the appearance of a red-
shifted feature in the transient absorption spectra at cryogenic temperatures to polaron
formation. They propose that the polaron may be formed in inter-ring connection points.
We will not address this problem specifically in the current work, but rather address the
global aspects of the polaron formation as a viable process in photosynthetic light-
harvesting systems.
2 THEORY
According to the Born-Oppenheimer approximation we separate the electronic degrees of
freedom from the nuclear motion. The electronic system will be treated quantum
mechanically in the molecular exciton picture213,while for the nuclear motion a classical
approach will be used. Furthermore, the nuclear system will be separated into explicit and
bath modes. The explicit modes are assumed to be coupled the electronic system via the
adiabatic potential energy surface. To take into account the nonadiabatic coupling of the
electronic system to the explicit modes, we use Tully's surface-hoppingmethod'. The other
modes will be described by use a Caldeira-Leggett" type of bath resulting in statistically
fluctuating external forces driving and damping the explicit modes.
2.1 Frenkel Excitons

According to the Born-Oppenheimer approximation'2 the molecular electronic
Hamiltonian is given by
where q and r ’, assign electronic coordinates and R,, the nuclear ones with index n denotes
the nuclear coordinate. h2V’ represents the electron momentum and m, the corresponding
mass. The nuclear charges are denoted as 2,.In what follows, it has been assumed that the
multi-electron problem is solved for the single molecule to obtain the single molecule
energies as h j , and the excitation creation and annihilation operators B: and B j ,
respectively, where j denotes the molecular sitei3. The creation and annihilation operators
are given by
where a: and a, are the creation and annihilation operators for electrons in the molecular
orbitals la) and 16) , and bjabare the corresponding expansion coefficients. Due to the
adiabatic approach the single molecule energies hi as well as the operators B: and B j
depend parametrically on the nuclear coordinates R,, . The electronic Hamiltonian can be
written in the excitonic representation as
where J j i describes the excitation transfer matrix elements between different molecules
(depending on nuclear coordinates R,, as well), For sake of simplicity we will restrict our
consideration to molecules that are well described as two level systems. Introducing the
1 I
molecular ground and excited states as #;,I) and @je)), respectively, we can expand the
exciton eigenstates as
These states can be obtained by solving the eigenvalue equation
The eigenstates la), as well as the expansion coefficients caj(R,,)= are

dependent on the nuclear coordinates R,, , The eigenvalues U,(R,,)yield the adiabatic
potential energy surfaces for the nuclear motions as
depending on the nuclear coordinates R,, via h~x)(R,,),

h y ) ( R , , ) ,J,(R,,) and c,,(R,,).
After transformation of the nuclear coordinates R,, to molecular normal coordinates qp,j as
performed in the next section, for an aggregate consisting of two-level system molecules
one can write down the diagonal matrix elements of the general electronic Hamiltonian in
the site representation as
H;y)(r) = E j - x u p , j d , , j q f l , j . (7)
CC
Herepdenotes the M different explicit modes with frequency belonging to the j’th
molecule. d,,j is the displacement between the minima of the excited and ground state
potential energy surfaces, and E j the energy gap between excited and ground state for
q,,j = 0 , means minimum of the ground state potential surface. This Hamiltonian is taken
in the Frank-Condon approximation and both, the ground and excited state potential energy
I I b
q
U
d
Figure 1. The two level schemes used for the individual molecules where the energy
of the direct Frank-Condon transition from the ground state is E’ and the energy
gap is H::).
surfaces are taken to have same slope means frequencies

In the present work, we restrict ourselves to one explicit mode per molecular site with a
unique frequency w and displacement d . If also the energy gap is site independent, Eq.
(7) simplifies to
2.2 Explicit Modes

In the gas-phase the molecular nuclear system can be described by three kinds of motion:
translation, rotation and vibration. In the condensed phase the surrounding matrix hinders
the first two. Nevertheless, it is useful to distinguish between inter-molecular motion,
which means that the whole molecule changes its position relative to the neighbors; and
intra-molecular motion, which means that the nuclei of a single molecule change their
position and orientation in respect to each other. Inter-molecular motion will mainly affect
the interaction matrix J,.(R,), while intra-molecular motion has also an effect on the site
energies hjg)(R,) and h;.‘)(R,). In what follows, we will neglect the inter-molecular
motion and focus on the intra-molecular motion. Expanding the ground state potential
energy surface h:g)(Z?fl)of sitej around its minimum at Rt’) one can write
By transformation to principal axis we obtain a multidimensional harmonic oscillator

potential in the normal mode representation, given by
using dimensionless coordinates qjs. The excited state potential hic)(&) of site j can be
expanded using the same coordinates as
where Wtf take into account a different curvature and djea different minimum of the
potential for the excited state. For sake of simplicity, the curvature of excited as well as
= J5{,.
ground state potential surface shall be unique too, i.e. Wtet Now one can deduce from
Eqs.(6), (10) and (11)the excitonic potential surface as
1
U,(. . . q j . . . ) =E+-Cttw,q; - ~ A u j q j d j c ;t j~ J u c , i c , j ,
2 i i i.j
One has to note that the cai still depends parametrically on qj . The potential surface given
in Eq. (12) will be used to determine the intrinsic forces for the nuclear motion with
respect to the site modej as
Thus Eq. (13) leads us to Hook's force law. This is the core of our theoretical model.
2.3 Bath Modes

The complete Hamiltonian for the nuclear motion on the potential surface U, (...q j ...) is
given by
Here the index 4 assigns the bath modes, while the indexj assigns the explicit site modes
as above. Following Caldeira-Leggett", the bath modes are assumed to be harmonic
oscillators and the coupling between bath and explicit modes to be bilinear in the
coordinates qj and qC with coupling parameter K i t . The dimensionless momentum pi.,
are related to the coordinates qj,{ by the canonical equations
Resolving these canonical equations for the explicit mode and treating the bath modes
statistically (see ref.10 for details), Eq. (16) results in the equation of dissipative nuclear
motion
where the intrinsic forces Fuj(...qj...)are given by Eq. (13). The stochastical bath forces
fj(t) are connected to the dissipation, which is given by yj(z), according to the
dissipation-fluctuationtheorem as
For a thermalized bath the momentum variance (p:) is given by
1
(P:) = +-1
exp[hoj / k,T] - 1 2
If one neglects the memory of the bath, i.e. invokes the Markov approximation, Eq.(17)
reduces to the Langevin equation12
where the integral yj(t -t’)pj(t’)dt’ is substituted by the damping term yjpj.The
dissipation-fluctuation theorem time-averaged gives the variance of external forces f j(t)
as
where r* is the correlation time of the fluctuations’.
3 NONADIABATICITY AND THE SURFACE HOPPING METHOD
In general, the time-dependent one-exciton wave function IY(t))can be expanded in the

adiabatic basis set la) and propagated by
’ For the numerical calculations we will use random f j( t , ) and set Z*equal to the time-step At = t, - f,-,
of the propagation. This is justified because the value of f j ( t ) will not change within the time interval
[t, , t , + At[ , which means that f j( t ) is fully correlated.
where C,(t) are time-dependent expansion coefficients. One has to note that the eigen-
functions la) as well as the eigen-energies U,(r) are time-dependent due to the
parametric dependency of the eigen-value problem Eq.(S) on the nuclear coordinates.
These, now represented by normal coordinatesq j are propagating in time according
Eq.(20). From Eqs.(4) and (15) one can deduce the identity
I
for the nonadiabatic coupling between two excitonic eigenstates a) and IB> . The left side
of the equality means that for a finite nuclear motion, i.e. pi f 0, the adiabatic excitonic
states are mixed by the nonadiabaticity of the electronic Hamiltonian beyond Born-
Oppenheimer approximation'*. On the right side the same coupling is represented by the
time evolution of the coefficients cai. The latter can be determined without explicit
knowledge of their dependency on the coordinate normal coordinates qj by
diagonalisation of the excitonic Hamiltonian in the site representation, with diagonal
elements H , as given in Eq.(8) and off-diagonal elements as H, = J , .
3.1 Surface Hopping Probability
Instead of a fully coherent propagation of the exciton, in this work we will apply the
surface hopping method introduced by Tully'. Therefore, the time-dependent Schrijdinger
equation for the exciton wave function, given as
is transformed into a system of differential equations for the coefficients C&) = (alY),
yielding as
using transport coefficients given by

where U,(z)- U,(z)= hw,(z). Since one can write down the probability to jump from a
exciton eigen-state la) to another1 a) within a time interval [ f , , t ] as
using the density matrix elements p,(z) = Ci(z)Cg(z),the relation k>(z) = -kb(z), Eq.
(25) and where to is the time of the last hop. According to Eqs.(26) and (23) the transport
coefficients k,(z) are mainly determined by the nonadiabatic coupling. With other words,
P ( a -+p) gives the probability for the nonadiabatic hopping between adiabatic potential
surfaces. Between the jumps, the exciton propagation is performed coherently on the
adiabatic potential surfaces due to the exponential term in Eq.(26).
3.2 The Surface Hopping Algorithm
In the surface hopping method, two sets of states are simultaneously followed, the
reference and the auxiliary states. The latter are propagated using Schrtjdinger equation as
discussed above and used for calculating the surface hopping probabilities. The
populations on the auxiliary states are propagated coherently. The populations on the
reference states are always 0 or 1; it will only change if the surface hopping occurs. In
figure 2 below, the reference states are shown verses time and the thick black solid line
corresponds to the populated state at that given point in time. The surface-hopping
Algorithm can be defined by following steps:
(i) The system is initialized, as the electronic Hamiltonian at time zero is generated
using a random distributed set of coordinates and momenta, {qi(0),~ ~ ( 0, )corresponding
)
to the given system temperature. Solving the SchrMinger equation for this configuration
one obtains the initial exciton energies and wave functions and the initial population can be
created in a predefined manor. Note that at time zero the reference states and the auxiliary
states are the same, which is not true for the following time steps.
(ii) The explicit mode oscillators are then propagated a time step At using the Langevin
equation, Eq.(20), and the new electronic Hamiltonian is generated. The eigen-value
problem is solved generating new states and energies. The populations on the reference
states are always integer, but the populations on the auxiliary states are propagated
according to Eq.(24).
(iii) Based on the auxiliary states the hopping probability for the reference configuration
is calculated using Eq.(27).A random number 6 is pulled from a uniform distribution [0,1]
and the population is moved from state la) to I,O) if the criterion
is fulfilled. Otherwise the decision is taken not to jump. Then one continues with step (ii).
3.2 Localization and Motion of Excitons
As discussed in the Introduction, the states in a molecular aggregate with a sufficiently

large intermolecular interaction will be more or less delocalized and one usually denotes
them as exciton states. If, however, the excitons couples strongly to the (intra- or inter-)
molecular vibrations, localization may occur, this means polaron formation. In our model
the polaron formation results from a feedback between excitonic and nuclear Hamiltonian,
as follows: Since the exciton states are delocalized to some degree by dipole-dipole
interaction, each individual molecular site will contribute only a certain amount to the
exciton wave-functions according to Eq.(4). However, the displacement d is weighted
with these contributions to give the internal forces in Eq.(13). For occupation of the lowest
excitonic state, these forces will drive the molecular site that contributes most to the
exciton wave-function to a lower molecular site energy. Consequently, this site will give a
still higher contribution to the lowest excitonic state than before. This results in a self-
amplifying increase of the exciton localization. Likewise the mobility of the exciton
decreases - eventually up to self-trapping- since it becomes harder for the excitation to slip
out of the trap to a neighboring site. Nevertheless due to Eq. (20) at finite temperature the
external forces from the thermal bath may statistically kick the system out of the self-
trapping configuration. On the other hand, for small displacements localized polarons will
not be formed. In this case only the center of the delocalized excitonic wave function may
fluctuate thermally, due to the kicking bath forces.
To study the localization quantitatively we use the customary inverse participation ratio,
defined as
where thecja(t) are the parametrically time dependent expansion coefficient of the exciton
eigen-state la).The diagoal density matrix element p,(t) is just the population of this
eigen-state and results from Eq.(25). Z represents the normalization factor5. In order to
analyze the mobility of the exciton we begin by introducing the exciton (polaron) position
operator Q = Z j B j B , ,following Mak et. al.I4, to obtain
i
This quantity gives information on which site is the center of gravity of a localized as well
as a delocalized excitation. However, it can trace the motion of an exciton only for single
-
13500
13000
I
5* 12500
E?
g 12000
w
-# 11500
5
5 11000
*
.I
0
10500
10000
0;o Oil 092 093 094
Time (ps)
Figure 2. The energies for the electronic states overlaid with the current reference
state (thick solid line). The Huang-Rhysfactor is 1,125and the temperature OK.
Monte-Carlo trajectory. For sampling over a huge amount of trajectories on homogeneous

aggregates it yields only a constant averaged value. In order to remedy this situation, we
define the velocity of an exciton as
which will be not canceled by the Monte-Carlo averaging. If the velocity has a high value,
the exciton moves rapidly through the system, probably on a random walk. However, if the
value of Z ( t ) is small the interpretation is ambiguous: either the exciton is trapped, e.g. by
“small” polaron formation, or it is delocalized in such a degree, that its motion does not
result in a significant change of the site populations. The second kind of interpretation can
be only excluded, if both, I ( t ) and bPr(t) are small, means the exciton is trapped and
localized on mainly one site.
The model system considered here is a linear aggregate of six identical two-level systems
with the molecular optical transition at 12400 cm-' and a value of the nearest neighbor
intermolecular interaction set to 342 cm-' for all sites. We restrict ourselves to a single
explicit intra-molecular uncorrelated vibrational mode per molecular site with a unique
200
Huang-Rhys factors c
3 150-I +0.125
- * -0.5
- -A- 1.125
-v-*2.0
034 096
Time (ps)
Figure 3. The exciton (polaron) velocity as a function of time, where panel A displays
Wvibequal to 250 em-', B 684 ern*', C 900 ern-'. The symbols are displayed in the legend of
panel C.
vibrational frequency and displacement d.

The electronic level dynamics is displayed in figure 2. The excitonic states are
displayed using different line styles; the thick solid line represents the currently populated
reference level. In the displayed case the population has relaxed to the lowest exciton state
in a few hundred femtoseconds. On the longer time scale the broadening of the energetic
band gap can be seen.
For detailed analyzing of the polaron formation we start to figure out the influence of
the displacement d (expressed as Huang-Rhys factors S = 0.5 d 2 ) on the localization
(L,,(t) ) and the exciton velocity (I(t)) of the exciton. Figures 3 and 4, where the Huang-
Rhys factor is varied between 0.125 and 2.0. shows the strong dependence of both
quantities on the mode frequency and displacement (Huang-Rhys factors). The influence of
the displacement on I(t) and L,,,is easy to understand since it controls the molecular
1: , , , , , , . , ,
0,o 092 0,4 0,s 0,s 1
h
i
Figure 4. The inverse participation ratio displayed as a function of time. The symbols
and parameters are the same as in figure 3.
energy gap according to Eq.(8). Due to the feedback channel in the Langevin equation, the
exciton-phonon coupling increasingly influences the localization.
We know from the dimer that the inter-level transfer rates are maximized if the
vibrational frequency is tuned to twice the inter-molecular interaction energy. In this case
the vibrational frequency matches exactly the energy gap between the levels and a
resonance occurs. We tuned the inter-molecular vibrational frequencies from 250 cm-'as
shown in Panel A, 642 cm-' Panel B up to 900 cm" as in Panel C. There is a fast initial rise
followed by a monotonic decrease in all kinetics in figure 3, panels A through C. Even
though the time-scales change, the dynamics behaves similarly in all three cases, i.e. the
velocity decrease as the vibrational frequency increase.
*I
35
30
A
v)
25
y. I
$ 20
CI
'5
v
15
g 10
5
0
B
Temperature
+OK
L 4
; a - * -100K
- *A - 200 K
--.300 K
P 3 \:.
Figure 5. The time-dependence exciton velocity and inverse participation ratio

displayed for four dflerent temperatures. Intermolecular coupling V= 100 em",
w=200cm-' and the Huang-Rhysfactor is 1.125.
In figure 4, there is a fast initial decrease of Lip* followed by a slower phase present in all
the three cases. From the frequency dependence, we conclude that the polaron formation
occurs faster at higher vibrational frequencies. Further, it is clear that even for very low
temperatures, were the vibrations are almost negligible, polaron formation depends on the
ratio of the inter-molecular interaction and the Huang-Rhys factor multiplied with the
vibrational frequency. Stated differently, polaron formation does not occur until the
6
B
5-
Temperature
+OK
4- - * -100K
i i -
- A - 200 K
i-3- --a300 K
2-
1 I I I 1 I 1 I I
Figure 6. The exciton velocity and inverse participation ratio, with a intermolecular
coupling V= 342 cm-l, u=684cm-' in the hexamer model system, as a function oj
temperature. The Huang-Rhysfactor is 1.125.
coupling between the nuclear and electronic degrees of freedom become sufficiently strong
to overcome the delocalization of the intermolecular dipolar coupling.
Next we turn our attention to the temperature dependence for the inverse participation
ratio (Lipr)and the exciton velocity (I(t)). The important parameter for the temperature
dependence is the ratio of the vibrational quantum and the thermal energy w,,, / kT . We
performed two sets of simulations; the first with V=100 cm-' and wv,=20O cm-'. The
temperature was varied from 0 to 300 K. The second set was done exactly in the same way,
but for V=342 cm-' and mvib=684 cm-'. Figure 5 displays the first case with panels of both,
the exciton velocity (see panel A) and the inverse participation ratio (panel B) and in
Figure 6 corresponding data is displayed for the second scenario. In Figure 5 , the ratio
between vibrational and thermal energy is close to unity, and temperature dependence is
clearly visible. The velocity (I(t)) at time zero is a factor of 1.5 larger for 300K than for
OK, but after Ips the velocity at 300K is nearly twice as large as at OK. In panel B, the
inverse participation ratio (Lip,)is displayed as a function of time and temperature. Here
the temperature dependence is slightly, but still clearly visible below 400fs, indicating the
importance of the ratio k , T / A o mentioned above. In Figure 6 , the ratio between the
intermolecular vibrational frequency and thermal energy is much larger than unity. There
is weak temperature dependence visible in panel A. The final velocity for the different
temperatures lies between 10 and 20 sites/ps.
For all temperatures the inverse participation ratio is between 1 and 2 molecules. However,
the time dependence of the velocity I(t) and the inverse participation ratio Lip is
qualitatively similar for all temperatures. The apparent difference between the panels in
figure 5 and 6 is deceiving. Due to the lower interaction energies and vibrational
frequencies only the time scale is shifted. If one takes this into consideration the two panels
show very similar time dependence.
Finally we want to turn the attention to the exciton band gap of the molecular aggregate.
The polaron formation will involve a lowering of the energy of a specific site due to the
feedback channel introduced in Eq.(13), hence causing the lowest exciton energy to
decrease and widening the exciton band. The band gap is here defined by the energy
difference between the lowest and highest exciton states and is displayed in the two panels
in Figure 7. Panel A shows the averaged exciton energies and panel B the time-dependent
band gap. In figure 7 two effects can be seen when the exciton-phonon coupling is
increased. The increase of the band gap due to the increase of the molecular site
fluctuations, i.e. the increase of the slope in panel A and the initial rise in panel B.
Secondly, the polaron formation due to the decrease of the lowest exciton state energy so
that it energetically deviates from the rest of the equally spaced exciton energy levels. The
first effect arises from the displacement dependent Hamiltonian in Eq.(8) which causes the
energy fluctuations of the excitonic states to become larger with increasing Huang-Rhys
factors, The decrease of the lowest exciton energy is a result of the polaron formation, i.e.
the energetic trap is formed by the interaction of the localized exciton and the molecular
vibrations on the occupied site. The second effect indicates that the polaron formation
process has come into balance with the delocalizing effect of the inter-molecular
interactions. Further, it is clear that after a few hundred femtoseconds the band gap stays
constant for all Huang-Rhys factors. This is caused by the exciton-vibrational coupling,
since an increase of Huang-Rhys factor does not only lower the energy on the populated
site but does also increase the amplitude of the molecular energy fluctuation.
To summarize, we have proposed a model, based on the surface-hopping method in the
real-space (site) representation, capable of calculating the excitation dynamics in molecular
aggregates where the molecular vibrations are incorporated on a real-time basis. We have
-n13500 4 Huang-Rhys factom

-m-O.125
- 0 -0.5
. *A.1.125
-r.2.0
5 10500
c,
'C 10000
la 9500 1 I I I I i
6 5 4 3 2 1
Exciton State
4000 I
B
I
2
3500 1
0 3000
P
E
# 2500
0
.-
0
2000
1500
Figure 7 . Panel A displayes the exciton energies as a function of state numberfor four
different realizations of the Huang-Rhysfactor. Panel B shows the time dependent exciton
band gap for the same Huang-Rhysfactors as panel A. The parameters are the same as in
figure 4.
shown that this model can capture such phenomena as polaron formation and self-trapping.
The dependency on temperature, exciton-phonon coupling, phonon frequency and
intermolecular interaction strength were investigated.
Reference:
1. V. Sundstriim; T. Pullerits; R. van Grondelle, J. Chem. Phys. B 1999, 103,2327.

2. J. I. Frenkel, Phys. Rev. 193 1,37, 17.
3. A. S.Davydov, Theory of Molecular Excitons, Plenum Press: New York, 1971.
4. M. Dahlbom; T. Minami; V. Chernyak; T. Pullerits; V. Sundstrtim; S . Mukamel, J.
Chem. Phys. B. 2000,104,3976.
5 . M. Dahlbom; T. Pullerits; S. Mukamel; V. Sundstriim,J. Phys. Chem. B. 2001,105,
5515.
6. R. Peierls, Ann. Phys. 1932,13,905.
7. J. I. Frenkel, Phys. 2. 1936,9, 158.
8. T. Polivka; T. Pullerits; J.L. Herek; V. Sundstriim, J. Chem. Phys. B 2000,104, 1088.
9. J. C. Tully, J. Chem. Phys. 1990,93, 1061.
10. U. Weiss, Quantum Dissipative Systems; World Scientific: Singapore, 1999.
11. S. Mukamel, Principles of Nonlinear Optical Spectroscopy; Oxford University Press:
New York, 1995.
12. I.B. Bersuker; V.Z., Polinger Vibronic Interactions in Molecules and Crystals;
Springer Verlag: Berlin, 1998.
13. F. C. Spano, Phys. Rev. k t t . 1991,67,3424.
14. C. Mak; R. Egger, Phys. Rev. E. 1994,49, 1997.
SURPRISING ELECTRO-MAGNETIC PROPERTIES OF CLOSE PACKED
ORGANIZED ORGANIC LAYERS- MAGNETIZATION OF CHIRAL
MONOLAYERS OF POLYPEPTYDE
Itai Carmeli', Viera Skakalova', Ron Naaman'*, Zeev Vager2
'Department of Chemical Physics

2Departmentof Particle Physics
Weizmann Institute, Rehovot 76100, Israel
1 INTRODUCTION
Close-packed, organized organic layers are the focus of substantial studies in recent years,
due to their abilities to modifl electronic properties of substrateslV2,metals or
semic~nductors~~~, and to serve as elements in modern optical' and electronic devi~es,6'~'*'~
light emitting diodes", solar cells", sensors,12etc. Closed packed molecular system are also
the main building blocks in biological membranes.
It is usually assumed that the electronic properties of the adsorbed molecules are similar
to that of the isolated molecule or of the molecule embedded in an isotropic medium. The
weak coupling between the molecules in a monolayer seems to support this notion. This is
taken as a justification to use molecular based calculations for predicting the properties of
the monolayer.13*'
In what follows we present theoretical and experimental results that point to the fact that
this assumption is generally not justified and that properties of molecules can vary
significantly upon adsorption to a close packed layer. This observation may be of
importance in understanding physical properties of natural membranes. In the present work,
by studying well-characterized monolayers of polyalanine we are able to obtain an insight
on the details of a mechanism that may account for the previously observed magnetic
behavior of biological membrane^'^,
It is known that adsorbed layers can affect the work function of the substrates by acting
as a dipole layer in which a force is exerted on the electrons while passing through it. Hence,
the energy required to remove an electron from the substrate (that is, the work function)
depends on the layer's dipole density.
When atoms are adsorbed on the surface, the dipole layer arises from either charge
transfer between the substrate and the adsorbate layer or an induced polarization of the atom.
It has been known for quite some time that for adsorbed atoms the size of the dipole of the
layer is limited due to the dipole-dipole interaction between the adsorbed species. Clearly
the atomic adsorbates are modified and their electronic structure differs from that of the
isolated atoms16. The dipole properties result directly from the adsorption process.
The situation is different when molecules are the adsorbates since the fiee molecules may
carry their own dipole moment. A simple picture is that the properties of the dipole layer
result solely fiom the molecular properties. Indeed, several studies””* find correlations
between the dipole moment of the isolated molecule and the dipole density of the adsorbed
layer. Furthermore, this approach looks as if this is a good way to achieve a very high dipole
density. The rationalization is as follows. Strong chemical binding between the adsorbed
molecule and the substrate provide enough energy to overcome the dipole-dipole repulsion,
keeping negative the total change in the free energy of the adsorption process.
However, it has been recognized already that when molecules are assembled into a dense
packed layer, their electronic structure varies. The surface potential theory of thin films on
water or metal substrates is probably best described by Taylor and Bayslg. They used a
perturbative approach to deal with the effects caused by the layer formation. Similar
approaches have been taken in more recent publications that show that indeed the dielectric
constant of an organized molecular monolayer varies with the layer density?”*’
The theoretical approach taken by Taylor and Bayslg deals with an effective molecular
dipole M which is modified from the free molecular dipole p by the self consistent electric
field in which the dipoles are immersed. The modification occurs through the Stark effect on
the free molecular states and is calculated by first order perturbation theory as
+ - + +
M=p+aE
where a is the electric molecular polarizability.
A relative permeability (dielectric constant), E, is defined by
&=- P
M
The condition for the first order perturbation to be valid is
I&- 1)<< 1 (3)
It is shown here that for packing of molecules with substantial dipoles into a monolayer,
the quantity IEI, defined by Eq. 2, is always much larger than one. Furthermore, the effective
dipoles within monolayers diminish drastically in a non-pertubative manner. Most
importantly, as in adsorbed atoms, the electronic states of the adsorbed molecules differ
considerably fkom the states of the free molecule. This electronic rearrangement by
adsorption to a monolayer, must be accompanied by a current and plays a substantial role in
defining the properties of the adsorbed layer. It alter significantly the electronic properties of
the molecule imbedded in the monolayer, as compared to the isolated molecule or a
molecule in an isotropic medium. Following this part, it will be shown that when the
molecules are homochiral, a splitting between their spin states occurs which may result in a
magnetic layer. This is due to the electronic rearrangement and the electric potential on the
layer.
2 THEORETICAL CONSIDERATIONS
2.1 Substantial charge rearrangements
We start by following the electrostatic formulation as in ref. 19, except that E will be treated
empirically. Denote by q the effective charge separation in the molecule before adsorption
and w the effective distance of the charge separation, such that p = qw is the electric dipole
of the unadsorbed molecule. Idealize the layer as an electrostatic bilayer with smoothly
distributed opposite charges on each side of the adsorbed layer. The displacement vector D
of a bilayer is confrned to the inner volume of the layer and it is zero elsewhere. Therefore,
by Gauss law (esu), it is given by
D=-=&
4F!
(4)
fl
where o is the effective area occupied by a single adsorbed molecule and E is the physical
electric field, also confined within the layer. A typical bilayer potentialjump of
V = w E = -4ZP
(5)
€0
occurs along the width w of the layer. This potential jump is compared with a fictitious
bilayer of effective electric dipole moments M such that
41tM
V=- (6)
0
Thus the effective electric dipole moment of an adsorbed molecule is
M = -P
&
(7)
which is the same as Eq. 2 with no reference to first order perturbation theory.
The conversion of Eqs. 4 and 6 to practical units (Debye, A, and eV) results in
38M( Debye)
E( Volts/ A) =
v(A3)
38M(Debye)
V (Volts)= (9)
0(A2)
where v=w o is the effective molecular volume.
Already here, the validity of first order pertubation theory for E is suspect, since for ~ = l
(Eq. (3)) unrealistically large electric fields are obtained.
The real obstacle for reaching large electric fields in thin films (as well as bulk material) is
the dielectric breakdown. For example, highest breakdown electric field in a very thin film
ofpolypropylene22is
ED* = 0.06 (V/A) (10)
Using Eq. 8 with this limitation, an upper bound is found for the effective dipole in
monolayers
A4 (Debye) 5 1.6.10” v (A3) (1 1)
The upper bound for the dipole moment obtained from this consideration, is not related to
the initial electric dipole moment of the molecule. As is pointed out in ref. 23, the breakdown
mechanism may be related to the wave-like character of the electrons when laterally
confined and therefore related to a universal phenomenon. In general, this upper bound is
much smaller than the unadsorbed molecular dipole moment, hence by Eq. 7, IE~>>I.Figure
1 presents the calculated effective dielectric constant and the effective electric field within a
layer as a hnction of the dipole moment of the adsorbed molecules. It is assumed that each
molecule occupies an area of 40 and that the thickness of the layer is of 20 A. Since the
electric field cannot exceed the breakdown limit, the effective dielectric constant of the layer
must increase linearly with the free molecule dipole moment.
n 0.05
W 15
J
c
c
m
c
v)
0.04 E
P
El0
0.03'p
5-
.P
II
0.01
0.00
0 1 2 3 4 5 6 7 8 9 10
Dipole moment (Oebye)
Figure 1: The relative dielectric constant (solid line) and the electric j e l d within a
monolayer (dashed line) as afinction of the dipole moment of the molecules, assuming that
each molecule occupies an area of 20d2and its length is 20 A.
2.2 Chiral thin films
As a result of the above discussion, it is clear that intensive charge rearrangement occurs
within the molecules of the layer. The rearrangement happens with the aid of the
surroundings. This surrounding can be a substrate, in the case of a layer adsorbed on solid,
or a solution in the case of Langmuir films or natural membranes. A formal way to express
charge rearrangement is by strongly mixing the ground state either with excited electronic
states of the molecule itself or of the surroundings. This process allows a finite probability
that the molecules will be left with unpaired electrons.
In most molecular layers, the electric field is just below the breakdown field (due to
previously existing dipole moments). This symmetry breaking huge field cannot split the
spin states of the electrons, unless another intrinsic symmetry breaking element exists in the
layer. Specifically, consider a state of a homochiral molecular layer and its mirror image.
Though the Hamiltonian is invariant under parity transformation, the states are characterized
by a handedness c, which changes sign upon a mirror reflection. Thus, the handedness c is
an isoscalar quantity. For simplicity, choose a mirror plane parallel to the field E (normal to
the surface of the layer). Such a transformation does not change E but does change c.
Therefore, one can define another operator B=cE which is an axial vector and can interact
with the spin of electrons in the same way as angular momentum does in atoms or as a
magnetic field does in a macroscopic system. Different handedness are expressed by c=l or
c=-1 and their correspondingB points either parallel or anti-parallel to E.
The upper bound for the electric potential, given in Eq. 10, corresponds to a value of B,
which is equivalent to a magnetic field of up to 20,000 gauss. This value is larger than the
fields found in common ferrites.
From the above considerations one can conclude that when chiral molecules form a close
packed layer, due to the charge rearrangement, the layer may gain magnetic properties. This
effect may be of relevance in biological membranes where homochiral molecules with
relatively large dipole moment are packed together.
3 EXPERIMANTAL
Self-assembled monolayers of either L or D polyalanine polypeptides in the form of c1

helices were prepared on a gold substrate and the orientation of the molecules was
monitored by IR spectroscopy. Monolayers of either L or D polyalanine polypeptides were
prepared on glass slides coated with 100 nm thick annealed gold film. By connecting a
sulfide group either at the C- or N- terminal of the peptide, the dipole moment of the
attached molecules is pointing either away fkom the substrate or towards the substrate. Three
types of films were investigated, two were of poly L-alanine and poly D-alanine, both
connected to the surface at the C-terminal (referred to as LC and DC respectively), and one
type of poly D-alanine connected to the surface at the N-terminal (referred to as DN).
CD spectra of L,D-polyalanine a helix
on gold covered quartz slides.
5 ,
-Poly 0 C-terminal
- - - - Poly L C-terminal
cv
v 'I-
6 '
= -'t
L
I
I
Figure 2: Circular Dichroism (CD) spectra of monolayers of L and D polyalanine attached

through the C-terminal to a I0 nm thick gold film coated quartz slide.
Polypeptide lengths between sixteen and twenty-two amino acid units were used. The
structure of the films was determined by their FTIR spectra24. The handedness of the
adsorbed films was verified by circular dichroism absorption as shown in Figure 2, and their
thickness was measured by ellipsommetry. Characterizations by AFM measurements were
carried out as well. All the results reported here were obtained at room temperature fiom
closed packed layers.
IR spectra of LC and DC monolayers were measured at magnetic field strengths of 0,
*900, and k4500 Gauss applied perpendicular to the layer. Henceforth, ‘‘North” indicates
magnetic field lines starting at the North Pole and penetrating through the gold surface to the
monolayer and vice versa for “South”.
For the electron transmission studies, the samples were inserted into an ultrahigh vacuum
chamber at <lo’* mbar. The polarized photoelectrons are ejected fiom the substrate by
applying a laser beam at 248 nm using a W4 plate to create either left- or right-handed
circular polarized light. It is known that right-handed circularly polarized light induces
positive helicity2’ in the photoelectrons ejected &om the gold substrate and the reverse for
the left-handed polarized light. The photoelectrons are known to be polarized by about
15%26. After passing through the or anic layers, the electrons energy distribution is analyzed
using a time-of-flight spectrometer.B
4 RESULTS
Figure 3 presents the grazing angle IR spectra obtained for the amide vibrations?* The
amide I vibration is parallel to the molecular axis (at about 1665 cm-’) while the amide I1
vibration at about 1550 cm-*is perpendicular to the axis. The spectra are normalized at the
peak of the 1665 cm-’ line. If the molecules are oriented normal to the sudace, the intensity
of the amide I1 component vanishes because of the metal substrate canceling of the transition
dipole moment. Hence, the ratio between the intensity of the two peaks provides a direct
measure of the tilt angle of the molecules relative to the surface n0rmal.2~
From Fig. 3 it is evident that there is a magnetic field effect on the relative intensities for
both LC and DC layers. For the two LC layers shown (Figs. 3A,3C) South fields increase
the average tilt angles while North fields decrease them. The reverse is true for the DC
layers (Figs. 3B, 3D); namely, South fields decrease the average tilt angles while North
fields increase them. Larger fields show larger tilt deviations. It takes 2 to 6 hours for the
effect of the magnetic field on the tilt angle to reach its equilibrium. This time depends on
the strength of the magnetic field and on the quality of the monolayer, namely the packing
density of the layer and the size of the average tilt angle. Better quality layers require more
time for equilibrium in magnetic fields. The above magnetic field orientation effects within
the layers provide clear evidence that the films have magnetic properties and that opposite
magnetization occur for different handedness. After all, the classic magnetization of iron is
interpreted as orientation of domains. Looking on either the LC layer or the DC layer, the
magnetic orientation of these artificial monolayers mimic orientational effects already found
in biological rnernbrane~.’~
Figure 3: The IR spectrum of the polyalanine layers of L and D molecules bound to the
surface through the carbon terminal. The amide I and amid II bands are shown at about
1665 cm-' and 1550 cm-' respectively. The spectra are normalized at the peak of the 1665
cm-I line. The spectra were taken either with a magnet having its North or South pole
pointing towards the Jilm (dashed and dotted lines respective&) or without magnetic field
(solid line). The spectra shown in A and B were taken with a magneticfield of 4500 Gauss,
while those at C and D were taken with 900 Gauss. Panels A and C correspond to the
spectra of L layers while B and D correspond to D layers.
Figure 4 presents the kinetic energy distributions for photoelectrons ejected with a left or
right circular polarized laser (solid and dashed lines respectively). The spectra in panels A
and C are obtained for the transmission of electrons through films of L- and D-polyalanine
respectively, both bound to the surface through the C-terminal. Panel B corresponds to a
film of D-polyalanine bound to the surface through the N-terminal. The results shown in Fig.
4A and 4C c o n f m earlier results that showed a large asymmetry for polarized electrons
transmission through organized film of chiral molecules?' The sign of the asymmetry
depends on the handedness of the molecules. Surprisingly, for a given handedness of the
molecules the sign of the asymmetry switches upon reversing the way the molecules are
adsorbed on the surface (from N to C terminated molecules). This is clearly seen in Figs. 4B
and 4C.The observed asymmetry31in the transmission through the layer of photoelectrons
produced by left and right circular polarization of photons changes from 0.0910.02 to -
0.10*0.02 (Fig. 4B and 4C respectively). Importantly, the observed 10% effect is induced by
merely 15% polarization of the photoelectrons26. Thus, the selectivity to the incoming
helicity of the electrons is as large as 70% and within experimental error could be even
higher.
F Poly L c tennid
0 .o 0.2 0.4 0 .B 0 .8 1 .o 1.2 1 .4
Poly D N term inal

I
B
-
-
-
L
0 .o 0.2 0.4 0 .6 0 .8 1 .o 1.2 1.4
Poly D C terminal
0 .o 0.2 0.4 0 .8 0 .8 1 .o 1.2 1.4
Electron Energy (eV)
Figure 4: The energy distribution for photoelectrons ejected with a lefi (negative spin
polarization-red,solid) or right circular &ositive spin polarization-blue,dashed) polarized
laser. The electrons are transmitted throughfilms of L and D-polyalanine both bound to the
surface through the carbon terminal (A and C respectivelyl, and through a film of D-
polyalanine bound to the surface through the N-terminal (B).
5 DISCUSSION
Several questions arise fiom the above experimental observations-

- Why are the layers magnetic?
- Why is there a relation between the direction of magnetism and the molecules’
handedness?
- Why is there a large spin-selectivity for electron transmission, and
- Why does it change sign with both change of handedness and change of the direction of
the molecular electric dipole?
Our model is based on the properties of closed packed monolayer of molecules having a
dipole moment, as described above. The polyalanine molecules in solutions have a large
electric dipole moment of about 50 D e b ~ e . 3 ~When ’ ~ ~ the molecules are adsorbed as
monolayers, by attaching either C or N terminals to the substrate, they must lose almost all
their dipole moment. The simple electrostatic arguments given above indicate that the
molecular dipole moment must be reduced by more than an order of magnitude. A direct
indication on the value of the adsorbed molecular dipole moment can be obtained from the
substrate’s work function. From the high-energy cutoff of the photoelectron spectra (Fig. 4)
it is evident that upon reversing the direction of the layer from being bound to the surface
through the carbon versus being bound through the nitrogen (Figs. 4A, 4C versus 4B), the
work h c t i o n increases by only about 0.3 eV, while for molecules with a dipole moment of
about 10 Debye or more the work function would have changed by many volts. This means
that the dipole moment of the molecules was reduced by two orders of magnitude due to
ad~orption.~~ Consequently, the electronic structure of the molecules is substantially
modified upon adsorption, probably by charge transfer between the metal substrate and the
molecule. Apparently, the charge redistribution process results in unpaired electrons on the
adsorbed molecules, namely the adsorbed molecules become paramagnetic.
Still, the above posed questions are not answered. Though a full description of these
phenomena does not yet exist, the following simple model is compatible with all the
observations and predicts correctly the direction of magnetization. Moreover, since the
reported phenomena exhibit parallel effects to some biological membranes, the model may
serve as a basis for some understanding of magnetic effect on biological systems.
Upon adsorption, there is a transient current through the helix, which discharges the
electric dipole while inducing magnetism, like a classical electro-magnetic coil. The spins of
simultaneously created unpaired electrons, adjust accordingly and stay polarized by support
of exchange forces with the metal and the already polarized electrons of attached neighbor
helices. The result of such an adsorption mechanism is a magnetic layer where the direction
of the magnetic field is given by Ampere’s law. For each adsorbed helix, the original electric
dipole polarity and the handedness of the helix determine the direction of the magnetic field
along the helix axis.
As can be seen in Figure 3, the helix axes are not well aligned normal to the surface.
Therefore, when an external magnetic field is applied in the normal direction, it forces these
axes to get either closer to the normal or away from it, depending on the direction of the
external magnetic field. The observed magnetic effects in Figure 3 are consistent with the
simple model described above.
The magnetic properties of the film are the reason for the large spin selectivity in the
electron transmission. This effect is consistent with the effect in inorganic thin magnetic
layers, known as the Giant Magneto Spin selectivity35y36 and the related colossal
magnetoresistance e f f e ~ t . 3 ”Hence,
~~ the magnetic properties of the organic films explain
their high selectivity for transmission of spin polarized electrons, an effect that results from
the cooperative nature of the film on the metal substrate. The seemingly surprising change in
the direction of magnetization by the change of the molecular terminals on the gold surface
is resolved by noticing the direction of the electric dipole discharge through the helices.
In the present work we find that the adsorption process of chiral species into layers
converts electric dipoles into magnetic dipoles. Some biological membranes, which show
similar magnetic orientation effect, are now suspected to have similar magnetic properties.
Acknowledgement
We are gratefbl to Prof. M. Fridkin and his group for helping us in the synthesis of the
polyalanine. Partial support fiom the Israel Science Foundation and the US-Israel Binational
Science Foundation is acknowledged.
REFERENCES
1. H. Ishii, K. Sugiyama, E. Ito, K. Seki, A h . Mater. 1999,11,605.

2. S.N. Yaliraki, A.E. Roitberg, C. Gonzalez, V. Mujica and M.A. Ratner, J. Chem. Phys.
1999,111,6997.
3. Vilan, A. Shanzer and D. Cahen, Nature, 2000,404,166.
4. I.H. Campbell, J.D. Kress, R.L. Martin, D.L. Smith, N.N. Barashkov, J.P. Ferraris, Appl.
Phys. Lett. 1997,71,3528.
5 . T. Verbiest, S. Van Elshocht, M. Kauranen, L. Hellemans, J. Snauwaert, C. Nuckolls, T.J.
Katz, A. Persoon, Science, 1998,282,9 13.
6. C.P. Collier, E.W. Wong, M. Belohradsky, F.M. Raymo, J.F. Stoddart, P.J. Kuekes, R. S.
Williams, J.R. Heath, Science, 1999,285,391.
7. J. Chen, M.A. Reed, A.M. Rawlett, J.M. Tour, Science, 1999,286,1550.
8. Vuillaume, B. Chen and R.M. Metzger, Lungmuir, 1999,15,4011.
9. E. Punkka and R.F. Rubner, J. Elect. Mat. 1992,21,1057.
10. See for example: A. Yamamori, S. Hayashi, T. Koyama, and Y.Taniguchi, App. Phys.
Lett. 78,3343 (2001); B. K. Crone, P. S.Davids, I. H. Campbell, D. L. Smith, J. Appl. Phys.
2000,87,1974.
11. See for example: P. Peumans, V.Bulovic, S. R. Forrest, Appl. Phys. Lett. 2000,76,
2650; U. Bach, D. Lupo, P. Comte, J.E. Moser, F. Weissortel, J. Salbeck, H. Spreitzer, and
M. Grtitzel, Nature, 1998,395,583.
12. D. G. Wu, D. Cahen, P. Graf, R.Naaman, A. Nitzan, D. Shvarts, Chem. Eur. J. 2001,
7,1743.
13. A. Nitzan and I. Benjamin, Acc. Chem. Res., 1999,32,854.
14. G.R. Hutchison, M.A. Ratner, T.J. Marks, R. Naaman, J. Phys. Chern. B. 2001,
105,288 1.
15 D.-Ch Neugebauer, A.E. Blaurock and D.L. Worcester, FEBS Lett. 1977,78,31.
16. See for example: W. Zhao, G. Kerner, and M. Asscher, X. M. Wilde, K. Al-Shamery,
and H.-J. Freund, V. Staemmler and M. Wieszbowska, Phys. Rev. B, 2000,62,7527.
17. M. Bruening, E. Moons, D. Cahen, A. Shanzer, J. Phys. Chem. 1995,99,8368.
18. S . Liao, Y. Shnidman, and A. Ulman, J. Am. Chem.Soc. 2000,122,3688.
19. D.M. Taylor and G.F.Bays, Phys. Rev. E 1994,49,1439.
20. M. Iwamoto, Y. Mitzutani, and A. Sugimura, Phys. Rev. B, 1996,54,8186.
2 1. C-X. Wu and M. Iwamoto, Phys. Rev. B 1997,55,10922.
22. D. Liufb, X. S. Wang, D. M. Tu, K. C. Kao, J. Appl. Phys. 1998,83,2209.
23. E. Miranda and J. Sune, Appl. Phys. Lett. 2001,78,255.
24. In helix peptides, the transition moment of amide I band lies nearly parallel to the helix
axis and that of amide I1 perpendicular. Transition moments, which lie parallel to the gold
surface, cannot be detected in grazing angle FTIR. The ratio between amide I band (1 665
cm-') and amide I1 band (1550 cm-') indicates to what extend the molecules in the
monolayer are oriented perpendicular to the gold surface.
25. The helicity is defined for particles with momentum p and spin s, as the expectation
valueof - s.P .
Is * PI
26. J. Kirschner, Polarized Electrons at Surfaces, (Springer-Verlag, 1985); F. Meier and D.
Pescia, Phys. Rev. Lett. 1981,47,374-377; F. Meier, G. L. Bona ,S. Huher, Phys. Rev.
Lett. 1984,52, 1152; G. Borstel, M. Wohlecke, Phys. Rev. B 1982,26, 1148.
27. R. Naaman, A. Haran, A. Nitzan, D. Evans, and M. Galperin, J. Phys. Chern. 1998, B
102,3658.
28. "PeptidesPolypeptides and Proteins" E.R. Blout et. al. eds, John Wiley&Sons 1974, p.
379.
29. Since polyalanine is hydrophobic, it is dificult to clean the sample using
chromatography and to obtain a single size polypeptide. Hence, the samples tend to vary in
the distribution of the peptides length &om batch to batch. This is expressed by the spread in
tilt angles of the different layers, which was found to be larger for less uniform batches.
However, the results obtained were consistent for all selected samples. Samples were
rejected if the average tilt angle exceeded 50'.
30. K. Ray, S.P. Ananthavel, D.H. Waldeck, R. Naaman, Science, 1999,283,814.
31. The asymmetry parameter is defined as A = I(+P) - I(-P) where I(+P) and I(-P) are the
I(+P) + I(-P)
transmission of the electron beam with spin angular momentum oriented parallel (+) and
antiparallel (-)to its velocity vector.
32. W.G.J. Hol, P.T. van Duijnen and H.J.C. Berendsen, Nature, 1978,273,443.
33. C . Park and W.A. Goddard 111, J. Phys. Chem. B 2000,104,7784.
34. The change in the workhnction (A&) is related to the dipole of the layer by the
equation A& = 47tD where D is the dipole density. We assumed that the layer density was
5x1014molecules/cm2.
35. E. Velu et. al. Phys. Rev. B 1988,37,668.
36. A. Filipe et. al. Phys. Rev. Lett. 1998,80,2425.
37. R. von Helmholt, et. al. Phys. Rev. Lett. 1993,71,2331.
38. S . Jin et. al. Science, 1994,264,413.
BARRIER CROSSING BY A FLEXIBLE LONG CHAIN MOLECULE - THE
KINK MECHANISM
K.L. Sebastian
Department of Inorganic and Physical Chemistry, Indian Institute of Science,

Bangalore 560012, India
1. INTRODUCTION
The calculation of the rate at which a particle trapped in a metastable state escapes is a
very important problem, having applications in several areas of physics and chemistry
and biology. Kramers' found solutions in the limit of weak friction and also in the
limit of moderate to strong damping. A classic reference to the problem is the article
by Chandrasekhar.2 The recent progress on this topic has been surveyed in a detailed
review by Hanggi et. aL3 The reason for this extensive activity is that this forms a
model for a chemical reaction occuring in a condensed medium. The Kramers problem
for few degrees of freedom has also been the topic of s t ~ d y Here
. ~ we consider a natural
extension of the problem to a case where the number of degrees of freedom, N is very
large ( N --+ 00). Further, the way these are connected (to form a long chain molecule)
leads to interesting new aspects to the problem that are not present in the case where
there are only finite number of degrees of freedom (see Fig. 1).
The problem is of importance in biology as quite a few biological processes involve
the translocation of a chain molecule from one side of a membrane to the other, through
a pore in the membrane. Some examples are:
1. The translocation of proteins from the cytosol into the endoplasmic reticulum, or
into mitochondria, or chloroplasts. Often, the proteins are hydrophilic and the
pore in the membrane forms a hydrophobic region, through which it has to pass
thro~gh,~ resulting
-~ in an increase in the free energy for the portion of the chain
inside the pore
2. In infection by bacteriophages, conjugative DNA transfer etc, long chain DNA
molecules snake through pores in membrane^.^^
In the above processes, the translocation of the chain molecule seems to occur
with ease, contrary to the expectation that one gets from the theoretical analysis
available in the literature. In an interesting experiment Kasianowicz et. al.lo forced
long polynucleotides to move through a pore in a membrane and studied the time that
it takes the molecule to cross the pore, as a function of the length of the molecule.
They found the time to be proportional to the length of the chain. More recently,
Han and Craighead" studied the forced motion of very long DNA molecules through
microfabricated channels and found that the activation energy for crossing shallow
regions in the channel was independent of the chain length.
1.1 The problem and related earlier work
All the above problems involve the passage of a long chain molecule, through a region in
space, where the free energy per segment is higher, thus effectively presenting a barrier
for the motion of the molecule. This is what we refer to as the Kramers problem for a
chain molecule. Muthukumar and Baumgartner12 studied the movement of self avoiding
polymer molecules between periodic cubic cavities separted by bottlenecks, the passage
through which presents an entropic barrier to the motion. They find that there is an
exponetial slowing down of diffusion with the number of segments N in the chain. Park
and Sung5 have studied the translocation through a pore. They analyze the passage
through a pore on a flat membrane, with the effects of entropy included. They refer to
The free energy barrier has the functional form F -

the side that the polymer is initially in as the cis side and the other as the trans side.
+
-kB'Tln ( n ( N - n ) )/2 nAp,
where n is the number of segments that have crossed from the cis side to the trans
side and Ap is the free energy change per segment in the process. As is obvious,
this barrier is rather broad, and its width is proportional to N . Consequently, they
trans
R->
Figure 1: The potential energy per segment of the chain, plotted as a function of
position
consider the translocation process as being equivalent to the motion of the center of
mass of the molecule. Using the result of the Rouse model that the diffusion coefficient
of the center of mass is proportional to 1/N, they effectively reduce the problem to
the barrier crossing of single particle having a diffusion coefficient proportional to 1/N.
As the translocation involves motion of N segments across the pore, the time taken
to cross in the case where A p = 0, t,,, scales as N 3 . They also show that in cases
where there is adsorption on the trans side (Ap < 0), translocation is favored and then
t,,,, scales as N 2 . Kumar and Sebastian13 have analysed the free energy profile for
translocation from the outer surface of a vesicle to the inner surface, for the case where
the molecule is adsorbed on both the surfaces, pointing out that adsorption on the
outer surface can lead to the non-existence of the entropic barrier, thereby facilitating
the process. Park and Sung14have given a detailed investigation of the Rouse dynamics
going over a broad potential barrier. They use multidimensional barrier crossing theory
to study the motion of a chain molecule over a barrier, in the limit where the width of
the barrier is much larger than the lateral dimension of the molecule. Lubensky and
Nelson15 study a case where the interaction of the segments of the polymer with the
pore is strong. In this case, it is the dynamics of the portion that is inside the pore
that is important and they show that this can lead to t,,,, proportional to N . In
recent papers, we have suggested16 a kink mechanism for the motion of the chain, and
we give details of this mechanism in this paper. We consider a polymer undergoing
activated crossing over a barrier whose width w is larger than the Kuhn length 1 of
the polymer, but small in comparison with the length NZ of the polymer. Thus, we
assume Z << 20 << Nl . This is the case in the experiments referred to above. For
example, in the experiments of Kasianowicz et. al.,l0 the length of the pore is about
100 A, while the Kuhn length for a single stranded DNA is perhaps around 15 A.15
Therefore, one is justified in using a continuum approach to the dynamics of the long
chain. (It is possible to retain the discrete approach, and develop the ideas based on
them, but this is more involved mathematically).
Our approach is the following: We describe the dynamics of the chain using the
continuum version of the Rouse model, discussed in detail in the book by Doi and
Edwards.lg The barrier exerts forces on the segments and inclusion of this in to the
Rouse model makes the equation a non-linear one. We refer to this equation as the
non-linear Rouse model. The portion of the chain inside the barrier would be distorted
in comparison with the portions that are outside and we refer to this distortion as
the kink. In the non-linear Rouse model, the distortion is a special solution and as is
usual in non-linear physics, we refer to it as a kink. Movement of the chain across the
barrier is equivalent to the motion of the kink in the reverse direction. In the presence
of a driving force (i.e. A p < 0) , the kink moves with a finite velocity and hence the
polymer would cross the barrier with t,,,, proportional to N .
1.2 Kinks pinned in space
Traditionally, the usual non-linear models (for example, the 44 or the sine-Gordon
have potentials that are translationally invariant, and hence the kink can
migrate freely in space. For example, the sine-Gordon model has the equation of motion
I50 Biophysical Chemistry: Membranes and Proteins
-2 + 9 = sin($). Note that the position co-ordinate enters the equation only
through the derivative terms, and all points in space are completely equivalent. Kink
solutions to this equation may be found by putting $ = q5(x - v t ) , and these represents
kinks that are free to move anywhere in space. In comparison, in the barrier crossing
problem, the position of the barrier is fixed in space. This breaks the translational
symmetry. As a result, the kink is pinned in space. However, the long chain molecule
(modelled as a stringlg) can move in space and hence the kink migrates, not in space,
but on the chain. This is the obvious and simple mechanism of translocation. Such a
suggestion is new for barrier crossing (somewhat similar ideas are used in reptation)
and we believe that this is a very useful idea in understanding polymer translocation.
In the following, we develop this idea based upon the one dimensional version of the
Rouse model.
transition state
d
initial state
\ f
\ I
NAV
Figure 2: The potential energy along the reaction co-ordinate. The Eactas independent
of the length of the chain. After the barrier is crossed, there is a region of width W ,
with W proportional to N , which is to be crossed. The time required to cross this
regions is t,,,,
2. THE FREE ENERGY LANDSCAPE

The considerations in this subsection are quite general and do not depend on the model
that one uses to describe the polymer dynamics. We assume only that the polymer
is flexible over a length scale comparable to the width of the barrier. We start by
considering the free energy landscape for the crossing of the barrier. The barrier and
the polymer stretched across it are shown in Fig. 1. The polymer has initially all its
units on the cis side, where its free energy per segment is taken to be zero. So the initial
state has a free energy zero in the free energy hypersurface shown in Fig. 2. In crossing
over to the trans side, it has to go over a barrier, as in the Fig. 1. The transition state
for the crossing can be easily found, from physical considerations, by remembering that
the transition state is a saddle point - i.e. it is a maximum on the free energy surface
in one direction, while in all the other directions it is a minimum. The transition state
is shown in the Fig. 3. In the transition state, the configuration of the polymer is such
that the free energy of the chain is an extremum, subject to the two constraints: (a)
the end of the polymer on the trans side is located exactly at the point at which its free
energy per segment is zero (b) the other end is on the cis side. This is the transition
Figure 3 The transition state

state, because if one moves the end at the trans side either in the forward or in the
backward direction (and the rest of the chain adjusted so that the free energy of the
chain as a whole is a minimum), then the total free energy of the chain would decrease.
Thus the transition state shown in the free energy hypersurface in the Fig. 2 has the
configuration shown in the Fig. 3. Once the system has crossed the transition state, the
chain is stretched across the barrier. The path of steepest descent then corresponds to
moving segments from the cis side to the trans side, without changing the configuration
of the polymer in the barrier region. As there is a free energy difference AV between
the two sides, this would lead to a lowering of the free energy by AV per segment,
and this leads to a path on the free energy surface with a constant slope, and of width
W proportional to N (see the Fig. 2 ). Such a landscape implies that the barrier-
crossing would involve two steps. The first step is going through the transition state
by overcoming the activation barrier. Once this is done, there is a rather wide region
of width proportional to the length of the chain. Traversing this is the second step.
As this region has a constant slope, the motion is driven and it is similar to that of
a Brownian particle subject to a constant force. Such a particle would take a time
t,,,,, proportional to N to cross this region. Till now, we considered the case where
one end of the molecule overcomes the barrier first, which we refer to as end-crossing.
It is also possible for a portion not at the end, to overcome the barrier, by forming a
hairpin. For this, the molecule has to be flexible and the pore/channel wide enough
to accommodate the hairpin. The scenario for hairpin crossing is similar, though the
activation energy is higher for hairpin crossing. Within the Rouse model discussed
below, for hairpin crossing, the transition state is equivalent to the one end crossing,
repeated two times.17 Hence the activation energy for the process is two times larger.17
Once a hairpin crossing occurs, a kink-antikink pair is formed and the kink and the
anti-kink separate on the chain, due to the driving force of the free energy gain. Then
further crossing occurs by the movement of these two on the chain, and this too leads to
a time of crossing proportional to N . In the following we make all these considerations
quantitative, using the Rouse model to describe the dynamics of the chain.
3. THE NON-LINER ROUSE MODEL
We consider the continuum limit of the Rouse model, discussed in detail in the book by
Doi and Edwards.lg The discrete nature of the chain is ignored and it is approximated
by a string. The position of a segment (bead) along the string is measured by the
variable n, which is taken to be a continuous variable, having values ranging from 0
to N . The chain moves only in one dimension and the position of the nth bead at the
time t is denoted by R(n,t ) . In the Rouse model, this position undergoes overdamped
Brownian motion. Its change with time is given by
In the above, ( is a friction coefficient for the nthsegment. The term ma2f$?)comes
from the fact that stretching the chain would lower its entropy and hence increase
its free energy. Consequently, the parameter m = 3knT/Z2 (see Doi and E d w a r d ~ , ' ~
equation (4.5). They use the symbol k for our m) . As the ends of the string are
free, the boundary conditions to be satisfied are {y}n=o = { T }=
aR(n,t) n 0. = N
V ( R )is the free energy per segment of the chain for a segment at the position R. In
the following, we shall take V ( R )to have a double well shape, as in Fig. 1. f(n,t)
are random forces acting on the nth segment. These have the correlation function
, = 25k~T6(n
(f(n,t ) f ( n ltl)) - n1)6(t- tl)(see Doi and Edwards,lg equation (4.12)),
so that the forces acting on different segments are uncorrelated. Further, the random
force on each segment is white noise. The deterministic part of the equation (l),which
will play a key role in our analysis, is obtained by neglecting the random noise term in
(1). It is:
aR(n,t ) a”R(n,t )
- V’(R(n,t ) ) . (2)
C 7 = m dn2
This may also be written as:
where E[R(n,t)]
is the free energy functional for the chain defined by:
In the following, we take
This potential has two minima at R = -a0 and at R = al and a maximum at R = 0.

If al > ao, then the minimum at --a0 is less stable than the one at ul. Hence, if the
polymer is initially in the metastable minimum, it will eventually cross over to the
more stable minimum.
3.1 The Activation Free Energy
The crossing of the polymer across the barrier involves two steps. The first is the
process by which one end of the polymer goes over the barrier as a result of which the
polymer gets stretched across the barrier. This process has a free energy of activation.
The second step involves the motion of the polymer across the barrier, once it has
been stretched across. The second step does not involve overcoming any barrier. In
this subsection, we consider the first step and calculate the activation free energy. The
activation free energy can be obtained from the free energy functional of Eq. (4).
This free energy functional implies that at equilibrium, the probability distribution
functional is exp [-& {J d n irn (e>2+
V ( R ( n ) ) } ] If
. the polymer is on the left
side of the barrier, then minimum free energy is attained by having R(n)= -uo. This
has a free energy that is equal to zero. If one end of the polymer is stretched across
the barrier, then V ( R ( n ) is
) non-zero and the extremum free energy configuration is
determined by = 0, which leads to the equation
d2R
m- = V‘(R)
dn2
Notice that this is just a Newton’s equation for a ficticious particle of mass m moving
in a potential -V(R). This equation has to be solved, subject to the condition that one
end of the polymer is on cis side of the barrier and the other end on the trans side. As we
are interested in the case where the polymer is very long, we can find the extremum free
energy configuration by finding a solution satisfying R(-oo) = -a0 and the other end
of the polymer to be at a point with R > R, where R,,, is the point where V ( R )has
its maximum value. For the specific form of the potential given above R,,, = 0. For the
Newton's Eq. (6) the conserved energy is E, = irn ( g ) 2 - V ( R ( n ) )For
. the extremum
path, Ec = 0. Thus, the particle starts at R(-oo) = --a0 and ends up at Rf where
R f ( > R,,), is the point such that V ( R f )= 0. The free energy of this configuration
is the activation free energy. As for this configuration, irn
the activation free energy to be given by
(e)2 = V ( R ( n ) )we
, find
As the parameter m is proportional to the temperature ( = 3 k ~ T / 1), ~we arrive at

the general conclusion that the activation energy Eact
-3KL
cue. The Boltzmann factor
e k g T for the crossing of one end of the polymer over the barrier thus has the form
e-cmstantlJSi. Further, we find that it is independent of N for large N.
For the barrier potential given above, for the forward crossing, there is a barrier
with height Vf = V ( 0 )- V(-ao) = $kao3 (a0 + 2 a l ) and for the reverse process,
+
the barrier height is Vb = V ( 0 )- V ( a l ) = $kaI3 (2ao al). The forward crossing,
would lead to a net change in free energy per unit of the polymer, which is equal to
+
AV = V(al) - V(-ao) = $ k (a0 - a l ) (a0 a1)3. Though our results are for this
particular form of the potential, the ideas and the conclusions are quite general and of
wide applicability.
For the potential of equation (5)) we find Rf = ao(y - d -) where y = (1 +
2:);. Further,
4. THE KINK AND ITS MOTION
4.1 The kink solution and its velocity
Having overcome the activation barrier, how much time would the polymer take to cross
it? We denote this time by ttrans.To calculate this, we first look at the mathematical
solutions of the deterministic equation (2). The simplest solutions of this equation are:
R(n,t ) = -a0 or with R(n,t ) = al. These correspond to the polymer being on either
side of the barrier. Thermal noise makes R(n,t ) fluctuate about the mean positions and
this may be analyzed using the normal co-ordinates for fluctuations about this mean
position. Each normal mode obeys a Langevin equation similar to that for a harmonic
oscillator, executing Brownian motion. Further, the mean position (center of mass)
itself executes Brownian motion.lg In addition to these two time independent solutions,
the above equation has a time dependent solution (a kink) too, which corresponds to
the polymer crossing the barrier.
As is usual in the theory of non-linear wave equations, a kink solution moving with
a velocity v may be found using the ansatz R(n,t ) = Rs(r)where r = n - ~ t . 'Then~
the equation (2) reduces to
d2& dR
m -dr2
+ v C L =dr V'(RS). (9)
If one imagines r as time, then this too is a simple Newtonian equation for the motion
of particle of mass m, moving in the upside down potential -V(R). However, in this
case, there is a frictional term too, and v [ / m is the coefficient of friction. This term
makes it possible for us to find a solution for quite general forms of potential. If there
was no friction, then total energy would be conserved and such a solution would not
exist for the potential given above.
We give results for the potential given in the Eq. (5). The solution of the equation
(9), obeying the conditions R,(T)= -a0 for T -+-00 and Rs(r)= a1 for r -+00 is
with w = JG . The solution exists only if the velocity v = 9 (a0 - al). This
solution is a kink, occurring in the portion of the chain inside the barrier. We shall refer
to the point with r = 0 as the center of the kink. (Actually one has a one-parameter
+
family of solutions of the form R,(T ro), where 70 is any arbitrary contant). As
r = n - vt, the center of the kink moves with a constant velocity w. Note that this
velocity depends on the shape of the barrier. Thus for our model potential, if a0 < a l ,
then Vf < V b , and this velocity is negative. This implies that the kink is moving in the
negative direction, which corresponds to the chain moving in the positive direction.
That is, the chain moves to the lower free energy region, with this velocity. If the
barrier is symmetric, then a0 = ul(i.e., Vf = 6 )the velocity of the kink is zero.
4.2 The crossing time t,,,,

For the polymer to cross the barrier, the kink has to go in the reverse direction, by a
distance proportional to the length of the chain. As the kink moves with a velocity v,
t,,,, N N/v.Further, as v is proportional a,
assuming V ( R )to be temperature
independent we find t,,,, N . This too is a general conclusion, independent of the
N
model that we assume for the potential.

If the barrier is symmetric, the kink velocity v = 0 - it does not move. This is
not surprising as there is no free energy gain, by the movement of the kink. The kink
can be anywhere on the chain, with the same total free energy. Thermal fluctuations,
ofcourse, would drive the kink randomly- which means that the kink would execute a
wfi.
random walk. We do not give here the detailed analysis,16but only the final result: the
result is that it executes a diffusive motion, with diffusion coefficient D =
In this case, the time required for the polymer to cross over the barrier is ttrans N 2 .
N
Park and Sung5 consider the passage of a polymer through a pore for which the
barrier is entropic in origin. Consequently it is very broad, the width being of the
order of N . In comparison, we take the barrier to be extrinsic in origin and assume
its width to be small in comparison with the length of the chain. The crossing occurs
by the motion of the kink, which is a localized non-linear object in the chain whose
width is of the same order as that of the barrier. As the kink is a localized object, its
diffusion coefficient has no N dependence and hence our results are different from those
of Park and Sung.5 In the case where there is no free energy difference, our crossing
time is proportional to N2(in contrast to N 3 of Park and Sung) , while if there is a free
energy difference, our crossing time is proportional to N (in contrast to N 2 of Park
and Sung). In a very recent paper,14 Park and Sung consider the Rouse dynamics of a
short polymer surmounting a barrier. The size of the polymer is assumed to be small
in comparison with the width of the potential barrier. Consequently, the transition
state has almost all the beads at the top of the barrier, leading to the prediction that
the activation energy is proportional to N . This leads to a crossing probability that
decreases exponentially with N . In comparison, as found earlier, the free energy of
activation does not depend on the length of the chain. Hence, the kink mechanism
must be the favoured one for long chains.
4.3 The net rate
As the actual crossing is a two step process, with activation as the first step and kink
motion as the second step, the net rate of the two has to be a harmonic mean of the
two rates. However, though we have estimated the free energy of activation, it seems
rather difficult to estimate a pre-factor for this process. Hence one can only make some
general observations: For a very long chain, the motion of the kink has to become rate
determining. In the experiments of Kasianowicz," one directly observes ttransand
hence our considerations on kink motion must be directly applicable. Also, in the case
of translocation of biological macromolecules (see the next section) there does not seem
to be any free energy of activation and then the rate is determined by ttransalone,
To verify the analysis given above, we performed computer simulations, the details
of which are given elsewhere.24 The simulations were done for a discrete model, for
a variety of barrier heights and widths. In all the cases, the average time at which a
given bead crosses, when plotted against the number of the bead, lead to a straight
line, thus verifying the operation of the kink mechanism for a flexible chain.
Further, even though, the analysis given above is for a one dimensional version of the
Rouse model, it is fairly straightforward to extend the analysis to a three dimensional
situation, and argue that the kink exists. The operation of the kink mechansim, even
in this three dimensional case too has been verified by sir nu la ti on^.^^
5. TRANSLOCATION IN BIOLOGICAL SYSTEMS

If there was a high activation energy (>> ~ B Tfor) the translocation, the process would
be unlikely. However, translocation seems to be very efficient in biological systems,
and hence it is probably a barrierless process. What is the mechanism by which this
reduction is achieved? One possibility is for the segments to have a lower free energy
when they are inside the membrane. But then, the chain would not leave the membrane
and perhaps would get implanted there. The final destination where a biological long
chain molecule ends up is determined by a sequence of units at the begining of the
chain, referred t o as the signal sequence. The way the sequence works is simple. If the
pore is hydrophobic and the chain hydrophilic, then the signal sequence is hydrophobic,
so that the signal sequence has a low free energy inside the pore. Analysis of this type of
problem, using the Rouse model17 is straightforward. One assumes that the segments
at the end of the chain has a lower potential energy if they are inside the pore.
.........
".. . hydrophobic
'I.. .. //
hydrophili c
f,,.,
..__....__...........
barrier due to hydrophobic pore
Figure 4: The transition state for a hydrophilic chain with a hydrophobic signal se-
quence, passing through a hydrophobic pore, Compare with figure 1-6-14 of the book bg
Alberts et. al.
Analysis of this model leads to a transition state that is shaped like a h00k.l~ The
hydrophobic part of the chain is completely in the short arm of the hook (see figure
3). The'activation energy would depend on the length of the hydrophobic part, and
if it is sufficiently long, the activation energy may become zero, so that the crossing
can become a barrierless process. In such a scenario, the rate is determined by t,,,,.
The transition state, though it seems likely to occur in crossing between liquid-liquid
interfaces, seems rather difficult to form in the case of passage through a pore as there
are two difficulties: (1) the chain has to bend to form the hook (2) the pore has to be
wide enough to accommodate the two strands of the hook simultaneously. In spite of
these, nature does seem to use this as an inspection of the figure 14-14 of the book by
Alberts et. aL6 shows.
6. CONCLUSIONS
We have considered the generalization of the Kramers escape over a barrier problem
to the case of a long chain molecule. It involves the motion of a chain molecule of N
segments across a region where the free energy per segment is higher, so that it has
to cross a barrier. We consider the limit where the width of the barrier w is large in
comparison with the Kuhn length I , but small in comparison with the total length Nl
of the molecule. The limit where NZ << w has been considered in a recent paper by
Park and Sung.I4 We use the Rouse model and find that the free energy of activation
has a square root dependence on the temperature T , leading to a non-Arrhenius form
for the rate. While in the short chain limit Park and Sung find the activation energy
to be linearly dependent on N, we find that for long chains, the activation energy is
independent of N. We also show that there is a special time dependent solution of the
model, which corresponds to a kink in the chain, confined to the region of the barrier. In
usual non-linear problems with a kink solution, the problem has translational invariance
and the kink can therefore migrate. In our problem, the translational invariance is not
there, due to the presence of the barrier and the kink solution is not free to move in
space. However, the polymer on which the kink exists can move, though the kink is
fixed in space. Thus, the polymer goes from one side to the other by the motion of
the kink in the reverse direction on the chain. If there is no free energy difference
between the two sides of the barrier, then the kink moves by diffusion and the time of
crossing t,,,, N N 2 . If there is a free energy difference, then the kink moves with a
non-zero velocity from the lower free energy side to the other, leading to t,,,, N N .
Our result that t,,,,N N is in agreement with the recent experiments of Kaisanowicze
et. allo where DNA molecules were drawn through a nanopore by the application of
a potential difference. We also consider the translocation of hydrophilic polypeptides
across hydrophobic pores. Biological systems accompolish this by having a hydrophobic
signal sequence at the end that goes in first. Our analysis leads to the conclusion that
for such a molecule, the configuration of the molecule in the transition state is similar
to a hook, and this is in agreement with presently accepted view in cell biology.6
7. ACKNOWLEDGEMENTS
I thank Professor S.Vasudevan for interesting discussions.
References
1 H.A. Kramers, Physica (Utrecht) 7,284 (1940).
2 S. Chandrasekhar, Rev. Mod. Phys., 15,1 (1943).
3 P. Hanggi, P. Talkner and M. Borkovec, Rev. Mod. Phys., 62, 251 (1990).
4 H. Riezman, Science, 278, 1728 (1997)
5 W.Sung and P.J. Park, Phys. Rev. Lett., 77,783 (1996), see also: P.J. Park and
W. Sung, J. Chem. Phys., 108, 3013 (1998) and P.J. Park and W. Sung, Phys.
Rev.E 57,730 (1998).
6 B. Alberts, D. Bray, A. Johnson, J. Lewis, M. Raff, K. Roberts and P. Walker,
Essential Cell Biology, Garland Publishing Inc, New York (1998).
7 S.M. Simon and G. Blobel, Cell, 65,371 (1991).
8 B. Dreiseikelmann, Microbiological Reviews, 58, 293 (1994)

9 V. Citovsky and P. Zambryski, Ann. Rev. Microbiol. 47,167 (1993).
10 J.J. Kasianowicz, E. Brandin, D. Branton and D.W. Deamer, Proc. Natl. Acad.
Sci. USA, 93, 13770 (1996).
11 J. Han, S.W. Turner and H.G. Craighead, Phys. Rev. Lett., 83, 1688 (1999).
12 M. Muthukumar and A. Baumgartner, Macromolecules, 22,1937 (1989).
13 K.Kiran Kumar and K.L. Sebastian, Phys. Rev. E62,7536 (2000).
14 P. Park and W. Sung, J. Chem. Phys., 111,5259 (1999).

15 D.K. Lubensky and D.R. Nelson, Biophysical Journal, 77,99005 (1999)
16 K.L. Sebastian, Phys. Rev. E 61, 3245 (2000). K.L. Sebastian, J. Am. Chem. SOC.
122,2972 (2000)
17 K.L. Sebastian and A.K.R. Paul, Phys. Rev. E 62, 927 (2000).
18 K.L. Sebastian, Phys. Rev. E 62, 1128 (2000).
19 M.Doi and S.F. Edwards, The theory of polymer dynamics, Clarendon Press, Ox-
ford, (1986). The model has the defect of not taking excluded volume interactions
in to account.
20 R. Rajaraman, Instantons and Solitons, North Holland, Amsterdam (1982)
21 A. Scott, Nonlinear Science, Oxford University Press, Oxford (1999).

22 P.M. Chaikin and T.C. Lubensky, Principles of Condensed Mutter Physics, Cam-
bridge University Press, Cambridge, (1998).
23 There is an extensive literature on this topic. The most relevant to our discussion
are: M. Buttiker and R. Landauer, Phys. Rev. A 37,235 (1988), P. Hanggi, F.
Marchesoni and P. Sodano, Phys. Rev. Lett., 60, 2563 (1988). F. Marchesoni,
Phys. Rev. Lett., 73,2394 (1994).
24 Alok K.R. Paul and K.L. Sebastian, to be published.
I1 Proteins, Lipids and Their Interactions
LIPID INTERACTION WITH CYTIDYLYLTRANSFERASE REGULATES
MEMBRANE SYNTHESIS
S. Jackowski and I. Baburinal
Protein Science Division, Department of Infectious Diseases, St. Jude Children's Research
Hospital, Memphis, TN 38105, USA
*Present address: Roche Diagnostics Corporation, Building D, 9115 Hague Road, P.O.
Box 50457, Indianapolis, IN 46250, USA
1 INTRODUCTION
CTP:phosphocholine cytidylyltransferase (CCT) is a major regulator of

phosphatidylcholine biosynthesis. Phosphatidylcholine is the major membrane
phospholipid in eukaryotic cells and also is a precursor to the two other major membrane
phospholipids, sphingomyelin and phosphatidylethanolamjne. The dominant metabolic
route for phosphatidylcholine biosynthesis in mammals is the cytidine diphosphocholine
(CDP-choline) pathway (Figure 1)'. The concentrations of the biochemical intermediates
in the CDP-chohe pathway are not equal, with the phosphocholine pool being the largest
under normal culture conditions and the CDP-choline pool being the smallest, thus
illustrating that the conversion of phosphocholine to CDP-choline by CCT is the slow step
and limits the overall rate of phosphatidylcholine headgroup formation. Changes in the
cellular CCT activity have a direct impact on the rate of phosphatidylcholine synthesis and
the rate of membrane formation.
2 DISTRIBUTION OF CCT ISOFORMS
2.1 Tissue Expression
Three isoforms have been identified: CCTa encoded by the PCYTIA gene in human2
(Pcytla in mouse), and CCTPl or CCTP2, splice variants arising fiom the human PCYTlB
(Pcytlb in mouse). The mouse Pcytla gene was originally named C@dY6 but was
later changed to agree with the human CCT gene nomenclature. The CCTa cDNA was
first cloned fiom rat liver using degenerate PCR primers complementary to the yeast CCT
gene' and an active CCTP cDNA clone was identified in the public expressed sequence
tagged database3. AU three isoforms can complement a Chinese hamster ovary cell line
with dekctive CCT activity which limits phosphatidylcholine production at elevated
temperatures', illustrating that CCTa, CCTPl and CCTP2 catalyze the same biochemical
, ~ . is ubiquitous and expressed in all tissue types examined thus far4.
activity in v ~ v o ~CCTa
The CCTP isofonns, on the other hand, are expressed at their highest levels in fetal tissue,
placenta, ovary, testis and brain. The shorter (31 isoform is the major CCTP expressed in
placenta, liver and fetal lung, whereas the P2 isoform is present in brain. Adult lung tissue
appears to have lost CCTP and normal wild-type macrophages do not
express signiticant levels of CCTf3". However, genetic deletion of the CCTa locus causes
CCTj32 gene expression to increase dramatically", thereby compensating for the loss of
CCTa.
Choline
CK = Choline Kinase
CM-ChoUmCytidylylbono~
cm = choline PhO@olmnrlsrarre
Figure 1 Pathway of Phosphatidylcholine Biosynthesis. Choline (4 is incorporated into

cells and phosphorylated by choline kinase,forming phosphocholine.
Phosphocholine and cytidine triphosphate (CTP) are substratesfor the next
enzyme, the CTP:phosphocholine cytidylyltransferase (CCT) whichproduces
cytidine diphoshocholine (CDP-choline). The choline phosphotransferase then
combines CDP-choline together with diacylglycerol (DAG) toform
phosphatidylcholine.
Lipid Interaction
M
Catalytic Region a-Helix C-Term
CCTa
Figure 2 Comparison of CCT Isoforms. Thefunctional domainsfound in the CCTa:

CCTPI and CCTP2proteins are indicated.
CCTPl and CCTj32 are identical proteins from the amino-terminus through residue
323, but differ l?om each other at their carboxyl termini (Figure 2)4. The predicted CCTP2
protein has 369 amino acids whereas the CCTPl protein is smaller, with 333 amino acids3.
The CCTQl represents a truncated version of CCTj32 and lacks most of the
phosphorylation sites that characterize the carboxyl terminal end of CCTP2. CCTa, which
has 367 amino acids, has a highly phosphorylated carboxyl terminus but the sequence
11 Proteins, Lipids and Their Interactions 165
varies considerably from that of CCTP2. The CCTa and the CCTP proteins also differ at
their respective amino termini, but are strikingly similar in the central catalytic domain and
the adjacent a-helical domain. Like CCTa, CCTPl and CCTP2 require the presence of
lipid regulators for maximum catalytic activigS4and the helical domains of each protein
mediate an interaction with
2.2 Cellular Location
A hctional nuclear localization signal characterized by a cluster of positive charges

distinguishes the amino terminus of CCTa16 fiom CCTP and, in fact, the majority of
CCTa protein can be found in the cell nucleus in many cell type^^,'^-'^. A smaller amount
of CCTa associates with the endoplasmic r e t i c u l ~ m ~ together
” ~ ’ ~ ~ with the CCTP
proteins394.The role of nuclear CCTa is currently beiig investigated by several groups and
two hypotheses are current. One hypothesis hvors the nucleus as a reservoir for inactive
CCTa, whereby the protein translocates to the endoplasmic reticulum when immediate1
needed for membrane biosynthesis, as in the IIC9 cellular response to serum stimulation*B.
However, translocation between the nucleus and the cytoplasm m a y not be a general
h c t i o n of further cell cycle progression” as originally postulated”. Overexpression of
CCTa in a variety of cell lines results in CCTa localization in the nucleus, without a
detectable immunofluorescent signal outside of this organelle despite perturbation of the
cell cycle”. The alternative hypothesis about the role of nuclear CCTa is based on the
observation that CCTa associates with chromatin2’. CCTa is active in the production of
highly saturated hospholipids which are unique among the cellular phosphatidylcholine
molecular speciesY2 . These phospholipids are actively synthesized in nuclei isolated during
the transition &om cell cycle arrest to growth in response to serum stimulation of IMR-32
neuroblastoma cells, indicating that at least a portion of the CCTa remains nuclear and
catalytically active during the Go to G1 transition.
We investigated the distribution of CCTa protein as a function of cell cycle
progression through two cell cycles in the BAC1.2F5 mouse macrophage cell line. The
cells were cultured on slides4 and synchronized by removal of growth factor, i.e., colony
stimulating factor-1 (CSF-1)23. Cell cycle progression was monitored using flow
cytofluorometry of permeabilized cells stained with propidium iodide for quantitation of
DNA content23. CCTa-specific antibodies were modzed with Oregon Greec-fluorescent
label and CCTP-specific antibodies were labeled with Texas Redm using a Molecular
Probes kit. Cells were fixed, permeabilized and stained with the antibodies as described4.
At least 100 cells per condition were examined using direct immunofluorescent confocal
microscopy and the intensity gain for the fluorescent signal was maintained at the same
setting during examination of cells. The use of direct immunofluorescence, rather than the
more common indirect method, together with preparation of labeled antibodies with a 1:1
molar ratio, increased the sensitivity of the fluorescent signal and allowed us to visualize
smaller quantities of CCT protein.
Asvnchronous cells statved Cells
Figure 3 CCT Localization by Direct Immunofuorescent Microscopy. Isoform-specific

antibodies were labelled directly withfluorescentprobes and used to detect the
distribution of CCTa and CCTpproteins in situ in either normally growing
asynchronous macrophage cells or afrr 18 hours of CSF-1 growthfactor
withdrawal.
As expected, the CCTa protein was distributed both in the nucleus and outside the
nucleus in asynchronous, proliferating cells (Figure 3). CCTP was exclusively located
outside the nucleus in these same cells. In contrast, the CCTa protein was located
exclusively outside the nuclei of cells arrested in Go/G1 due to removal of CSF-1 fiom the
culture rnedi~m~~(Figure 3). Also, the amount of CCTa protein per cell was significantly
reduced when cells were arrested by starvation for CSF-1, consistent with the observation
that CCTa gene expression is dependent on CSF-1 in the medium24.
CCTa protein accumulated and translocated to the cell nuclei with time following re-
addition of CSF-1 (Figure 4). Accumulation in the nucleus was not significant until 36
hours after CSF-1 re-addition, however, which was well after the synchronized population
progressed through two complete cycles of growth. As a follow-up experiment, normally
proliferating asynchronous cells were examined for their cell cycle distribution which
indicated that 66.3% of the cells were in GI, 28.6% in S and 5.1% in G2 plus M phases. In
all cases, CCTa was located both within and outside the nucleus, supporting the view that
CCTa translocation is not associated with a cell cycle event. Rather, the data suggest that
the accumulation of CCTa protein in the nucleus may be a h c t i o n of the quantity of
CCTa protein. The amount of CCTa protein is reduced during CSF-1 starvation, and then
new CCTa protein is synthesized following return of CSF-1 to the growth medium and
stimulation of CCTa gene expressionz4. The endoplasmic reticulum is the site of new
protein synthesis and thus, new CCTa is likely to first accumulate at this site and then
move to the nucleus as the protein quantity increases. Interestingly, the extended absence
of CSF-1 causes the cells to undergo a program of cell death called apoptosis. In turn, one
might postulate that the absence of CCTa from the nucleus for an extended period may
contribute to the characteristic chromatin condensation and fragmentation associated with
apoptosis.
I1 Proteins, Lipids and Their Interactions 167
Q I0 20 30 40 Asyn.
HWIS PIUS CSF-I
Figure 4 Quantitation of CCT Positive Nuclei. Macrophage cells were synchronized in
GdGl by CSF-1 withdrawal and then were examinedfor CCTa distribution
following re-addition of CSF-1. Cell cycle progression was determined by flow
cytofluorometry.
2 BIOCHEMICAL REGULATION BY LIPIDS
Lipid interaction with CCT regulates enzymatic activity. Lipids bind tightly to CCT, even
when the protein is purified in soluble form, and a detergent wash is necessary to reliably
assay the lipid-activation or lipid-inhibition of the enzyme'423925ag . The catalytic fiagment
is not activated by lipid nor inhibited by delipidation, and CCT catalytic activity is
compromised when the lipid regulatory domains are in contrast to the full-
length protein. While the activity ofthe truncated protein is greater compared with the
delipidated full-length protein, it is less than the full-length protein in the presence of
lipid3', particularly when the CCT constructs are assayed under identical condition^^^. The
first model for lipid regulation of CCT proposed that the helical domain was inhibitory to
activity13, perhaps due to steric interference of critical reaction sites3', and suggested that it
was the removal of this inhibition that was responsible for elevated activity. A second
early model proposed that the re-alignment of the regulatory domain by lipid-bind'
caused a conformational change in the CTP substrate site, thereby increasing its afhityY.
The key to the development of the second model was not only a comparison between the
fill-length, lipid-activated CCT and the truncated catalytic fiagment, but also comparison
with the delipidated, full-length protein. Close examination of the data in subsequent
studies from the two laboratory groups reveals that both models are correct and are not
mutually exclusive, and that the regulatory domain(s) perform both functions (Figure 5)31.
Removal of the lipid regulatory domain(s) results in a dysregulated protein that lacks both
positive and negative control over CCT catalytic activity based on the data fiom kinetic
analyses3' and carefbl specific activity measurements '.
r -1
400 I-
100 €2
Cat Cat+20 W+39 Helix C-Term Corn-

PWho:deicacid (50 mow)
Figure 5 Lipid Responsiveness of CCT molecular constructs. 7he catalytic region alone
(Cat) was assayed using high CTP (I 0 mM) and was not stimulated by lipid.
Two longer constructs with 20 additional residues (Cat +20) or 39 additional
residues (Cat +39) at the carboxyl terminus also have the same activity and do
not respond to lipid. Expression of either the helical domain (He0 or the
authentic carboxyl terminal domain (C-Term) confer slight inhibition in the
absence of lipid or lipid activation, similar to the full-lengthprotein
(Complete). ( D r m porn tabular data in reference3'.)
3.1 Regulation by the Helical Domain
Two lipid interaction domains have been identified in xnammah CCTa and are contained
*
within the helical region spanning residues 257 through 290, and Within the carboxyl
terminal region fiom residues 310 through 3673'. The helical region has long been
recognized as a site for membrane lipid binding and has been studied and reviewed
e x t e n s i ~ e l $ ~ ~As
~ ~a. result, the limits of the membrane interaction site within the helical
domain are delineated more precisely. The helical region is characterized by three 11-mer
repeats, each of which is sufficient to form an a-helix and one individual 11-mer Unit is
capable of binding to phosphatidylcholine vesicles34. However, the tandem repeats
actually form one contiguous 33-residue helix when associated with an optimizing lipid
mixture35and the entire 30+ helical domain is required to activate3' or inhibit the enzyme25
in the absence of the carboxyl terminus. A peptide corresponding to the helical region
lacks structure in solution and formation of the helical conformation requires addition of
phosphatidylcholine vesicles35, suggesting that the regulatory domain may be a random
coil in the delipidated form of CCT and then convert to an a-helical conformation when
lipid mixtures are added. One can envision the bulk of a random coil interfering with the
reactive sites of the catalytic domain, accounting for the inhibitory nature of the delipidated
helical region (Figure 5). The structuring of the helical region in the presence of lipid,
then, likely influences the structure of the adjacent catalytic domain, slightly rearranging
the CTP site to enhance binding of the ~ubstrate'~ or to promote the chemical events
associated with the reaction3', for example, CTP dephosphorylation, phosphocholine
esterscation of CMP, or exit of inorganic phosphate or CDP-choline. The CTP reactive
site appears to be the relevant target for regulation since the activity of the delipidated
enzyme can be revealed by very high concentrations of this n u c l e ~ t i d e ' ~ ~ ~ ~ ~ ~ ~ .
6.
Figure 7 CCT response to membrane curvature elastic stress. Panel A: The tendency of
a monolayer to bend toward the polar aspect of the interface is influenced by
type 2 lipids. The CCT amphipathic helix inserts into the monolayer half of the
bilayer and the stress is relieved. Panel B: The insertion of type I lipids
increases the tendency of a monolayer to bend awayfrom the polar aspect of
the interface. CCT does not insert into the type 1 monolayer.
The helix is amphipathic and features hydrophobic amino acids clustered along one
side of its long axis, with charged amino acids distributed along the opposite side. The
hydrophobic portion of the helix is stabilized by interaction with the acyl chains of a
phosphatidylcholine monolayer, positioning the charged residues of the helix among the
charged phosphate and choline moieties of the phospholipid headgroups. The insertion of
the CCT helix into one leaflet of a phosphatidylcholine bilayer is predicted eom the
alignment of its amino acids, although phosphatidylcholine alone will not activate the
enzyme3' and other phospholipids cannot substitute. Mixtures of type 2 lipids with
phosphatidylcholine maximally activate CCT. Type 2 lipids include the zwitterionic
phosphatidylethanols, fatty acids which in the presence of phosphatidylcholine are
only partially deprotonated and neutral lipids, such as the dia~ylgycerols~~. These lipids
increase the curvature elastic stress of the lipid monolayer which, in turn enhances CCT
binding to the membrane (Figure 6). The enhanced binding arises out of the increased
amount of curvature elastic stress that is released by the splay of the lipid chains upon
insertion of the hydrophobic strip of the binding helix?6. Type 1 lipids thereby disallow
the CCT helix fiom inserting into the lipid monolayer due to their tendency to bend the
phospholipid structure away &om the polar interfa~e~~. Type 1 lipids thereby oppose the
activation of CCT and compete kinetically with the type 2 lipid when enzyme activity is
assayed in vitro2'.
alkylphospholipid antineoplastic agents3 . r
Type 1 li ids include lysophosphatidylchohe and the
The membrane curvature elastic stress hypothesis incorporates many of the features of
previous hypotheses for CCT activation by lipids and, in doing so, resolves many of the
associated concerns. CCT binds to phosphatidylcholine vesicles that contain negatively
charged lipids, such as phosphatidylserhe or fatty acid, and electrostatic interaction
between the anionic lipids and the positively charged amino acid residues in the
amphipathic helix has been suggested previously as an underlying mechanism37938.
However, electrostatic interactions cannot account for enzyme activation by uncharged
170 Biophysical Chemistry:Membranes and Proteins
lipids, such as diacylglycerol. Similarly, CCT inhibitors such as lysophosphatidylcholine

and the synthetic alkyl analogs ET-18-OCH:5 or HexPC26, are characterized by
phosphocholine moieties and thus argue against inhibition due to alteration of the
phospholipid headgroup packing of the phosphatidylcholine vesicle2'. Rather, activators or
inhibitors can be grouped into two distinct classes which either promote negative or
positive elastic stress, respectively, of a phosphatidylcholine bilayer and the magnitude of
the effect is a h c t i o n of the mole percent of the incorporated am~hiphile~~.
Lateral stress
of a phospholipid monolayer is determined by the repulsion of headgroups, the interfacial
tension and the outward chain pressure of the acyl groups. The curvature elastic stress of a
bilayer, in turn, is determined by the relative degree of lateral stress of the back to back
monolayers. A subsequent study conhns the basic tenets of the membrane curvature
elastic stress hypothesis whereby CCT binding and activation correlate very well with
relaxation of the curvature strain energy of several chemically distinct type 2 lipid
systems39. In this study, however, cholesterol does not activate CCT and the authors
attribute this to its inability to alter the surface hydrophobicity despite its tendency to
increase the negative strain on a phosphatidylcholine m ~ n o l a y e r ~Alternatively,
~. if one
considers cholesterol as a rigid lath shaped amphiphilic molecule, then it cannot splay and
fill the regions beneath the CCT. This aspect would explain cholesterol's poor activation
of CCT, despite cholesterol's ability to change the rigidity of the membrane so that any
curvature of the membrane would be sensitive to its mechanical properties.
3.2 Regulation by the Carboxyl Terminal Domain
The carboxyl-terminal region is also known as the phosphorylation domain and only
recently has it been demonstrated to interact with lipids3'. The helical domain and the
carboxyl-terminal region interact with lipids in dserent ways (Figure 5 ) , as indicated by
lipid protection of the hll-length protein fiom proteolysis12 and using CCTa molecular
constructs3*. The helical domain is hlly protected fiom chymotrypsin proteolysis in the
presence of phosphatidylcholine vesicles, while the carboxyl terminus is exposed and
susceptible to degradation12. Comparative analysis of CCTa proteins with only one
domain tethered to the catalytic core reveals that the helical domain is responsive to both
neutral and anionic lipids embedded in phosphatidylcholine vesicles, and the carboxyl
domain is responsive to anionic lipids only, preferably mixtures of phosphatidic acid and
Triton X-100detergent3'. Either domain can hlly activate CCTa when the other is absent
Erom the protein, and either domain is inhibitory when delipidated3'. The fact that either
domain confers the same range and degree of regulation on CCT activity (Figure 5 ) argues
against a specific inhibitory interaction between either regulatory domain and the catalytic
core in the absence of lipid, and supports a general steric hindrance by an unstructured
peptide. One would postulate that the carboxyl terminal domain, then, takes on organized
structure in the presence of lipid although there is no experimental evidence as yet. The
two domains can regulate CCTa independentlg' or coordinately3*,depending on the lipids
that are present. The carboxyl domain can also interfere with lipid activation via the
helical domain31, by conferring negative cooperativity on the kinetics of the lipid
activation2'. Either interaction between the two regulatory domains, or competition for
binding to a c o m o n anionic lipid may account for the apparent dampening of the enzyme
response by the carboxyl terminal region. The high degree of similarity of the carboxyl
termini of CCTP2 and CCTa suggests that CCTP2 would be regulated in an analogous
II Proteins, Lipids and Their Interactions 171
40
30
8z! 20
10
Figure 8 The CCT Carboxyl Terminus Prefers Phosphatidic Acid. CCTaprotein

lacking the helical domain but retaining the authentic carboxyl terminus was
assayedfor the response to various anionic lipids (20 mol %) mixed with
TritonX-I00 or phosphatidylcholine. (Redrawnfrom tabular data 3'.)
manner, and that CCTPl would not be responsive to low concentrations of phosphatidic
acid due to the absence of this unique domain.
The rodent CCTa carboxyl terminus is phosphorylated reversibly at 19 serine
residues4' whereas the humm CCTa domain has 13 serines, all of which are potential
phosphorylation sites and several kinases are capable of utilizing CCTa as s~bstrate'~*~'.
CCTP2 is also highly phosphorylated at the carboxyl-terminus and has 20 potential
phosphorylation sites, whereas CCTPl has only 5 phosphorylation sites4. CCTa is
phosphorylated in a cell cycle-dependent manner23: the population of CCTa molecules is
phosphorylated least in early GI, and attains a higher phosphorylation state as cells
progress fbrther through GI, S and G2M phases. Increased phosphorylation coincides with
a protein shift on denaturing polyacrylamide gels to a slower-migrating molecular species
and there are at least three phosphorylated species that have been identified by gel-~hifl~~.
When cells are arrested in M phase by a cell-cycle blocker, only hyperphosphorylated
CCTa is evident as characterized by the slowest migrating gel forms. Phosphorylation is
inversely proportional to CCTa activity in immune precipitate^^^ and in in vitro assays15y42.
The enzyme gains activity after G1 initiation and peaks at mid-Gl phase, then becomes
progressively less active as phosphorylation increases through the remainder of the cell
cycle. Cellular CCT activity is at its lowest level in M phase, corresponding to the
hyperphosphorylated state. In cells that are proliferating asynchronously, decreased CCTa
phosphorylation often correlates with a higher rate of phosphatidylcholine biosynthesis and
a tighter association between CCT and the particulate tiaction in permeabilized cells43.
Dephosphorylation does not drive the apparent membrane translocation, however, but
rather, it has been suggested that the CCT protein may be a substrate for phosphatases
following membrane association4. Dephosphorylated CCT may, in turn, be a substrate for
protease
4 CELLULAR REGULATION OF CCT ACTIVITY
4.1 Feedback Regulation by Cell-Associated Lipids
Type 2 and type 1 lipids regulate CCT in vivo (Figure 10). Exogenous fatty acids
immediately stimulate CCT and phosphatidylcholine biosynthesis in a variety of cells and
thsUes43,4649 . Stimulation of phosphatidylcholine biosynthesis aiso follows treatment of
cells with phospholipases C or D, generating diacylglycerol or phosphatidic acid,
respectively, in situ at the cell membrane43248'so.It is dficult to prove, however, that
enhanced membrane phosphatidylcholine synthesis is a direct result of type 2 lipid
invasion of the CCT membrane environment rather than a cellular signaling event, such as
activation of a CCT phosphatase. Since membrane phosphatidylcholine synthesis is
essential for cell survival, a basal rate of CCT activity is always operating and hrther
activation may be limited by the availability of an inactive CCT pool, thus dampening the
maximal CCT activity achievable while probing the system experimentally.
Overproduction of phosphatidylcholine due to increased CCT activity also does not yield
an overt accumulation of except in a few specific cell types under unique
circumstances. For example, the loadin of macrophages with free cholesterol stimulates
formation of lamellar membrane bodies?4 . Rather, the cellular phospholipid content is
usually controlled by phospholipase degradation which o f k t s any biosynthetic increase to
maintain homeostatic control over the absolute amount of cellular membrane52y53.Lastly,
the type 2 activators identified thus far are readily metabolized and converted to inactive
molecular species. The intracellular fkee fatty acid pool is small and endogenous fatty
acids are either modified to acyl-coenzyme A molecules, bound to fatty acid binding
proteins, or incorporated into phosphatidylcholine, sphingomyelin or triacylglycerol, all of
which are inactive in regulating CCT activity. However, exogenous unsaturated fatty acids
can remodel pre-existing p h o s p h a t i d y l e t h o l , which also can, in turn, stimulate
CCT. Similarly, phosphatidic acid or diacylglycerol, either from de novo synthesis or
generated in situ in the membrane by phospholipase treatment of cells, can become part of
the phospholipid pool or the triacylglycerol poolss and thus provide only transient,
localized stimulation of CCT activity. The plasticity of the system is ultimately beneficial
for cells, allowing the population of CCT proteins to respond immediately near sites of
membrane modification or damage.
Negative regulation by treatment of cells with type 1 lipid, such as
lysophosphatidylcholine, clearlr6 causes inhibition of CDP-choline and de novo
phosphatidylcholine production . Inhibition by this lysolipid does not interrupt cell
growth, however, because of its conversion to phosphatidylcholine. On the other hand, the
non-metabolizable type 1 lipids, namely the alkylphospholipid antineoplastic agents, reveal
persistent CCT inhibition and foster the development of a quantifiable cellular phenotype.
The synthetic alkylphospholipids, such as ET-18-OCH3 (edelfosine), are structural analogs
of lysophosphatidylcholine, a natural CCT inhibitor25. Lywphosphatidylchohe does not
cause apoptosis because it bypasses the CCT inhibition of the CDP-choline pathway and
provides an alternate source for the membrane phospholipid necessary for survivals'. Both
a genetic model and a molecular model empower the hypothesis. A mutant Chinese
hamster ovary cell line with a conditional defect in CCT8 undergoes apoptosis under the
non-permissive conditions8. The apoptosis caused by genetic inactivation of CCT can be
prevented by supplementing cells with a source of phosphatidylcholine or
lysophosphatidylcholine to maintain membrane integrity5'. Induced overexpression of
CCTa in an engineered human HeLa cell line prevents the apoptosis caused by exogenous
alkylphospholipids, proving that the target is indeed CCT. Addition of supplementing
lysophospholipid to the medium does not increase survival further59,indicating that both
methods rescue the same intracellular process. Altogether, the data strongly support the
concept that inhibition of CCT, or the CDP-choline pathway, is a primary effector
mechanism in apoptosis. Thus future investigation of CCT as a cellular target in ligand-
initiated apoptosis, for example in response to FAS ligand or TNFa, is warranted.
Inactivation of CCT by proteolytic degradati~n"~~~, ',
reduced expression' or inhibition of
the CCT-lipid interaction, all would decreased membrane phosphatidylcholine synthesis.
P-Cho
Figure 10 Feedback Regulation of CCT The CDP-choline pathway of

phosphatidylcholine biosynthesis is regulated by CCT. Phosphatidylcholine is
degraded by phospholipase A (PLA),phospholipase C (PLC) or phospholipase
D (PLD), releasing lipid products (a)which either activate or inhibit CCT.
4.2 Regulation of CCT Expression
While it is clear that CCT is strictly dependent on lipid for maximum activity, high-level
overexpression of the truncated, dysregulated and compromised catalytic fragment can
complement cells with defective endogenous CCT13. These data demonstrate that despite a
need for higher cellular CTP14330, despite reduced catalytic efficiency'sJo, and without the
ability to associated with cellular lipids'4y3o,CCTa can still be active in cells13931.
Expression levels of CCT, then, become another factor in regulating the rate of
phosphatidylcholine production, in addition to the nature of CCT association with lipid.
When CCTa expression is knocked-out in mouse macrophages by genetic inactivation of
the Pcytla locus, CCTP2 expression increases dramatically to compensate for the loss of
CCTa, allowing for normal macrophage development and apparently normal overall
mouse development". However, the macrophage CCTP2 total activity is signiticantly
lower compared to that otherwise attributed to CCTa, and the macrophages are restricted
in their response to the challenge of fiee cholesterol loading. Downregulation of CCTa
expression by proteolytic degradation has been observed in pancreatic acini treated with
cholecy~tokinin~~ and in alveolar type I1 cells treated with TNFa''. Cholecystokinin
stimulates the release of digestive enzymes and alveolar cells specialize in secreting
disaturated phosphatidylcholine, a major component of lung surfactant. In both cases,
cellular phosphatidylcholine synthesis is reduced sigdkantly when CCT protein levels are
reduced, supporting a role for CCT expression in the regulation of cellular phospholipid
content and function.
5 CONCLUSION
Alteration of CCT activity changes the flux through the phosphatidylcholine biosynthetic
pathway and CCT is regulated biochemically by different lipid mixtures. CCT is activated
by anionic lipids or neutral lipids when presented to the enzyme in the context of a
phosphatidylcholine vesicle. The CCT helical domain mediates interaction with this lipid
structure. CCT activity is inhibited by inclusion of lysolipids in the phosphatidylcholine
vesicles by subverting enzyme insertion into the vesicle. CCT is also activated by
phosphatidic acid when presented to the enzyme in a mixture with a non-ionic detergent.
The carbxyl terminus mediates the second type of lipid interaction and suggests that
CCTa and CCTP2 regulation in vivo is not strictly dependent on association with
membrane bilayers. The distribution of the CCTa, CCTPl and CCTP2 isoforms in tissues
likely contributes to the membrane characteristics and h c t i o n of cellular compartments,
particularly CCTa in the nucleus. Evidence argues against cell-cycle regulation of CCTa
translocation outside of the nucleus. Changes in the cellular lipid environment govern the
response of CCT through physical association between the protein and the membrane,
allowing the cells to produce more or less phosphatidylcholine as needed.
Abbreviations: CCT, CTP:phosphocholine cytidylyltransferase; CTP, cytidine

diphosphate; CDP-choline, cytidine diphosphocholine; ET- 18-OCH3,
octaethyleneglycolmonohexadecxy1 ether; HexPC, hexadecylphosphocholine;CK, choline
h e ; CPT, choline phosphotransferase;DAG or DG, diacylglycerol; P-Cho,
phosphocholine; PtdCho,pho sphatidylcholine;PtdEtn, phosphatidylethanolamine; Sph,
sphmgomyelin, LPC, lysophosphatidylcholine;PtdOH, phosphatidic acid
Acknowledgements
We thank George Attard (University of Southampton) and Richard Templer (Imperial

College) for valuable discussion and keen interest. The research is supported by the
National Institutes of Health Grant GM45737, Cancer Center (CORE) Support Grant
CA21765, the American Lebanese and Syrian Associated Charities, the Royal Society, and
the Engineering and Physical Sciences Research Council.
References
1 C. Kent, Annu. Rev, Biochem., 1995,64,315.

2 G.B. Kalmar, R. J. Kay, A.C. LaChce and R.B.Cornell, Biochim. Biophys. Acta,
1994,1219,328.
3 A. Lykidis, K.G. Murti and S. Jackowski, 3: Biol. Chem., 1998,273, 14022.
4 A. Lykidis, I. Baburina and S. Jackowski, J. Biol. Chem., 1999,274,26992.

5 M.S. Rutherford, C.O. Rock, N.A. Jenkins, D.J. Gilbert,T.G. Tessner, N.G. Copeland
and S. Jackowski, Genomics, 1993, 18,698.
6 W. Tang, G.A. Keesler and I. Tabas, J. Biol. Chem., 1997,272, 13 146.
7 G.B. Kalmar, R.J. Kay, A. LaChance, R. Aebersold and R.B. Cornell, Proc. Natl.
Acad. Sci. U. S. A., 1990,87, 6029.
8 J.D. Esko, M.M. Wermuth and C.R.H. Raetz, J. Biol. Chem., 1981,256, 7388.
9 T.D. Sweitzer and C. Kent, Arch, Biochem. Biophys., 1994,311,107.
10 R.K. Mdampalli, A.J. Ryan, RG. Salome and S. Jackowski, J. Biol. Chem., 2000,
275,9699.
11 D. Zhang, W. Tang, P.M. Yao, C. Yang, B. Xie, S. Jackowski and I. Tabas, J. Biol.
Chem., 2000,275,35368.
12 L. Craig, J.E. Johnson and R.B. Cornell, J. Biol. Chem., 1994,269,33 1 1.
13 Y. Wang and C. Kent. J. Biol. Chem., 1995,270,18948.
14 W. Yang, K.P. Boggs and S. Jackowski, J. Biol. Chem., 1995,270,23951.
15 R.B. Cornell, G.B. Kalmar, R.J. Kay, M.A. Johnson, J.S. Shanghera and S. L. Pelech,
Biochem. J., 1995,310,699.
16 Y. Wang, J.I.S. MacDonald and C. Kent, J. Biol. Chem., 1995, 270,354.
17 C.J. DeLong, L. Qin and 2.Cui, J. Biol. Chem., 2000,275, 32325.
18 Y. Wang, T.D. Sweitzer, P.A. Weinhold and C. Kent, J. Biol. Chem., 1993,268,
5899.
19 I.C. Northwood, A.H. Tong, B. Crawford, A.E. Drobnies and R.B. Cornell. J. Biol.
Chem., 1999,274,26240.
20 M. Houweling, Z. C c C.D. Anfuso,M. Bussiere, M.H. Chen and D.E. Vance, Eur. J
Cell Biol., 1996,69, 55.
21 A.N. Hunt and G.C. Burdge, Biochem. SOC,Trans., 1998,26, S223.
22 A.N. Hunt, G.T. Clark, G.S. Attard and A.D. Postle, J. Biol. Chem., 2001,276,8492.
23 S . Jackowslu,J. Biol. Chem., 1994,269,3858.
24 T.G. Tessner, C.O. Rock, G.B. Kahnar, R.B. Cornell and S. Jackowski, J Biol.
Chem., 1991,266, 16261.
25 K.P. Boggs, C.O. Rock and S. Jackowski, J. Biol. Chem., 1995,270,7757.
26 K.P. Boggs, C.O. Rock and S. Jackowski, Biochim. Biophys. Acta, 1998,1389, 1.
27 R.B. Cornell, Biochemistry, 1991,30,5881.
28 M.M. Luche, C.O. Rock and S. Jackowski, Arch. Biochem. Biophys., 1993,301,114.
29 W. Yang and S. Jackowski, J. Biol. Chem., 1995,270, 16503.
30 J.A. Friesen, H.A. Campbell and C. Kent, J. Biol. Chem., 1999,274, 13384.
31 A. Lykidis, P. Jackson and S. Jackowski, Biochemistry, 2001,40,494.
32 J.E. Johnson and R.B. Cornell, Mol. Membr. Biol., 1999,16,217.
33 R.B. Cornell and I.C. Northwood, Trends Biochem. Sci., 2000,25,441.
34 J. Yang,J. Wang, I. Tseu, M. Kuliszewski, W. Lee and M. Post, Biochem. J., 1997,
325,29.
35 J.E. Johnson and R.B. Cornell, Biochern. J. , 1994,33,4327.
36 G.S. Attard, R.H. Templer, W.S. Smith, A.N. Hunt and S. Jackowski Proc. Natl.
Acad. Sci. U.S.A., 2000,97,9032.
37 R.B. Cornell, Biochemistry, 1991,30,5873.
38 R.S. Arnold and R.B. Cornell, Biochemistry, 1996,35,9917.
39 S.M. Davies, R.M. Epand, R. Kraayenhof and R.B. Cornell, Biochemistry, 2001,40,
10522.
40 J.I.S. MacDonald and C. Kent, J. Biol. Chem., 1994,269, 10529.
41 M. Wieprecht, T. Wieder, C. Paul, C.C. Geilen and C.E. Orfanos, J. Biol. Chem.,
1996,271,9955.
42 Y . Wang and C. Kent, J. Biol. Chem., 1995,270, 17843.
43 H. Tronchere, M. Record, F. Terce and H. Chap, Biochim. Biophys. Acta, 1994,1212,
137.
44 M. Houweling, H. Jamil, G.M. Hatch and D.E. Vance, J. Biol. Chem., 1994,269,
7544.
45 G.E. Groblewski, Y. Wang, S.A. Emst, C. Kent and J.A. Williams, . IBiol. Chem.,
1995,270,1437.
46 R.K. Mallampalli, R.G. Salome, C.H. Li, M. VanRollins and G.W. Hunninghake, J.
Cell Physiol., 1995, 162,410.
47 H. Tronchere, F. Terce, M. Record, G. Ribbes and H. Chap, Biochern. Biophys. Res.
Commun., 1991,176,157.
48 C. Kent, Biochim. Biophys. Acta, 1997, 1348,79.
49 J.M. Clement and C. Kent, Biochem. Biophys. Res. Commun., 1999,257 643.
50 F. Terce, M. Record, G. Ribbes, H. Chap and L. Douste-Blq, J. Biol. Chem., 1988,
263,3 142.
5 1 F. Terce, M. Record, H. Tronchere, G. Ribbes and H. Chap, Biochim. Biophys. Acta,
1991,1084,69.
52 C.J. Walkey, G.B. Kalmar and R.B. Cornell, J. Biol. Chem., 1994,269,5742.
53 I. Baburina and S. Jackowski, J. Biol. Chern., 1999,274,9400.
54 Y. Shiatori, M. Houweling, X. Zha and I. Tabas, J. Biol. Chem., 1995,270,29894.
55 S. Jackowski, J. Wang and I. Baburina, Biochim. Biophys. Acta, 2000,1483,301.
56 K.P. Boggs, C.O. Rock and S. Jackowski, J. Biol. Chem., 1995,270,11612.
57 J.D. Esko, M. Nishijima and C.R.H. Raetz, Proc. Natl. Acad. Sci. US.A., 1982,79,
1698.
58 Z. Cui, M. Houweling, M.H. Chen, M. Record, H. Chap, D.E. Vance and F. Terce, J.
Biol. Chem., 1996,271, 14668.
59 I. Baburina and S. Jackowski. J. Biol. Chem., 1998,273,2169.
MODELS AND MEASUREMENTS ON THE MONOLAYER BENDING ENERGY OF
INVERSE LYOTROPIC MESOPHASES
A. M. Squires, J. M. Seddon and R. H. Templer
Department of Chemistry, Imperial College, London SW7 2AY, UK
1 INTRODUCTION
Aqueous dispersions of amphiphilic molecules such as lipids can form a rich array of
mesophase structures, both lamellar and non-lamellar. A given lipid system can form
different mesophase geometries, depending on the water content, as well as other fbctors
such as temperature, pressure and pH. Here we present work carried out on a lipid system
which is a mixture of lauric acid, LA, and dilauroylphosphatidylcholine,DLPC, in the molar
ratio of 2:l LA:DLPC. The mesophase structures formed by this system are shown in
Figure 1. The lamellar phase simply consists of a stack of flat bdayers separated by layers of
water. In the inverse hexagonal phase, cylinders of water surrounded by lipid monolayers
stack to make a 2-dimensional hexagonal array. The three inverse bicontinuous cubic phases
each consist of a single continuous lipid bilayer at the centre of which runs a mathematical
surface known as a Triply Periodic Minimal Surface (TPMS).
If we change the geometry of a structure by adding or removing water, we have to
bend the constituent lipid monolayers into different shapes, and this changes the energy of
the system Our research models and measures this as a fimction of monolayer curvature.
An understanding of the bending energy of lipid monolayers enables us to predict
certain properties of a lipid system, such as the mesophase geometry and lattice parameter
adopted in excess water. In addition, certain biological processes involving changes in
membrane topology require intermediate structures of higher curvature, so a study of the
energy associated with curvature will help us to understand the energetics of such
transformations in biological systems. Furthermore, there is considerable evidence that
bending energy itself is used in the regulation of the activity of a number of membrane-
bound proteins.
Figure 1. Structures of the liquid crystalline mesophes formed by 2LA/DLPChvater.

Top left: Lamellar (La ). Top right: Inverse Hexagonal (HI$. Bottom: Inverse
Bicontimous Cubic (Qlj,@om left to right: Gyroid (QIF), Diamond (Ql:) and Primitive
(QI3-
2 BACKGROUND
2.1 Theory
Our treatment assumes that lipid molecules have a preferred chain splay, and therefore that
a given lipid would prefer to form monolayers of a certain constant curvature. If the
monolayer is forced to adopt a structure of a different geometry, then its curvature would
have to deviate away fiom this value, and in each lipid molecule the chains regions would be
squashed or splayed into a different shape. This incurs an energetic cost known as the
Curvature elastic energy. We would expect the curvature elastic energy to increase in the
sequence spheres < cylinders < saddles < flat planes’.
At any point on a surface, the curvature can be described by two principle curvatures,
c1 and c2. In this work we combine these two curvatures to give the mean curvature, H = ?4
(cI+c2), and the gaussian curvature, K = cI x c2. The dependence of curvature elastic energy
on the curvature of the monolayer is assumed to take the following quadratic form, taken
fiom Hefiich’s Ansatz’
where gcis the curvature elastic energy per unit area of the monolayer. The curvature elastic
behaviour of the lipid system is characterised by three constants, Ho, K and Q. These are,
respectively, the spontaneous mean curvature, and the bending stiflhess moduli associated
with mean and gaussian curvature. Here, we present work seeking to measure the values of
these three quantities for the 2LA/DLPC system at 42OC.
The surfwe at which we make our measurements of monolayer curvature is known
as the 'pivotal' (or 'area-neutral') s h e . This surface corresponds to the position on the
lipid molecules where the area per molecule does not change as the monolayer is bent at a
constant temperature. 3* '.
2.2 Experimental Techniques
In this work we present results obtained using three different ways of probing the curvature
elastic behaviow of 2LA/DLPC. These are as follows.
2.2.I Excess water point. Over a certain range of temperatures and water contents, the
2LA/DLPC system forms one of three inverse bicontinuous cubic (&) phases. Below
excess water conditions, a QI1 phase is geometrically constrained to adopt a certain lattice
parameter, where the monolayers have a greater curvature than their preferred value. The
addition of water allows the lattice to swell, increasing the lattice parameter and so lowering
the monolayer curvatures and thus the curvature elastic energy of the system. As more
water is added, the lattice parameter continues to increase until the curvature elastic energy
reaches a minimum. Beyond this point, there is no longer an energetic incentive for the
mesophase to swell any further. Thus at higher hydrations, the lattice parameter of the
mesophase remains constant, and the mesophase co-exists with excess water. Measurement
of the lattice parameter in excess water enables us to determine the geometry and
curvatures of the system at which the curvature elastic energy is minimised'.
2.2.2 Addition of a hydrophobic liquid. The use of a hydrophobic liquid to relieve

packing stress follows an approach first used by Rand et al' to determine
the bending energy of DOPE in the HIIphase.
2LNDLPC forms QIJphases in excess water only because of the packing fiustration
energy associated with the formation of the H,I phase; this is the energy associated with the
variations in lipid chain extension in order to satisrj. the requirement that the mesophase fills
all space6.Purely fiom curvature elastic considerations,the Hi,phase is more stable than the
QII phases, a general result for systems where O>Q /~>-l'. (The value of KG k f o r
2LA/DLPC does indeed fsll in this range, as we shall show later.) When a hydrophobic
liquid such as an alkene is added to a type I1 lyotropic liquid crystal, it preferentially
partitions into regions of higher packing stress. This can lower the packing hstration
energy enough to induce the formation of the HIIphase instead of the inverse bicontinuous
cubic. If enough alkene is added, the system reaches an energetic minimum, and no firther
change in lattice parameter is observed as still more alkene is added. This method allows us
to calculate Ho,which is the value of the mean curvature at this point.
2.2.3 Osmotic stress measurements. The osmotic stress technique gives a direct
measurement of the amount of energy required to remove water fiom a mesophase. The
mesophase is placed in contact with water of a (known) lower chemical potential. This
sucks water out of the mesophase until the energetic advantage in moving water to a lower
chemical potential is halanced by the increase in energy of the mesophase as it is
dehydrated. The technique has been used to measure curvature elastic parameters in the HII
phase in DOPE' and 1-MO', and in QIIphases in 1-M08. Osmotic stress measurements give
information on curvature elastic parameters in these phases because the energetic cost of
dehydrating the mesophase directly reflects the energy required to increase the monolayer
curvature around progressively thinner water channels.
2.3 Previous Work on the ZLA/DLPC System
The phase behaviour of the 2LA/DLPC/water system has already been studied’. Its
Temperature-Compositionphase diagram is shown in Figure 2.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Figure 2. Temperature-Compositionphase diagram of ZLALDLPC/water.
The position of the pivotal surface for 2LA/DLPC has been determined fiom lattice
parameter / water content data’. The area per molecule at the pivotal surface and the
volume of the chain regions below it were found to take the values of A,=108k2A2 and
v,=1230-+30A3respectively. We note that the ‘molecule’ in this stage refers to an ‘average’
unit consisting of one DLPC and two LA molecules. The total volume per molecule, v has
also been determined by density measurements, and was found to be 1735 A3.
3 EXPERIMENTAL METHODS
3.1 Sample Preparation
DLPC and LA were obtained fiom Avanti Polar Lipids (Alabaster, AL) and Fluka
respectively, each at >99%purity, according to the manufacturers. The purity of both lipids
had been c o b e d by thin-layer chromatography. The lipids were left overnight in a
!I Proteins, Lipids and Their Interactions 181
desiccator before use to remove residual moisture. 2LA/DLPC mixtures were prepared by
co-dissolving the appropriate ratio of LA and DLPC in cyclohexane (BDH chemical
supplies, Dorset, UK), and removing the solvent by fieezing to liquid nitrogen temperature
and then freeze-drying overnight. To each was added an appropriate amount of milipore-
filtered distilled water. The sample was then incubated overnight at 42"C, which is
approximately 10°C above the chain melting temperature. It was then allowed to cool to
room temperature before being homogenised by centrifuging to force the components
through a lmm diameter hole in a Teflon 'hourglass'. The sample was centrifbged through
this device approximately 10 times, until it appeared homogeneous.
3.1.1 Addition of a hydrophobic liquid For the samples containing alkene, the lipids
were co-dissolved in an exact weight of cyclohexane. To this was added a weighed amount
of the alkene. The alkene used was 1-octadecene(Sigma chemical company, St Louis, MO)
in a cyclohexane solution of known concentration (approximately 10% by weight). The
mixture was frozen in liquid nitrogen and fieeze-dried overnight. The following day, the
sample was re-weighed to ensure that no lipids or alkene had been removed by the fkeeze-
drying procedure. The sample was then hydrated to 67% water by weight, incubated
overnight at 42"C, allowed to cool, then homogenised using the centfigation procedure
described above.
This experiment requires that there is no significant change in monolayer properties
caused by penetration of the alkene into the lipid monolayer. It has been shown that using
alkene molecules of chain length greater than those of the amphiphiles minimiSes the
possibility of such penetration'. Since we use the 18-C alkene 1-octadecene while the LA
and DLPC lipid c h a i i are both 1242, we assume that no significant cham penetration
occurs.
3.1.2 Osmotic stress method. The osmotic stress method began with hydrated lipid
samples of 60% water by weight. Lipid samples of around 0.lml were osmotically stressed
overnight at 42°C using a solution of polyethyleneglycol (PEG)6000 (BDH, Dorset, UK),
through a semi-permeable membrane made of Visking tubing (Medicell, London, UK)
which had been pre-treated by boiling three times in micro-pure water. After osmotic
stress, the sample was rapidly transferred to a sealed container. Some was then analysed
using x-ray dfiaction. The remainder was weighed, both before and after fieeze-drying,to
determine the water content.
3.2 Sample Analysis
3.2.1 Mass anaZysis. In order to determine the water content of a sample after osmotic
stress, the sample was 6.ust weighed. It was then fiozen at -20°C for >3hours, and fieeze-
dried in ice/water/salt overnight in order to remove all water, and the dry sample re-
weighed.
3.2.2 X-ray dzfiaction. The mesophase and lattice parameter of a sample were
determined by x-ray crystallography. The sample was analysed either in a glass capillary
tube of diameter 1.5mm, wall thickness 0.Olmm (Pantak ltd., Reading, UK) which was
sealed with heatshrink tubing (RS supplies, UK), or else in a sealed cell made fiom a Teflon
spacer of thickness lmm, with Mylar windows attached using double-sided tape, clamped
between two aluminium spacers in a brass sample cell holder. The sample was kept at the
required temperature in the x-ray machine for 20 minutes before the start of the dfiaction
experiment, and several successive x-ray dfiaction exposures were taken, in order to
ensure equilibration. Exposure times of between 2 and 20 minutes were used.
X-rays were produced on a GX-20 rotating anode generator (Entaf-Nonius,
Netherlands) run at 20kV and 25mA, monochromated and focused US& a nickel filter and
Franks double mirror optics to select for Cu K, X-rays (h=1.54 A). The beam was isolated
and fkther attenuated with a set of adjustable tungsten slits in the sample chamber, to
produce a beam of dimensions 0.15 mm x 0.1 mm, with a typical flux of 1 million photons
per second.
The sample temperature was regulated using peltier heating. This was computer-
controlled via a Scorpion K4 micro-controller, which also enabled automated shutter
operation. The equipment provided temperature control up to 80 "C, to an accuracy of
k0.03 "C. The sample chamber and optics were kept under vacuum, to minimiSe both loss in
beam intensity and scatter due to air. The sample chamber included a variable-length flight
tube with a lead beam. stop at the end. This set-up allowed the X-ray detector to be placed
between 100 and 300 mm fiom the sample, keeping an evacuated X-ray path throughout.
The x-ray dijEaction pattern was collected on a (Zn,Cd)S:Ag phosphorescent screen,
intensified using a high-voltage image intensifier, and then focused onto a CCD camera
(Wright instruments) which relayed the image back to the control computer. The dfiaction
pattern was analysed using the TV4 software program, originally written by Prof. S. M.
Gruner (Cornell University) and Prof. E. F. Eikenberry (Paul Schemer Institute,
Switzerland), with more specific alterations made for our purposes by Neil Warrender and
Prof R. Templer, at Imperial College.
The lipid samples were powder-like, and therefore gave dfiaction patterns which
appeared as a series of concentric rings. The radius of each ring can be converted to a
scattering angle for a single reflection, and therefore into a d-spacing for a particular set of
planes within the mesophase. A given mesophase gives a set of reflections where the d-
spacing ratios are characteristic of the symmetry of the mesophase, and the actual values
may be used to calculate its lattice parameter. This was carried out by the TV4 software
package. The HII phase was characterised by the 41, 43, 44 reflections, the Q,: phase by
the 42, 43, 46, 48 and 49 reflections, the Q,: phase by 46, 48 and a broad peak fiom the
420, 422, 424 and 426 reflections, and the QI; phase by the 42,44, and 46 reflections. The
beamline was calibrated with silver behenate, and could measure lattice parameters to an
accuray of*O. 1A.
4 RESULTS, CALCULATIONS AND DISCUSSIONS
4.1 Excess Water Measurements
4.1.1 Results. Previous measurements at fixed hydration on 2LA/DLPC indicate that

the excess water point lies close to a water volume fiaction of 4,=0.55 at 35OC, and that the
water volume b t i o n at the excess water point decreases with increasing temperature (see
Figure 2). A sample of 2LA/DLPC was therefore prepared with a water volume fiaction of
4,=0.595 (60% water by weight). A temperature-scan was carried out on this sample, with
x-ray diffraction patterns taken at intervals of 2°C. The phase sequence and lattice
parameters determined fiom the dfiaction patterns are shown in Figure 3.
-
0
0
10 20 30 40 50 60 70 80
Temperature I OC
Figure 3, Temperature-scan of 2LA/DLPC/60% water by weight. Diflaction patterns were

obtained using 10-mimte exposures, with an equilibrium time of 20 minutes at each 2-
degree increment. m e iemperature-scan was carried out in order of increasing
temperature. Between 28 and 3 4 T , no well&@ted refectrions were observed. ?Itis was
probabb because the system was not forming a SUfJiciently ordered structure.
The strong temperature-dependenceof the lattice parameter indicates that the sample is co-
existing with excess water throughout.
4.1.2 Calculations In order to model the lattice parameters of the Q1l phases, we use a
parallel-interface model where the pivotal surface for each monolayer lies a uniform distance
tnaway &om the underlying minimal s h . Throughout this section we locate the pivotal
surface by using values for molecular dimensions quoted in Section 2. These are
An=108*28iZ,vn=123W30A3and v =1735 A3.
cn
We can calculate for each lattice parameter by solving the following equation at each
different lattice parameter9
The water volume &tion within the mesophase, +w, is calculated using the following
equation, whose derivation is outlined elsewhere":
+ ,/v'~o'(a6A:a3- ( a 2 A t a- 6m,3)'))'
#w=l- (
4v303 a'An3v606
2
a2A,,2v4a4
- 6m4v,'a3x + ,/d2o9(a6A:d - (a2A:a - 6m;x)'))'
+ a3An3v6d
In these equations, 0 is the dimensionless surface area of the TPMS in the unit cell, and has
values of 3.0910, 1.9189 and 2.3451 for Q I: , QJ: and Q/ respectively?. The
corresponding values of x ,the Euler characteristics of the TPMS, are -8,-2 and 4.
Equation 1 can be re-written using values for the mean and gaussian curvatures for a
QJ,phase, assuming a pivotal surface lying parallel to the TPMS'. This gives a first-order
expression for the surface-averaged curvature elastic energy per unit area of the monolayer,
<gc>, in terms of 5" .
The excess water point corresponds to an energetic minimum. Thus differentiating Equation
4 with respect to 5. and equating to zero gives us a relationship between Q/K and Ho at the
excess water point, where en+
For example, at 42"C, the system is in a Q / phase, with lattice parameter

a(Q/)=147.OkO.SA. From Equation 3 we can calculate the water volume fraction, $,
=0.518k0.001. We note that this is the water volume fiaction within the mesophase, and
that the value is smaller than 0.595, the water volume fiaction of the sample overall. This
confirms that the mesophase is coexisting with excess water. From these values of $w and
the lattice parameter a, we can calculate 5. in excess water (go), and thus fiom Equation 5
determine the quantity Ho(rdm). This treatment yields the following values at 42°C.
We can perform the same calculations on the rest of the data shown in Figure 3, to see
how the quantity Ho(K/KG)varies with temperature. The estimated values are shown in
Figure 4.
0.0246
0.0244
0.0242
-
Q,,P Ql,D
0.0238 + +
0.0236
__
00.0232
.023414
0.023 I -
30 40 50 60 70 80
T 1O C
Figure 4. Variation of &(id@) with temperaturefor 2LA/DLPC.
The graph shown a linear increase in Ho(K/KG) with temperature up to 48OC, given by
Ho(K/Q) = 0.0204 + 8.0 x 10" T("C). Above 48"C, the graph shown a change in
behaviour; HO(K/IQ)still shows a linear increase with temperature, but with a different
gradient, and the behaviour is described by H0(d~)(A-')=0.0234+ 1.7 x lo-' T(*C). The
sudden change in behaviour occurs at the point at which the H;r phase &st appears. The
most likely explanation for this is that a slight partitioning of LA and DLPC is occurring,
which leads to a difference in lipid compositionbetween the two phases.
4.2 Addition of Alkene
4.2.I Results. Throughout this section we refer to the alkene content of a sample in
terms of 4,the number of molecules of 1-octadeceneper 2LA/DLPC,calculated fiom the
masses of the constituent components using values of md2LA/DLPC) = 1022.5g mol-' and
al-octadecene) = 252.5g mol-*.The lattice parameter of the fi1phase formed with
samples of different alkene contents in excess water varies with the alkene content as shown
in Figure 5.
As we can see, the lattice parameter, a(fi,), increases with akene content, rg, up to a
certain point, the 'excess alkene point'. This corresponds to an energetic minimum where
the mean curvature H is equal to the spontaneous mean curvature, &. Beyond this point,
a(&) remains constant. From Figure 5, we calculate that the excess alkene point
corresponds to values of ~=0.42&0.06 and 4&,) =79.6*0.6A.
4.2.2 Calculations. We use the values quoted before, of A,=l 08*2A2,v,=l 23W30A3
and v =1735 A3. We also define a new quantity, v,', which represents an average value for
the overall volume of the hydrophobic region below the pivotal surface, including both the
chain regions and the alkene. Given that for 1-octadecene, mdl-octadecene) = 252.56 mol-'
and p(1-octadecene) = 0.788g d-', and that ~ 4 . 4 a2t the excess akene point, we can
calculate that, at this point, vn'=145&40A3.
We can now calculate the radius of the pivotal surface at the excess alkene point, Rp,
using Equation 6. The derivation of this equation is given elsewhere".
(6)
From our experimental results, this gives a value of Rp = 3WlA. We assume that the
spontaneous mean curvature for 2LA/DLPC, Ho, is equal to the mean curvature at this
point, which is given by H = 1424). From this we estimate &= -0.0167k0.0006~-'.
We can now combine this with our value of 0.0238 for the quantity Ho(K/KG) at this
temperature, to give a value for the ratio (KG /K) of K G /K = -0.7W0.03. We note that this
does indeed fall in the predicted range of O>Q k>-1 mentioned earlier.
88 -
86 -
84- 0
82 - 0
80 -
4 78-
2, 761
.i_l
721
70
, El
62
60
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
molecules of alkene per lipid
Figure5 Effect of alkene content on the lattice parameter of the &phase in 2 M D L P C at
42°C.
4.3 Osmotic Stress Experiments
4.3.I Results. The analysis of each sample following osmotic stress both by weight and
by x-ray difliaction yields two sets of data; both the water content and the phase and lattice
parameter can be plotted as a h c t i o n of the osmotic pressure applied by the polymer
solution. The polymer solution, PEG 6000, has been characterised at a range of
concentrations". Online datahttp://aqueous.labs.brocku.ca/data/peg6000is available for the
osmotic pressure applied, TI, as a k c t i o n of the weight% of polymer, and it can be
described by the empirical fit Zog(l3 /Pa) =a + b(wt%)", with the parmeters a=5.12, b 0 . 2 8
and c=0.59.
This equation was used to calculate the osmotic pressure applied, and plots for the
water content and lattice parameters of the 2LA/DLPC samples following this osmotic
stress are shown in Figure 6.
Water contents at different Osmotic Pmsrurer

6 0 1 . . ' i '' " " ' ' ' I 1 ' ' ' .1 . ' . " ' ' ''1
* *
* *
.*
Figure 6. Water contents fleji) and lattice parameters (righr) of osmotically stressed
2LA/DLPC at 42°C.
4.3.2 Calculation of enera. The energy of the mesophase after osmotic stressing to
different levels of dehydrationmay be calculated fiom the osmotic stress data as follows.
The weight% of water in the sample is converted to a volume of water per
2LA/DLPC, v,.
The osmotic pressure ll is plotted against this quantity. This is shown in Figure 7.
A smooth curve is fitted through the data in order to give us a function which we
can integrate. The function chosen in this case has the form n / P a = a x (vJA3)".
The fit is shown in Figure 7,with the parameters a = 3 . 1 2 1 5 ~ 1 0and
~ ~b = 4.936.
500 1000 1500 2000 2500

vw I
Figure 7 . 1 7 ~water contentfor osmotic stress of 2LA/DLPC at 42°C.

The h c t i o n is integrated to give the energy required to dehydrate to a given water

content, per 2LA/DLPC. (In this case we assume that this is the increase in
curvature elastic energy.)
E= - [ndu,,, (7)
Using the curve fit chosen, this gives
a
E=E, + 1 O - ~ O -v,,-(*-’)=E0+ ( 7 . 9 2 9 6 ~~ O - ’ ’ ) V ~ - ~ . ~ ~
6-1
Here, E is the energy per 2LA/DLPC, in J, Eo is the integration constant, and v, the volume
of water per ~LA/DLPC,in Hi3.
4.3.3 Calculation of K and a. While it is possible to extract values of K and Q fiom

the data for the integrated energy calculated as described in the previous section 13, there
are statistical disadvantages to analysing the data in this way”. It is more appropriate to fit
the data for I7 directly, rather than E.
Before this can be done, we need a model for the actual geometry of the pivotal Surface
at which we will be measuring the mean and gaussian curvatures of the monolayers. There
is a choice of two possible ge~metries’~. The first is a ‘parallel interface’ model where the
pivotal surface lies a uniform distance from the underlying TPMS, as described in Section
4.1.2. The second is a ‘constant mean curvature’ interface, where the distance of the pivotal
surface fiom the TPMS may vary over the mesophase, but where its mean curvature H is
assumed to be uniform.
ParaNeZ interface model. Using the parallel interface model, we can obtain an
expression for ll in terms of 5”. The derivation for this is contained elsewhere”, and it
yields the following result.
We use the parameters for v, v, and A, given earlier, and a value of HO=-0.0167 f
0.0006A-*measured in the experiments involving the addition of akene, described in the
previous section. The curve fit can thus provide estimes of the two parameters K and KG.
This has been carried out for the osmotic stress data fiom the el," phase. The data and
curve fit are shown in Figure 8. The curve fit shown yielded values of ~=1.4+-0.4~1 0-19J,
and ~ = -.W0.3~10-'~
l J.
-h
500000
I
\ *
XxXIo.
ma-
t . , I 1
m 11al 1200 1m 1400
10.- 10.7 10.75 10.8 10.85 '0.9
Figure 8. Left: Osmotic pressure H a s a function of the distance 4;( of a @arallel)pivotal

surface fiom the underlying minimal surface, with curve fit. Right: Osmoticpressure L!as
a function of the volume of water per lipid vw , and curve fit for a pivotal suflace oj
constant mean curvature. Both data sets calculated from measurements taken on
osmotically stressed 2 M D L P C at 42°C.
Constant mean curvature model. Assuming the constant mean curvature model, we
can generate an equation for the energy per lipid, E, as a h c t i o n of the volume of water
per lipid, v,. Again, the derivation of this is contained elsewhere". The expression
generated is as follows.
-H, -
i=O
J
Values for the coefficients 4 and Oi for the different Q,, phases are given elsewhere 14. This
equation can be differentiated with respect to vw to give an expression for the osmotic
pressure n. We can then use this to fit data of n against vw to generate K and Q. In
practice, the expression for l2 is too cumbersome to copy down and then input into a
separate data analysis package, so the expression is generated in a software package such as
Mathematica, which can use this output directly as a curve fit. We used this to fit the same
data as in the previous section, with the same values for v, v, and A,, and still using a value
of HO = - 0 . 0 1 6 7 ~ . 0 0 0 6 ~The
' . data and curve fit are shown in Figure 8. This yielded
values of ~=3flxlO-~' J. - ~ ~
J, and ~ ~ = - % 2 x l O
We can compare the values we have obtained in this way with bending moduli
measured on DLPC bilayer vesicles'5316, which provide a value of 0 . 9 1 ~ 1 0 "J ~for bilayer
bending, or 4.6~10'~' J for the curvature elastic modulus of the monolayer (which we can
equate to K).Our estimate for K is three times as great as this if we calculate it assuming a
parallel pivotal surface, but smaller by one-third if we assume a pivotal surface of constant
mean curvature.
Our results therefore indicate that is is possible to measure approximate values for
curvature elastic parameters using the techniques described, and that these agree with
values obtained using other techniques to within an order of magnitude. However, our
results clearly suggest that we should introduce a note of caution. Our interpretation of the
osmotic stress data is heavily model-dependent; the estimated value of K varies by a factor
of 5, depending on whether we assume a parallel pivotal surface or one of constant mean
curvature.
Acknowledgements. We wish to acknowledge funding fiom the EPSRC.
References
1 R. H. Templer, B. J. Khoo, and J. M.Seddon, Langmuir, 1998,14,7427.

2 W. Helfiich, 2. Naturforsch, 1973,28c, 693.
3 M. M. Kozlov, S. Leikin, and R. P. Rand, Biophysical Journal, 1994,67, 1603.
4 R. H. Templer, Langmuir, 1995,11,334.
5 R. P. Rand, N. L. Fuller, S. M. Gruner, and V. A. Parsegian, Biochemistry, 1990,27,
76.
6 G . L. Kirk, S. M. Gruner, and D. L. Stein, Biochemistry, 1984,23, 1093.
7 H. Vacklin, B. J. Khoo, K. H. Madan, J. M.Seddon, and R. H. Templer, Langmuir,
2000,16,4741.
8 H. Chung and M. CafEey, Nature, 1994,368,224.
9 R. H. Templer, J. M. Seddon, N. A. Warrender, A. Syrykh, 2. Huang, R. Winter, and J.
Erbes, Journal ofphysical Chemistry B, 1998,102,725 1.
10 A. Squires, 'PhD Thesis', Imperial College, 2001.
11 R. P.Parsegian, R. P. Rand, N. L. Fuller, and D. C. Rau, Methods in Enzymology,
1986,127,400.
12 http://aqueous.labs.brocku .ca/data/peg6000
13 S. C. Roberts,PhD thesis, Imperial College, London, 1998.
14 R. H. Templer, J. M. Seddon, P. Duesing, R. Winter, and J. Erbes, Journal of Physical
Chemistry: B, 1998,102,7262.
15 L. Fernandez-Puente, I. Bivas, M.D. Mitov, and P. Meleard, Europhysics Letters,
1994,28, 181.
16 P . Meleard, C. Gerbeaud, P. Bardusco, N. Jeandaine, M.D. Mitov, and L. Fernandez-
Puente, Biochimie, 1998,80,401.
HEMOLYTIC AND ANTIBACTERIAL ACTIVITIES OF LK PEPTIDES OF
VARIOUS TOPOLOGIES : A MONOLAYER AND PM-IRRAS APPROACH.
S. Castano'; B. Desbat'; H. Wr6blewski' and J. Dufourcq3
LPCM-CRCM, U M R CNRS 5803, University Bordeaux I, 33405 Talence, France

U M R CNRS 6026, University Rennes I, 35042 Rennes, France
CRPP-CNRS, 33600 Pessac, France
1 INTRODUCTION
Most living species secrete a wide range of cytotoxic peptides which play a critical role
against external agression either with specific antimicrobial activity or aspecific
cytotoxicity. Most of these natural peptides enhance the permeability of the biological
membranes generally by a direct perturbation of the lipid Intensive studies have
demonstrated that i) peptides generally display a positive net charge and a high
amphipathic character which allow them to adopt mainly amphipathic a-helical or P-sheet
secondary structures in the membrane ii) the enhancement of these
parameters increases their binding ability and hence their membrane activity; iii) binding is
only the first step: for the same amount of peptide bound, very different activities are
obsered, i.e. an efficiency for increasing permeability can be defined. Therefore numerous
rationalized amphipathic peptides have been designed to mimic these natural active
compounds. The very minimal requirement to get an amphipath being to associate properly
Peptide Putative Sequence

amphipathicity 1 5 10 15
i.a. LK,, a helix K L L K L L L K L L L K L W K
Alterned (KL,)K p sheet Dns-K L K L K L K L K L K L K L KCONH2
Regular LK,, 3,,, helix K L L K L L K L L K L L K L W R

Scrambled LK,, no L K L L L L K L L K L K L W K
Figure 1 :Representation of the peptide, from top to bottom and left to right:
Schiffer-Edmundson 's helical wheels of i.a.LK,, and scrambled LK,,; vertical projection of
the putative 3,,, helix of regular LK,,; P-sheet representations of alterned (KL),K and
scrambled LK,,. White circles: L, leucine residue; grey circles: K, lysine residue.
two residues of opposite properties, this led to the design of several families of peptides out
of which those composed of only leucine (L), and lysine (K), representative of apolar
residues or polar and charged ones respectively. Such LK peptides proved to be efficient,
versatile and relevant to the problem of getting new active compounds which further
improve our understanding of their mechanism^^-'^. Ideally secondary amphipathic
structures were obtained upon folding such peptides only playing with the composition, i.e.
the L K molar ratio, and the charged residue periodicity, either 2 , 3 or 3.6'-17.
Here we compare the role of the amphipathic topology of a series of LK peptides on their
hemolytic and antibacterial activities. Four peptides of 15 or 16 residues were then
designed with nearly identical composition (WK == 2 molar ratio) to generate an ideally
amphipathic a-helix (i.a.LKI5),a 3,, helix (regular LK,,) and a non amphipathic peptide
(scrambled LK,,); the ideally P-sheeted analog (alternated (KL),K) differs since it requires
a 1:l L:K composition (tablel, figurel). We choosed Spiroplasma rneliflerum mollicutes as
bacterial targets because they proved to be sensitive to many membrane active natural
peptides and allowed monitoring of toxin activity at different levelsI8. To adress the
problem of the mechanism of interaction and perturbation of lipids by these peptides, they
were allowed to interact with lipid monolayers whose compositions mimic biological
environments. The peptides secondary structures and orientation, bound to the lipids were
determined in situ using the Polarization Modulation Infra-Red Reflection Absorption
Spectroscopy (PM-IRRAS) t e ~ h n i q u e ' ~ ' ~ ~ .
2.1 Hemolytic and antibacterial activities of the peptides
The hemolytic activity of the peptides was measured on human erythrocytes as already
described24.The dose-response curves show very similar sigmoi'dal shapes, total lysis
occuring for concentrations <2OpM. The lethal doses for 50% of lysis, LDso values, listed
in table 2, vary in a ten fold range with the following hierarchy:
i.a.LK15 < alternated (KLhK < regular LK16 < scrambled LK,,.
The antibacterial activity of the peptides, characterized by minimal inhibition
concentrations (MICs) on cultured S. melliferurn BC3 mo11icutes2s, were determined by
growing the bacteria in the presence of peptide. All assays were performed in triplicate and
the results are summarized in table 2. All the peptides are active with MICsllOOpM. The
hierarchy of MICs differs from that observed for hemolysis:
i.a.LK15 = regular LK,, c< alternated (KL)7K = scrambled LK,,.
The concentration range of bacterial activity is 2 to 66 fold higher than for hemolysis. This
could be due to the different lipid composition of the two cell membranes and to the
growth medium which leads to peptide loss by aspecific interactions.
2.2 Peptide insertion into lipid monolayers
Since both biological activities vary differently according to the peptide topology, the
interaction of the peptides with model membranes such as lipid monolayers at the aidwater
interface of a Langmuir trough was investigated. Monolayers of two different compositions
were formed to mimic the two different biological systems: a zwiterionic monolayer of
DMPC to mimic the erythrocyte membrane, since it was already found that PC lipids can
-1
account for a parallel variation of cytotoxic lytic and a monolayer of natural
Peptide Biological activity Li id affinit (lo6M-')

LD,,(pM) M E (PM)
i.a. LK,, 095 6,25
Alternated 195 100
(=l)K
Regular LK,, 3 90 6,25
Scrambled 5 ,o 100
LK,,
Table 2: Hemolytic and antibacterial activities of the peptides on erythrocytes (LD,,) and
on Spiroplasma melliferum BC3 (MIC)respectively, and peptide afinities for lipid
monolayers (DMPCand Spiroplasma lipids (Sm lip)). n.d.: not determined.
Figure 2: Relative film surface increases due to peptide insertion at l7=30mN/m in
-.-
preformed DMPC (left) and Sm lipids (right) monolayers on increasing pepdide bulk
concentration in the subphase:( 20mM Tris, I30mM NaCl, HCI, pH=7.5, T=25"C :
i.a.LK,,, -+- alternated (KL)7K, -V-regular LK,, , -A-scrambled LK15,
lipids extracted from the spiroplasma membrane (Sm lipids). The ability of the peptides to
insert into the compressed lipid monolayers were determined by measuring the relative
film surface increase (AS/S,) after peptide injection into the subphase. Data in figure 2
shows the different abilities of the peptides to insert into the lipid monolayers according to
the lipid environment and the peptide topology. The hierarchy of insertion into the DMPC
monolayer follows that of hemolytic activity. In the case of the Sm lipids, the hierarchy
differs: i.a.LK,, > regular LK,, = scrambled LKl5.
The most striking difference is for (KL),K which induces a very large surface increase on
DMPC while there is no change on the Sm lipids. But it is important to notice that if the
lack of surface increase observed for this peptide characterizes a lack of insertion, it does
not rule out the potentiality of peptide adsorption underneath the Sm lipid monolayer.
2.3 PM-IRRAS structural study of the peptides in interaction with lipid monolayers
PM-IRRAS spectroscopy is a powerful technique to determine the secondary structure and

the orientation of the peptides in situ in the lipid envir~nment'~-*~.
The 'subtracted spectra'
presented in figure 3 result from the subtraction of the pure lipid spectrum from that of the
mixed peptide/lipid one and then allow one to extract the characteristic contributions of the
peptide bound. Clearly the peptide structure and orientation depends both on the lipid
environment and on the peptide topology. In the case of i.a.LK,, the main amide I
absorption band, centered around 1655 cm-', whatever the lipid environment, is attributed
to an a-helix as expected from the peptide design. Simulation of orientations
proves an orientation of the a-helix that is flat at the interface. These conclusions are in
1I Proteins, Lipids and Their Interactions 195
0.4 4
0.3 1
i.a. LK,,
-0.2 ~
-o.1~~"~"""""'~"''""'''"'""~ J
1800 1700 1600 1500 .,I400 1300 1200 1100 1800 1700 1600 1500 -1400 1300 1200 1100
cm crn
0.25
1800 1700 1600 1500 .,I400 1300 1200 1100

cm
Figure 3: PM-IRRAS subtracted spectra of the peptides in interaction with a lipid

monolayer composed of either DMPC (-) or Sm (---)lipids.
agreement with previous s t ~ d i e s ~The ~ ~ * ~I .mode of the alternated (KL)7K is split

~ * amide
into an intense band around 1622 cm-' and a weaker one around 1693 cm", both
characteristic of antiparallel P - ~ h e e t s ~oriented
~ - ' ~ flat to the interface whatever the lipid
~ ~ ~ ~ . spectroscopy is only sensitive to molecules located
~ ' ~ ~ PMIRRAS
e n v i r ~ n r n e n t ' ~ Since
at the interface, the spectrum obtained on Sm lipid environment proves the interaction of
alternated (KL)7K with the monolayer even if no lateral surface increase was observed.
The different spectra obtained for the scrambled LK,, indicate a versatile structure and
orientation dependent upon the lipid environment. Inserted into the DMPC monolayer, the
peptide folds into antiparallel P-sheets (amide 1 bands around 1625 cm-' and 1695 c ~ - ' ) ~ ' - ~ O
flat oriented at the i n t e r f a ~ e ' ~,,while
~ ' , ~ inserted
~ ~ ~ ~ into the Sm lipid environment, if folds
into an a-helix (amide 1 band around 1655 with a tilted orientation (=30°)21*23,24,
both structures displaying no amphipathicity. Similarly for regular LK16the spectra vary
dependent upon the lipid environment. On the DMPC monolayer, the amide 1 band of the
peptide is mainly characteristic of a tilted (=40°)a-helix but not of the expected
amphipathic 3,, On Sm lipids the broad amide 1 band may contain some 310 helix
contrib~tion~ but ~ -the
~ ~structure
, is not well defined.
2.4 Peptide affmities for lipid monolayers
The determination of the peptide secondary structure and orientation allows an estimate of
the molecular area for each peptide in lipid environment and then to calculate the affinity
constants for lipid monolayers using a previously proposed equation and AS& data
obtained in 0 2.2l3,I7(table 2). No affinity can be estimated for the alternated (KL),K on
Sm lipids, since it does not insert into the monolayer. In similar conditions, the affinities
for Sm lipids are 20 to 1500 fold higher than those measured for DMPC. This well known
effect differentiating interactions with erythrocytes and bacteria is first related to the strong
electrostatic interaction due to the presence of negative charge in Sm lipids. It is also
noteworthy that if a strong affinity for lipids is required to get the aspecific hemolytic and
antibacterial activities, this is not the sole parameter of importance since there is no simple
correlation between lipid affinities and biological activities.
3 CONCLUSION
We show that all the rationally designed LK peptides have cytotoxic activities which
strongly differ according to the peptide topology without correlation between the
hierarchies of hemolytic and antibacterial activities. As anticipated the estimated affinities
for the negatively charged bacterial lipids are significantly higher than for DMPC, then it is
not peptides affinity but peptides structures when bound which govern activities on Sm.
Indeed such structures, orientations and the affinities vary according to the peptide
topology and are modulated by the lipid environment which implies the need to define
such structures in the most relevant conditions compared to those of the target cells.
Nevertheless, new conclusions clearly appear:
i) The ideally amphipathic a-helix of i.a.LK,, is the most active structure both in hemolysis
and in antibacterial assays, the membrane destabilization is then probably due to
accumulation of flat oriented a helices on the outer leaflet whatever the lipid environment.
ii) The ideally amphipathic &sheet of alternated (KL),K is as efficient on erythrocytes as
i.a.LK15, due to its flat insertion mainly governed by hydrophobic interactions3*.However
it is demontrated here that the same peptide is almost inactive on bacteria where it only
adsorbs on the outer leaflet of the membrane due to the main contribution of the
electrostatic effect. This striking difference is at odds with what might be expected since
numerous natural peptides essentially structured with a 6-hairpin are strongly
antiba~terial~”~~. We propose that it is the ability to generate large antiparallel
intermolecular aggregates on the membrane which is unfavorable for antibacterial activity
due to the inability of these large aggregates to penetrate into the membrane.
iii) Both the scrambled LK,, and regular LK16display very versatile structures, varying
according to the lipids, they penetrate into the membranes and better in charged bacterial
lipids. Their very weak hemolytic activity could be related to their very low amphipathicity
(the hydrophibic moment pH= 0.04 and 0.15 for their a-helical and P-sheeted actual
structures respectively).
iiii) The regular LK,, was designed to fold into an ideally amphipathic 3,, helix but it does
not behave as expected. However some 3,, helical content did occur but only when bound
to bacterial lipids. This could be due to electrostatic interactions which force K residues to
be in register with charged polar head groups, it results in pH= 0.58 , a value higher than
that of the i.aLK,,. peptide in Ha (pH= 0.46).It then fits with the previously defined
criteria for being active and, indeed, it is strongly antibacterial. This is at odd with the
conclusions on the first designed (LRL), peptides, which were perhaps too but it
reinforces the more recent findings4.
Finally in generating rationally designed de novo active compounds, more structural data
are required to fully ensure the main conclusions that it is the intermolecular packing
which should be controlled to get highly antimicrobial p-sheeted peptides and better
characterization is needed to understand the 3 helix antibacterial mechanisms.
Aknowledgements
This work was supported in part by the GDR 790 from the CNRS.
References
1 I. Cornut, E.Thiaudikre and J.Dufourcq, in: The Amphiputhic Helix. Epand RM (ed),
CRC Press, Boca Raton, FL, 1993,pp 173.
2 G. Sabenval and R. Nagaraj, Biochim. Biophys. Actu, 1994,1197,109.
3 W.L.Maloy, U.P. Kari, Biopolymers, 1995,37,105.
4 P.Nicolas and A. Mor, Annu. Rev. Microbo1.,1995,49,277.
5 R.M. Kini andH.J. Evans, Intern. J. Pept. Prot. Res., 1989,34,277.
6 K.Matsuzaki, Biochim. Biophys. Acta, 1999,1462,1.
7 6 W.F.De Grado and J. D. Lear,J. Am. Chern. SOC.,1985,107,7684.
8 S.E.Blondelle and R. A. Houghten, Biochemistry, 1992,31,12688.
9 J. A.Reynaud, J. P. Grivet, D. Sy, and Y. Trudelle, Biochemistry, 1993,32,4997.
10 I. Cornut, K.Buttner, J. L. Dasseux and J. Dufourcq, FEBS Lett., 1994,349,29.
1 1 A.Brack and G.Spach, J. Am. Chem. SOC., 1981,103,6319.
12 S.E. Blondelle, J. M. Ostrech, R. A-Houghten and E. Ptrez-Paya, Biophys. J.,1995,
68,351.
13 S.Castano, B. Desbat, and J. Dufourcq, Biochim.Biophys.Actu, 2000,1463,65.
14 T. Kiyota, S.Lee and G. V. Sugihara, Biochemistry, 1996,35,13196.
15 A. Kitamura, T. Kiyota, M. Tomohiro, A. Umeda, S. Lee, T. Inoue, G. Sugihara,
Biophys. J., 1999,76,1457.
16 M. Dathe, T. Wieprecht, H. Nikolenko, L. Handel, M. L. Maloy, D. L. Mc Donald,

M. Beyermann and M. Bienert, FEBS Lett., 1997,403,208.
17 S. Castano, B. Desbat, M. Laguerre, and J. Dufourcq, Biochim. Biophys. Acta, 1999,
1416,176.
18 L. Beven and H. Wrbblewski, Res-Microbiol, 1997,148, 163.
19 D. Blaudez, T. Buffeteau, J.C. Cornut, B. Desbat, N. Escafre, M. Pezolet and J.M.
Turlet, Thin Solid Films ,1994,242,146.
20 D. Blaudez, J.M. Turlet, J. Dufourcq, D.Bard, T. Buffeteau and B . Desbat, J. Chem.
SOC.Faraday Trans., 1996,92,525.
21 I. Cornut, B. Desbat, J.M. Turlet and J . Dufourcq, Biophys. J., 1996,70,305.
22 M. Boncheva and H. Vogel, Biophys. J., 1997,73, 1056.
23 S . Castano, B. Desbat, I. Cornut, P. M616ard and J . Dufourcq, Lett. Pept. Sci., 1997,
4, 195.
24 S . Castano, I . Cornut, K. Biittner, J.L. Dasseux and J. Dufourcq, Biochim. Biophys.
Acta, 1999,1416, 161.
25 L. Bkven and H. Wrbblewski, Res. Microbiol., 1997,148, 163.
26 T. Miyazawa, in: PoZy-a-amino acids, Fasman GD (ed), vol 1, Marcel Dekker Inc
New York, 1967, p 69.
27 W.K. Surewicz, H.H. Mantsch and D. Chapman, Biochemistry, 1993,32,389.
28 E. Goormaghtigh, V. Cabiaux and J.M. Ruysschaert, in: Subcellular Biochemistry:
Physicochemical Methods in the Study of Biomembranes, Hilderson HJ, Ralston GB
(eds), vol23, Plenum Press, New York, 1994, pp 363.
29 S . Krimm and W.C. Reisdorf, Faraday Discuss., 1994,99, 181.
30 S.M mick., G.V. Martinez, W.R. Fiori, A.P. Todd and G.L. Milhauser, Nature, 1992,
359,653.
3 1 W.R. Fiori, S.M. Miick and G.L. Milhauser, Biochemistry, 1993,32, 1957.
32 G.V. Martinez and G.L. Milhauser, J.Struct.BioZ., 1995,114,23.
33 L.K.Tamm and S.A. Tatuiian, Quarterly Rev. Biophys., 1997,30, 365.
34 S . Castano, B. Desbat and J. Dufourcq, Biochim. Biophys. Acta, 2000,1463,65.
35 L.H Kondejewski., S.W. Farmer, D.S. Wishart, C.M. Kay, R.E. Hancock, R.S Hodges.
JBiol Chem,1996,271,25261,
36 L.H Kondejews
37 ki., S.W. Farmer, D.S. Wishart, R.E. Hancock, R.S Hodges, Znt .J .Pept Protein Res,
1996,47,460,
38 S . Thennarasu, R. Nagaraj, Biochem. Biophys. Res. Cummun., 1999,254,281.
39 2. Oren, J . Hong, Y. Shai, Eur. J. Biochem., 1999,259,360.
40 T. Iwata, S. Lee, 0. Oishi, H. Aoyagi, M. Ohno, K. Anzai, Y.Kirino, G. Sugihara,
J.Biol.Chem., 1994,269,4928.
41 K. H . Mayo, J. Haseman, H. C. Young and J. W. Mayo, Biochem J., 2000,349,717.
A NOVEL APPROACH FOR PROBING PROTEIN-LIPID INTERACTIONS OF MscL,
A MEMBRANE-TENSION-GATED CHANNEL
P. C. Moe and P.Blount
Department of Physiology, University of Texas-Southwestem Medical Center

Dallas, TX 75390-9040
U.S.A.
1 ABSTRACT
The bacterial mechanosensitive channel MscL, a homo-pentameric protein that resides in
the cytoplasmic membrane, protects the cell from acute hypo-osmotic stress by opening in
response to increased membrane tension. Early work demonstratedthat only the structural
protein and a lipid membrane are required for the activity of this channel. Subsequent
characterization, much of it directed by the crystallization of the M.tuberculosis protein,
focused on a genetic dissection of internal contributionsto the channel activity and yielded
an intramoleculargating model based on hydrophobic interactions. However, because the
membrane matrix provides the medium through which the ‘agonist’, tension, is delivered
to the channel we have inaugurated work to investigate the role of the lipid environment
on the activation of this channel. Here we show that the MscL channel activity can be
reconstituted in lipid membranes of discrete composition, and that the gating-tension
threshold can be determined for these systems. This technique now enables the dissection
of the various lipid effects such as lateral pressure, bilayer thickness, and head-group
chemistry, on the fbnction of a membrane protein. Here we describe the methodology
required to address this question and present preliminary data demonstrating a specific
lipid effect on the finction of this mechanosensitive channel.
2 TNTRODUCTION
The bacterial mechanosensitive(MS) channel, MscL, was originally isolated fkom E. coli
membranes. Subsequent investigation found orthologues of MscL scattered across a wide
evolutionary range of the eubacteria, and even into the Archaea.’. Numerous lines of
evidence suggest that this channel serves as a bacterial “emergency valve” protecting the
organism fiom acute hypotonic stress3-’ The channel itself is exquisitely simple:
consisting solely of a homo-pentamer comprised of two-transmembrane subunits.6In the
E. coli channel these subunits are only 136 amino acids long providing a facile model for a
dissection of the mechanosensation phenomenon. In fact, at the time of this work, MscL
remains the only mechanosensitive channel to be cloned, reconstituted and available for
molecular characterization. Early work toward such characterization concentrated on

intramolecular manipulation through random mutagenesis. Ablation of the MscL channel
evoked no obvious phenotype because of redundant functions within the cell; therefore
channel-mutants were screened for a gain-of -hnction (GOF) phenotype resulting fiom
channels that gated inappropriately. These experiments quickly yielded a number of
mutants, nearly all of which clustered to the proximal portion of the first transmembrane
element, and comprised changes to charged and/or more hydrophilic r e s i d ~ e sIn . ~ fact,
when these residues are projected on a helical wheel model, the most severe are found to
line one face.g
After a structure was solved for an orthologue fiom the bacterium, Mycobacterium
tuberculosis these genetic results, and other evidence from biochemical experiments,
- ' ~ GOF mutations were projected
coalesced to yield a gating model for the ~ h a n n e l . ~When
onto the M.tuberculosis structure they were noted to line the channel lumen along the first
transmembrane element. In fact, among the most severe mutant recovered, valine 21 in E.
culi, mapped to a residue resting at the closest stricture of the lumen. It was hypothesized
that this valine, and several proximal residues, constituted a key element in the MscL
gating mechanism: a hydrophobic lock. If this lock was disrupted by the substitution of a
charged or more hydrophilic residue, the consistent result was the creation of a GOF
mutant. Although this part of the puzzle appears incontrovertible, there remains the
question of the ultimate conformational state of the open channel, and equally prickly;
how does the membrane transfer the tension to the channel? For this answer we have
turned to the membrane itself
For many years the lipid bilayer that delimits a cell fiom its environ was referred to
as a ''sea of lipids" in which the "important players" i.e. proteins, were suspended and
carried out their various vital tasks. This invokes an image of the bilayer as a vaguely
homogeneous structure that, at best, serves to define the boundaries of the cell, and
provides a milieu for hydrophobic proteins whose wont it is to reside in such an
environment. We are now beginning to find evidence that this outlook was entirely too
facile. The sheer complexity of the lipid bilayer, even as found in bacteria belies this
supposed simplicity. Additionally, it is also known that most organisms closely regulate
the constituency, and architecture of their lipid membranes.
Why would organisms, at great energetic cost, manipulate their membrane?
Perhaps the answer lies in the complex diversity of lipid species available to most
organisms. Biophysical studies have clearly shown the impact on bilayer mechanics of a
head group substitution of glycerol for ethanolamine. The former readily adopts a bilayer
conformation under physiological conditions, while the latter adopts an inverse hexagonal
(non-bilayer) conformation. Surprisingly, phosphatidylethanolaine, at -6O%, constitutes
the single most prevalent lipid in E. coli membranes. What is the role of such an apparent
non-sequitur? Also,how is the tension or stress within a lipid membrane transduced to a
channel opening as in the case of MscL? What effects do the lipid headgroups and
pressures internal to the membrane have on the activity of such channels?
Here, using the bacterial mechanosensitive channel - MscL- as a reporter of
membrane energetics we explore a possible modulatory role for regulation of membrane
lipid composition by means of a refined technical approach. This technique includes
reconstitution of the channel into defined lipid systems coupled with the biophysical
analysis of the membrane under tension. Not only is this approach ideally suited for
probing the lipid milieu itself but also for systematically dissecting the critical aspects of
the protein-lipid interface that govern gating of MscL. At present, a handhl of seemingly
disparate theories compete to define the nature of these interactions: headgroup chemistry,
I1 Proteins, Lipids and Their Interactions 20 I
bilayer lateral pressure, and hydrophobic mismatch induced by bilayer thinning have all
been invoked. As we are now poised to winnow these theories, we present preliminary
data showing a marked effect of different lipid compositions on the activity of the MS
channel MscL
3.1 Molecular Biology and Protein Purification
To facilitate purification of the E. coli MscL protein the structural gene was modified by
PCR to encode a string of six histidine residues appended to the carboxyl-terminus
(MscLK). This construct was expressed from the moderate copy-number vector pBlOb,
under the control of an inducible promoter, in the mscl-null E. coii strain PB104.8
Previous work has demonstrated that the moderate expression level fiom this vector does
not perturb the ultimate activity of the protein or induce the formation of inclusion bodies,
both indicative of protein misfolding.
Host cells were grown at 37°C in Luria-Bertani medium containing ampicillin
(100 &ml) to maintain the plasmid and IPTG (1 mM) to induce expression of the
channel protein (Figure 1). Cells were harvested while in logarithmic growth (-0.6
ODm), and washed with KMg buffer (50 m M Kpi pH 7.2, 5 mM MgSO4, and 1 mM
DTT). Cells, resuspended to 20% w/v in KMg buffer plus 1 mM PMSF, 1 mg/ml
lysozyme and 0.5 pg/ml DnaseI, were incubated at room temperature with gentle mixing.
The resultant lysate was then subjected to 2 passages through a French pressure cell at
16,000 psi to complete the lysis. This and most further steps were pe~ormedat 4°C.
The crude lysate was cleared by centrifbgation for 10 min at 6,000 x g. The total
cell membrane fiaction was collected by subsequent centrikgation at 200,000 x g for 30
min. This membrane fiaction was solubilized on ice with the detergent octyl-fb
glucopyranoside (OG) to facilitate the extraction of MscL&. OG was selected because it
is a fairly mild detergent, and, at least for MscL, allows for the recovery of a fbnctional
protein (Figure 1). It should be noted however, that some membrane proteins are not
compatible with OG, fairing much better with detergents such as n-dodecyl-P-D-maltoside
(DDM); the only requirement is that the detergent be easily removed by dialysis (see
below). The extraction buffer consisted of 300 mM NaCl, 50 mM KPi pH 8.0, 20 mM
imidazole pH 8.0, plus 3% OG. The membrane suspension was thoroughly, but gently,
homogenized in a glass dounse, and cleared by centrifirgation for 20 min at 230,000 x g.
MscL& was purified fiom the membrane homogenate by metal-chelation
chromatography with Ni-NTA resin (Qiagen) using the batch column method. Briefly, 500
p1 of the washed resin, equilibrated with extraction buffer plus 1% OG, was incubated, at
room temperature, for 30 min with the homogenate. The suspension was then transferred
to a gravity column and washed first with extraction buffer plus 1% OG and finally with
elution buffer (50 mM NaPi pH 7.2,300mM NaCl, 1 mM PMSF and 1% OG) and a step-
gradient of histidine. Elution of the M s c L h was effected at 200 mM histidine and found
to be essentially homogeneous by SDS-PAGE. However, the specific activity of the
M s c L b protein was not of concern because earlier works had demonstrated
unequivocally that MscL, alone, is necessary and sufficient for the observed channel
activity.6 Total protein in the eluted fiaction was quantified by modified Bradford assay
(Pierce).
3.2 Reconstitution
Various lipids, and lipid mixtures, were assayed for their ability to support the
reconstitution of the MscL& as evidenced by patch-clamp. In this inaugural study, to
establish a baseline for future experimentation, we chose two common, biologically
relevant phospholipids: dioleoyl phosphatidylcholine (DOPC), and dioleoyl
phosphatidylethanolamine (DOPE). These lipids are found in membranes, have the
biologically favored 18-carbon unsaturated acyl chains, and are well-characterized
biochemically. DOPE is found as a predominant lipid in the E. coli plasma membrane;
ironic given its proclivity for assuming a non-bilayer contipration in solution. DOPC,
although not represented in the E. coli envelope, was selected for these experiments
because of its relatively neutral nature; it contributes little to the lateral pressure within the
membrane, or charge to the sucface of the bilayer. A stepwise increase in the ratio of
DOPC:DOPE was pdormed to determine the effect of DOPE on MscW activity. The
--
maximum ratio achieved was 1:3, ostensibly due to disruption of the bilayer configuration
at DOPE concentrationsabove this ratio.
Detergent
e
@Q
/
0
O0
O
Vesicles
Lipid
qjpP t+Mg”
Detergent $r
Figure 1 Functional reconstitution of hem-His tagged proteins into a &flned
phospholipidmembrane.
Lipid vesicles suitable for reconstitution (Figure 1) were prepared as described

previously with rnodificati~ns.'~'~ Lipids, dissolved in chloroform and stored under argon
at -2OoC, were aliquoted to a 13 mm glass tube. When lipids were to be mixed, they were
mixed as a chloroform solution at this step. The chloroform was evaporated under a
stream of argon for 30 min at room temperature. In the course of evaporation, the tube was
rotated to yield a thin film of lipid that was deposited as the chloroform passed off. This
lipid film was resuspended to 20 mg/ml in buffer (10 m M TrisCl pH 7.2, 1 mM EDTA, 1
m M EGTA) by vortex until it presented as a milky suspension. Lipid vesicles were formed
by bath sonication to near-clarity. The time required to accomplish this varied from 10
min to greater than 45 min, but could be accelerated by addition of a few grains of OG.
Aliquots of the MscL& fiaction were combined with lipid vesicles at a 1500
protein-to-lipid mass ratio. Removal of detergent and reconstitution of MscL into the lipid
was effected by dialysis against 2x 1000 vols of buffer (100 mM NaC1,0.2 m M NaEDTA,
Tris C1 5 mM pH 7.2, NaN3 0.02%) containing a few microliters of Calbiosorb beads
(Calbiochem) to sequester the detergent. The resultant proteoliposomes were collected by
a 20 minute centrifugation at 30 psi in an Airfbge (Beckman-Coulter Instruments). The
pellet was resuspended to 1.3 mg/pl in 5% ethylene glycol, 10 m M MOPS pH 7.4,
separated to 5 PI aliquots on a plastic cover-slip, and desiccated overnight at 4°C.
Desiccated proteoliposomes were rehydrated for at least 2 h in Buffer A (150 mM
KCl, 0.1 m M EDTA, 10 plkl CaCl,, 5 mM HEPES pH 7.2) at a lipid concentration of 90
mg/ml. Unilamellar blisters, suitable for access by patch clamp electrode were induced in
the proteoliposomes by incubation in Buffer A plus 30 mM MgCl2 (Figure 1).
Although this technique has proven quite repeatable and trouble-free with the
lipids used to date, some difficulties would be anticipated as the range of lipid species
widens. Fundamental to this technique is the formation of accessible blisters for patch-
clamp. Some lipids, or mixtures of lipids, may prove disinclined to form these unilamellar
structures. Varying the protein:lipid ratios, or the Mg2' concentration in the buffer, are two
approaches that have been employed with success. Phase transition behavior may also
present a problem, especially as one canvasses lipids with transition temperatures well
above the 0 - 4°C at which these reconstitutions normally take place. While an obvious
approach might be to raise the temperature the welfare of the MscL protein must be kept
in mind. Finally, although the lipid may blister readily, formation of a gigaohm resistance
seal with the glass electrode (an imperative for the patch-clamp technique, described
below) may not be possible. This seal is thought to involve the interaction of the lipids
with the glass so such a problem might be addressed by selecting electrodes of different
glass composition, pretreatment of the electrode with acid, or other chemical
modifications. These potential difficulties should not adversely affect the efficacy of this
technique .
3.3 Biophysical Characterization
Excised, air-cleared, patches were examined at room temperature under symmetrical

conditions in Buffer A plus 30 mM MgC12. Recordings were performed at +20 mV
(electrode). Data were acquired at a sampling rate of 20 lrHz with a 5 kHz filtration using
an AxoPatch 2OOB amplifier in conjunction with Axoscope software (Axon Instruments).
A piezoelectric pressure transducer F l d Precision Instruments) was used to measure
the pressure response ofthe channels. '
Biophysicai Chemistry Membranes and Proteins
Figure 2 Luplace 's Law defines the tension (T) &eloped wiihin a bilayer as afunciion of
the radius of curvature (r), andpressure (p) across the membrane.
Because the MscL channel is thought to gate in response to tension in the bilayer,
rather than pressure across the membrane, it was necessary to find a way to derive the
actual tension of the membrane patch. As depicted in Figure 2, Laplace's law shows that
there is an intimate relationship between pressure and tension that is dictated by the
curvature of ~t surface. Therefore, for any given pressure indicated by the manometer,
there could be a very different curvature to the membrane patch depending on its nature
and its position in the electrode, thus generating differing tension loads in the patch.
Fortunately Laplace's law provides the solution: directly measure the patch radius and the
pressure and solve for tension!
To facilitate visualization of the membrane patch the tip of the electrode was fused
to an angle that placed it parallel to the focal plane while in the recording chamber. A 60x
DIC oil-immersion objective, coupled with a Nikon DXM CCD camera and 5x relay lens,
was used to capture images of the membrane patch (Figure 3). The patch radius was
determined fiom the captured images using a commercial computer assisted design
package IntelliCAD 2000 (CADopia).
Figure 3 Morphometric analysis of ihe membrane patch pennits the calculation of tension
by Laplace 's Law.The scale bar represenis 10 microns.
Boltmurnn data sets were constructed by increasing the pressure in a stepwise

fashion until all MscL channels present in the patch were in the open state, or saturated
(Figure 4). At this saturating pressure the open probability (PO) is unity. As the pressure
was held at each step, the average current was noted (an indicator of the number of
channels active at that pressure), and an image of the patch was captured for fiuther
analysis.
1200 PA
2s
0mm Hg -5 m m Hg -37 m m Eg
"1
0,
P,=O Po= 0.002 Po= 0.026
-50 mm Ffg -61 m m Hg -73m m Hg

Po= 0.14 Po= 0.5 Po= 0.92
5
Figure 4 Electrophysiological record of E.coli McL reconstituted in DOPC membranes

highlighting the increase in openprobability (PO)as afinction of pressure.
The two lipids used in this study (see above) were selected with an eye to some
current theories explaining MscL gating. For instance, one approach holds that membrane
thinning, due to lipid bilayer strain, and resultant hydrophobic mismatch between lipid and
protein is the salient stimulus.' If true, then changing the ratio of DOPE in the bilayer
might mimic this condition thus resulting in a lower gating threshold for the reporter -
MscL.Alternatively, manipulating possible headgroup interactionsmay provide the key. It
has been suggested that some DOPE might be required by MscL to satisfy hydrogen
bonds with extra-membranous loop regions thus stabilizing the channel in the bilayer, and
perhaps even facilitating the flow of membrane tension into the channel? Finally, by
virtue of its inherent geometry, addition of DOPE will promote a negative curvaturestrain
in the bilayer, thus increasing the lateral pressure within the membrane. This increased
energy within the bilayer is expected to affect the dynamics of lipid-protein interaction as
well as impinge on the function of the proteins themselves.
Figure 5 depicts two data sets that are each representative of the lipid conditions
presented. In parallel experiments (n=6 each lipid ratio) it was noted empirically, and
consistently, that at a DOPC: DOPE ratio of 3:1 (e) far less tension was required to gate
MscL than at a 1:3 ratio (m). Hence, higher complements of DOPE apparently result in
more global effects on the membrane that are manifested in the higher gating tension. One
interpretation of these data is that MscL responds to the decrease in lateral pressure in a
stressed membrane as approximated in the 3:l ratio. Alternatively there may be direct
interactions between DOPC and MscL that increase the sensitivity of the protein to
membrane tension or reduce the energy of the gating transition.
One should bear in mind that these experiments are yet in their infancy and fbture
experimentation is necessary to unravel these clues. However, the observation that
variations in lipid composition so notably affect mechanosensitive channel gating reflects
the importance of the membrane and its influence on membrane-protein interactions; even
suggesting the possibility that membrane composition, known to change during cell
growth and in response to different environmental conditions, could modulate the fbnction
ofthis and other proteins.
I.o
0.8
0.6
n?
0.4
0.2
0.0
2 4 6 8 I0 12 14
T (dyneskm)
Figure 5 Boltmann analysis of E.coli MscL reconstituted in DOPCLDOPE membranes at
two ratios; 3:1 (0) and 1:3 (m).
4 CONCLUSION
What may we take away from these preliminary experiments? Firstly, The bacterial
mechanosensitive channel MscL provides an exquisitely simple, yet robust, reporter for
studying the biophysical properties of lipid bilayers. Here we describe a system that
enables a rigorous, step-wise examination of the various lipid constituents of biological
membranes. Any molecule that will support the functional reconstitution of the MscL
channel, and subsequent visualization, can be studied with this technique. Further, as the
biophysical characterizationof lipid bilayer systems becomes more defined, the technique
can be reversed to effectively study the intramolecular gating of MscL and other
membrane-bound channels and transporter systems.
Secondly, these preliminary results show the dramatic effect wrought by merely
adjusting the ratio of two common lipids in a binary system. It is well known that
biological systems manipulate the lipid constituency of the membrane in response to
homeostatic demands. These early results also show that DOPE, a major lipid fraction of
bacterial membranes, though not a bilayer forming lipid, may serve a modulatory role in
the bacterial envelope.
Finally, this technique lends itself to elucidation of the contributions made to the
total energetic landscape of a biological membrane by lipid head-group moieties, acyl
chain length, and even the degree of acyl chain saturation. Which, or what combinations
of these influence the flow of tension-energy from lipid to protein? Put another way, what
do mechanosensitive channels like MscL actually sense? Future experiments are being
designed with these questions in mind.
Acknowledgements
This research was supported by grants &om Robert A. Welch Foundation Grant 1-1420,
U.S.Air Force Grant F49620-01-1-0503, and NIH Grants GM61028 and DK60818.
References
1. 0.P.Hamill and B.Martinac, Physiol. Rev. 81,685(2001).

2. A,Kloda and B. Martinac, Biqhys. J 80,229(2001).
3. I. R Booth and P.Louis,Curr. Opin. Micro. 2, 166 (1999).
4. P. Blount, S. I. Sukharev, P. C. Moe, S. K.Nagle and C . Kung, Biol. Cell 87, 1
(19%).
5. P. Blount and P. C . Moe, TrendsMicrobiol. 7,420(1 999).
6. S. I. Sukharev, P.Blount, B.Martinac, F. R.Blamer and C . Kung, Nature 368,
265 (1994).
7. X . Ou, P.Blount, R.J. Hoffman and C . Kung, Proc. Ndl. Acud Sci. USA. 95,
11471 (1998).
8. P. Blount, S. I. Sukharev, M.J. Schroeder, S.K. Nagle and C. Kung, froc. Nutl.
Acad Sci. U.S.A. 93,11652 (1996).
9. P.C. Moe, G. Levin and P.Blount, J Biol. Chem. 275,3 1121 (2000).
10. B. Martinac, Cell. PhysioZ. Biuchem. 11,61 (2001).
1 1. S. Sukharev, M.Betanzos, C.S. Chiang and H.R. Guy, Nature 409,720(2001).
12. G. Chang, R.H. Spencer, A. T. Lee, M.T. Barclay and D. C . Rees,Science 282,
2220 (1 998).
13. S. I. Sukharev, B. Martinac, P. Blount and C. Kung, Methods: A Companion to
Methods in Enzymology 6, 5 1 (1994).
14. P. C. Moe, P.Blount and C. Kung,Mol. Microbiol. 28,583 (1998).
15. P. Blount, S. I. Sukharev, P. C. Moe, B. Martinac and C . Kung, Methods in
Enzymology 294,458-482(199).
16. D. E . Elmore and D.A. Dougherty, Biophys. J. 81,1345 (2001).
FOLDING OF THE a-HELICAL MEMBRANE PROTEINS DsbB AND NhaA
Daniel E. Otzen
Department of Life Sciences, Aalborg University, Sohngaardsholmsvej 49, DK - 9000

Aalborg C
1 INTRODUCTION
The biophysical properties of integral membrane proteins are not as straightforward to

study as those of their soluble counterparts, due to the difficulties in obtaining
sufficient quantities for characterization, This is linked to their absolute need for a
membrane-mimicking environment to stabilize and solubilize the native structure. A
closer study of the mechanisms by which membrane proteins are inserted into the
membrane may shed more light on the unique structural and functional facets
underlying this class of proteins and allow us to manipulate conditions to improve
yields of native protein. Pioneering studies on the a-helical integral membrane protein
bacteriorhodopsin by Booth and co-workers 1-3 have demonstrated that in vitru folding
of membrane proteins can be carried out and followed spectroscopically.
Bacteriorhodopsin is a robust protein, resistant to 8M guanidinium chloride 4, but it
remains denatured in SDS when transferred from an acid-denatured state 4. The
protein is then refolded by mixing with an excess of lipid vesicles. Both folding yields
and the kinetics of folding are modulated by the properties of the membrane
environment, such as bending rigidity 5 and lateral pressure 6 .
To assess the generality of these observations, it is important to extend such studies to
other proteins. Here I describe initial folding studies on two integral membrane
proteins from the inner membrane of E. coli, namely the 176-residue disulfide bond
reducing protein B (DsbB) and the 388-residue sodium-hydrogen antiporter A (NhaA).
DsbB oxidizes the periplasmic protein DsbA 7, allowing transfer of oxidative potential
to other reduced proteins in the periplasm, while NhaA exploits the proton gradient
across the inner membrane to pump sodium ions out of the cell 8. Unlike
bacteriorhodopsin, the structures of these proteins are not known in atomic detail, but
DsbB is predicted to contain 4 transmembrane helices, while 2-D crystals of NhaA
have permitted electron cryo-microscopy analysis, revealing 12 transmembrane
helicesg. None of the proteins contains an exogenous chromophore, precluding
absorption spectroscopy, but their aromatic residues provide convenient structural
probes for fluorescence spectroscopy. Both proteins can be at least partially
reconstituted from the SDS-denatured state into vesicles and the insertion reaction
followed in real time. The kinetics are clearly modulated by environmental
parameters such as ionic strength and pH as well as vesicle size.
2 METHODS
DsbB and NhaA were expressed and purified as described 7310. The genes for both
proteins were cloned into expression vectors which inserted a His6 tail onto the C-
terminus of the protein, allowing a one-step purification on a Ni-NTA column
(Amersham-Pharmacia). Both proteins were eluted using a 30-300 mM imidazol
gradient and concentrated with Centriprep tubes (Millipore). The concentrated protein
solution was twice diluted with buffer (50 mM sodium phosphate pH 8.0, 300 mM
NaCl and 0.02% (w/v) dodecyl-maltoside) and concentrated again, in order to remove
imidazol which interferes with far-UV CD spectroscopy. Unless otherwise stated, all
experiments were carried out in 50 mM sodium phosphate pH 8.0 and 300 mM NaCl.
Emission spectra were recorded on a PTI QM-2000-7 fluorometer with 1 pM protein
and circular dichroism spectra on a Jasco 5-715 spectropolarimeter with 10 pM protein
in a 1 rnm cuvette. Activity measurements of DsbB were carried out with 3.5 pM
DsbA 1 1 and 30 pM decyl-ubiquinone in 50 mM sodium phosphate pH 6.0 and 300
mM NaCl at 25°C. Excitation was at 298 nm and emission at 350 nm.
Stopped-flow experiments were carried out on an SX-I 8 MV stopped-flow
spectrometer (Applied Photophysics) using an excitation wavelength of 280 nm and a
320-nm cut-off glass filter. 1 volume of protein in detergent (typically 5 mM SDS)
was mixed with 10 volumes of lipids to give a final concentration of 5 mg/ml lipid (ca.
8 mM), 0.45 rnM SDS and 1 pM protein.
All lipids were from Avanti Polar Lipids. Vesicles of 50-200 nm size were prepared
using an extruder (Northern Lipids) according to the manufacturer’s instructions. All
extrusions were carried out at least 10°C above the lipid’s T,. Unless otherwise stated,
all experiments were done with 50 nm vesicles.
3.1 Unfolding DsbB and NhaA in SDS and reconstitution in vesicles
Since membrane proteins are difficult to refold once denatured, the first issue to
address is whether the denatured protein can be reconstituted into vesicles.
Denaturation is best carried out in SDS, since low concentrations are required, which
do not significantly affect vesicle structure upon mixing. In addition, membrane
proteins generally retain significant a-helical structure in SDS, and this may crudely
mimic in vivo folding, in which a-helices appear to be formed prior to membrane
insertion, when released through the translocon 12.
DsbB and NhaA are purified in 0.4 mM dodecyl maltoside (DM). This detergent is
preferable to e.g. decyl maltoside and n-octyl-glucoside, which lead to substantial loss
of activity of DsbB (M. Bader and J. Bardwell, personal communication). DsbB and
NhaA are both easily refolded from 5 mM SDS when mixed with 10-fold molar excess
DM. Even boiling in SDS for 10 minutes followed by refolding into DM at room
temperature does not destroy DsbB activity (data not shown).
I $2 1 I I I
I
$
B
8 -
a
E . 6
"
2 4
U
0 . 5 m M SDS
2
300 320 340 360 380 400

0.0Sf'
0
" "
IW
" " '
ZOO
. " '
300
' ' . '
400
" "
JM)
'
Wavelcngih (nm) Time (I)
Figure 1 (A) Fluorescence spectra of NhaA. The spectrum in DOPC and SDS was
recorded after the protein was unfolded in 5 mM SDS and then diluted 10-
fold into 8 mM DOPC.(B) Time-profile of NhaA folding when mixed with
excess DOPC vesicles at 25 C. The inset depicts the first 0.5 seconds in
more detail.
There are clear structural differences between the SDS- and DM-state. When NhaA is
transferred from 0.4 mM DM to 5 mM SDS, the fluorescence spectrum blue-shifts
from 330 to 327 nm and increases in intensity (Figure lA), suggesting alterations in
tertiary structure'. CD spectra also indicate an increase in secondary structure (data
not shown). Both sets of data are consistent with increased solvation of the aromatic
residues by the apolar interior of SDS micelles, accompanied by increased a-helix
formation. SDS is known to induce helical structure to facilitate micellar solvation of
the protein backbone l4,
When the SDS-denatured NhaA is transferred to the lipid DOPC (giving 0.45 mM
SDS and 8 mM DOPC), the spectral maximum shifts back to 329.5 nm and the
intensity drops to slightly below that in DM, suggesting that at least partial
renaturation has taken place. The slight drop in intensity may reflect the
environmental differences between DM and DOPC. Fluorescence studies with DsbB
yield similar results. In addition, activity assays on DsbB (the oxidation of DsbA,
leading to a drop in Trp-fluorescence 11) show a complete loss of activity in SDS and
a 60% regain after mixing with vesicles under the above conditions. The 40% gap
may reflect a misfolded fraction, similar to the observations for diacyl glycerate kinase
15. It could also represent a fraction of the protein population with its active site (the
two periplasmic loops) oriented towards the interior of the vesicles, which would
isolate it from the water-soluble DsbA.
3.2 Following the folding kinetics of DsbB and NhaA
Given that some renaturation of DsbB and NhaA takes place under equilibrium
conditions, we would expect to be able to follow the renaturation kinetically. Indeed,
there is a substantial increase in the fluorescence signal when SDS-denatured DsbB
'The critical micelle concentration of SDS is 7 m M in water, but this falls to below 1 m M in 300 m M NaCl
due to electrostatic screening of the head groups 3.
II Proteins, Lipids and Their Interactions 21 1
and NhaA are mixed with excess vesicles in a stopped-flow apparatus (Figure IB).
For both proteins, three well-separated relaxation phases are observed over a period of
several hundred seconds, with rate constants kl around 2-20 s”, k2 around 0.1-2 s-’ and
k3 around 0.02-0.1 s-l. These signals represent some interaction between protein and
vesicles, since the control experiment (SDS mixed with vesicles) shows no significant
relaxation signal. Increasing the SDS concentration from 5 mM to 10 mM prior to
mixing with vesicles does not change the kinetics significantly; however, higher
concentrations of SDS (giving a final mole fraction of more than ca. 20% in the SDS-
lipid mixture) leads to vesicle disrupture and a loss of protein folding signal (data not
shown). Therefore we use 5 mM SDS as a standard denaturing condition. It does not
appear possible to denature the proteins further in SDS; heating the SDS-protein
solution to 90°C prior to carrying out the refolding experiment did not change the
refolding kinetics (data not shown).
If the proteins are not denatured, but retain their native structure in a buffer containing
DM prior to mixing with vesicles, only 1-2 phases are seen with much smaller
amplitudes (data not shown). This suggests that some structural rearrangement has to
take place for the protein to insert into the membrane even from the native state, but
the conformational changes are largest when the protein starts from a more disordered
state. We therefore initiate all folding reactions from the SDS-denatured state.
The existence of several phases in the refolding time profile indicates that the proteins
have to pass through several structural rearrangements to reach the final state. Multi-
state folding is not without precedence for a-helical membrane proteins.
Bacteriorhodopsin passes through three intermediate states en route to the native
retinal-bound state 1. Native-like a-helical content is only attained in the second
bacteriorhodopsin intermediate 3, whereas essentially all helical structure of DsbB and
NhaA appears to be gained within the deadtime of mixing (data not shown). Thus, for
the latter two proteins, the rate limiting folding steps seem to involve alterations in
tertiary interactions within the membrane (probably through helical docking steps),
rather than insertion of helices into the membrane.
3.3 Modulation of the folding kinetics
The kinetics of folding of both DsbB and NhaA can be modulated by various
parameters. Vesicle size, controlled by the pore size of the filters used in the extrusion
process, plays a substantial role. As the vesicle size increases from 50 nm to 200 nm,
the folding rates of DsbB decline substantially while those of NhaA increase (Figure 2
A). Solvent conditions also influence folding. Increasing the ionic strength slows
down DsbB’s refolding rates (both the intermediate and slow phases) but has little
effect on those of NhaA (Figure 2B). However, all rates decline uniformly with pH
(Figure 2C). In contrast, folding rates for bacteriorhodopsin peak at pH 6 when the
protein is refolded into mixed DMPCDHPC micelles 5 ; in this case, the recovery
yield also decreases at extreme pH, unlike DsbB and NhaA (judging from the
amplitudes of the refolding reaction).
Why does vesicle size affect the folding kinetics? Diminishing the vesicle size will
increase surface curvature, exposing chinks to the hydrocarbon region and increasing
the hydration of the interfacial region. This may affect folding reactions adversely or
not, depending on the properties of the membrane protein in question. For example,
those membrane proteins, which contain a significant number of polar groups in the
parts residing in the interfacial region, may undergo rearrangements associated with
folding more easily if the groups are well hydrated. Based on this interpretation, DsbB
requires extensive hydration whereas NhaA does not.
More clues may be obtained from the two proteins’ response to solvent changes,
namely increasing ionic strength and pH, The effect of increasing ionic strength is
complex. Obviously, there will be increased screening of charges, which will
attenuate electrostatic interactions on the protein surface as well as between the protein
and the vesicle. Screening phenomena will give rise to a linear dependence between
the logarithm of the folding rate and the square root of the ionic strength, and such
linearity appears to exist for DsbB (Figure 2B). In addition, small ions lead to a
certain amount of dehydration, because their own solvation by a large shell of water
molecules effectively lower the concentration of water molecules available for binding
to lipid molecules. The ions can also interact directly with the lipid headgroups, and
this in turn stabilizes more ordered phases 16. As a consequence, melting transition
temperatures are elevated, though this will not affect the order of the liquid phase
significantly.
The results from the ionic strength dependence suggest that DsbB relies on extensive
hydration andor electrostatic interactions internally andor with the vesicle to facilitate
folding, whereas NhaA does not. This is consistent with the data on vesicle size
dependence.
Altering the pH will not affect screening significantly. Protons mainly affect lipid
order in the gel phase like other ions, so that lowering the pH increases Tm 17. All our
experiments with DOPC and DLPC are well above T, (-18°C for DOPC and -1°C for
DLPC), so vesicle phase order is probably not affected much. Furthermore, only
negligible changes in lipid head group hydration are seen over the pH range 3-9 I*.
Therefore a direct pH effect on the membrane protein is more likely. The linear
decline of log k with pH suggests a simple protonation model, in which titratable
groups are repelled by the vesicle headgroups in the deprotonated form. The most
likely candidates for this scenario are Asp and Glu; if this is indeed the case, the
ionized side chains will be repelled by the negatively charged phosphate groups.
Ionization of Asp and Glu probably also underlies the pH-dependence of SDS
denaturation, in which unfolding kinetics are much more rapid at low pH (D.E.O.,
unpublished observations), where repulsion by the sulfate groups does not occur.
4 CONCLUSIONS
It appears possible to follow the renaturation of the a-helical proteins DsbB and NhaA
from the SDS-denatured state by conventional stopped-flow analysis. The kinetics
probably involve several consecutive structural transitions, which for both proteins are
sensitive to changes in the solvent and vesicle, but to different degrees. Thus, the
response of membrane proteins to manipulations in the folding environment is dictated
Figure 2 The folding kinetics of NhaA (empty and filled circles) and DsbB (empty
andfilled squares) in DLPC are modulated by (A) extrusion size, (B) ionic
strength and (C) pH. For clarity, only the rate constants of the two
slowest phases are shown and data are fitted to exponentialfunctions.
10
Rates
(s-')
0.1
5 10 15 20
[NaCI]" (mM")
0.1
0.0 1
by the properties of the individual protein. These properties may include the amino
acid composition in the interfacial region and the hydration and electrostatic properties
in this area.
Acknowledgements
I am most grateful to Drs. J. Bardwell (DsbB) and W. Kiihlbrandt and C. Ziegler

(NhaA) for providing expression plasmids and providing very helpful advice on
protein expression and purification. This work is supported by a grant from the
Danish Technical Science Research Council.
Abbreviations
DLPC, L-a-dilauroylphosphatidylcholine,DM, dodecyl maltoside, DOPC, L-a-

dioleoylphosphatidylcholine,SDS, sodium dodecyl sulfate.
References
1. P.J. Booth, Biochim. Biophys, Acta, 2000, 1460,4.

2. S.J. Allen, J.-M. Kim, H.G. Khorana, H. Lu and P.J. Booth, J. Mol. Biol., 2001,
308,423.
3. M.L. Riley, B.A. Wallance, S.L. Flitsch and P.J. Booth, Biochemistry, 1997, 36,
192.
4. K.-S. Huang, H. Bayley, M.-J. Liao, E. London and H.G. Khorana, J. Biol.
Chem., 1981,256,3802.
5. P.J. Booth et al., Biochemistry, 1997,36, 197.
6. A.R. Curran, R.H. Templer and P.J. Booth, Biochemistry, 1999,38,9328.
7. M. Bader, W. Muse, D.P. Ballou, C. Gassner and J.C.A. Bardwell, Cell, 1999,98,
217.
8. E. Padan and S. Schuldiner, J. Exp. Biol., 1994, 196,443.
9. K.A. Williams, Nature, 2000,403, 112.
10. Y. Gerchman, A. Rimon and E. Padan, J. Biol. Chem., 1999,274,24617.
11. M. Wunderlich and R. Glockshuber, Prot. Sci., 1993,2,717,
12. A.E. Johnson and M.A. van Waes, Ann. Rev. Cell Dev. Biol., 1999,15,799.
13. B. Jonsson, B. Lindman, K, Holmberg and B. Kronberg. Surfactants and
polymers in aqueous solutions, (Wiley & Sons, New York, 1998).
14. W.L. Mattice, J.M. Riser and D.S. Clark, Biochemkty, 1976, 15,4264.
15. J.K. Nagy, W.L. Lonzer and C.R. Sander, Biochemistry, 2001,40,897 1 .
16. G. Cevc and D. Marsh, J. Phys. Chem,, 1983,87,376.
17. G. Cevc, Biochemistry, 1987,26,6305.
18. J.M. Seddon, G. Cevc and D. Marsh, Biochemistry, 1983,22, 1280.
FhuA, AN ESCHERICHIA COLI IRON TRANSPORTER AND PHAGE RECEPTOR
P. Boulanger, L. Planqon, M. Bonhivers and L. Letellier
Institut de Biochimie et Biophysique MolCculaire et Cellulaire, UMR CNRS 8619

Universite Paris Sud, B5t 430,91 405 Orsay cedex, France
1 INTRODUCTION
Gram-negative bacteria are surrounded by two membranes, which codine the periplasm
and the peptidoglycan, a structure responsible for the cell shape and rigidity. The inner
(cytoplasmic) membrane is constituted of a phospholipid bilayer in which proteins are
embedded. Ions and solutes are actively transported through the inner membrane by means
of channels, pumps and transporters. ATP and the electrochemical gradient of protons
generated by the electron transfer chain are used as the driving force for these transports.
The outer membrane constitutes a permeability barrier which protects the cell against
noxious agents and allows exchanges of solutes with the external medium. Its lipid
composition is atypical since the inner and outer leaflets contain phospholipids and
lipopolysaccharides respectively. Non-selective porins are the most abundant class of
outer membrane proteins. These pore-forming proteins allow the passive diffusion of small
(< 600 Da) hydrophilic molecules. Selective porins, also permit the diffusion of small
hydrophilic solutes but they contain specific binding sites for the ligands. The high affinity
transporters (also called ligand-gated porins or To&-receptors) are involved in the uptake
of molecules that are present at very low concentration in the growth medium and which
are too large to diffuse through the porins?
Among the essential nutrients that are transported by these transporters is ferric iron.
Gram-negative bacteria seeking to colonize aerobic environments at physiological pH are
faced with the practical insolubility of ferric iron. To cope with this shortage they
synthesize and secrete iron chelators (siderophores)that bind ferric iron with high affinity
or use siderophores present in the growth medium. Iron-siderophores are transported
across the outer membrane by high affinity transporters, each of them recognizing a
specific siderophore. Outer membrane transport requires energy provided by the
electrochemical gradient of protons generated by the electron transfer chain in the
cytoplasmic membrane. This energy is transduced to the outer membrane transporters by a
protein complex (TonB-ExbB-ExbD)anchored in the cytoplasmic membrane? FhuA, is
one of the seven ferric iron outer membrane transporters that are expressed in Escherichia
coli. We present recent data obtained by our group on its in vitro functional and structural
properties. We discuss these properties in light of its three dimensional structure.
2 FhuA: FROM STRUCTURE TO IN VITRO FUNCTION
1.1 in vitro Functionality of FhuA
FhuA is a 78.9 kDa protein that transports iron coupled to the hydroxamate siderophore
ferrichrome (MM 688 Da). Besides its physiological function, FhuA is also the receptor
for different phages (T1,T5, UC-1, @80)and for the antibiotic albomycin and the toxin
colicin M.4 FhuA is the first outer membrane transporter which was purified in a functional
form and characterized in vitro.*Using a fluorescent DNA intercalator YO-PRO-1, we
demonstrated that the interaction of phage T5 with purified FhuA, solubilized in non ionic
detergent, resulted in the release of its double-stranded DNA (121 kbp) into the
surrounding medium within a few s e ~ n d sFurthermore,
.~ FhuA which shows no channel
activity when incorporated into planar lipid bilayers was converted into an open channel
upon phage binding.6 Ferrichrome prevented channel opening and DNA ejection, indicating
that it interacted with FhuA in vitro. From equilibrium dialysis experiments we calculated a
Kd of ferrichrome binding to FhuA of 2.1 +/- 0.5 nM.7
1.2 Three-Dimensional Structure of FhuA
Our view of how iron-siderophore transporters fhction has considerably improved since
the recent determination of the 3D structure (2.5A0 resolution) by X ray diffraction of
FhuA by two different group^.^?^ FhuA, as well as FepA,” another member of the iron
transporters family, share the same unexpected and unique structural organization. FhuA is
composed of a barrel domain consisting of 22 antiparallel g-strands (residues 161 to 7 14)
lodged in the membrane and of an amino-terminal globular domain (residues 1 to 160) that
folds inside the barrel and occludes it. This “plug” or “cork” domain, which spans most of
the interior of the fbbarrel, consists of a four-stranded 6-sheet and four short helices. It is
connected to the f3-barrel and to hydrophilic loops facing the external medium and involved
in ligand binding. Ferrichrome binds to the top of the cork domain and phage T5 to one of
the largest external loop. Few structural changes are observed within the cork domain upon
ferrichrome binding except for a short helix (residues 24-29) located in the periplasmic
pocket in the ligand-he conformation of FhuA which completely unwinds upon
ferrichrome binding.8.9
1.3 Role of the N-terminal “Cork” Domain of FhuA in Structure and Function
The complex and very unusual structure of FhuA raises the question of the connectivities
of the f3-barreland cork domains and of their respective role in the protein function. To
address these questions we expressed and purified a recombinant FhuA A02 1- 128 (FhuA
A) missing almost all the cork domain and compared its properties to those of wild type
FhuA (FhuA WT).’
1.1.1 The Cork Domain is not Essential for the Function of FhuA. Previous in vivo
studies had shown that FhuA A could bind phage T5 and transport ferrichrome indicating
that the mutated protein was targeted to the E. coli outer membrane and folded so as to
retain functionality." In vitro studies indicated that the purified protein was also
functional. However phage T5 binding and DNA release required concentrations of FhuA A
two orders of magnitude higher than those required for FhuA WT. Phage T5 binds to one
of the hydrophilic loop of FhuA facing the external medium. The connectivities between
the loop and the cork are therefore important for the phage receptor activity of FhuA.
Ferrichrome could also bind to FhuA A since its addition prior the phage prevented DNA
release. However, the affinity of the protein for ferrichrome was strongly decreased since
its binding to FhuA A was not detectable in equilibrium dialysis experiments even when
concentrations of ferrichrome up to 30 pM were used. These results are in accordance with
the fact that residues belonging to the binding site of ferrichrome are deleted in FhuA Ae8l9
1.1.2 The Cork Domain Contributesto the Stability of the Protein. Studies of the effects
of proteases, denaturants and temperature gave strong indications of the reduced stability
of FhuA A compared to FhuA WT. Indeed, the part of the cork remaining in FhuA A was
fully cleaved by trypsin within 5 min whereas FhuA WT remained protected from cleavage
for at least 1 h. Furthermore, and in contrast to FhuA WT, FhuA A was not protected from
proteolysis by ferrichrome. The two proteins also showed a different pattern of migration
on PAGE and of accessibility to monoclonal antibodies. FhuA WT remained folded up to
60°C whereas FhuA A was denaturated at all temperatures studied between 4' and 100°C.
Differential scanning calorimetry experiments corroborated these results. The pattern of
thermal denaturation of FhuA WT was indicative of the presence of two well-resolved
structural domains which corresponded to the unfolding of the cork and loops (TI = 65 "C)
and of the barrel (Tz = 75°C). Ferrichrome had a strong stabilizing effect on the loops and
cork since it shifted the first transition to 71.4"C. Removal of the cork destabilized the
protein since a unique transition at 61.6"C was observed even in the presence of
ferrichrome. The cork is connected to the p-barrel and to the hydrophilic loops by nine salt
bridges and about 60 hydrogen bonds. Given the data obtained on FhuA A it is likely that
they contribute significantly to the stability of the &barrel.
These results are corroborated by recent data showing that the N-terminal domains of
FhuA and FepA can be genetically exchanged or deleted without loss of the siderophores
and phage binding capacity of the Altogether they indicate that the cork and
the f3-barrel of high affinity transporters behave as autonomous domains. What is then the
function of the cork? It could enhance the binding capacity of the siderophores. In addition
it may constitute a physical barrier preserving the cell from entry of noxious compounds
which are excluded from porins.
3 INVOLVEMENT OF FhuA IN THE FIRST STEPS OF INFECTION OF E.COLl

BY PHAGE T5
Infection of Gram-negative bacteria by phage starts by recognition of the host and binding
to a specific outer membrane receptor. Almost all outer membrane proteins are potential
receptors for phages.14 In the case of tailed phage, this binding triggers conformational
changes that are transmitted along the tail to the capsid, allowing its opening and the release
of the viral genome, which is then transferred, base pair after base pair via the tail through
the host envelope. The tailed phage T5 has proved to be a well-suited model to study
phage-host interactions and DNA transfer since almost all these steps can be reconstituted
in vitro.
3.1 FhuA forms a Functional Complex with Phage T5 Receptor-Binding Protein pb5
Binding of phage T5 to E.co2i cells is mediated by specific interactions between FhuA and
pb5, a 67.8 kDa protein located upstream of the central straight tail fiber at the distal part
of the phage tail. A histidine-tagged form of pb5 was overproduced, purified and
characteri~ed.'~ The purified protein is monomeric and mostly organized as f3-sheets (51
%). pb5 hnctionality was attested in vivo by its ability to impair infection of E.coZi cells
by phage T5 and Q, 80 when added before the phages. pb5 also prevented growth of
bacteria on iron-ferrichromesuggesting that it alters either the binding of iron-ferrichrometo
FhuA or its transport.
pb5 was also functional in vitro since addition of equimolar concentration of pb5 to
purified FhuA prevented DNA release fiom phage T5.Direct interaction of pb5 with
FhuA was demonstrated by isolating a pbS/FhuA complex using size exclusion
chromatography. Its stoichiometry, 1 mol pb51 1 mol FhuA, was deduced from its
molecular mass determined by analytical ultracentrifugation and confirmed by SDS-PAGE
analysis. SDS-PAGE and differential scanning calorimetry experiments highlighted the
strong stability of the complex. i) the thermal denaturation of the complex occurred at 85
"C while pb5 and FhuA were denatured at 45OC and 74 "Crespectively. ii) the complex
was not dissociated by 2% SDS even when the temperature was raised up to 70 "C. The
strength of the association between pb5 and FhuA is reminiscent of the irreversible nature
of the phage adsorption step in the infectious process. Such stabilization is probably
correlated to conformational changes occurring in both proteins. The availability of a stable
and well defined complex opens the way to crystallographic studies and to the
investigations of conformational changes occurring into siderophore receptors and phage
receptor-binding proteins upon binding of a phage to its host.
3.2 Delivering Phage T5 Genome into Liposomes
In an attempt to decipher the mechanisms of phage DNA transport FhuA was

reconstituted into unilamellar liposomes containing the fluorescent DNA intercalator YO
PRO 1. Addition of phage T5 resulted in an increase of fluorescence consistent with the
binding of the phage to reconstituted FhuA and transport of part of the DNA into the
vesicles.16 The high resolution of cryo-electron microscopy allowed the visualization of the
phage-proteoliposome interactions and demonstrated unequivocally that the phage genome
was delivered into the liposome." Figure 1 (left) shows a typical cryo-electron micrograph.
T5 phages are bound to the proteoliposomes by the tip of their tail. The capsid of one
phage is filled with a dark gray striated material that corresponds to densely packed DNA.
A second capsid is partially filled with DNA. Another is empty as evidenced by its lower
electron density. The proteoliposome contains a dark gray material similar to that found in
the capsids, an indication that the DNA has been transferred into the vesicle. The
morphology of the liposome is not disturbed although the concentration of DNA within the
vesicles reached values as high as 130 mg/ml.
Figure 1 Phage T5 DNA Transfer into FhuA-containing Liposomes @om reference 18 and
19)
Furthermore, many DNA strands could be transferred from several phages and condensed
into a unique compact toroidal structure when the proteoliposomes contained the DNA-
condensing agent spermine (Figure 1,
The route used by phage DNA to cross membranes is still a matter of debate. A model was
initially proposed which supposed that the naked DNA would diffuse directly through the
FhuA channel.6 The 3D structure of FhuA makes now this model less realistic. In vivo
studies argued in favor of a central role of pb2, the protein forming phage T5 straight tail
fiber, in channelling DNA through the envelope.*’ Furthermore, cryo-electron microscopy
and recent electron tomography images unambiguously show that the straight tail fiber
crosses the lipid bilayer.” A reasonable hypothesis would be that the interaction of pb5
with FhuA triggers conformational changes in pb2, allowing its insertion in the membranes
and the formation of a DNA channel. In vifro studies are now in progress taking advantage
of the recent expression of pb2 in E.coIi and of its purification.
3 CONCLUSIONS
Although major advances have been made in deciphering the mechanism by which high
affinity transporters bind ligands, we are far from understanding how iron-siderophores are
transported across the outer membrane. This transport is an energy-dependent process
requiring a cytoplasmic membrane-anchored complex consisting of three proteins TonB,
ExbB and ExbD. (Ton complex). Ton-dependent transport systems are widely spread. To
date more than 20 outer membrane proteins whose function depends upon TonB have been
identified in Gram-negative bacteria. However we are missing structural as well as

functional data on these proteins. The next challenge will certainly be the expression,
purification and functional reconstitution of these proteins together with the transporters
in an energized in vitro system.
References
W.Cramer and DKnaff. (1990). In Energy transduction in biological membranes

(Springer-Verlag,ed.), pp. 246, New York.
H. Nikaido, Science, 1994,264,382.
G.S. Moeck and J.W. Coulton, Mol Microbiol, 1998,28,675.
V. Braun, Biol Chem, 1997,378,779.
P. Boulanger, M. Le Maire, M. Bonhivers, S. Dubois, M. Desmadril and L. Letellier,
Biochem., 1996,35, 14216.
M. Bonhivers, A. Ghazi, P. Boulanger and L. Letellier, Embo J, 1996,15,1850.
M. Bonhivers, M. Desmadril, G.S. Moeck, P. Boulanger, A. Colomer-Pallas and L.
Letellier, Biochemist?y, 2001,40,2606.
A.D. Ferguson, E. Hohann, J.W. Coulton, K. Diederichs and W. Welte, Science, 1998,
282,22 15.
K.P. Locher, B. Rees, R. Koebnik, A. Mitschler, L. Moulinier, J.P. Rosenbusch and D.
Moras, Cell, 1998,95,77 1.
10 S.K. Buchanan, B.S. Smith, L. Venkatramani, D. Xia, L. Esser, M. Palnitkar, R.
Chakraborty, D. Van Der Helm and J. Deisenhofer, Nat Struct Biol, 1999,6,56.
11 G. Camel and J.W. Coulton, J Bacteriol, 1991,173,4394.
12 H. Killmann, M. Braun, C. Herrmann and V. Braun, JBacterioZ, 2001,183,3476.
13 D.C. Scott, 2. Cao, 2. Qi, M. Bauler, J.D. Igo, S.M. Newton and P.E. Klebba, J Biol
Chem, 2001,276,13025.
14 M. Bonhivers, L. Plancon, A. Ghazi, P. Boulanger, M. Le Maire, 0. Lambert, J.L.
Rigaud and L. Letellier, Biochimie, 1998,80,363.
15 L. Planqon., M. Le Maire, M. Desmadril, M. Bonhivers, L. Letellier and P. Boulanger,
submitted.
16 L. Planqon, M. Chami and L. Letellier, J Biol Chem. 1997,272, 16868.
17 0. Lambert, L. Planqon, J.L. Rigaud and L. Letellier, MoZ Microbiol, 1998,30,761.
18 0. Lambert, L. Letellier, W.M. Gelbart and J.L. Rigaud, Proc Natl Acad Sci U S A ,
2000,97,7248.
19 J. BBhm, 0. Lambert, A.S.Fangakis, L. Letellier, W. Baumeister and J.L. Rigaud,
Current Biology, 2001,11, 1168.
20 G. Guihard, P. Boulanger and L. Letellier, J Biol Chem, 1992,267,3173.
MORPHOLOGICAL, ASPECTS OF IN CUB0 MEMBRANE PROTEIN
CRYSTALLISATION
C. Sennoga', B. Hankamer2, A. Heron', J. M. Seddon', J. Barber2and R. H. Templer'

1
Department of Chemistry, Imperial College, London SW7 2AY, UK
2WolfsonLaboratories, Department of Biochemistry, Imperial College, London SW7
2AY, UK
1 INTRODUCTION
Comprehensive understanding of the molecular mechanisms underlying membrane

protein function(s) is often obscured by the dimensions, at which such processes
occur. Comprising as they do specific receptors, signal and energy transducers,
channel-forming proteins, active transport pumps, and electron transport systems,
membrane proteins play key roles in a variety of cellular processes. They are believed
to comprise ca. 25-40% of all proteins in the cell.' In order to have the greatest
possible impact on applications that rely on membrane protein functionality, for
example in the pharmaceutical formulation of highly specific drugs, three-
dimensional (3D) structural information at atomic resolution is required as a starting
point. Despite their importance and diversity less than 40 membrane proteins
structures have been determined to atomic or near atomic resolution. This is due,
primarily to the rate-limiting step of producing 2D or 3D crystals of sufficient
electron dfiaction or X-ray crystallographic quality. One of the essential challenges
of post-genomic biology therefore is to determine to high resolution the structure of
membrane proteins.
Conventionally, membrane protein crystallisation has involved lipid-based 2-
dimensional (2D) or detergent-based 3-dimensional (3D) methods2. In 2D
crystallization both the purified and detergent solubilized membrane proteins are
mixed with specially selected lipids. The resultant protefipiddetergent blend is then
depleted of the solubilizing detergent in a controlled manner thereby inducing
formation of lipid bilayers in which 2D membrane protein crystals are embedded. In
contrast, detergent-based 3D crystallization systems are aimed at inducing crystal
formation by the controlled addition of precipitants to pure detergent/protein
complexes. In trials utilising either method, hydrophobic membrane proteins have
proven to be particularly intractable to crystallization. Perhaps the reason for this
stems fiom the fact that hydrophobic membrane proteins are characterised by
hydrophilic surface domains that are insd5ciently large to favour protein-protein
crystal contacts, due to the steric hindrance of the bound detergent micelles. A novel
method based on crystallization of the membrane protein in lipid-based, liquid
crystalline bicontinuous cubic phases, in cubo3 crystallisation, has the potential to
overcome this and a number of related problems. Indeed one of the attractions of the
cubic phase method is that in theory it combines almost all of the advantages of 2D
crystallization with those of detergent based 3D crystallization methods. As its name
suggests the process is based on the incorporation of the target protein into the
complex 3D structure of liquid crystalline bicontinuous cubic phases (Figure 1). The
power of this method is clearly demonstrated by the fact that it has yielded two of the
highest resolution structures of any membrane protein published to date.4
1.1 in cubo Crystallisation Hypothesis
For brevity, we have schematically illustrated the proposed incorporation,

stabilisation and crystallisation mechanism in Figure 1. In a marked departure from
conventional methods, the pioneers of the in cubo crystallisation method argued that
membrane proteins would more readily crystallise in lipid bilayers than in nonbdayer
environments (2D and detergent 3D crystallisation) provided they were incorporated,
stabilised and were capable of isotropic diffusion in an appropriate lipid phase. The
systems they chose are the well studied inverse bicontinuous cubic phases5-*formed
by the mono-glyceride, 1-monooleoyl-rac-glycerol, or monoolein (MO) in water
(Figure 2), known to form a stable, structured matrix in which d f i s i o n of both water
and soluble membrane proteins could take place'. The premise underlying in cub0
crystallisation is that labile membrane proteins once incorporated into continuous
lipid bilayers are stabilised by the near native environment and &se fieely along
the bdayer in an analogous way to the lateral diasion of lipids. Upon nucleation, this
leads to well-ordered crystals.
Figure 1 Proposed mechanism of cubic phase crystallisation based on the work of

Landau and co-workers. A) Dry monoolein (MO) is mixed with detergent solubilized
protein. B) A hydration gradient is established through the monoolein wherein
lamellar and cubic phases are proposed to coexist and the solubilized protein inserts
into the near native environment of the lamellar phase bilayer. C) At equilibrium
60% (w/w)MO (see Figure 2), the system is purely cubic and the protein is then able
to readily diffuse through it. 0)Addition of salt or suitable crystallisation solution
induces localised cubic to lamellar phase transition in the bulk cubic phase. E)
Protein domains phase separate out into the lamellar phase and crystallise in the
presence of suitable counter-ions.
Whilst this method has facilitated macromolecular crystallization of 5 well ordered

3D crystals with different structural characteristics, namely: the light-driven proton
pump bacteriorhodopsin (bR)3T'0,halorhodopsin (hR)", sensory rhodopsin I1
(SRII)''~'~and the reaction centres of the purple bacteria Rhodobacter sphaeroides
(RCsph) and Rhodopseudomonas viridis (RCvir) as well as the light-harvesting
complex 2 from Rhodopseudomonas acidophila (LH2),1° little is known about the

underlying mechanism(s) of this method, although a working hypothesis has been
proposed (see Figure ly3. The role played by the cubic phase remains unclear
although it exists both as the only phase prior to crystallization and as a major phase
subsequent to protein crystal nucleation and early growth14. Early events subsequent
to protein incorporation into the lipid bilayer might be important, although these
remain largely unexplored. Clearly a number of physical and chemical parameters
(i.e., protein concentration, temperature, solubilizing detergents, salts, pH and lipid
type) need to be screened if a high crystallisation throughput is to be realised, in
addition to extending our understanding of the crystallisation mechanism.
120
100
\ L,+Water
2 80
E
Y
E!
60
E
F
40
20
0
0 10 20 30 40 50 60
Water %(wlw)
Figure 2 The equilibrium temperature-composition phase diagram of the

monoloeidwater system indicating the composition and temperature used in the
crystallisation of bR. The binary phase diagram of MO/H20 was determined by
Lutton”, extended by Hyde16 and later confirmed by Caflrey17. Thus at conditions
above 25OC and 4-5 (w/w)% hydration, the MO/H20 equilibriumphase diagram is
characterised by a reversed micellar phase (L2). Upon further hydration 8-22
(w/w)% or an increase in temperature, the L2 phase undergoes a phase transition to
a fluid lamellar phase (La). In the hydration region 25-40 (w/w)% H20, two
bicontinuous cubic structures, the gyroid (Ia3d) and the double diamond (Pn3m),
occur at low and high water content, respectively. At compositions with a water
content above 40 (w/w)% the Pn3m phase co-exists with excess water while in the
region 20-30 (w/w)% H20 the two bicontinuous phases undergf phase transition to
form a reversed hexagonal (HI$phase at temperatures above 80 C.
One of the essential factors relevant to the understanding of how and what the in
cubo crystallisation mechanisms might involve, is how the protein is localised in the
lipid matrix and what effect its presence has on the structure and stability of the lipid
membrane. We have carried out studies with bacteriorhodopsin (bR) as the model
system in order to address some of these questions, and our results are summarised
here. The arrangement, distribution and transformations were investigated usiig
polarizing microscopy (PM) and small angle X-ray difFraction (SAXD).
2.1 Preparation and Crystallisation of bR as the Model System
A major control for our work was the ability to produce bacteriorhodopsin (bR)
crystals using the approach of Landau and Rosenbusch3 as the “model” system.
Thence, aqueous solutions of the purified detergent solubilized bR were mixed with
dry monoolein (MO) in composition ratios based on the phase diagram of the pure
lipid water system using the centmgation approach of Rummel et all3. Purified bR
was concentrated by centrifugal extrusion (10 m g / d in 0.025 M NaPj, pH 5.6) using
an Amicon (PM-10) filter, and admixed with dry MO in the ratio 40:60%(w/w). In
an eppendorf tube crystallisation admixtures were homogenised by temperature
controlled (20°C) centrifbgal miXing in a desktop microfbge (3000 g) for a period of
10 minutes. Subsequent mixing, was effected by rotating the eppendorf through 180”
about its long axis, and repeating the mixing process with increments (1OOOrpm) to
reach 13000rpm and giving a uniformly coloured purple gel. Crystallization was
initiated by layering a solution of 1 M NaP, (PH 6) onto the purple gel. Control
experiments were conducted in which either 1 M NaPi (PH 6) or 2-methyl-4-
pentanediol (MPD) was used as the precipitant. The resultant admixture was sealed
and incubated at 20°C in the dark. Crystallisation progress was monitored on
duplicate samples by way of polarising microscopy. After a period of five weeks
small purple crystals were observed to grow within the cubic phase (Figure. 3).
iA
Figure 3 Lipidic cubic-based crystallisation of bacteriorhodopsin. A) A

Crystallisation admixture of 60% (w/w) MO lipid and 40% (w/w) aqueous protein
suspension prepared by the centrifugation approach. All operations were conducted
at 20°C. Bacteriorhodopsin solutions contained: n-octyl-P-D-glucopyranoside(OG)
solubilized monomeric bR protein, N d K Pi CpH 4.6) saltfiuffeer. Additional NaPi I A4
@H 4.6) was used as the precipitant. B) bR crystals were observed to grow within
the cubic matrix, reaching their full size afterflve weeks incubation of the dispersion
admixture.
II Proteins, Lipids and Their Interartrnns 225
2.2 Multi-WeU Batch Crystallisation
The centfigation approach of mixing membrane proteins with MO has two main
limitations. Firstly, the method is extremely time-consuming, as dry MO has to be
weighed out in individual samples with the subsequent miXing step taking a m h e r 2
hours. Secondly, weighing out the required amount of MO with any degree of
accuracy renders the process difficult for trials having a volume of less than 15 pl,
making the process expensive in terms of protein requirements. To address these
problems we have built a syringe-mixing device, which we use in combination with
an IMP& automatic micro-dispenser to screen a large number of crystallisation
parameters. Figure 4 shows such a device with the green chlorophyll binding protein
(LHC-11) loaded into the left syringe (Figure. 4A) with the appropriate amount of
solid MO (white powder) in the right. A screw fit needle with a constriction connects
the two syringes. During mixing, the protein sample is slowly injected through the
needle connector into the syringe containing the MO (Figure 4B).The resultant gel is
then gently pushed back and forth between the syringes until a homogeneously
coloured gel (Figure 4C) is obtained. The resultant protein based admixture is then
dispensed into wells of nonbirefimgent multi-well sitting drop plates with different
precipitants (Figure 4D). Using this approach we have reduced the mixing times
itom 2 hours to 10 minutes and achieved a 15-fold reduction in protein requirement
due to bulk mixing. We are now able to routinely set up in excess of 300
crystallisation trials per day. By using nonbirehgent plates, we are able to routinely
examine the crystallisation progress as a h c t i o n of added precipitant on the cubic
phase or physical conditions by way of polarising microscopy. The images are
digitally captured both with and without cross-polarizers and then transferred to an
in-house database. This approach has allowed us to screen a large number of
crystallisationparameters.
Figure 4 Micro-batch mixing and set up. A) Solubilized and buflered chlorophyll
binding protein i.e., Light Harvesting Complex II (LHC-II) loaded into syringe (lef2,
and dry MO lipid loaded in right syringe. B) LHC-II is slowly injected into the
syringe containing dry MO lipid, C) After gentle mixing (to and fro) through a small
constriction in the syringe connector, a uniform green cubic phase gel is obtained
0) Using a dispensing device, lpl batches of cubic phase gel are injected into ideally
silica based nonbirefringent sitting drop plates. 6p1 of a crystallisation solution is
subsequently dispensed onto each lpl gel batch. A further I O O p l of solution is
pipetted into the equilibration reservoir before sealing the wells with clear
nonbirefringent tape.
2.3 Bulk Morphological Changes that accompany bR Protein Crystallisation
Small angle X-ray d s a c t i o n (SAXD) was used to determine the phase changes that
accompany the crystallisation of bR. In so doing we aimed to answer a number of
questions: Is the transition from the cubic to the lamellar phase the key stage in
crystal production? Is the protein size relative to the cubic phase unit cell dimensions
important? And finally, could the cubic phase be accessed at low temperatures?
J8
35
28
300 400 500
Pixels
Figure 5 The changes to the small-angle X-ray diffi.action pattern with time that
accompany detergent solubilized bR incorporation into the cubic bilayer, and
subsequent crystallisation (continual image capture over CI duration of six weeh). It
is noteworthy that only a few crystals were observed with this run that quickly
dissolved as the system transformed to a purely lamellar phase. The corresponding
lattice parameters and mean curvature of the bilayer are given in Table 1.
2.3.1 SAXD Sample Preparation. Purple membranes were isolated fiom

Halobacteria Salinaria S9 strain according to reference18. Partially denatured
bacterio-opsin (bo) the de-retinalized form of bacteriorhodopsin (ca.2mghl) was
purified in 2% sodium dodecyl sulphoxide (SDS) at a 5:l SDS:protein (w/w)19.
Purified bacterio-opsin (b0) (200~1) was then supplemented with solid
dimyristoylphosphatidylcholine (DMPC) and OG together with 9M of all trans
ethanolic retinal, to yield a protein suspension comprising 0G:DMPC:retinalprotein
in the ratio 20:5:1:1 all buffered with 50mM NaPj pH 6.0. Under these conditions the
retinal is essentially in excess as only 75% of the protein refolds. Protein refolding
and thus concentration was determined spectrophotometrically by comparing the
absorption peaks at 280nm (total protein, ~280nm= 66,000 M-' cm-') and 555nm
(refolded bR, ~ 5 5 5 n m= 55,300 IW' cm-'). Residual SDS was eluted fiom
reconstituted bR by sequential washing and centrifbgation with an elution buffer of
1.2% OG, 50mM NaP, pH 5.6 using an Amicon 10 microconcentrator to give a final
protein concentration of 4.2mg/ml. Known aliquots of the resultant protein stock
were mixed with dry M O in X-ray transparent capillaries (W. Miiller, Berlin,
Germany) using the centfigation method of Rummel and coworker^'^.
Crystallisation was initiated by adding 1M NaPi to yield a lipid-protein dispersion
with 40% (v/w) protein dispersion medium. Sample homogeneity was monitored
using polarizing optical microscopy to ensure the formation of a uniform cubic phase
before the tube was sealed. All samples were optically isotropic, nonbirefiiingent and
exhibited a uniform microstructure. Crystallisation progress was monitored by time
resolved low-angle X-ray measurements (see 52.3.2). Control experiments were
conducted in which one sample was used to monitor progress by polarizing
microscopy at the 20°C isotherm or merely incubated in the dark thus allowing us to
eliminate any possibility of radiation damage.
2.3.2 SAXD Measurements. X-ray diffiaction measurements were conducted in a

time-resolved mode using nickel-mered Cu Ka.lines (h=1.542 A) fiom an Elliot
GX20 rotating anode generator (Enraf-Nonius, Netherlands). The generator utilised a
l o o p focus cup with a typical loading of 30kV x 25mA = 75OWatts. X-ray focusing
was achieved by using Frank's optics in a Kirkpatrick-Baez arrangement with 200
mm focal length. The X-ray focal beam spot was 16@m in height and 1l o p m
width (111 width at half-maximum). A set of adjustable tungsten slits placed
approximately 15 mm before the sample reduced parasitic scatter to levels where we
were able to measure lattice spacings up to 300 A. X-ray capillaries were placed in a
copper sample holder with thermoelectric temperature control. The temperature was
maintained at 20°C (kO.03)OC. The optics and the sample cell were both held under
vacuum to minimize air scatter. The sample chamber included a variable length,
evacuated fight path, which allowed the X-ray detector to be placed between 120 and
300 mm fiom the sample, thereby taking advantage of the long focal depth of the
optics. The design was limited to wide-angle measurements down to a minimum
spacing of 3.5 A. Time-resolved dfiacted intensity was recorded on a two-
dimensional CCD-based image-intensified detector. The entire X-ray system was
placed under computer control, allowing us to automate the acquisition and analysis
of data. The acquired powder dfiaction images were radially integrated and
subsequently indexed using our TV4 fitting program. TV4, was designed by
Professors E. Eikenberry and S. h e r (Cornell, USA). Modified to suit our exacting
requirements, TV4 has been tailored to operate our home-built detector and
communicate with the off-board control system we use to operate our thermoelectric
heaters. All dfiaction measurements were performed at 20°C in the dark. Samples
were allowed to equilibrate for at least 180 minutes before acquiring an X-ray pattern.
Repeated runs on a single sample using this protocol gave results, which were
reproducible to f l . O A in the lattice parameter, and no observable radiation damage
was recorded. Phase transition temperatures measured on a similar sample, using this
protocol were reproduced to *1 "C.
Our findings reveal that the phase transitions accompanying in cub0
crystallization of bR are dominated by two cubic structures, one of which exists both
prior and subsequent to crystal growth. The two structures with inverse bicontinuity
are closely related to the Schoen G (gyroid) and Schwurz D (double diamond) infinite
periodic minimal sux%uxs (IPMS), and occur at low and high aqueous content,
respectively. The two cubic phases belong to space group Ia3d (QZ30) for the G and
space group Pn3m (Q224) for the D surface. The structure of the Pn3m cubic phase is
characterised by two independent networks of aqueous tubes joined four-by-four at
tetrahedral angles. Each tube forms a continuous channel system, which is surrounded
by a lipid bilayer separating the aqueous phase fiom the hydrocarbon continuum of
the supramolecular lipid organization (hence double diamond). The Pn3m cubic
Iattice is characterised by X-ray dithction peaks (Figure 5) spaced in the ratio
&:&:A:&:&:& .... In contrast the structure of the Ia3d cubic phase is
characterised by two independent networks of aqueous tubes joined three-by-three.
Adjacent networks of the aqueous channels are mirror images of each other. The Ia3d
cubic lattice is characterised by X-ray diffkxtion peaks spaced in the ratio
& : 6: :J16 :&6 :&...while lamellar (La) phases are characterised by
Bragg peaks with s values spaced in the ratio 1:2 :3 :4 :5 :6.... The repeat d -spacing
of lamellar structure are determined as a mean value of the spacing dn=n/soon,
against n ,where n is the Bragg peak reflection order number.
Prior to salt addition, the observed Q224phase is characterized by lattice
parameters, uQ of approximately 98A. The Q224 lattice parameter was observed to
initially increase to 1 1 4 4 before decreasing as the osmotic dehydration took effect
upon adding the precipitant in solution form. After three days the bulk crystallization
system was observed to undergo a phase transition from the Q224 to the QZ3'phase,
with a lattice parameter of 16881. The Q230 lattice parameters were observed to
decrease fiom 168A down to 138A before undergoing a gradual phase transition to
the L hase (d =50A). The sequence of morphological transformations was: Q224+
salt +a Q
p224
/ Q230-+ QZ3'-+ Q230/L,-+La.
2.4 Detergent Effects and Crystallisationat Lower Temperatures
For the purposes of in cub0 crystallization, detergent solubilized membrane proteins

are mixed with dry MO, with the aim of incorporating them into the cubic phase. The
resultant structural topology and stability thereof is dependent on both the detergent
type and aqueous concentration present. As a h c t i o n of their composition, therefore,
detergedwater mixtures give rise to specific phase behaviour, which may include
cubic phases of the normal type2'. Because of their tendency to form interfaces curved
away from water, high concentrations of protein solubilizing detergents are
incompatible with stabilisation of the inverse cubic phases. Indeed, it has been
reported that at high concentrations, n-octyl-P-D-glucopyranoside2'or n-dodecyl-P-D-
maltoside22 when mixed with MO induces formation of lamellar phases. Since the
stability of the cubic phase in the crystallisation system is dependent on both the
detergent type and concentration present we have examined the temperature-
dependence of glycoside-induced phase transitions on the MO cubic stability for a
number of commonly used mild detergents, under equilibrium and metastable
conditions at the relevant detergent concentration for crystallisation of 1 and 3 times
the critical micelle concentration (CMC). Figure 6 shows the phase behaviour for the
systems MO/n-dodecyl-p-D-maltoside/H20 at the CMC (Figure 6A) and n-heptyl-P-
D-thio-glucopyranoside at 3 x CMC (Figure 6B) as a h c t i o n of temperature. In both
cases, noncubic-to-cubic and cubic-to-cubic phase transitions are observed between -
15°C and 35OoC in the heating direction (equilibrium). In the cooling direction
(metastable) however, no cubic-to-noncubic phase! transition is observed over the
temperature range 35 to -5°C suggesting that through carefbl exploitation of the
lipids' metastability, the cubic phase can be accessed at temperatures above 4°C in
some MO/detergent blends. We have examined the MO/detergent metastability on a
wide range of commonly used detergents by gradual cooling of the cubiddetergent
blends to 4OC. The blends were held at 4OC for one month prior to re-examining the
structural form of the MO/detergent mesophase. In the majority of systems studied the
cubic structure was maintained. Depending on the detergent type and concentration,
we find this phenomenon to be long lived to well over a month. Our fidl findings are
described elsewhere23.
It has been demonstrated that the crystallisation process can be carried out
using detergent-fiee blends24. Indeed this might suggest that the influence of the
detergent on the crystallisation process is a minor one. However, it is noteworthy that
firstly, this approach is only feasible in cases where the membrane protein can be
obtained in an enriched form. Secondly, detergents are used because they act as a
conduit to maintaining the proteins integrity prior to incorporation into the cubic
bilayer. To this end they have to be selected in such a way that the concentration used
in no way destab&s but maintains both, the integrity of the protein and topology of
the cubic matrix. Finally, based on these findings we believe, that low temperature
crystallisation can be accessed by exploiting the lipid's mtastabiky. A low
temperature crystabation strategy (for example at 4OC) has the potential to improve
the crystal quality of temperature sensitive proteins.
i
04 I
Figure 6 Temperature dependent phase behaviour at 6U%(w/w) M0:40%(w/w)

aqueous detergent A) I x CMC, n-dodecyl-P-D-maltopyranoside and, B) 3 x CMC,
n-heptyl-8-D-thioglucopyranosidemild detergents. The red and blue data set indicate
phase changes induced by sample heating (equilibrium) and cooling (metastable),
respectively.
2.5 SpectroscopicModifications of PSI1 and LCH-II Macromolecules
We have characterized the events following the incorporation of Light Harvesting

Complex (LHC-11) and Photosystem I1 (PSII) membrane proteins into the
bicontinuous cubic phases of monoolein. We find that the phase behaviour of the
crystallisation matrix is well correlated to that of the model system, but find that the
spectral properties of both protein candidates to be modified by the incorporation
process (data not shown). In both cases the structure of the cubic mesophase was not
modified by the incorporation of either protein. The incorporation of the PSII core
complex for example, into the MO cubic phase was accompanied by a 5 nm blue shift
in the chlorophyll absorption bands. On the assumption that the curvature of the cubic
phase might be responsible for the observed spectroscopic mod~cations,we have

incorporated both proteins in lamellar phases. In both cases, the spectral modifications
observed were similar to those observed in cubic mesophases. Given that the same
modifications were observed both in cubic and lamellar phases, this points to the
possibility that the spectral modifications originate fiom a dimensional mismatch
between protein moieties and the MO bilayer, and not fiom the strain exerted by the
bilayer curvature on the protein. The 5 nm blue shift in PSII core complexes is known
to be associated with a decoupling of the integral manganese-cluster fiom the core
complex. Based on previous studies, we know that PSII is relatively labile and thus
readily undergoes spectral modification owing to loss of the manganese-cluster from
the complex. We believe the loss of the manganese-cluster upon incorporation in the
cubic phase has its origins in the said mismatch and thus exposure of the membranes
hydrophobic subunits to the aqueous cores of the cubic matrix.
Based on correlations of the estimated lateral (trans-membrane) hydrophobic
diensions between the target protein and the estimated parameters of the in cubo
mesophases (see $2.6), we argue, that the observed macromolecular spectral
modifications have their origins in the bilayedprotein dimensional mismatch (see
2.7). We argue further that the mismatch will be obviated by using lipid blends with a
maximal swelling of the cubic mesophase, doped with additional lipids that promote
reverse hexagonal and lamellar structures23. A suitable blend for the PSII core
cornpiex might involve doping with synthetic phytanyl-chained glycolipids, which
have shown successll functional reconstitution of PSII core complex26. We are
exploring the use of this blend and will report its potential in another publication.
2.6 Parametric Estimation of in cubo Mesopbases
The structure of lipid aggregates depend on the bending energy of the lipidic
monolayers within the aggregates. The bending energy therefore leads to the
formation of lipid mesophases with different radii of curvature and geometry. Here
we use observed SAXD data for aqueous bR hydrated MO and apply an established
formalism to estimate the mean curvature of the phases observed. A prerequisite of
using this approach is that the location and indeed existence of the pivotal surface has
to be established as a starting Before defining the pivotal surface we shall first
restate the structural nature of bicontinous cubic mesophases. As outlined earlier
inverse bicontinuous cubic phases are best envisaged as consisting of an amphiphilic
bilayer draped onto an infinite periodic minimal surface (IPMS) with the minimal
surface occupying the average location of the bilayer mid-plane. In applying this
formalism27, all molecularly defined surfaces, for example the polar/nonpolar
interface and the pivotal surface are defined as occupying parallel surfaces to the
minimal surf8ce. Thus we define the pivotal surface as a particular cross-sectional
area, A,of a molecule within a curved bilayer which lies parallel to the underlying
infinite periodc minimal surface (IPMS) and remains invariant to isothermal bending
i.e., the mass of the sample on either side of the pivotal surface remains contant
during molayer bending. Therefore any change in molecular shape due to isothermal
change is accompanied by a change in the distance, 6, fiom the IPMS to this surface.
To derive the relationships relating the lattice parameter to the water volume
&action we have to define the geometry of the “pivotal surfhce”. The cross-sectional
area per iipid at this surfhce, and the molecular volume between this surface and the
ends of the hydrophobic chains are given by A,, and v,, respectively. The
i I Proteins, Lipids and Their interactions 23 1
“stoichiometric’ ’ lipid geometry therefore is characterised by these two parameters,

and by the total molecular volume of the lipid, v . The location of the ivotal surface
can be determined by applying the rules of differential geometq?P and for our
purposes it is sufficient to use values determined earlier, which place the pivotal
surface at or about the C3 position on the hydrocarbon chain 27, 28. For MO we have
(v) =612A3;(v,) = 465A3;(An)= 33A2 (1)
Our X-ray data for the crystallizationprocess are summarized in Table 1. The location
of the pivotal surface,t, was established using the experimentally determined lattice
parameter, aQ, and the aqueous volume fiaction, 9, via equation (1). Here o ,is the
ratio of the IPMS area to the (unit cell vol~rne)~”,and x is a topology index of the
IPMS known as the Euler-PoincarC characteristic for a given cubic phase. The
constants o and x are 1.919 and -2 for the Q224 phase, and 3.091 and -8 for the Q230
P k .
of the surface averaged mean curvature defined to the pivotal

was determined for the Q224 and Q230 cubic mesophases of the
crystallisation matrix based on the lattice parameter, aQ, and the pivotal surface, 5 at
the limiting hydration in each phase according to the relations hi^^^
The average radius of cubic aqueous network defined at the poladapolar interface was
determined via equation (4)
(R,)= -- (4)
J&i I
where, the surface average Gaussian curvature, (K)= 2nx / oaQ2and I , the length of
the lipid mphiphile is determined via the lipid volume fiaction, according to +,
equation (5>30
The diameter of the aqueous channel within Q244 and Q230cubic mesophase is then
determined by using the expression in (6)
Table 1 Selected Structure and Lattice Parameters from bR crystallisation mixtures

prepared as described earlier in X-ray transparent tubes. Column 2 and 3 show
recorded structural transformations, cubic lattice parameters, and lamellar repeat
spacing in the bulk mesophasefollowing crystallization induction at 20°C. Column 7
shows the corresponding variation in mean curvature due to osmotic pressure (see
$2.6 Parametric Estimation of in cub0 Mesophases) exerted by the precipitant salt.
Column 6 shows the variation in lipid length (I), and thus trans-membrane bilayer
thickness (21) while column 9 shows the local variation in the average diameter of the
cubic aqueous network Finally, the time (-I) refers to the duration of time prior to
salt addition. The dimension in parentheses refer to the water separation in the
lamellarphase.
Time Space
Duration Group a k
7q
. 5" I )w) I<Rv>l l(0)l
(days) [A] [A] ~-'x10-~ [A] [A]
-1 Pn3m 97 0.399 11.9354 16.17 - 9.74 21.75 43.51
1 Pn3m 104 0.404 12.6826 17.17 - 7.68 23.49 46.99
3.5 Pn3m 92 0.337 12.5899 17.21 - 8.31 18.76 37.52
7 - - - - - - - -
21 Ia3d 138 0.289 12.6332 17.36 -10.79 16.86 33.73
, 35 La , 42 , 0.158 - , 17.68 - - , (6.64)
The crystallization dependence of the mean curvature, ( H p ), determined at

the MO pivotal surface is presented in column 7 of Table 1. The trend shows that the
curvature of the lipid monolayer (defined at the pivotal surface) in the bicontinuous
crystallization matrix at 20°C is initially constant but as expected decreases as the
aqueous additives are added. Once the sample achieves structural homogeneity (1
day) as judged by the difbction, the mean curvature reaches a maxjrnum.
Subsequently the system undergoes a cubic-to-cubic (Q224 to Q230) phase transition (1
week). After three weeks the system undergoes a cubic to noncubic (Q230to La) phase
transition (Figure 5). That the curvature is decreasing indicates that the additives have
an osmotic effect on the cubic matrix, dehydrating the cubic phase and ultimately
leading to a lamellar phase transition. Thus we find that the crystallization process
occurs via an increase in the bilayer membrane spanning thickness and a decrease in
the aqueous space available to the protein. One would expect that protein-lipid and
protekprotein interactions might be least favorable under conditions of high
curvature compared to those of zero curvature. Part of the fiee energy driving force
that leads to the dehydration of the cubic carrier phase might therefore be related to
this. Phase diagrams for potential lipid systems for crystallizing membrane proteins
may therefore require that high water composition cubic phases border on to a lower
water composition lamellar phase.
2.7 Effects of Proteinflipid Hydrophobic Mismatch on Crystallization
The dimensional matching between the hydrophobic length of an integral membrane

protein and the hydrophobic thickness of the target membrane lipid bilayer is an
important &tor in protein-lipid interactionz8. In addition to the well-documented
protein sorting, hydrophobic mismatch plays important roles in other membrane
processes Iike reconstitution. Given that membrane protein reconstitution forms a key
step in the in cub0 membrane protein crystallization method, the effects of
hydrophobic matching or mismatch between an embedded membrane protein and the
supporting lipid bilayer might be central to the early stages of protein stabilizationand
as a consequence crystallization in the cubic matrix. The lateral hydrophobic
dimension in the case of bacteriorhodopsin is estimated to be ca. 30-31 A31 a value
that M s within the range of that determined for the MO cubic-based bdayer. In Table
1 we have that the pivotal surface has a width of approximately 12.6 A. This is the
width to the C3 carbon on MO, so if we add on 2 for the C1 and C carbons, the
hydrophobic thickness of the bhyer will be 29.2 A. This is just somewhat thinner
than the estimate of the hydrophobic thickness of bacteriorhodopsin.
The exact unitary dimensions of a bR monomer are unknown, however, given
its relative molecular size and a comparison of this to the other proteins investigated
here, it can be argued that bR is an unrepresentative system given that many proteins
are characterized by much larger dimensions. Based on our cubic parametric estimates
(Table 1) it would appear that bR monomers are reasonably readily accommodated
both within the MO bdayer and the cubic aqueous channels, which take diameters of
the order 31-42 A. The two other proteins studied, the light-harvesting complex
(LHC-11) and photosystem I1 (PS 11) are characterised by dimensional length scales of
the order (ca. 33.9 x 38.29 x 53.41 A) and (ca. 220 x 150 x 95 A), respectively.
Based on correlations between our determined estimates of the membrane-spanning
parts of bR, LCH-I1 or PSII and the hydrophobic thickness of MO we are drawn to
conclude that while the hydrophobic thickness of bR is reasonably well matched to
that of MO, the reverse is true in the case of the LCH-I1 and PSII. It might therefore
be reasonable to attribute the observed spectral modifications in both the LCH-I1 and
PSII protein complexes to hydrophobic mismatch. Figure 7 B-C illustrates the two
main forms of hydrophobic mismatch. In both LCH-I1 and PSII the case illustrated in
(Figure 7C) applies.
In planar mesophases (lamellar phases), hydrophobic mismatch between the
lipid alkyl chain region and the protein hydrophobic region may lead to bdayer
deformations that carry a fiee energy penalty. The same will hold true in non-planar
structures (such as the bicontinuous cubics). Noting that the hydrophobic thickness in
the cubic phases of MO are thinner than that of the lamellar phase it is conceivable
that proteins with a narrow hydrophobic region might drive the system to stabilise a
bicontinuous cubic structure. In cases where the protein has a thicker hydrophobic
domain the latter is not possible and where the mismatch is extreme it is quite
possible that the protein will simply not be stable within the MO matrix. Under such
circumStances it would be quite probable that one would observe the aggregation of
protein into amorphous forms. In the case of PSII, we believe that something of this
sort may be occurring, leading to the loss of Mg in the PSI1 submit and the observed
spectroscopic modifications. The observed blue shift is diagnostic of PSII

denaturation. This and related issues will be addressed in a fbture publication.
A B C
Figure 7 A schematic illustration of the proteidipid hydrophobic matching. A) The

hydrophobic thickness of the protein and that of the lipid membrane (db are well
matched. In this case, the interfacial and chain pressure of the two opposing
monolayers are well matched and as a consequence the net moment is zero. To this
end, both opposing monolayers have no desire to bend B) The protein-lipid
hydrophobic thickness is mismatched, i.e., the protein hydrophobic dimension is
shorter than that of the host membrane lipid bilayer. C) Here the protein hydrophobic
dimension is much larger than the dimensional length of the membrane lipid alkyl
chain.
3 CONCLUSION
In experiments based on the pioneering work of Landau and Rosenbusch we have

used a combination of optical microscopy and time resolved SAXD to identify the
structural phase transitions that accompany the crystallisation of bR as a model
system. This work was aimed at facilitating continuous monitoring of the membrane
lipid structural properties during crystallisation, in such a way that crystallisation
conditions, and their parameters could be optimised to produce protein crystals with a
high structural resolution. To our knowledge this is the first time that it has been
possible to follow the crystallisation of membrane proteins in this way. In our efforts
to screen a whole series of crystalliition parameters, an increase in the number of
trials undertaken has made automation a necessity. We have therefore implemented
Multi-Well batch crystallisation methods. Using this approach we have screened a
number of in cubo crystallisation conditions. New insights into the necessary
crystallisation conditions has been achieved, although more research is required
before a detailed analysis is complete, and thus providing a full understanding of the
mechanism of in cubo membrane protein crystallisation. In addition we have
characterized the events following Light Harvesting Complex (LHC-11) and
Photosystem I1 (PSII) membrane protein incorporation into bicontinuous cubic phases
of MO. Our experiments show that whilst, the spectral properties in both proteins
were modified by the incorporation process (data not shown) the structure of the cubic
phase matrix was stabilised by the incorporation of either protein, as long as the
protein to cubic ratio was kept to that required by the standard crystallisationprotocol.
On the assumption that the curvature of the cubic phase might be responsible for the
observed spectroscopic modifications, we have incorporated both proteins in lamellar
phases. The resultant spectral analyses prove that the modifications are similar to
those observed in cubic phases. These observations point to the possibility that the
spectral modifications originate from a dimensional mismatch between protein
moieties and the MO bilayer, and not &om the strain exerted by the bilayer curvature
on the protein. Based on lipid packing calculations, we have shown that the observed
macromolecular spectral modifications might indeed originate kom the
bilayer/protein hydrophobic mismatch. We argue M h e r that the mismatch can be
obviated by using lipid blends with a maximal swelling of the cubic mesophase,
doped with additional lipids that promote reverse hexagonal and lamellar structures.
C.S acknowledges support in form of a studentship from the EPSRC and Unilever Research, Port
SunI ight.
References
1. D.T. Jones, FEBS Lett. 1998,423,281.

2. H. Michel and D. Oesterhelt, Proc. Natl. Acad. Sci. USA, 1980,77, 1283.
3. E. M. Landau and J.P. Rosenbusch, Proc. Natl. Acad. Sci. USA, 1996,93, 14532.
4. H. Luecke, B. Schobert, H.-T. Richter, J-P. Cartailler and J. K. Lanyi, J Mol.
Biol. 1999,291,899
5 . G. Lindblom and L. Rilfors, Biochim. Biophys. Acta. 1989,988,221.
6. R.H. Templer, Curr. Opin. Coll. Inter$ Sci. 1998,3,255.
7. J.M. Seddon and R.H. Templer, Phil. Trans. R. SOC.Lond. A 1993,344,377
8. V. Luzzati, Cum. Opin. Siruct. Biol., 1997,7,661;
9. S . Cribier, A. Gulik, P. Fellman, R. Vargas, P. Devaux and V. Luzzati J Mul.
Biol. 1993,229,517.
10. M.L. Chiu, P. Nollert, M.C. Loewen, H. Belrhali, E. Pebay-Peyroula, J.P.
Rosenbusch and E.M. Landau, Actu Ctyst. 0,2000,56,781.
1 1 . A. Royant, P. Nollert, K. Edman, R. Neutze, E.M. Landau, E. Pebay-Peyroula, J.
Navarro, Proc. Natl. Acud Sci. USA, 2001 ,98, 1013 1.
12. H. Luecke, B. Schobert, J. K. Lanyi, E. N. Spudich and J. L. Spudich, Science,
2001,293,1499.
13. G. Rummel, A. Hardmeyer, C. Widmer, M.L.Chiu, P. Nollert, K.P. Locher, 1.1.
Pedruzzi, E.M. Landau and J.P. Rosenbusch, J. Struct. Biol., 1998,121,82.
14. P. Nollert, H. Qiu, M. Caf€rey, J.P. Rosenbusch, and E.M. Landau, FEBS Lett.
2001 ,504,179.
15. E.S. Lutton, J Am. Oil Chern. SOC.1965,42,1068
16. S. Hyde, S. Andersson, B. Ericsson and K. Larsson, ZKristallugr. 1984,168,213
17. H.Qui and M. Cafl?ey, Biomaterials, 2000,21,223
18. D. Oesterhelt, and W. Stockenius, Methods Enzymol1974,31,667
19. M. S. Braiman, et al. J Biol. Chern. 1987,262,9271
20. P. Sakya, J.M. Seddon and RH. Templer, J Phys. II. 1994,4, 13 1 1
21. B. Angelov, M. Ollivon and A. Angelova, Langmuir, 1999,15,8225
22. X . Ai and M. Cafltkey, Biophys J, 2000,79,394
23. C. Sennoga, B. Hankamer, J.M. Seddon and R.H. Templer, (in preparation)
24. P. Nollert, A. Royant, E. Pebay-Peyroula and E.M. Landau FEBS Lett. 1999,457,
205
25. C. Sennoga, Ph.D. thesis, Imperial College, London, (in preparation)
26.T. Baba, H. Minamikawa, M. Hato, A. Motoki, M. Hirano, D. Zhou and K.

Kawasaki Biochem. Biophys. Res. Commun., 1999,265,734
27. R. H. ‘Templer, Langmuir, 1995,11, 334
28. B. J. Khoo, Ph.D. thesis, Imperial College, London, 1996
29. R. H. Templer, J. M. Seddon and N. A. Warrender Biophys. Chem. 1994,49,1
30. D.C. Turner, 2.-G. Wang, S.M. Gruner, D.A. Mannock, R.N. McElhaney, J Phys.
II France 1992,2,2039
31. B. P h o v h , E. Perochon and J.-F, Tocanne, Eur. J. Biochem., 1993,218,385
MOBILITY OF PROTEINS AND LIPIDS IN THE PHOTOSYNTHETIC MEMBRANES
OF CYANOBACTERIA
C.W. Mullineaux and M. Sarcina
Department of Biology, University College London, Darwin Building, Gower St. ,London
WClE 6BT, UK
1 INTRODUCTION
In recent years there has been tremendous progress in understanding the structures of
photosynthetic light-harvesting antennae and reaction centres. However, we know much
less about the dynamics of these complexes in vivo. How do light-harvesting complexes
and reaction centres interact in the intact photosynthetic membrane? Are the interactions
permanent or transient, and how are they affected by regulatory mechanisms? We have
been probing these questions using a variant of Fluorescence Recovery after Photobleachhg
(FRAP) which allows us to observe the diffusion of fluorescent pigment protein complexes
in photosynthetic membranes in vivo.' Our currently preferred model organism is the
cyanobacterium Synechococcus sp PCC7942. In common with some other cyanobacteria,
Synechococcus 7942 has elongated cells with the thylakoid membranes arranged as regular
concentric cylinders aligned along the long axis of the cell. The cells may be fbrther
elongated by growth in the presence of cell division inhibitors, without any detectable side-
effects in terms of photosynthetic h c t i o n or membrane structure.' Cyanobacterial
thylakoid membranes have a rather uniform composition, with no significant lateral
heterogeneity. Their regular geometry is in contrast to the photosynthetic membranes of
virtually all other photosynthetic organisms, which tend to exhibit lateral heterogeneity
and/or intricate and irregular fine-scale membrane structure. These properties make
Synechococcus 7942 ideal for FRAP measurements. In addition, Synechococcus 7942 is
well-characterised and transformable, and numerous mutants are available.
Since FRAP is an optical technique it has limited spatial resolution. Thus, we require a
regular membrane geometry for quantitative measurements. We use a one-dimensional
FRAP technique which exploits the cylindrical geometry of the Synechococcus thylakoids
(Fig. 1). As in other cyanobacteria, the principal light-harvesting complexes in
Synechococcus are phycobilisomes, large and highly-ordered protein aggregates coupled to
the cytoplasmic surface of the thylakoid The Photoystem I and Photosystem II
reaction centres are integral membrane protein complexes. Energy transfer measurements
indicate that phycobilisomes can interact with both kinds of reaction centree4 We have used
FRAP to measure the diffusion rates of the phycobilisomes and Photosystem I1 reaction
23 8 Biophysical Chemistry:Membranes and Proteins
centres. In Synechococcus, as in all the other cyanobacteria that we have examined, we find
the Photosystem I1 is essentially immobile. However, phycobilisomes diffuse rapidly on the
surface of the thylakoid membrane. This shows that the interaction between phycobilisomes
and reaction centres is transient and unstable. We report the use of F W measurements on
mutants to fiuther explore the nature of the interaction between phycobilisomes and
thylakoid membrane components, and we discuss the possible physiological role(s) of
phyco bilisome mobility.
Figure 1
Geometry of a one-dimensional FRA P
measurement (adaptedfrom Ref I). A.
A. A cell aligned in the y-direction is
selected. A highly-focused confocal
laser spot is scanned rapidly across the
cell in the x-direction, bleaching the
pigments in a line across the cell.
B. The laser power is reduced to prevent
further bleaching, and the spot is
scanned in the XYplane to record a
series of two-dimensionalfluorescence
images of the cell.
C. The images are integrated in the x-
direction to produce plots of
fluorescence intensity versus position
along the long axis of the cell. A series
of plots at different times after the
bleach shows the diffusion of the
fluorescent complex.
2 MATERIALS AND METHODS
Synechococcus sp. PCC7942 was grown in BGl1 medium supplemented with 10 mM

NaHCO3 and appropriate antibiotics for mutants. Liquid cultures were grown in an orbital
shaking incubator at 30 "C with white illumination at about 10 pE. m-2s". For use in FRAP
measurements the cells were elongated by treatment with thiobendazole. This resulted in
increased mean cell length without any detectable alteration in photosynthetic hction?
FRAP experiments were carried out at CLRC Daresbury Laboratory (Warrington, Cheshire,
UK) using the scanning confocal microscope Syclops with a 633 nm Helium-Neon laser or a
442 nm Helium-Cadmium laser. Fluorescence was selected using a Schott RG665 red glass
filter, transmitting light above about 665nm. Under these conditions excitation with 442 nm
light allows observation of fluorescence predominantly Photosystem XI, and excitation with
633 nm light allows observation of fluorescence predominantly fiom phycobilisome cores'.
Cells were spread on 1.5% agar containing growth medium, covered with a glass cover slip
and placed on a temperature-controlled stage under the microscope objective lens. A 40 x
oil immersion lens (numerical aperture 1.3) was used with 20 pm pinholes to create a
confocal spot with FWHM dimensions of about 0.9 pm in the Z-direction and 0.3 pm in the
XY plane. The confocal spot was scanned for about 1 second in the X-direction to create
the bleach. The confocal spot was then scanned in the XY plane to record a sequence of
images of the cell at 3 s intervals. Images were analysed and diffusion coefficients
calculated as described by Mullineaux et al. '
3.1 Mobility of phycobilisomes and Photosystem I1 in Synechococcus 7942
Figure 2 is a FRAP image sequence showing the mobility of phycobilisomes. We found that
the phycobilisomes diffused rapidly. At 30 "C, the average diffusion coefficient for
phycobilisomes was (3.1 f 1.0) x lo-'* cm2.s-'. However, in the case of Photosystem I1 we
could detect no dfision on the timescale of the measurement (not shown). As in the other
cyanobacteria we have examined, it appears that the association between phycobilisomes
and reaction centres is transient and unstable.
Figure 2
FRAP image sequence showing phycobilisomeJluorescence. The scale bar is 3 microns.
Note that the bleached line spreads and becomes shallower with lime, indicating diffusion
of the phycobilisomes
3.2 Effect of phycobilisomesize
We have explored the effect of phycobilisome size by measuring the diffusion coefficient in
a mutant lacking the phycobilisome rod elements. The mutant, NHECAT, lacks genes
coding for the a-and p-subunits of phycocyanin and rod linker polypeptides. However the
phycobilisome cores are still assembled and fimctionaL6 The phycobilisome cores have a
molecular mass of 1200-1300 kDa and dimensions of about 22 x 11 x 12 nm. The intact
phycobilisomes of wild-type cells are hemidiscoidal structures with a typical molecular mass
of about 6000 kDa and a longest diameter typically about 60 nm.' At 30 OC, the mean
diffusion coefficient for the phycobilisome cores in R2HECAT was (7.1 k 0.8) x 1 0 ' O
cm2s-' . This compares to a mean dfiion coefficient of (3.1 +_ 1.O) x lo-'' cm2.s" in the
wild-type. Thus phycobilisome diffusion at growth temperature is faster by a factor of 2.3 f
0.7 in WHECAT. This suggests that cytosolic crowding plays a role in limiting the rate of
dfision of the phycobilisomes.
3.3 Diffusion of intact phycobilisomes or detached rod elements?
We have interpreted our FRAP results in terms of the movement of intact phycobilisomes,
since we excite the phycobilisomes with short wavelength light predominantly absorbed by
phycocyanin in the phycobilisome rods, and observe long-wavelength fluorescence
predominantly from the phycobilisome cores.' However, spectral overlap makes it hard to
completely exclude an alternative possibility, that the phycobilisome cores are immobile and
the diffusion we see is of rod elements that may not be stably coupled to the phycobilisome
cores in vivo. Our studies with the R2HECAT mutant (see above) shed m h e r light on
this problem. We find that the phycobilisomes are mobile in this mutant. Since the rod
elements are lacking, the cores must be moving. Thus the diffusion we observe in the wild-
type is most probably of intact, hlly assembled phycobilisomes.
3.4 How do phycobilisomes interact with the membrane?
Phycobilisomes are assembled and are membrane-associated even in the absence of

Photosystem I1 and Photosystem I reaction centres.' Thus, when phycobilisomes diffuse, we
imagine them decoupling from a reaction centre, but remaining attached to the membrane
surface. The phycobilisome will then diffuse freely on the membrane surfbce before
coupling to another reaction centre. However, the nature of the interaction with the
membrane is unclear. The ApcE protein of the phycobilisome core is implicated. Proposals
for the association of ApcE with the membrane have included an integral membrane
domain" or a covalently attached acyl group." We have explored this problem by
measuring the dfision coefficient for phycobilisomes at a range of temperatures. We
found that cooling below the phase transition temperature of the membrane had no
significant effect on the mobility of phycobilisomes. Under the same conditions, the
diffusion coefficient of a lipid-soluble fluorescent marker was reduced by a factor of six
(data not shown). This strongly suggests that there is no integral membrane component in
the phycobilisome. Instead, we propose that phycobilisomes interact with lipid head-groups
at the membrane surface. A precedent is spectrin, a component of the erythrocyte
cytoskeleton. Spectrin is proposed to interact with the membrane via multiple weak
interactions with lipid head-groups.l 2 As with phycobilisomes, spectrin can d a s e rapidly
on the membrane surface, and the dfision coefficient is not strongly affected by cooling to
the phase transition temperature of the membrane.
3.5 Effect of lipid desaturation - role of lipids in controlling phycobilisome-reaction

centre interaction?
Mutants in which the thylakoid membrane lipid composition is altered provide a fiuther
opportunity to explore the interaction between phycobilisomes and membranes. We have
used desA', a transformant of Synechococcus 7942 which contains desA, the A12 fatty acid
desaturase gene fiom Synechocystis 6803.l 3 DesA' cells have a much higher proportion of
unsaturated fatty acids than the wild-type. As would be expected, the thylakoid membranes
are more fluid in desA+than in the wild-type (unpublished data). Unexpectedly, we found
that phycobilisome dfision was far slower in desA' than in the wild-type. At 30 "C, the
mean phycobilisome diffusion coefficient in desA” was (2.5 2 1.2) x cm2 s-’, slower
than in the wild-type by a factor of 120 f 70. The most likely explanation is that the
interaction with the reaction centres is s t a b i d in desA’. We know that Photosystem I1 is
immobile (Fig. 3). Therefore, if the binding of phycobilisomes to Photosystem I1 is
stabiiised, the diffusion coefficient for phycobilisomes will be reduced. How could lipid
desaturation alter phycobilisome-reaction centre interaction? Maybe specific lipids, or the
general lipid environment of membrane, play a crucial role in mediating phycobilisome-
reaction centre interaction. Alternatively, it could be an indirect effect. Phycobilisome-
reaction coupling may be influenced by the redox state of electron transport cofactors (see
below) and this may differ in wild-type and desA’ cells under our measuring conditions.
3.6 Physiological role(s) of phycobilisome mobility?

Phycobilisome mobility is characteristic of all the cyanobacteria that we have examined.
What physiological role(s) could it play? Three possible explanations are explored below:
a. Phycobilisome mobility is requiredfor regulation of light-harvesting through state

transitions.
The physiological adaptation mechanism known as state transitions involves the
redistribution of phycobilisomes between Photosystem I1 and Photosystem I. l 4 This
presumably requires movement of the phycobilisomes. However, state transitions occur on
a timescale of a few seconds to about a minute. At the diffusion rates we observe, we can
estimate that a phycobilisome could diffuse fiom Photosystem I1 to Photosystem I in about
15 milliseconds. Thus it is likely that the rate at which state transitions occur is controlled
by the signal transduction pathway, rather than by the diffusion of the complexes. We could
predict that state transitions could still occur if the diffusion rate of phycobilisomes were
hundreds of times slower. In fact we find that state transitions occur normally in the desA”
transformant, in which phycobiliwme diflfUsion is about 120 times slower than in the wild-
type (see above).
b. Phycobilisome mobility is required for synthesis and turnover of thylabid membrane

components
Phycobilisomes are large complexes, which normally occupy much of the cytoplasmic
swface of the thylakoid membrane.” It could be argued that phycobilisome mobility is
necessary to allow access of ribosomes, proteases, and regulatory enzymes to the membrane
surface, in order to allow synthesis, turnover and regulation of thylakoid membrane
components. One prediction of this idea would be that the turnover of the D1 polypeptide
should be slower in the desA’ transformant, in which phycobilisome mobility is greatly
reduced (see above). However, it appears that D1 turnover is actually faster in desA’ than
in the wild-type.I6
c. Phycobilisome mobility increases the efJiciency of light-harvesting

Phycobilisomes are mobile on the same timescale as the secondary electron transport
reactions. Could phycobilisomes decouple from photochemically “closed” reaction centres
and re-associate with open reaction centres, thus minimising the wastefbl transfer of
excitons to closed reaction centres? In this model, phycobilisome mobility would be a way
to allow a limited pool of phycobilisomes to act as efficient light-harvesting antennae for a
much larger pool of reaction centres. Experiments to test this idea are in progress.
Acknowledgments
We thank Mark Tobin (CLRC Daresbury Laboratory, Warrington, UK) for his assistance
with the FRAP measurements, and Abosede Felix and Jessamy Findlater for electron
microscopy. We also thank Norio Murata (National Institute for Basic Biology, Okazaki,
Japan) and Petter Gustafsson (University of Umek Sweden) for providing mutants.
Supported by a BBSRC research grant to CWM.
References
1 . C. W. Mullineaux, M.J. Tobin and G.R. Jones, Nature, 1997,390,42 1 .

2. M. Sarcina and C.W. Mullineaux, FEMS Microbiol. Lett., 2000, 191,25.
3. R. MacColl J Struct. Biol., 1998,124,311.
4. C .W. Mullineaux, Aust J Plant Physiol., 1999,26,67 1.
5. R.W. CastenhoIz in Methods in Enzymology, eds L. Packer and A.N. Glazer,
Academic Press, San Diego, 1988, Vol167,68.
6. R.P. Bhalerao, T. Gillbro and P. Gustafsson Photosynth. Rex, 1995,45,61.
7. A.N. Glazer, Biochim. Biophys. Acta, 1984,768,29.
8. R.J. Ellis, Cum Opin. Struct. Biol., 2001, 11, 1 14.
9. J. Yu, Q. Wu, H. Mao, N. Zhao and W.F.J. Vermaas, IUBMB Life, 1999,48,625.
10. T. Redlinger and E. Gantt, Proc. Nut. Acad. Sci. USA, 1982,79,5542.
1 1 . D. Bald, J. Kruip and M. Rogner, Photosynfh.Res., 1996,49, 103.
12. P.J. O’Toole, C. Wore, S. Ladha and R.J. Cherry, Biochim. Biophys. Acta, 1999,
1419, 64.
13. Z. Gombos, E. Kanervo, N. Tszvetkova, T. Sakamoto, E.-M. Aro and N. Murata
Plant Physiol., 1997, 115, 55 1,
14. J. J. van Thor, C.W. Mullineaux, H.C.P. Matthijs and K.J. Hellingwerf, Botanica Acta,
1998,111,430.
15. L. Mustardy, F.X. Cunningham and E. Gantt, Proc. Nut. Acad. Sci. USA ,1992,89,
10021.
16. K. Sippola, E. Kanervo, N. Murata and E.-M. Aro, Eur. J. Biochern., 1998,251,641.
PARTITIONING AND THERMODYNAMICS OF CHLORDIAZEPOXIDEIN n-
OCTANOUBUFFER AND LIPOSOME SYSTEM
Catarina Rodriguesa, Paula Gameiroa*, Salette Reisb, J.L.F.C. Limab and Baltazar de
Castroa
a CEQUPDepartamento de Quimica, Faculdade de Cihcias, Universidade do Porto, 4169-

007 Porto, Portugal
CEQUPLaborat6rio de Quimica-Fisica, Faculdade de Farmicia, Universidade do Porto,
4050-047Porto, Portugal
1 INTRODUCTION
Chlordiazepoxide is a benzodiazepine and belongs to a class of neuroactive drugs that

enjoy a wide administration because of their muscle relaxant, hypnotic anticonvulsivant
and ansiolytic characteristics. The ability of a drug to elicit a pharmacological action in a
biological system requires, in most cases, an interaction between the drug and the
biological membrane. The permeability of these membranes is determined by physical
factors including the partition coefficient, diffusion coefficient, membrane thickness and
interfacial barriers. '
Thermodynamic studies have been employed to gain an insight into drug transport and
relative biological activity and are also used to predict drug incorporation into
phospholipid vehicles like liposomes, which are considered to be useful drug carriers?
Recently, interest has focused on determination of the partition coefficients of
drugs between lipid bilayer vesicles and the aqueous phase in order to investigate the
behaviour of drugs towards biomembranes. Partition coefficients are normally determined
in an isotropic two-phase solvent system like water/octanol but as membranes possess an
ordered molecular arrangement, with structural characteristics varying from a liquid
crystalline state to a rigid gel and are able to exert electrostatic influences, liposomes are
believed to be a better model membrane to predicted drugmembrane partition coefficients.
This is because they mimic better the hydrophobic part and the outer polar and negatively
charged surface of the phospholipids of natural membranes3
The present study compares the partitioning behaviour of chlordiazepoxide in n-
octanolhuffer and buffered DMPC and DPPC multilamellar liposome using a
thermodynamic approach and, from this, evaluates the interaction of chlordiazepoxide with
lipid bilayers in comparison with the n-octanolhuffer system?
2 MATERIAL AND METHODS
Dimyristoylphosphatidylcholine (DMPC) and Dipalmitoylphosphatidylcholine (DPPC)

were obtained from Avanti Polar Lipids Inc. Chlordiazepoxide was a gift from Hoffman-
L a Roche. n-Octanol was from Fluka and chloroform from Merck (pro-analysi). All lipid
suspensions were prepared with aqueous10 mM Hepes buffer solution ( I d . 1 M NaCl; pH
7.4).
2.1 Distribution studies in n-octanolhuffer systems
Convenient volumes of aqueous phase (5ml Hepes buffer) containing the appropriate
concentration of drug (23.6 pM) and n-octanol (0.1 mL) were equilibrated for 4h at
constant temperature (M. 1°C) in a shaking water-bath. Concentrations of chlordiazepoxide
in the aqueous phase were determined by UV spectrophotometry in an UNICAM UV-300
spectrophotometer equipped with a constant-temperature cell holder at A d x (261nm). The
distribution of chlordiazepoxide was obtained from the average of duplicate determinations
at each temperature over the range 27-50 "C.
2.2 Distribution studies in liposome systems
Liposomes of DMPC and DPPC were prepared by evaporation to dryness of a lipid

solution in chloroform (2mg/mL) under a stream of argon and then left under vacuum
overnight to remove all traces of the organic solvent. The resultant dried lipid films were
dispersed with 5mL Hepes buffer containing 23.3kM chlordiazepoxide and the mixtures
were vortexed above the phase transition temperature to produce multilamellar liposomes
(MLV). The distribution of the drug was determined using UV analysis over the range 4-
50°C. Determinations were made in duplicate and the results averaged.
1.0-
0.8 -
8
(d
0.6-
0.4-
a
a
0.2 -
0.0 -
4 ) . , . I . I . , . I . ) . , . l i
220 240 260 280 300 320 340 360 380
h Inm
Figure 1 Absorption spectra of chlordiazepoxide (23.3pM) in Hepes bufler solution (1)

and in DMPC liposorne suspension afer incubation at di'erent temperatures:
(2) 4"CI(3) 7"CI(4) 23"C, (5) 2SoC, (6)30°C (7) 40°C.
3 CALCULATION OF PARTITION COEFFICIENTS
The molar partition coefficients, Kp, were calculated from the distribution results by:
where Ct is the total initial concentration of chlordiazepoxide in the aqueous buffer phase
before equilibration, C, is the final aqueous phase concentration of chlordiazepoxide, W 1
is the molar concentration of water and W2 is the molar concentration of lipid or n-octanol.
The partition coefficient is related to the standard transfer Gibbs energy, AGowjL, that is
the change in Gibbs energy when one mole of solute is transferred from water to lipid at
infinite dilution, by:
Considering the standard transfer enthalpy and standard transfer entropy constants in
the range of temperature used in this work it is possible to calculate their values by a linear
plot of the partition coefficients vs temperature, logKpvs T-'.
logK, =-+M- O W 4 AS0W+L (3)

2.303RT 2.303R
The linear plot of equation 3 enables calculation of AHow,,L from the slope and ASowjL
from the intercept. AHowjL and ASowjL have the physical meaning of the change in
enthalpy and in entropy, respectively, when one mole of solute is transferred from water to
lipid at infinite dilution.
However, the values of the change in entropy of partitioning, ASow-,L, were more
conveniently obtained from
Table 1 Effect of temperature on partition coeflcients of chlordiazepoxide in DMPC

and DPPC liposome suspensions.
(DMPC) (DPPC) (DMPC) (DPPC)

277 3.83 297 4.29
280 3.99 303 4.30
295 3.20 4.28 3 13 3.33 4.33 3.84
303 3.28 3.53 318 4.02
307 3.80 322 3.40 3.83 4.13
a) Kp (w/o) is partition coeficient in water/octanol
Partition coefficients were higher in liposomes than in n-octanol and for the liposomes they
are higher in DMPC than in DPPC but in both systems they increase with temperature
below and above T, (Table 1). From figure 2 it is also evident that the change of Kp with
temperature is similar for waterloctanol and for DMPC above the Tc, this observation can
be explained by the less-ordered structure, of DMPC, above phase transition.
As liposomes mimic better the hydrophobic part and the outer polar and negatively
charged surface of the phospholipids of natural membranes, the Kp values obtained in these
systems characterise better the drugmembrane interaction.
The values of enthalpy and entropy obtained for the three systems are positive but are
lower for DMPC liposomes. This latter result shows that the partitioning of
chlordiazepoxide depends on the rigidity of lipid bilayers and once more it can be
concluded that liposomes constitute a more selective partitioning system than does the n-
octanol.
The observed positive entropy changes arise from the loss in water structure
surrounding the drug molecules on transfer to the bilayer. The partitioning disrupts the
ordering of the bilayer and the magnitude of the effect is dependent on the structure of the
membrane phase (Table 2).
- 3.0-
2.5-
Figure 2 Log K p values vs T'forch1ordiazepoxide:a-DMPC, b-DPPC and c- n-octanol.
Analysing the enthalpy of partitioning of chlordiazepoxide in the three systems it is

evident that the least endottermic process occurs in DMPC above the phase transition
which shows that less energy is required to insert a drug molecule into its fluid bilayer
state.
All the AG W-+L values are negative, with positive AS" and AH", which show that in
liposomes and n-octanollwater systems partitioning is, as expected, entropy dominated?
The higher Kp values obtained for the DPPC liposome system are clearly a consequence of
the high standard transfer entropy obtained when the drug is transferred from water to
DPPC liposomes. The even higher Kp values obtained for DMPC liposomes are not only a
consequence of the high standard transfer entropy but also of the small standard transfer
enthalpy compared with the other two systems.
4.1 Concluding Remarks
A comparison of the K,,values determined in liposomes and determined in water/octanol

media reveals that liposomes are a different and (better) membrane model than an isotropic
two-phase solvent system, such as water-octanol. They allow for a better prediction of
drug-membrane partition coefficients, since they mimic better the hydrophobic part and the
outer, polar and negatively charged, surface of the phospholipids in natural membranes.
The analysis of the thermodynamic functions shows that partition is a process where
entropy dominates, but as is usual when there is a system with more favourable enthalpy
and entropy, the Gibbs energy of transfer (and consequently the partition coefficient) is
greater. This is observed for DMPC liposome system.
Table 2 Thermodynamics of partitioning of chlordiazepoxide in the n-octanol or

phospholipiaWepes bufler systems at pH 7.4.
T=400C AGOW+L ~ O W + L ASow+L

(KJmol-') (KJmol-') (Jmol-')
n-octanol 3.33 -19.95 15.03 111.74
DMPC 4.30 -26.01 3.54 93.70
DPPC 3.84 -23.17 52.98 243.31
References
1 A. R.James and S . D. Stanley, Biochimica et Biophysica Acta, 1980,598,392.
2 T. Tomita, M. Watanabe, T. Takahashi, K. Kumai,T. Tadakuma and T. Yasuda,

Biochimica et Biophysica Acta, 1989,978,l85.
3 A. A. Omran, K. Kitamura, S. Takegami, A. Y. El-Sayed and M. Abdel-Mottaleb,

Journal of Pharmaceutical and Biomedical Analysis, 2001,25,3 19.
4 G.V.Betageri and S.R. Dipali, J.Phurm. Pharmaco1.,1993,45,931.
5 M. Arrowsmith, J. Hadgraft and I.W.Kellaway, Biochimica et Biophysica Acta, 1983,

750, 149.
DISTRIBUTION OF VITAMIN E IN MODEL MEMBRANES
P.J. Quinn
Division of Life Sciences, King's College, London SE1 9NN, U.K.
1 INTRODUCTION
Vitamin E is found in all mammalian cell membranes where it represents a minor

component of the polar lipid fraction. Although it is classified as a fat-soluble vitamin it
possesses a hydroxyl group attached to the ring structure that confers a weakly
amphipathic character on the molecule. This physical property is believed to influence
the way vitamin E interpolates into the membrane and hence how it performs its
biochemical functions.
Despite the relatively low concentration of vitamin E compared to other
membrane lipids it is believed to play an important part in preserving the integrity of
membranes. Foremost amongst its putative hnctions is its ability to protect
polyunsaturated lipids of the lipid bilayer matrix of membranes against free radical
'
oxidation. Another important function in membranes is the formation of compIexes
between vitamin E and products of membrane lipid hydrolysis such as lysophospholipids
and free fatty acids.2 The complexes thus formed tend to stabilize membranes and
prevent the detergent-like actions of lipid hydrolytic products on the membrane.
One of the key factors in understanding how vitamin E fulfils its functions in
membranes is how it is localized in the lipid bilayer matrix and what effect its presence
has on the structure and stability of membranes. Studies with model membrane systems
containing vitamin E are described which address these questions.
2 METHOD AND RESULTS
2.1 Preparation and Characterisation of Model Membranes
The arrangement and distribution of vitamin E in model membranes has been

investigated using aqueous dispersions of mixtures of vitamin E with defined
phospholipids. The principle method used to characterise these model membranes was
synchrotron X-ray diffraction in which X-ray scattering intensity in the small-angle

(SAXS) and wide-angle (WAXS) scattering regions were recorded simultaneously during
heating and cooling scans.3 The dependence of mesophase structural parameters
observed in dispersions containing varying proportions of vitamin E was used to
determine the mode of interaction of vitamin E with the phospholipid and its distribution
in the particular mesophase. Differential scanning calorimetry was used to assess the
effect of vitamin E on the thermotropic phase behaviour of the phospholipid.
2.2 Effect of vitamin E on phospholipid phase behaviour
Examination of the thermal properties of mixed aqueous dispersions a-tocopherol

(vitamin E) with distearoyl-, dipalmitoyl-, dimyristoyl- and dilauroyl-derivatives of
phosphatidylcholine (DSPC, DPPC, DMPC,DLPC) by differential scanning calorimetry
showed that increasing proportions of a-tocopherol cause a progressive broadening of the
gel to liquid-crystalline phase transition and a decrease of enthalpy change of the
tran~ition.~'~ Similar results were obtained by Fourier transform infrared and Raman
spectroscopy.lo Proportions of a-tocopherol greater than about 20mol% resulted in
almost complete loss of transition enthalpy.
A consistent observation in the thermal studies of mixtures of a-tocopherol with
saturated phosphatidylcholines is that the presence of a-tocopherol appears to eliminate
the pretransition enthalpy. This does not mean that the ripple structure itself is
eliminated. Evidence from synchrotron X-ray diffraction and fieeze-fracture electron
microscopy have shown that different ripple structures are induced in mixed aqueous
dispersions of a-tocopherol and PC." Figure 1 shows the SAXS patterns of
codispersions of DSPC containing up to lOmol% a-tocopherol at 25OC. It can be seen
0 0.12 0.24 0.36 0.48 0.60

S(nm-'
Figure 1 SAXS intensity patterns recorded from DSPC containing different amounts of
a-tocophe rol.
that the number and intensity of the additional peaks increase with increasing proportions
of a-tocopherol, suggesting that these peaks originate from a-tocopherol enriched
domains. Together with freeze-fracture electron microscopy it was concluded that one
effect of a-tocopherol on phosphatidylcholine was to form disordered ripples of large
periodicity (about 50-150nm). When the proportion of a-tocopherol is increased the
ripples of large periodicity are replaced by ordered ripples of a periodicity of 16nm,
which produce at least 6 reflections in the SAXS region (see Fig. 1). Comparison of
enthalpy changes and SAXS intensity values of DPPC containing varying proportions of
vitamin E indicate that a complex of stoichiometry 1 vitamin E:10 phospholipid
molecules forms in both gel and fluid phases in equilibrium with domains of pure
phospholipid.'
Figure 2 Partial phase diagrams of aqueous dispersions of saturated

phosphatidylcholines and a-tocopherol over temperature ranges about the gel to liquid-
crystal phase boundary. L, lamellar liquid-crystal; Lp; lamellar gel; Pp; ripple phase; *,
phases enriched in a-tocopherol.
Partial phase diagrams of vitamin E and saturated phosphatidylcholines dispersed

in excess water have been constructed from calorimetric and X-ray scattering data. These
are presented in Figure 2.
II Proteins, Lipids and Their Interactions 25 I
Unlike bilayer-forming phospholipids, the effect of a-tocopherol on the phase

behaviour of nonbilayer forming lipids appears to be different. X-ray diffraction
experiments have given strong evidence of phase separation in codispersion of a-
tocopherol and phosphatidylethanolamines.'2~'4There are three interesting aspects of the
a-tocopherol enriched domain formed in the mixture of a-tocopherol and
phosphatidylethanolamine. Firstly, they form a lamellar crystal phase during low
temperature equilibrium. Secondly, they form an HIIphase at temperatures much lower
than that of the La to HI' transition of the pure PE, even below the main transition of the
pure PE. Finally, they can form Pn3m cubic phases at higher temperatures. Increasing
the hydrocarbon chain length favours the formation of the lamellar crystal phase, but not
the formation of the Pn3m cubic phase. Moreover, the cubic phase was only observed in
mixtures of PE containing a-tocopherol between 5 and 20mol%. In the mixture
containing 20mol% a-tocopherol, however, only the HI[ phase was observed, which is
independent of the hydrocarbon chain length of the phospholipid.
2.2 Specific Interactions Between Vitamin E and Phospholipids
The effect of a-tocopherol on the phase behaviour of mixed diacylphosphatidylcholines

to determine whether any preferential interactions take place has been reported. Ortiz et
al have studied the effect of a-tocopherol on the equimolar mixtures of DPPC/DSPC
and DMPUDSPC. The equimolar DPPC/DSPC mixture only showed a single
endotherm, which was modified by the presence of a-tocopherol in a similar manner to
the individual PCs. The equimolar DMPUDSPC mixture, however, showed monotectic
behaviour. With increasing proportions of a-tocopherol the highest temperature
endotherm was broadened and shifted towards lower temperatures, while the lowest
temperature endotherm was weakened and disappeared in the mixture containing
20mol% a-tocopherol. This was interpreted such that a-tocopherol preferentially affects
the lower melting component which corresponds to DMPC. Increasing proportions of a-
tocopherol in an equimolar mixture of 18:0/18:1- and 18:0/22:6-PC also results in a two
component transition enthalpy. The lower temperature transition is progressively
broadened as the a-tocopherol content of the mixture increases but the higher
temperature component is virtually unaffected by the presence of up to lOmol% a-
tocopherol. This was interpreted as a phase separation of a-tocopherol within the mixed
phospholipid dispersion with a preference for an association with the more unsaturated
molecular species of phospholipid.
The inverted hexagonal phase induced by a-tocopherol in PEs can be used as a
marker for (PE-ta-tocopherol) domain in PC/PE r n i ~ t u r e s . ' ~X-ray
" ~ diffraction study of
mixtures of a-tocopherol with DOPC and DOPE showed that the effect of a-tocopherol
on unsaturated phospholipids follows a similar pattern to that observed for their saturated
counterparts. However, in the X-ray diffraction study of mixed aqueous dispersions of a-
tocopherol with DOPE/DOPC (1 :1) and DOPE/DMPC (1 :l), it was found that lamellar
gel and liquid-crystalline phases dominated the phase structure. This is consistent with
the fact that a-tocopherol does not preferentially interact with PE. The changes in the
SAXS patterns are similar to those of mixtures of a-tocopherol and phosphatidylcholines
or of equimolar mixtures of PE/PC.16 This again suggests that a-tocopherol
25 2 Biophysical Chemistry: Membranes and Proteins
preferentially interacts with the PC in the mixture or distributes randomly in domains

containing of both PE and PC regardless of the saturation and the length of hydrocarbon
chains of the two phospholipids.
The effect of a-tocopherol on equimolar mixtures of saturated PE and PC has
been investigated using DSC* and X-ray diffra~tion.'~ The DSC results were interpreted
that a-tocopherol preferentially partitioned in the most fluid phase irrespective of
whether this was PE or PC. However, the conclusions drawn from the X-ray data appears
to be different.17 The change in the WAXS intensity of the sharp peak at 0.43nm to
temperature was plotted in Figure 3a of the equimolar mixtures of DMPC and DPPE
containing up to 20mol% a-tocopherol. The curves show maxima corresponding to the
midpoint of the phase transition of the DMPC (Peakl) and DPPE (Peak2) components of
the mixture. The ratio in height, Peak2:Peakl, is found to increase with increasing a-
tocopherol in the mixture. This is consistent with a preferential partition of ct-tocopherol
into DMPC so as to reduce the contribution of change in the intensity of the wide-angle
Peakl Rak2
10 20 30 40 50 60 70 80
Temperature("C)
DLPE/DSPC ( 1 1 )
0%
2 5%
5%
2c%
15 25 35 45 55 65 75
Temperature(%)
Figure 3 Rate of change of normalised X-ray scattering intensiv, I, of the WAXS

peak (dI/dJ recordedfrom heating scans at 2'/min of codispersions of DMPCiDPPE (a)
and DLPE/DSPC (b) containing indicated mol% a-tocopherol plotted as a function of
temperature. Inset shows aplot of the relative heights, Peak2:Peakl, plotted as a
function of a-tocopherol in the mixture.
reflection due to DMPC relative to DPPE. A similar analysis of the WAXS intensity data
of mixtures of DLPE and DSPC containing different proportions of a-tocopherol are
presented in Figure 3b. The inset to the figure shows the relationship between the
relative heights of Peak 2 and 1 and the mol% a-tocopherol in the mixture. This shows
that as the proportion of a-tocopherol in the mixture increases the contribution to the
change in scattering intensity of the WAXS peak from DSPC decreases relative to that
from DLPE. This can be interpreted as a preferential partitioning of a-tocopherol into
the high melting point phospholipid component of the mixture, which is again
pho sphatidylcholine, DSPC .
3 CONCLUSIONS
Present evidence indicates that, although it is found in relatively minor proportions in

biological membranes, a-tocopherol segregates in membranes and forms complexes with
specific lipid constituents, for example, phosphatidylcholines. These interactions have
the effect of stabilizing the lipid bilayer matrix since a-tocopherol is prevented from
destabilizing non-bilayer forming lipids. The wider significance of these preferential
interactions, especially in regard to the putative antioxidant function of a-tocopherol,
remains to be established. Moreover, where the distribution of phospholipid classes
across membranes is asymmetric the resultant asymmetric distribution of vitamin E may
also have implications in the function of the vitamin.
References
1 E.B. Burlakova, S.A. Krashakov N.G. Khrapova, Membr. Cell Biol. 1998,12,173.
2 V.E. Kagan, Anna1 N.Y. Acad. Sc., 1989,570,121.
3 P.J. Quinn, J. Appl. Cryst., 1997,30,733.
4 B. DeKruijff, P.W.M., Van Dijck, R.A. Demel, A. Schuijff, F. Brants, and L.L.M.
van Deenen, Biochim. Biophys. Acta, 1974,356,l.
5 J.B. Massey, H.S. She and H.J. Pownall, Biochem. Biophys. Res. Commun. 1982,
106,842.
6 E.J. McMurchie. and G.H. McIntosh, J. Nutr. Sci. Vitaminol. 1982, 32, 551.
7 J. Villalain, F.J. Aranda, and J.C. G6mez-Fernhndez, Eur. J. Biochem. 1986, 158,
141.
8 A. Ortiz, F.J. Aranda, J.C. Gomez-Fernhdez, Biochim. Biophys. Acta, 1987, 898,
214.
9 P.J. Quinn, Eur. J. Biochem. 1995,233,916.
10 T, Lefevre, and M. Picquart, Biospectroscopy, 1996,2,391.
11 X. Wang K. Semmler, W. Richter and P.J. Quinn, Arch. Biochem. Biophys., 2000,
377,304.
12 X. Wang and P.J. Quinn, Prog, Lipid Res., 1999,38,309.
13 X. Wang and P.J. Quinn, Biophys. Chem., 1999,80,93.
14 X. Wang, H. Takahashi, I. Hatta and P.J. Quinn, Biochim. Biophys. Acta, 1999,
1418,335.
15 W. Stillwell, T. Dallman, A.C. Dumaual, F.T. Crump and L.J. Jenski,
Biochemistry, 1996,35,13353.
16 X. Wang and P.J. Quinn, Biochim. Biophys. Acta, 2000, 1509, 361.
17 X. Wang and P.J. Quinn, Eur. J. Biochem., 2000,267, 1.
THEORY ON OPENING-UP OF LIPOSOMAL MEMBRANES BY ADSORPTION
OF TALIN
Y. Suezaki
Physics Laboratory, Department of General Education, Saga Medical School, Saga

8498501 Japan. E-mail: suezaki@post.saga-med.ac.jp
1 INTRODUCTION
The lipid bilayer membrane is a good model system of biological cell membranes and
has been studied extensively. Lipid membranes in aqueous solution form various
morphologies such as spherical vesicles, tubules, lamellar structures and others
depending on the materials, preparations, and other factors. By mixing with another
kind of lipid or additive molecules, a wide variety of fiuther morphologies of
membrane are found.
Saitoh and others showed that spherical lipid vesicles opened up to cup-like
vesicles and other vesicles with more complicated topologies by adding protein'. The
protein, talin, was adsorbed by the orifices of the cup-like vesicles. Figure 1 shows the
concentration dependence of the change of the shape of cup-like vesicles. In Figure 1,
an increase of concentration of talin from A to G, makes a cup-like vesicle open up
more and more, and a reduction of the concentration, G to L, recovers the original
spherical vesicle reversibly. The reversibility as a function of concentration shows that
these vesicles are thermodynamically stable at least for the time scale of the observation
of the system. Other vesicles with more than single orifices were observed at higher
concentrations. Sheet-like vesicles as shown in H in Figure 1 were also observed.
Above a threshold concentration cup-like vesicles appeared, and cup-like vesicles and
sheet-like vesicles coexist above a certain concentration.
Figure 1 A sequence of morphological change of vesicles at talin concentrations from

0 to 2 pM (A to H);AJier talin dilution (H to L) the spherical vesicle is recovered'
Later, we analyzed the adsorption isotherm of talin between aqueous solution

and the vesicle membrane by taking account of the bending energy of the membrane
and the line tension induced by talin at the periphery of cup-like vesicles2.The shape of
cup-like vesicles was determined by the balance of the energy gain for the adsorption of
talin to the periphery of the cup-like vesicles and the change of the bending energy of
the lipid membrane. Observed coexistence of cups and sheet-like vesicles was
reproduced. Vesicles with two orifices were also analyzed and theoretically reproduced.
In this paper, the above theoretical analysis will be improved and further analysis will
be done. First, in the next section the effect of Gaussian bending modulus will be
discussed. Next, the theory will be improved to be more consistent with observed facts
than the previous paper. Analysis of concentration dependence in the neighborhood of
the threshold concentration will show that very small orifices begin to form just above
the threshold concentration, while previous theory predicted a frnite size of cup-like
vesicle at that concentration. Lastly, in the previous work2, we modeled the cup-like
vesicle as a partial sphere cut by a plane. However, this model does not satis@ the
mechanical force balance and mimics realistic shapes only in an approximate way. In
the section to the following, we will show the result of Monte Car10 simulation to
determine the shape of cup-like vesicles and vesicles with two orifices. The resulting
shapes resemble those observed in experiment.
2 THEORETICAL ANALYSIS
2.1 Bending energy of membrane
Here, we briefly review our previous theory and replace it with an improved version.
According to Helfiich’s model for the bending energy of membranes3,the general forms
of the bending energy, that of a sphere, and that of a cup, &end , and Ecup ,
respectively, are given as follows
K
Ecup= - l ( c x + c,, - c o y dS = 2 n ~ ( 2 -
x coR,)* + 27r~’(1 + cos6) (3)
2
where K, d,cX, c,, and co are the bending modulus of the membrane, Gaussian bending
modulus, two principal curvatures and the spontaneous curvature, respectively. Also, Ro
and x(=RdR, R is the radius of curvature of cup-like vesicle) and Bare the radius of the
initial spherical vesicle, the relative curvature of the cup with respect to sphere, and the
angle between the normal to the axis of the cup and the tangent of the cup periphery as
shown in Figure 2. Then the bending energies of a spherical vesicle and cup-like vesicle
follow the forms of Eqs. (2) and (3). The spontaneous curvature of the bilayer
membrane is usually assumed to be zero. However, this idea is based on the assumption
that the membrane system is in equilibrium. Vesicles were not necessarily in
equilibrium but were made by mechanical agitation (sonication) or were under some
specific boundary conditions or initial condition of the system. As the turnover time of a
lipid molecule fi-om one monolayer to the other is very long, artificially made vesicles
could initially possess non-zero spontaneous curvatures. The last terms of Eqs. (l), (2),
and (3) had not been considered in the previous paper.
Figure 2 Schematic drawing of a cup-like vesicle. The cup is mi-symmetric with the
z-axis.
Now, we will discuss the problem of the Gaussian bending modulus. For a thin elastic
plate, the value of the Gaussian bending modulus, K ’ should be the same order of
magnitude as K. For liquid membranes, however, Peliti and Nelson showed that the
Gaussian bending modulus should be zero from the shear fiee condition of a liquid
membrane4. This result is reasonable because the Gaussian bending modulus itself is the
modulus for shear deformation in the membrane surface. For real membranes, there
could exist frustrated stresses along the perpendicular direction to the surface. In this
case, a frnite value of the Gaussian bending modulus could exist. To simulate the
frustration, the author analyzed the surface elasticity model devised by Petrov and
others’. By inserting an intermediate surface to their dumbbell surface elasticity model,
we could reproduce a frustrated liquid membrane6. By analyzing this model correctly,
we could show that the Gaussian bending modulus can take both positive and negative
values depending on the hstration of the membrane inside, while that of the elastic
plate should always be negative. However, the value of the Gaussian bending modulus
is two orders of magnitude smaller than that of the usual bending modulus6.
Some experimental workers have tried to evaluate the Gaussian bending
modulus indire~tly’~~. However, there has been no direct evidence of the quantitative
observation of a Gaussian bending modulus so far. For lipid membranes, the magnitude
of the Gaussian bending modulus is expected to be small. This is due to the fact that the
frustration of a membrane is not large because the cross sectional area of the head group
is not much different from that of two a w l chains. Actually, cell fusion and fission
often occur in biological cell systems. This might be due to the fact that the energy gain
or loss during the hsion or fission is small. To explain, for instance, bi-continuous
structures seen in microemulsions or lipid vesicles, the role of Gaussian bending energy
term has been considered’. It might be possible where delicate and small energy balance
works in the conformational change of the system. For our fiuther analysis in this report,
therefore, we will neglect the term of Gaussian bending modulus.
2.2 Statistical mechanics of adsorption of talin
The actual shape of the cup is a little different from our model, but we employ the
partial sphere for cup as a first approximation. To evaluate the adsorption equilibrium
of talin, we define the total number, X,and the adsorbed number, N , of talin to orifices
of the vesicles. The partition function 2 of the total system is written as
Z=Za@)Zbu&-N) and respective partition h c t i o n s are represented as
where, @&lk and are entropic parts of the molecular partition h c t i o n s of

dissolved talin in the aqueous solution and an adsorbed talin, respectively. We assumed,
a,as an affinity fiee energy of talin from water to the orifice of the vesicle. By
defmition, the fiee energy, F, of the total system, becomes F=-RT In 2. In Eq. (4), the
factor no is the total number of original spherical vesicles and n is the number of
cup-like vesicles. By minimizing F with respect to N and n, we obtain the thermal
equilibrium and the mechanical force balance. The equations, d F / d N and d F / d n
determine the chemical potential balance of talin and a number, n, of cups as follows
where
&,,lk ?),
= kT In( X - N pd = - E ~- kT In(ed 3 ) + 8nnK(2x - c&)- ax
aN
(7)
and
I! Proteins, Lipids and Their Interactions 259
l+Roco - x * --- (8)

X 81cK
From Eq. (7), we derived a threshold concentration of talin. Eq. (8) determines the
number n, in other words, it shows the mechanical force balance of cup-like vesicles.
By our model calculation the observed shape change of vesicles was well reproduced,
and the observed coexistence of cup and sheet-like vesicle' was also reproduced2.In that
case, however, we should assume a finite spontaneous curvature of membranes to solve a
non-trivial solution. Also, we expected a finite size for the orifices of cups at a threshold
concentration. Experimentally, however, very small orifices appeared at just above
threshold concentration'. In this report, we improve and reanalyze our previous theory
that is not consistent with observed data.
By analyzing the concentration dependence of Eq. (8) carefully, we obtained an
improved result. A factor ndn in Eq. (8) becomes as follows
where CL,CX,and C t h are concentrations of lipid and talin and threshold concentration,
respectively. By carefully evaluating the right-hand side of Eq. (8) by use of Eq. (9), the
relative curvature, x, is determined as a cross section as shown schematically in Figure
3 for concentration CX that is close to the threshold concentration Cfh. Figure 3 is
different from Figure 2 in Ref. [23.
The minimized fkee energy per molecule, f, scaled by kT becomes
The value of Eq. (1 0) is always negative and thus the formation of cup-like vesicles has
been shown to be stable3. Here, stable formation of cup-like vesicle, in reality, is
metastable due to the fact that spherical vesicles themselves are metastable and realized
only in the finite time scale of experiments. Observations' were made around several
tens of minutes and many colloidal systems suffer from change by the aging effect as is
well known.
Figure 3 Schematic drawing of lep and right hand sides of Eq. (7) as finction of x
The energy profile of Eq. (10) is a sigmoidal curve with an inflection point.
Although we have not shown it here, the energy calculation of sheet-like vesicles with
plane ellipses showed that a common tangent exists for both energy profiles of cup-like
vesicles and plane elliptical vesicles2. This fact shows that there exists coexistence of
cup-like vesicles and sheet-like vesicles. This coexistence was also observed in
experiment made by Saitoh and others'.
3. THE MONTE CARL0 SIMULATION
3.1 The methodology of the Monte Carlo simulation
The model for cup-like vesicles in the previous section could explain the observed
adsorption isotherm qualitatively. However, the partial sphere model does not satisfy
the local mechanical force balance within the vesicle. Actually, observed shapes of
various vesicles are not a partial sphere, but deformed significantly from a spherical
shape'. To reproduce realistic shapes of vesicles, we performed Monte Carlo simulation
at equilibrium. We assume that the deformation proceeds by keeping axial symmetry
around the z-axis as shown in Figure 4. Then the vesicle shape is described by the
line contour 4, where 6 is the angle made by the contour's normal and the z-axis. In the
Monte Carlo simulation, we discretize the vesicle contour into joint segments. The i-th
joint is described by the distance ri from the z-axis and the coordinate Zi as shown in
Figure 4. Two consecutive joints i and i+l are connected by a segment of length si.
Local self-avoidance can be achieved by confining d 2 < Si < 2 a , where a is a unit of

length. The procedure of the Monte Carlo simulation is similar to that made by
Morikawa and others".
Figure 4 Schema of the joint-segment model of a cup-like vesicle. The Bi is the angle
variable between the segment i and r-axis, and S i is the segment length between the
joint i and i + l .
The model energy, E, of the Monte Carlo simulation is written down as
E =q ( C , + c , - c*)2ct4+ 2 n r ( l ) &( + 2 A r ( N ) & ) (11 1

2
where E is the line tension energy of the orifice of the cup-like vesicle. The factors
r(1) and r(N) are the radii of the orifice of the cup and that of the other orifice when the
vesicle has two orifices. The factor, N is the number of joints, which changes during the
course of the simulation. In the Monte Carlo simulation, the first term of Eq. (1 1) is
replaced by
where
(14)
7r
uj = -[(ri-l + 3ri)~i-l+ (3ri + rj+l ) ~ j ]
4
In Eq. (14), 6 is the defined as
The line tension energy in Eq. (1 1) is represented as
EL = 2 q ~ (+2rrN&when two orifices)
The total energy E is written as
E = E~ + E , + Y ( A - A,)*
where A and A0 are the total area and the initial total area, and y is the Lagrange
multiplier to conserve the surface area of the vesicle.
The Monte Carlo simulation consists of the following sequences of procedures. A
joint i is chosen randomly with a probability to its associated local area of Ai, and we
try to shift its coordinates randomly in the (2,r) plane according to
zi +zi + Azj
ri + ri + Ari
We calculate energy change dE associated with this position change and accept or reject
the trial according to the standard Metropolis algorithm. If dE is negative the trial is
automatically accepted, and if dE is positive it is accepted with a probability
proportional to exp(-dE /kT) at temperatwe T. In the following subsection, we will
show the results of the Monte Carlo simulations of cup-like vesicles and vesicles with
two orifices for each line tension.
3.2 Numerical result of the simulation
The preliminary result of the Monte Carlo simulation for the cup-like vesicle and
vesicle with two orifices are given in the following. The initial number, N was 100 and
200. The radius of the spontaneous curvature was assumed to be twice of that of the
original spherical vesicle as a preliminary trial of the analysis. In future work, we will
calculate the system with many conditions. In Figure 5a and 5b, we show the shapes of
cup-like vesicles with line tension energy d ~ 0 . and1 0.7 respectively. The shapes of
the cups are different from the partial sphere and well reproduce the observed shapes. In
Figure 6a and 6b, the shapes of vesicles with two orifices of the line tension energy
d ~ 0 . and
1 0.7, respectively, are shown in Figures 5 and 6.
2 r
-2
0 1 2 3
0 1 2
Figure 5 Shapes of cup-like vesicle obtained by MC simulation. a; E/F. 1,

b; d ~ O . 7
Similar vesicles with shapes of Figure 6a have also been observed'. The shapes
of cups are different fiom a partial sphere especially when the line tension is large and the
shapes by this simulation well reproduce the observed shapes of vesicles'. Although we
do not show the data of the local curvatures of respective vesicles, the sum of c1 and c2 is
increasing b c t i o n when the contour proceeds from the orifice to the bottom of the cup.
1 r
0.5
0
-0.5
-1
0 1 2 3
-2
0 1 2 3
Figure 6 Shapes of vesicles with two or$ces ouiained by. 4C simulation. a; E/F. I ,
b; d p . 7
And the sum has a maximum value at about the center in the case of vesicles with two
orifices. This is consistent with the fact that a freely sustained beam at both ends
possesses maximum curvature at the center while it is bent. In Table 1, the values of
energies Eb and EL of respective vesicles obtained by the Monte Carlo simulation are
given.
Table 1 Values of energiesfor cup-like vesicles and vesicles with two orijices
Type of vesicle &/K Eb / K EL /K (Eb + E~/K
Cup-like vesicle 0.1 0.015 1.067 1.082

0.7 0.870 5.856 6.726
Vesicles with two orifices 0.1 0.026 1.152 1.178
0.7 0.332 6.857 7.189
From Table 1, The total energies of vesicles with two orifices are larger than
those of cup-like vesicles, which is reasonable. For both vesicles, the ratio, EL /& , is
larger for vesicles with d ~ O . 1than those of d ~ 0 . 7 This
. fact shows that the line
tension and the bending of membrane are mechanically similar to the series connection
of two springs with different spring constants. The weaker spring stores more energy
than the stronger when it is pulled. More detailed analysis will be done in fhture work.
The obtained shapes of vesicles with two orifices are not symmetrical although the
symmetrical shape is expected fiom topological considerations. In this case, the
symmetric point might be a saddle point or a summit of the energy profile. Actually, the
simulation by assuming an axially symmetric shape cannot lower the total energy more
than those given in Figure 6. At this point, fbther study is left for future work.
4 CONCLUSION AND DISCUSSION
In the previous sections, we reviewed the previous theoretical paper on the formation of
cup-like vesicles, and a more reasonable discussion on the energy term of Gaussian
bending modulus was made. The previous theoretical analysis of the formation of
cup-like vesicles just above the threshold concentration was reanalyzed in more detail,
and the improved calculation showed that the calculated shape and the observed shape
of cup-like vesicles at just above that concentration came closer to the observed one
than the previous analysis*. To realize local force balance in the cup-like vesicle and
vesicles with two orifices, we performed the Monte Carlo simulation by assuming the
line tension energy and bending energy as well. Starting fiom the partial sphere, the
Monte Carlo simulation by random fluctuation of z and r resulted in the realistic vesicle
shapes and they approximately reproduced observed shapes. The obtained energies of
bending and line tension possess mechanically reasonable meaning.
Although the author believes in the physical nature of Gaussian bending
modulus as pointed out in the theoretical section, as far as biological membranes or their
model membranes are concerned, many people think that the essential nature remains
unclear. To resolve this point on the Gaussian bending modulus, it is useful to refine the
experimental accuracy of observing cup-like vesicles and vesicles with more than single
orifices. Furthermore, it is important to theoretically analyze the problem in more detail
by using our model including Gaussian bending energy. We are continuing the Monte
Carlo simulation more systematically to resolve the problem stated above. In this case,
the effect of changing the spontaneous curvature of the membrane is now being taking
into account.
Acknowledgment
The author would like to thank to Drs. Umeda and Morikawa for their useful
discussion on the problem. He also thanks Drs. Takefb and Horhoto and Mr. Ichinose
for their technical support of the numerical calculation. Mr. B. Andersen kindly assisted
the author with the English presentation, Financial support came f?om the Ministry of
Education and Science, Japan (grant # 10640372).
References
A. Saitoh, K. Takiguchi, Y. Tanaka, and H. Hotani, Proc. Natl. Acad. Sci. USA 1998,
95, 1026.
Y. Suezaki, H. Ichinose, K. Takiguchi and H. Hotani, Biohys. Chem. 1999,80,119.
W. Helfkich, Z. Naturforsch. 1973,28c, 693.
Peliti, L and Nelson, D., in Physics of Amphiphilic Layers, 1987,106,
Springer-Verlag, Berlin.
A.G. Petrov and A.J. Derzhanski, Phys. Coll., 1976,37, C3.
Y. Suezaki and H. Ichinose, J. Phys. I France, 1995,5, 1469.
A. Fogden, S.T. Hyde and G.J. Lundberg, Chem. SOC.Faruday Trans. 1991, 87
949.
B. Farago and D. Richter, Phys. Rev. Lett., 1990,65,3348.
S. Leibler, Statistical mechanics of membrane and surfaces, 1988,5 (World Science,
Singapore), 46.
10 R. Morikawa, Y. Saito and H. Hyuga, J.Phys. SOC.Japan, 1999,68,1760.
DIFFERENTIAL SCANNING CALORIMETRY AND X-RAY DIFFRACTION
STUDIES OF GLYCOLIPID MEMBRANES
0. Ces 1,J . M. Seddon1,R.H.Templer1, D. A. Mannock2 and R.N. McElhaney2
‘Department of Chemistry, Imperial College, London SW7 2AY, UK

2Department of Biochemistry, University of Alberta, Edmonton, Canada T6G 2H7
1 INTRODUCTION
Using X-ray diftiaction and differential scanning calorimetry (DSC) we have studied the
phase behaviour of 1,2-di-U-tetradecyl-3-O-~glucopyranosyl)-sn-glycerol(di-14:o-p-
D-GlcDAG) and 1,2-di-U-tetradecyl-3-O-(pgalactopyranosyl)-sn-glycerol(di-14:O-P-D-
GalDAG) as a function of temperature and water content. In this paper we present an
introduction into the observed phase behaviour of these synthetic dialkyl glycolipids.
The X-ray diffkaction data has been used to calculate the lattice repeat vector, d and the
limiting hydrations of the various phases. The behaviour of the two systems upon
heating is similar as they both adopt a fluid lamellar phase La and an inverse hexagonal
phase HII. Upon cooling however, striking differences are seen with di-14:O-P-D-
GlcDAG forming a metastable Q gel phase below the chain-melting transition in
contrast to di-14:O-j3-D-GalDAG which forms only crystalline lamellar phases on the
timescale of our experiments. Bicontinouos cubic phases are not observed in either
system. We compare the findings from these glycolipids with those fiom our previous
studies of the phospholipid didodecyl phosphatidylethanolamhe (di- 12:O-PE).
2 BACKGROUND
Glycolipids, biological amphiphiles whose polar headgroups contain one

(monosaccharide) or more (oligosaccharide) cyclic sugar groups, are vitally important
components of biological membranes, being involved in many regulatory and
recognition processes. There is increasing interest in their structural roles in membranes,
since they are the major membrane lipids of all plants (mainly galactose-based) and
many micro-organisms (mainly glucose-based)’. The reason why such species utilise
glycolipids rather than phospholipids may be to do with conserving phosphorus, which
is usually a precious commodity in the environment. Frequently, monosaccharide-based

lipids are the major glycolipid component of such membranes. The reasons why some
organisms utilise galactose-based lipids, and some glucose-based lipids, as the main
structural component of their membranes is poorly understood. It is therefore important
to establish whether there are any differences in the lyotropic phase behaviour of these
two classes of glycolipid.
If one compares the two sugars, D-glucose and D-galactose, they have rather
similar solubilities in water of 45.1 wt% and 40.6 wt% at 20°C, respectivelq. They also
have similar hydration numbers (8.4 and 8.7), but slightly different partial molar
3
volumes at infinite dilution (I 1 1.7 and 1 10.2 cm mol-'). These differences are due to
the change fiom an equatorial (glucose) to an axial (galactose) configuration of the
hydroxyl group on carbon C(4) of the sugar ring. This changes the arrangement of
hydrophilic sites around the sugar group, and will lead to differences in the pattern of
both headgroup-headgroup and headgroup-water hydrogen bonding. A further striking
effect is the large change in the distance fiom the hydroxyl oxygen atom on C(4) to the
hydroxyl oxygen on C(2), which is reduced from approximately 4.8 A for glucose to
approximately 4.4 A for galactose. There is evidence fiom compressibility studies that
galactose is less well accommodated within the structure of water, and consequently has
a stronger perturbing effect3.
When these two sugars are attached to diacyl (or dialkyl) glycerol, these
differences may be expected to lead to differences in the phase behaviour for the two
glycolipids. A well-established and important finding is that fully-hydrated
monoglycosyl diacylglycerols and dialkylglycerols, with either glucose-based or
galactose-based headgroups, have a strong tendency to adopt inverse, non-lamellar
phases such as the inverse hexagonal HI^ and bicontinuous cubic phases. This is found
both in naturally-occuming glycolipids extracted fiom biological membranes, and in
purely synthetic glycolipid systems4. Non-lamellar structures may exist within cells,
both animal (smooth endoplasmic reticulum) and plant (etiolated chloroplast^)^. The
former contain primarily phospholipids, whereas the latter contain glycolipids.
Furthermore, non-lamellar phases are topologically and geometrically related to defects
such as fusion channels between bilayers, and thus are also highly relevant to dynamic
processes in membranes.
The phase behaviour of monosaccharide-based glycolipids is strikingly similar to
that of the phosphatidylethanolamines (PE), one of the major classes of phospholipid in
animal cell membranes. For synthetic dialkyl (saturated, ether-linked) PE's it was found
that the effect of reducing the chainlength was first to raise the fluid bilayer - inverse
hexagonal (La - Hn) phase transiton temperature, and then when the chainlength was
reduced below C14, to induce inverse bicontinuous cubic phases to form, between the
La and HII phase^^'^,*.^.
This behaviour can be understood in terms ofthe chain packing hstration which
exists in the HIIphase, and which becomes progressively more severe as the chainlength
is shortened. This frustration can be partially relieved by the bilayer deforming to a
saddle surface, with negative gaussian curvature (K < 0), allowing each monolayer to
bend towards the water, but without creating voids within the hydrocarbon chain region.
For a symmetric bilayer, the two monolayers should curve in opposite directions but by
equal amounts, leading to the mid-plane having zero mean curvature (H = 0). Thus the
mid-plane should correspond to a minimal surface, which can be extended to form an
infinite periodic minimal surface (IPMS)with cubic symmetry. The three simplest of
these are the P, D and G minimal surfaces, which are related to each other by a Bonnet
transformation, a mathematical transformation which preserves the mean and gaussian
curvature at each point on the surfaces. Draping a lipid bilayer onto these three LPMS
leads to the formation of the bicontinuous cubic phases of spacegroups, Im3m, Pn3m
and Ia3d, respectively". It should be noted that the interfacial area is actually maximal
on the minimal surfhce at the bilayer midplane, and diminishes on moving towards the
polar headgroup regions. The system is still fnrstrated, because it is impossible to satistjr
both uniform interfacial mean curvature and uniform bilayer thickness. However, it has
been shown that the frustration in an inverse bicontinuous cubic phase is smaller than in
the H ~phase".
I
It was subsequently found that the above-mentioned effects of chainlength also
applied to glycolipids, which was to be expected since the underlying physical
mechanisms are rather general in nature, and not specific to phospholipid systems. In
fact the glycolipids have a stronger tendency to form inverse phases than the
phospholipid PE system, the La- HI[ phase transition in excess water occurring at 55 -
56 "C and 62 - 63 "C for di-14:O-~-D-GlcDAG'** l 3 and di-14:O-P-GalDAGI4, but at 96
"C for the di-14:0-PE6. Changing the chain linkage from ether to ester is sufficient, for
the di-14:O glycolipids, to cause the appearance of inverse bicontinuous cubic phases
above the La phase, but at higher temperatures of 72 "C for the ester-linked GlcI5and 8 1
"C for the ester-linked Gall6, compared to the La - HI^ transition temperatures of the
corresponding dialkyl compounds.
Many membrane lipids contain chiral centres, and so it is natural to ask whether
lipid membranes exhibit chiral recognition, or whether the molecular chirality manifests
itself in any detectable way, in terms of the membrane structure or interactions. Very
little evidence of chiral discrimination has so fhr been found in phospholipid bilayers,
where the molecules are usually enantiomeric, with a single chiral centre at carbon atom
C2 in the glycerol backbone. For glycolipids, chirality may be expected to play a greater
role than in phospholipids, since the 1,2-sn and 2,3-sn glycolipid diastereomers are not
mirror images of one another and so can have different physical properties. Indeed,
studies in excess water of ether-linked di-l2:O-P-D-GlcDAG found significant
differences in phase behaviour between the 1,2-sn and 2,3-sn For the longer
chainlength P-D-glucosyl compounds, however, the effects of chirality wefe very small.
For galactose-based lipids, on the other hand, the effects of chirality ate still quite
marked for the 14 carbon r ha in length'^.'^, and ate very striking for the di-l2:O and di-
I 3 :O compounds*'.
We have been interested in comparing the detailed phase behaviour of glucose-
and galactose-based glycolipids as models for glycolipid membranes. We have chosen to
study the dialkyl compounds because the ether linkages (of the saturated hydrocarbon
chains to the glycerol group) confer enhanced chemical stability compared to the ester-
linked compounds. The particular homologues we chose for detailed study are di-14:O-
270 Biophysical Chemistty: Membranes and Proteins
P-D-GlcDAG and di-l4:O-P-D-GalDAG which are diastereomers having the natural 1,2-
sn glyceroI configuration.
Ho
u-u GAL
Figure 1 Chemical structures of the synthetic dialkyl glycolipids di-14:O-j3D-GlcDAG

and di-i 4:0-PD-GalDAG.
This chainlength of C14 is just long enough to suppress the formation of any
cubic phases. For the corresponding PE phospholipid system, we have previously found
that the effect of hydrostatic pressure is to induce the formation of inverse bicontinuous
cubic phases (of spacegroups Pn3m, and Im3m) between the fluid lamellar La and HII
phases for pressures in excess of 600 ba?'. We plan to extend these high-pressure
structural studies to the di-14:O-P-D-GlcDAG and di-l4:O-P-D-GalDAG glycolipid
systems, to determine whether this pressure-induced formation of inverse bicontinuous
cubic phases, is a universal effect for systems which exhibit such phases at atmospheric
pressure at slightly shorter chainlengths. However, the first stage is to establish the
binary temperature-composition phase diagrams for these two compounds at
atmospheric pressure, and some of our preliminary findings are reported here.
3 EXPERIMENTAL
The lipids 1,2-di-O-tetradecyl-3-O#glucopyranosyl)-sn-glycerol (di-l4:O-P-D-

GlcDAG) and 1,2di-U-tetradecyl-3-O-(pgalactopyranosyl)-sn-glyceroI(di- 149-P-D-
GalDAG) were synthesized according to literature methods which have been previously
described22.The compounds were pure as determined by TLC, using a solvent system of
CHCL$CH30H (9:l v/v). Differential scanning calorimetry measurements were
performed on a TA Instruments DSC 2910 equipped with a data station or a Perkin-
Elmer DSC7 calorimeter. Heating and cooling scans were typically repeated three times
to ensure homogeneity of the samples and were recorded with heating-/-cooling rates of
0.5 and 1 "C per minute. Fixed hydration samples were made by weighing out exact
amounts of lipid and HPLC grade water into stainless steel DSC pans which were
hermetically sealed. The uncertainty in the concentration Ac/c is estimated to be I wt%.
Low and wide angle X-ray patterns were obtained using two different systems,
one based on a high intensity point source and the other around a line source. In the
former case the X-rays were produced by a Philips PW 221 3/20 generator operated at 40
kV and 30 mA with a fine focus tube with a copper target. The diffraction patterns were
obtained using a Guinier camera (Huber Difiaktionstechnik) which was equipped with a
quartz crystal monochromator set to isolate the CuK,t (h=l S405 A) radiation. This was
operated under vacuum to reduce air scatter with a fixed sampledetector distance of
114.6 mm. The samples were held in position by an electrically heated copper block
which was able to control the sample temperature to an accuracy o f f 0.5 "C by
employing an electronic controller. The temperature probe was embedded as close as
possible to the sample (within 1 cm). The electronic controller allowed the temperature
of the sample to be varied linearly with time whilst the X-ray film holder was scanned
vertically thereby producing a continuous dieaction pattern. Kodak Scientific Imaging
Film was used and developed according to standard techniques. The scanned films were
then analysed using software written within the IDL 5.4 (Research Systems) data
processing package. The estimated accuracy of the X-ray spacings is k0.5 A (low-angle)
and k0.05 8, (wide-angle).
The X-ray point source was provided by an Elliot GX20 rotating anode X-ray
generator operating at 30 kV and 25 mA, with a 100 pm focus cap, and was focused by
Franks optics to a point of dimensions 160 x 110 p2. The diffiction patterns were
captured by a custom-built electronic two-dimensional CCD detector. Photometrically
accurate images could be captured in a matter of minutes and then analysed and indexed
fiom a single computer terminal. The estimated accuracy of the layer spacing is *0.5 A.
To determine the phase behaviour of both compounds significant numbers of fixed

hydration samples were prepared with the percentage of water ranging from 0 wt% to 80
wt% at intervals of 1-2 wt%. DSC scans were measured for each of these samples.
Figure 2 shows representative scans for the dry samples and samples containing excess
water.
DSC measurements conducted upon a dry sample of the di-14:O-P-D-GlcDAG
compound show a single highly energetic phase transition at 54.5 "C upon heating. In
the case of the excess water sample two peaks are instead observed corresponding to a
highly energetic phase transition at 53.2 "C and a low-enthalpy transition at 59.5 "C. The
cooling scans indicate that these transitions are indeed reversible.
The behaviour of the di-14:O-P-D-GalDAG on the other hand is somewhat
different. On heating, samples at all water contents exhibit a single phase transition
which progressively drops in temperature with increasing hydration from 78.9 "C in the
case of the dry sample to a limiting value of 70.2 "C. Cooling scans revealed that this
behaviour is not reversible indicating that the Gal compound forms metastable phases.
The single exothermic phase transition observed in the cooling scans of the dry sample
at 64 "C is replaced in samples containing excess water by a poorly exothermic phase

transition at 6 1.5 "C, and a highly exothermic phase transition at 54.2 "C.
The thermodynamic data are summarised in Table 1 .
Sample Heating Phase Transition Enthalpy AH Entropy AS

Sequence Temperature("C) (kl mol-') (Jmol'K-')
G lc (Dry) Heating 54.5 43.0 131.3
Cooling 53.2 42.6 130.6
Glc (Excess Water) Heating 53.2 21.6 84.6
59.5 6.9 20.8
Cooling 55.7 7.3 22.2
49.7 27.8 86.2
Gal (Dry) Heating 78.9 80.9 229.9
Cooling 64 .O 63.3 187.8
Gal (Excess Water) Heating 70.2 86.6 252.3
Cooling 61.5 6.2 18.5
Table 1 Thermodynamic data (iransition temperatures, enthalpies and

entropies) relating to the DSC scans shown in figure 2.
I DV Sample I Excess Water Sample 1
Ti -c
Dry Sample I Excess Water Sample

I
TI 'c TI 'C
Figure 2 DSC scans of (a) di-14:O-PD-GlcDAG and (5) di-I 4:0-ED-GalDAG.

In order to understand the DSC results, the two glycolipids were studied by X-
ray dieaction. In addition to establishing the structures and symmetries of the phases,
these measurements also yielded the lattice parameters of each phase. Fig. 3 shows
representative low-angle X-ray patterns f b m the various lyotropic phases observed for
these two glycolipids.
The stable low temperature phase of the di-14:O-P-D-GlcDAG compound is a
lamellar crystalline phase (Lc~).Four orders of diffhction are visible in the low-angle
region, their difhcted orders being in the ratio 1:2:3:4. The interlamellar repeat distance
d of the Lcl phase is 52.581 and two strong reflections are seen in the wide-region at 4.6
and 3.90 A which demonstrate a highly ordered arrangement of the hydrocarbon chains.
The structural characteristics of this phase do not change upon heating. After heating
above the chain-melting transition the low-temperature phase becomes an untilted Lp
lamellar gel phase. The two peaks at 4.6 and 3.90 A are replaced by a single, quite sharp
symmetrical wide-angle peak at 4.1 5%1along with an increase in the layer spacing which
rises to 55.7 8, in excess water. At low hydrations, the chain-melting transition occurs
directly to the inverse hexagonal HI^ phase, whereas in excess water the transition at 53.2
"C is to the fluid lamellar La phase (d = 5 1.8 A), followed by a hrther transition at 59.5
O C to the HJIphase (lattice parameter a = 67.8 A).
The behaviour of the di-14:O-P-D-GalDAG compound is rather different. At low

temperatures, depending on the sample history, it can form either the Lcl phase, with a
layer spacing of 54.0 8, and strong wide-angle peaks at 4.54, 3.87 and 3.50 A, or a
second tilted lamellar crystalline phase L c ~with
, a smaller layer spacing of 47.2 8, and a
very different wide-angle pattern, containing numerous sharp peaks (Fig.3 (e)). It has not
yet been possible to identi@the packing modes tkom these wide-angle powder patterns,
but it has been suggested that monoclinic chain subcells may be involved14. On the other
hand, Hinz and co-workers suggest that the wide-angle patterns indicate hybrid
orthorhombic / triclinic subcell packing What does seem clear h m both our
data and these earlier results is that the packing within the Lcl phase is essentially the
same for both the Glc and the Gal compounds. The wide-angle difiaction pattern we
observe fmm the Lc2 phase appears to be different fiom any previously reported,
possibly indicating the formation of a new lamellar crystalline phase, with a different
subcell packing, not previously seen. At all hydrations, the lamellar crystal phase melts
directly to the HIIphase upon heating, generally fiom the LCZphase at low hydration (at
78.9 "C), but from the LO phase in excess water (at 70.2 "C). The lattice parameter of
the HI]phase in excess water is 62.3 A.
On the timescale of our experiments, we see no evidence for gel phase formation
in the di-l4:O-P-D-GalDAG compound. However, previous X-ray experiments did
observe a transient Lp lamellar gel phase, which reverted rapidly (less than one hour) to
a lamellar crystalline p h a ~ e ' ~The
? ~ ~rate
. of reversion fkom the gel phase to the more
ordered crystalline lamellar phases is known to be faster for P-D-Gal than for the j3-D-
Glc lipids, and also to depend on the chirality of the glycerol backbone24.Generally
speaking, the rate is fastest for the 1,2-sn compounds, intermediate for the racemate, and
slowest for the 2,3-sn compounds. The explanation for these differences is still unclear,
but must lie in the different patterns of headgroup-headgroup and headgroup-water
hydrogen bonding.
I
RaiprocPl Spacing
RecproCa Spacing
Reciprocal Spacing
Figure 3 X-ray powder dtflaction patterns fiom the various phases of di-14:O-PD-
GlcDAG (a: LCI; b: L s c: HII), and di-14:O-PD-GalDAG (d: Lcl; e: L c ~ 8; HI$ in
excess water.
For hydration levels in excess of 20 wt% water, we observe a fluid lamellar La

phase, for di-14:O-P-D-GalDAG upon cooling, at 61.5 "C, followed by atransition to the
lamellar crystal LCIphase at 54.2 O C . We infer that not only the Lp gel phase, but also
the fluid lamellar La phase is metastable for this lipid, although this has not as yet been
directly observed.
By plotting the lattice parameters as a function ofwater content (data not shown),
we have determined the limiting hydration of each phase. We find the following values
for the Lp gel, La fluid bilayer, and HII phases (these are nearly independent of
temperature within each phase):
System J-B L, Hi1
di-14 :O-P -DGIcDAG 7 13 21
di-14:O-fl-DGalDAC - 11 19
di-12 :0-PE' 6 12 16
'Reference [7]
Table 2 The limiting hydrations (waters per lipid molecule) of the Lpge1,fruid lamellar
L , and inverse hexagonal HI] phases.
The limiting hydrations in the gel phase are similar to the hydration numbers of 8.4 and
8.7 for fiee glucose and h e galactose in solution3. Our results are in quite good
agreement with the hydration values for dialkyl glycolipids inferred indirectly by Hinz
~ ~ , demonstrate that the limiting hydrations of the three phases for
and c o - w o r k e r ~and
the glycolipid systems are rather similar to those established earlier for the same phases
of the phospholipid system didodecyl phosphatidylethanolamine (di-1 2:O-PE)7. It has
been established that, as in the case of PE, the lamellar crystalline phases of the
glycolipids are essentially anhydrous (less than one water per lipid23).
The binary tempemtmwomposition phase diagrams of these two synthetic
glycolipid systems will be published at a later date.
Acknowledgments
This work was supported by grant GRL48065 fiom the EPSRC (UK)and by a grant
from the Canadian Institute of Health Research.
References
1 I. Ishizuka and T. Yamakawa, "Glycoglycerolipids", in Glycolipids, ed H.

Wiegandt (Elsevier, Amsterdam), 1985, 101.
2 H. Stephen, Solubilities ofInorganic and Organic Compounds, Pergamon Press,
1963.
3 S. A. Galema and H. Hoiland, J. Phys. Chem, 1991,95,532 1.
4 R.N. A. H. Lewis, D. A. Mannock and R. N. McElhaney, in Lipid Polymorphism
andMembrane Properties, ed. R. M. Epand, 1997,25.
5 Y. R. Deng and T. Landh, Zoo1 Stud, 1995,34 (l), 175.
6 J. M.Seddon, G. Cevc and D. Marsh, Biochemistry, 1983,22,1280.
7 J. M. Seddon, G. Cevc, R. D. Kaye and D. Marsh, Biochemistv, 1984,23,2634.
8 J. M. Seddon, J. L. Hogan, N. A. Warrender and E. Pebay-Peyroula, Prog. Colloid
Polyrn. Sci., 1990, 81, 189.
9 R. N. A. H. Lewis, D. A. Mannock, R. N. McElhaney, D. C.Turner and S. M.
Gruner, Biochemistry, 1 989,28,54 1.
10 J. M. Seddon and R. H. Templer, Phil. Trans. Ro. SOC.A., 1993,344,377.
I 1 D. M. Anderson, S. M. Gmner and S. Leibler, Proc. Nut. Acad. Sci. U.S. A., 1988,
85,5364.
12 R.Tenchova, B. Tenchov, H.-J. Hinz and P. J. Quinn, Liquid Crystals, 1996,20,
469.
13 H.-J. Hinz, H. Kuttenreich, R. Meyer, M. Renner and R. Freund, Biochemistry, 1991,
30,5 125.
14 D. A. Mannock, R. N. McElhaney, P. E. Harper and S. M. Gruner, Biophys. J., 1994,
66,734.
15 D. A. Mannock, R. N. A. H. Lewis, A. Sen and R. N. McElhaney,
Biochemistry, 1988,27,6852.
16 D. A. Mannock, P. E. Harper, S. M. Gruner and R. N. McElhaney, Chem. Phys.
Lipids, 2001, 111, 139.
17 D. A. Mannock, R. N. A. H. Lewis, R. N. McElhaney, M. Akiyama, H. Yamada, D.

C. Turner and S. M. Gruner, Biophys. J., 1992,63, 1355.
18 D. C. Turner, Z.-G. Wang, S. M. Gruner, D. A. Mannock and R. N. McElhaney, J.
Phys. II France, 1992,2,2039.
19 H. Kuttenreich, H.-J. Hinz, R. D. Koynova and B. G. Tenchov, Chem. Phys. Lipids,
1993,66,55.
20 N. Zeb, PhD Thesis, Imperial College, London, 1998.
21 P. M, Duesing, J. M. Seddon, R. H. Templer and D. A. Mannock, Langmuir, 1997,
13 (lo), 2655.
22 D. A. Mannock, M. Kreichbaum, P. E. Harper, S. M Gmner and R. N. Mcelhaney,
Synthesis and Thermotropic Characterization of an Homologous Series of Racemic
PD-Galactosyl Dialkylglycerols. Manuscript in preparation, 2002.
23 M. Koberl, H.-J. Hinz and G . Rapp, Chem. Phys. Lipids, 1998,91, 13.
24 D. A. Mannock, M. Akiyama, R. N. A. H. Lewis and R. N. McElhaney, Biochim.
Biophys. A d a , 2000,1509,203.
Subject Index
Ab initio methods, 23 CD54,63,65

Activation free energy, 153 CD86,65
Acylated protein, 61 CHARMM, 7
p2-Adrenergic receptors, 86 Chitin, 53, 56
Albmycin, 2 16 Chlordiazepoxide, 243, 244, 245
Alcohol dehydrogenase, 26 Chlorophyll, 229
AMBER-95,79 Chloroplasts, 147
Amide, 22, 141 Cholera toxin, 61,62
Amphipathic helix, 169, 191, 197 Cholesterol, 60,67, 170
Amyloidosis, 103, 115, 116 Chiral thin film, 139
Amyloids, 94, 103, 104 Chromatin, 165, 166
Apoptosis, 166 Chymotrypsin, 26, 170
Artificial lipid membranes, 3 Chymotrypsinogen, 26
Atomic force microscopy ( A M ) , 31,33, Classical molecular interaction potential
50-54 (CMP), 82
ATP, 64 CNDO/S, 27
Colicin M, 216
B cell receptors, 61 Collective excitation dynamics, 118
B850,119 Concanavalin A, 26
Bacteriophage, 147 Confocal microscopy, 31, 59
Bacteriorhodopsin, 26,208,211,222, Confocal spot, 238
224,226 Constant mean curvature, 189
Barrier crossing, 147 Cork domain, 216,217
Bath modes, 123 Crossing time, 155
Benzodiazepine, 243 Cryo-electron microscopy, 218
Bicontinuous cubic phase, 177, 178 Crystallisation, 221, 225,227,228
Bilayer, 138,200,206,215,230,251, C-type lectin, 58
254 Curvature elastic energy, 178
Binding force measurements, 35 Cyanobacteria, 237
Bioimaging, 61 Cyclohexane, 181
Bioluminescence resonance energy Cytidylyltransferase (CCT), 163
transfer ( B E T ) , 86 Cytochalasin D, 62
Brownian motion, 65, 155 Cytochrome c, 4,15,26
Butanedione monoxime, 62 Cytoskeleton, 35,40,41,42,43,60,62,
64,65,240
Calcium-sensing receptors, 85 Cytotoxic peptide, 191
Carbonic anhydrase, 26 Cytotoxic T cells, 65
Cardiomyocytes, 41
CASFT2,23 Dendritic cells, 62,65
CASSCF2,23,27 Density functional theory, 23
Catalase, 78, 81, 83 Detergent effects, 228
CD spectroscopy, 15, 18,20, 21,26, 140 Diamide, 25
. CD4,73 Differential scanning calorimetry, 218,
CD45,59 249,267,270,272,273
Dilauroylphosphatidylcholine, 177 Gangliosides, 6 1

Dioleoylphosphatidylcholine,202 Gaussian bending modulus, 257,265
Direct immunofluorescent microscopy, Gaussian curvature, 23 1
166 Gene therapy, 4
DNA, 147, 149, 158,218,219 Glutamate receptors, 85
DsbB, 209 Glutathione reductase, 26
Glycolipids, 267-269,275
Ectodomain shedding, 64, 65 Glycoprotein, 44
Elastase, 26 Glycosylation, 50, 57
Elastic stress, 169 GPI anchor, 61
Electro-magnetic properties, 136 G-protein coupled receptors (GPCRs),
Electron tomography, 2 19 85-89
Electrostatic free energy, 101 Gramicidin, 4, 12, 15
Electrostatic interactions, 22 Green fluorescent protein (GFP), 59,63-
Electrostatic stability, 94 67
Ellipsommetry, 141 GROMACS, 73,95105
Endoplasmic reticulum, 147, 165 GTPyS, 88
Entropy analysis, 86, 89, 90
Erabutoxin, 26 Halorhodopsin, 222
Erythrocytes, 193 Harmonic oscillators, 159
Escherichia coli, 199,2 15 a-Helix, 11,20,21, 140, 168, 191, 194,
Evolutionary trace (ET) method, 86, 87 196,208,209,213
Excess water measurements, 182 Heme, 81
Exciton, 118, 129, 132 Hemoglobin, 26
Exosomes, 64,65 Hemolytic activity, 191, 193
HIV-1,72
F-actin, 41,43,62 Huang-Rhys factor, 128, 130-132, 134
Familia1 amyloidotic pol yneuropathy Hydrogen bonds, 81,217
(FAP), 94 Hydrogen peroxide, 78,79,81-83
Fascin, 62
Feedback regulation, 172, 173 ICAM-1,59,62,63
FhuA, 216 Immune synapse, 58-60,62,63,65,66
Fibrilogenesis, 115 Immunoregulatory tyrosine activating
Flavodoxin, 26 motifs (ITAMs), 61, 62
Flow linear dichroism, 3, 18 Influenza virus, 72
Fluid lamellar phase, 274 Integrin, 38,39
Fluorescence imaging, 58 Inverse hexagonal phase, 177,251, 274
Fluorescence lifetime imaging (FLIM), Inverse lyotropic mesophases, 177
58,6648 Ion channel, 42,72
Fluorescence recovery, 237 IR spectroscopy, 140, 141,
Fluorescence resonance energy transfer Iron transporter, 215
(FRET), 61,86
Fluorescence spectroscopy, 208 Kinases, 61
Formamide, 23 Kmkmechanism, 147,149,150, 154, 156
FRAP, 238-240 Kramers problem, 147, 148
Frenkel excitons, 119
Friction image, 52 Lactate dehydrogenase, 26
FTIR spectra, 141 Lamellar phase, 226,232
Langevin equation, 119, 124, 126, 159
GABABreceptor, 85,87, 89, 90 Langmuir films, 139
Subject Index 279
Laplace’s law, 204 Multi-reference configuration interaction

Laser scanning confocal microscopy, 59, (MRCI), 23,27
63 Myosin, 62
Lateral force microscopy, 5 1 Myristoylated protein, 61
Lattice parameter, 183
Lauric acid, 177 Natural killer (NK) cells, 58,62-65
Lck,61 Nl3 protein, 73-76
LD cell, 7 NhaA, 209
LD spectroscopy, 3, 15, 18 Nonadiabaticity, 124
LFA-1,59,63
LHl, 119 Octadecene, 181
Light emitting diodes, 136 Octylglucosides, 61
Light harvesting complexes, 222,225, Oligosaccharides, 267
229,233,234 Opioid receptors, 85
Light harvesting systems, 118 Opsin receptors, 86
Light microscopy, 32 Oregon Green, 165
Lipid bilayer, 18,72,221 Osmotic stress, 179, 181, 186
Lipid rafts, 5 9 , 6 0 4 2 Osteoblasts, 42
Lipid vesicles, 203 Osteoclasts, 35, 39
Lipids, 167
Liposome, 3,4, 18,218, 243-245,254 Papain, 26
Lymphocytes, 65 Parametric estimation, 230
Lysophosphatidylcholine, 169, 172 Patch-clamp, 202
Lysozyme, 26, 103, 104 Pauli repulsion, 24
Periplasm, 215
Macrophages, 173 Peroxisome, 79
Mahalanobis distance, 106 Phage receptors, 215
MD simulation, 73, 114 Phase behaviour, 180
Mean curvature, 232 Phosphatidic acid, 171, 172
Mechanosensation, 199 Phosphatidylcholine, 163, 164, 168, 172
Mechanosensitive channel (MscL), 199, Phosphatidylethanolamine, 163,200,202
206 Phospholipase, 172
Membrane bending energy, 256 Phospholipid, 60, 243,25 1,275
Membrane curvature, 169 Phosphorylation, 61
Membrane dynamics, 118 Photosynthetic membranes, 237
Membrane pore, 147 Photosystem I, 240,241
Membrane proteins, 208,217,221 Photosystem 11, 229, 237,240, 241
Membrane synthesis, 163 Phycobilisomes, 237-239,241
Membrane-tension-gated channel, 199 Plasma membrane, 65
Mesophase, 184,230,231 Plastocyanin, 26
Metropolis algorithm, 262 Poisson-Boltzmann equation, 96,97
MHC, 58,59,63-65 Polarization modulation infra-red
Mitochondria, 147 reflection absorption spectroscopy
Molecular aggregates, 118 (PM-IRRAS), 192
Molecular dynamics, 20,78-80 Polarizing microscopy, 223
Monolayer, 139, 141, 177, 191, 193-196 Polaron, 118, 128,129, 132
Monoloein, 222, 223 Polyalanine, 136, 140, 142, 143
Monosaccharides, 267 Polyethylene glycol, 35
Monte Carlo simulation, 260,26 1, 262, Polypropylene, 138
263 Pore, 156
Porin, 26,215 Sphingolipids, 60

Prealbuim, 94 Sphingomyelin, 163, 172
Prealbumin, 26 Spiroplasma melifferum, 192, 193
pre-PsbW, 4, 15 Stark effect, 137
Protease inhibitors, 65 Stickler-Berg formula, 66-68
Protein folding, 208 Subtilisin, 26
Protein translocation, 147 Superoxide dismutase, 26
Protein-lipid interactions, 199 Supramolecular activating clusters
Proteoliposomes, 219 (SMAC), 59
Pyrene, 10 Surface hopping, 125,126
Surface roughness, 56
Ras, 89 Synchrotron radiation, 20
Reaction centre, 222,237, 240 Synechococcus sp., 237-239
Receptor activity modifying protein
(RAMP), 90,91 T cell receptors, 59,61,62
Renaturation, 2 13 Talin, 254,255,258,259
Retinol, 94 TDDFT, 23
Rhodanese, 26 Texas Red, 165
Rhodobacter sphaeroides, 222 Thermolysin, 26
Rhodopseudornonas acidophila, 223 Thylakoid membrane, 17,237,238,240,
Rhodopseudornonas viridis, 222 24 1
Rhodopsin, 86, 87 TIP3P, 79,82
Ribonuclease, 26 Ton complex, 219
Rosenfeld equation, 2 1 Transferrin receptors, 6 1
Rouse model, 148, 149,152, 157 Transition polarisation, 11
Rydberg states, 23 Translocation, 157
Transmembrane helix, 208
Saccharomyces cerev:Tiae, 5 0 , 5 1,57 Transmembrane segment (TM), 73,73,
Scanning Electron Microscopy, 53 75
Scanning probe microscopy, 3 1 Transthyretin (TTR), 94,95
Schizosaccharornycespombe, 5 1 Triose phosphate isomerase, 26
SDS micelles, 17 Triton X-100,61
Secondary structure, 80 Trypsin, 26
Senile systematic amyliodosis (SSA), 94 Tubulin, 65
Shear distorted liposomes, 4 Tyrosine lunase, 42
P-Sheet, 11,20, 191, 196, 197,216
Siderophores, 215 Vitamin E, 248,250,251,252
Single photon counting, 66 Vpu protein, 73-76
Site-directed mutagenesis, 81
Small angle X-ray diffraction, 223, 226 Xplor, 73
Solar cells, 136 X-ray diffraction, 181,216,227,249,
Solvent accessible surface area (SASA), 267,271,273
108,114 X-ray scattering, 250
Soret band, 17
Spectrin, 240 Zinc, 64

Biophysical Chemistry

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Biophysical Chemistry

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Biophysical Chemistry

Uploaded by

Copyright:

Available Formats

Biophysical Chemistry

Membranes and Proteins

ROYAL SOCIETY OF CHEMISTRY

Special Publication No. 283

0 The Royal Society of Chemistry 2002

All rights reserved.

Published by The Royal Society of Chemistry,

For further information see our web site at www.rsc.org

Printed by George Over Ltd., London and Rugby

I Probing Biological Molecules: Theory and Experiment

Quantitative Protein Circular Dichroism Calculations 20

Probing Cellular Structure and Function by Atomic Force Microscopy 31

Physical Characterization of Wild Type and mnn9 Mutant Cells of

Probing Supramolecular Organisation at Immune Synapses 58

Probing the Structure of Viral Ion Channel Proteins: A Computational

The Impact of H202 on the Structure of Catalases by Molecular

Entropy in the Alignment and Dimerization of Class C G-Protein Coupled

Electrostatic Stability of Wild Type and Mutant Transthyretin Oligomers 94

Collective Excitation Dynamics in Molecular Aggregates: Exciton Relaxation,

Surprising Electro-magnetic Properties of Close Packed Organized Organic

I1 Proteins, Lipids and Their Interactions

Models and Measurements on the Monolayer Bending Energy of Inverse

Hemolytic and Antibacterial Activities of LK Peptides of Various Topologies:

A Novel Approach for Probing Protein-Lipid Interactions of MscL,

Folding of The a-Helical Membrane Proteins DsbB and NhaA 208

FhuA, an Escherichia coli Transporter and Phage Receptor 215

Morpholgical Aspects of in cubo Protein Crystallisation 22 1

Mobility of Proteins and Lipids in the Photosynthetic Membranes of

Partitioning and Thermodynamics of Chlordiazepoxide in n-OctanoVBuffer

Distribution of Vitamin E in Model Membranes 248

Theory on Opening-up of Liposomal Membranes by Adsorption of Talin 254

Differential Scanning Calorimetry and X-Ray Diffraction Studies of

Subject Index 277

Alison Rodger:* Jascindra Rajendra; Rhoderick Morther; Terrence Andrews; Jonathan

a Department of Chemistry, University of Warwick, Coventry, CV4 7AL, UK

* email: a.rodger@warwick.ac.uk Fax (+44) 24 76524112

2 LINEAR DICHROISM OF ANALYTES IN SHEAR DISTORTED LIPOSOMES

Liposome Shear deformed Model of liposome

Figure 1 Schematic illustration of a shear deformed liposome.

3 MATERIALS AND METHODS

4 DESIGN AND CONSTRUCTION OF LD CELL

Main Body: Stainless steel construction.

To design and fabricate an instrument that rotates in a sealed environment where

Before attempting to incorporate proteins into lipid structures, we considered different

250 300 350 400 250 300 350 400

Soybean lipid structures prepared by vortexing (Figure 4b) gave LD ratios of

6 PROTEIN MOTIF TRANSITION POLARISATIONSAND EXPECTED LD

(d) Wavelength/nm Padau Oscillator strength

Figure 7 Schematic illustration of the proposed thylakoid membrane insertion mechanism

7 PROTEIN CD AND LD SPECTRA

The proteins we chose to study were: gramicidin, cytochrome c, a thylakoid membrane

Figure 9 Absorbance, CL)(averaged over 64 scans), and LD spectra of gramicidin (0.5 mg

Figure 10 LD spectrum of cytochrome c (1 mg in 1 mL of 2.5 mg/mL lipid in phosphate

200 210 220 230 240 250

Figure 1 1 LD spectrum of pre-PsbW (1 mg in 2.5 mg/mL lipid in phosphate buffer) in

200 220 240 260 280 300 320

Figure 12 LD spectrum of a monoclonal antibody (1 mg in 2.5 m g h L lipid in phosphate

Professor Malcolm MacMey and Sirilak Wannaborworn, Cambridge University, are