Vesicle aggregation by multivalent ligands: Relating crosslinking ability to surface affinity

Article abstract of DOI:10.1039/b706662g

In an effort to improve the stability of our tissue-mimetic vesicle aggregates, we have investigated how increasing the valency of our multivalent crosslinking ligand, poly-l-histidine, affected both the extent of vesicle aggregation and the affinity of t

Full text of DOI:10.1039/b706662g

View Article Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/obc | Organic & Biomolecular Chemistry
Vesicle aggregation by multivalent ligands: relating crosslinking ability to
surface affinity†
Xi Wang, Robert J. Mart and Simon J. Webb*
Received 2nd May 2007, Accepted 15th June 2007
First published as an Advance Article on the web 2nd July 2007
DOI: 10.1039/b706662g
In an effort to improve the stability of our tissue-mimetic vesicle aggregates, we have investigated how
increasing the valency of our multivalent crosslinking ligand, poly-L-histidine, affected both the extent
of vesicle aggregation and the affinity of the multivalent ligand for the synthetic receptor Cu(1)
embedded in the vesicle membranes. Although increasing ligand valency gave the anticipated increase
in the size of the vesicle aggregates, isothermal calorimetric studies did not show the expected increase
in the valence-corrected binding constant for the embedded receptors. To explain both observations, we
have developed a simple new binding model that encompasses both multivalent binding to receptors on
a single vesicle surface (intramembrane binding) and vesicle crosslinking (intermembrane binding).
due to the complexity of biological systems. To circumvent this
Introduction
problem, vesicles with synthetic receptors embedded in their
Nature uses cooperative multivalent binding between external
membranes have been employed as simple cell mimics and used to
develop systems that emulate cell signaling4 and cell adhesion.5
In particular, mimicking cell adhesion is an area of much
current interest due to the desirability of synthetic tissue-mimetic
materials.6 Several researchers have used multivalent chemical
“glues” to aggregate receptor-doped vesicles; these “glues” have
included tetravalent streptavidin (crosslinking biotin lipids),7,8
iron(II) (crosslinking terpyridine lipids),9 poly-L-arginine10 and
guanidinylated dendrimers (both crosslinking phosphate lipids).11
We are also interested in the aggregation of vesicles by multivalent
ligands, in particular determining how ligand valency affects the
balance between intermembrane crosslinking and intramembrane
ligand binding. Curiously, the effect of increasing ligand valency
on the extent of vesicle aggregation has been little explored,
though Sideratou et al. reported that doubling the valency of
guanidinylated dendrimers from 32 to 64 gave an increase in vesicle
crosslinking, even when compared on a per binding unit basis.11
In recent studies of vesicle adhesion, we used a highly multiva-
lent ligand, a 39mer of histidine (2, Fig. 1), to crosslink distearoyl
ligands and membrane-bound receptors to mediate key cellular
processes like adhesion and signaling. These processes are initiated
by highly specific binding of multivalent ligands to receptors on
the cell surface, resulting in the formation of multiple inter- and/or
intramembrane links. Although the binding of most natural
multivalent ligands to cell surface receptors is tightly controlled,
immunoglobulins are an interesting class of natural multivalent
ligands able to form both inter- and intramembrane links. For
example, bivalent IgG can activate the immune response via two
different binding modes; it can either bind two antigens on the
same cell surface or agglutinate cells by crosslinking antigens
on different surfaces.1 Moreover, increasing the valency of these
natural ligands improves both binding and crosslinking ability;
the larger decavalent antibody IgM not only has high avidity
for multivalent antigens like viruses but is a particularly strong
agglutinating agent, the latter effect ascribed to the greater distance
between the antigen binding sites.1 This behaviour is not restricted
to viral or cell surface antigens since IgG antibodies will also bind
vesicle-embedded antigens in both modes.2 Much like antibodies,
many synthetic multivalent ligands have been found to bind cells
in either mode,3 although often in an unpredictable fashion.3c Un-
derstanding the relationship between these competing processes
is important because developing multivalent ligands designed to
crosslink cells could lead to drugs with new modes of action, for
example drugs that agglutinate pathogens to form large aggregates,
or drugs that form “cancer-nets” around solid tumors and prevent
tumor growth.3b,3c
Understanding intermembrane (cell adhesion) or intramem-
brane binding (signaling) of natural multivalent ligands is difficult
Manchester Interdisciplinary Biocentre and the School of Chemistry, Uni-
versity of Manchester, 131 Princess St., Manchester, M1 7DN, UK. E-mail:
S.Webb@manchester.ac.uk; Fax: +44 161-306-5201; Tel: +44 161-306-
4524
† Electronic supplementary information (ESI) available: ITC mea-
surements and derivation of mathematical relationships. See DOI:
10.1039/b706662g
Fig. 1 Synthetic receptor Cu(1) and histidine-containing ligands 2, 3
and 4.
2498 | Org. Biomol. Chem., 2007, 5, 2498–2505
This journal is
The Royal Society of Chemistry 2007
©

View Article Online
phosphatidylcholine (DSPC) vesicles doped with 5% mol/mol of
a synthetic receptor Cu(1) (Fig. 1).12 The relatively weak histidine–
Cu(IDA) interaction, with a binding constant around 103 −1, was
M
chosen because it is a good model of the weak interactions used
in cell adhesion (K ∼ 104
M
−1).13 Addition of 2 to a suspension of
Cu(1)-coated vesicles caused an increase in turbidity, and stable
vesicle aggregates were observed by fluorescence microscopy, but
these studies did not give any indication of the extent or strength
of binding to Cu(1).14 We then speculated that, as observed with
immunoglobulins, increasing the valency of the multivalent glue
would enhance both inter- and intramembrane binding; allowing
us to use the former to increase the extent of vesicle aggregation
and the latter to improve the stability of the resulting aggregates.
Presented here are our investigations into how changing ligand
multivalency affected vesicle aggregation and the strength of
binding to vesicle-embedded receptors. It was possible to change
the valency of the poly-L-histidine ligand 2 easily, allowing
binding strength and crosslinking ability to be correlated with
the degree of multivalency. (His)226 (3) and N-acetyl-L-histidine
4 are commercially available variants of 2, which are longer
and shorter respectively; furthermore, monovalent N-acetyl-L-
histidine should be a binding, but non-crosslinking, control
compound. As previously, fluorescence microscopy and changes
in solution turbidity would reveal the ability of ligands 2, 3, and
4 to crosslink vesicles, whilst the binding affinity of the histidine
ligands for the membrane-bound lipid Cu(1) could be determined
by isothermal calorimetry (ITC), a technique widely used to
study ligand binding to vesicle surfaces.10,15 This ability to study
crosslinking separately from all other binding events allowed us
to elucidate the relationship between inter- and intramembrane
binding.
Scheme 1 Synthesis of lipid H21.
removed quantitatively within 10 hours to afford lipid H21 as
the trifluoroacetic acid salt; provided water was strictly excluded,
the ester linkage in H21 remained intact. This salt was suitable
for further use, but free-base H21 was obtained by reverse-phase
chromatography on octadecyl-functionalised silica gel, eluting
with acetone–water (60 : 40).
Results and discussion
Synthesis of lipid H21 and Cu(1)/DSPC vesicles
The iminodiacetate-capped lipid H21 has several features that
make it ideal for these studies. The iminodiacetate groups chelate
copper tightly to give receptor Cu(1) quantitatively,16 and Cu(1)
binds monovalent histidine ligands with affinities similar to those
between selectins and monovalent carbohydrates.13 The pyrene
group is the membrane anchor, which also allows visualization of
Cu(1)-vesicles by fluorescence microscopy and indicates any Cu(1)
clustering in the membrane.
The monotosylation of tetraethylene glycol by para-
toluenesulfonyl chloride, a key step in the synthesis of lipid
H21 (Scheme 1), was carried out in the presence of freshly
prepared silver(I) oxide and potassium iodide, a method used
to desymmetrise a range of diols.17a The monofunctionalised
tosylate then required heating at reflux for an extended period
with di(tert-butyl)-protected iminodiacetic acid in acetonitrile to
displace the tosylate and afford the protected binding group
7.17b The fluorescent membrane anchor was then appended to
7 through dicyclohexylcarbodiimide-mediated esterification with
1-pyrenebutyric acid. Although neat trifluoroacetic acid was
required to remove the tert-butyl protecting groups from 8,
which limited the transferability of this methodology to other
syntheses, this protecting group proved to be adequate for the
synthesis of lipid H21. The tert-butyl protecting groups were
To solely examine the effect of changing the ligand valency on
vesicle binding and aggregation, the membrane loading of receptor
Cu(1) was fixed at 5% mol/mol, a membrane concentration
known to give vesicle aggregation with ligand 2.12 Unilamellar
vesicles (approx. 800 nm diameter, 20 mM lipid in 20 mM MOPS
buffer, 100 mM NaCl, pH 7.4) composed of DSPC doped with
H21 were prepared by extrusion of a phospholipid suspension
◦
through 800 nm pore size polycarbonate membranes at 60 C, a
temperature above the phase transition temperature of DSPC. The
fluorescence spectra of these vesicle suspensions showed only
pyrene monomer emission, indicating dispersal of lipid H21
over the vesicle surface. Lipid H21 in DSPC bilayers chelates
12
copper(II) with an average binding constant of 4 × 107 M−1
,
so addition of one equivalent of Cu(II) to H21 at millimolar
concentrations gave the amine-binding synthetic receptor Cu(1)
quantitatively. This addition of copper(II) to H21/DSPC vesicles
gave no significant increase in solution turbidity, and fluorescence
microscopy confirmed no vesicle aggregation during formation
of Cu(1)/DSPC vesicles.18 Titration of copper(II) into H21/DSPC
vesicles (0.1 lM lipid H21) has shown that the entire cohort of lipid
H21 is available for complexation, and flip-flop between bilayer
leaflets is rapid (t1/2 < 30 s).12
This journal is
The Royal Society of Chemistry 2007
Org. Biomol. Chem., 2007, 5, 2498–2505 | 2499
©

View Article Online
In the turbidity and ITC studies, the key comparison is
between ligands 2 and 3, which will show whether increasing
poly-L-histidine valency has given a commensurate increase in
crosslinking efficiency and affinity for membrane-bound Cu(1).
The monomeric ligand 4 should not crosslink vesicles, but it will
give the affinity of a solution-phase histidine group for membrane-
embedded Cu(1). This value will be an approximation of the
equilibrium constant of the first binding event (K1) that leads
to immobilization of poly-L-histidine on the vesicle surface.
Inspection of the turbid mixtures by fluorescence microscopy
showed that addition of 2.5 lM 2 or 0.45 lM 3 (one eq. His
residues in each case) gave large clusters of intact aggregated
vesicles (Fig. 3).19 There was a wide distribution of aggregate sizes
in each sample, but for (His)39 the average cross-sectional area of
the aggregates was 99
33 lm2, whilst larger aggregates were
observed for (His)226 with an average cross-sectional area of 180
81 lm2. The larger size of the aggregates observed in mixtures of 3
and Cu(1)/DSPC vesicles correlates well with the higher turbidity
of these mixtures.6b,11,14
Vesicle aggregation by poly-L-histidine
The 39mer of histidine, (His)39 2 (∼14 nm extended length),
had previously been shown to aggregate Cu(1)/DSPC vesicles,
so the same conditions were employed to test the ability of
(His)226 3 (∼82 nm extended length) and N-acetyl-L-histidine 4
to aggregate 800 nm diameter vesicles. Titration of N-acetyl-
L-histidine 4 (2 mM) into suspensions of Cu(1)/DSPC vesicles
(2 mM in lipid, 0.1 mM in Cu(1)) gave little change in turbidity, as
monitored at 700 nm in the UV-visible spectrum, showing that
the morphology of the vesicles remained unchanged (Fig. 2).
Fluorescence microscopy showed only individual vesicles and
no aggregates. As observed previously, titration of the shorter
poly-L-histidine 2 into Cu(1)/DSPC vesicles resulted in a strong
increase in turbidity (Fig. 2), which started to plateau after 0.125
equivalents of 2 had been added (5 histidine residues per Cu(1)).
In comparison, titration of the longer poly-L-histidine 3 into
Cu(1)/DSPC vesicles resulted in a stronger increase in turbidity
earlier in the titration (Fig. 2) which peaked at 0.005 equivalents
of 3 (1.1 His residues per Cu(1)). Receptor Cu(1) was always
required for aggregation to occur; no aggregation was observed
when any of ligands 2, 3 or 4 were mixed with undoped DSPC
vesicle suspensions.
Fig. 3 Representative fluorescence micrographs showing vesicle aggre-
gates caused by addition of (a) 2.5 lM ligand 2 or (b) 0.45 lM 3 to a
suspension of Cu(1)/DSPC vesicles (5% mol/mol Cu(1) and 2 mM in total
lipid).
It was tempting to conclude that the greater ability of 3 to
aggregate vesicles was due to stronger cooperative binding of this
more multivalent ligand to membrane-embedded Cu(1). However,
this would be misleading since changes in turbidity do not reflect
the strength and extent of intramembrane binding within a single
vesicle. Therefore, to measure affinity for Cu(1) we used ITC
to measure heat flow to and from Cu(1)/DSPC suspensions
during the addition of 2, 3 or 4, which allowed us to evaluate
their respective valence-corrected binding constants to membrane-
bound Cu(1).
Measurement of binding by isothermal calorimetry
Isothermal calorimetry has been used to study a range of binding
events at phospholipid bilayer surfaces,10,15 but only rarely used to
study the formation of vesicle aggregates. Vesicles in suspension do
not normally aggregate because of repulsion between the hydrated
bilayer surfaces (often referred to as the “hydration force”);15b,20
a specific crosslinking interaction is necessary for vesicles to
aggregate. Furthermore, this repulsion between vesicle surfaces
usually requires a spacer group to be present to distance the vesicle
surface from the crosslinking groups.5b,15c,21 In our system a tetra-
ethylene glycol spacer was incorporated into lipid H21; we found
the copper(II) complex of N-nonyliminodiacetate embedded at 5%
mol/mol in DSPC vesicles was unable to mediate vesicle adhesion
by poly-L-histidine. The tetra-ethylene glycol linker minimises
unfavourable vesicle–vesicle interactions, allowing DG for the
formation of crosslinking bonds to overcome the unfavourable DG
for vesicle aggregation. This vesicle–vesicle repulsion will weaken
the crosslinking interactions (represented by the binding constant
Kiinter) relative to the solution control (K1) (vide infra, Fig. 4).
Fig. 2 The change in solution turbidity resulting from aggregation of
Cu(1)/DSPC vesicles (5% mol/mol Cu(1) and 2 mM total lipid) due to
addition of histidine-containing ligands 2 (ꢀ), 3 (᭺) or 4 (ꢀ). The lines
are smooth curve fits to guide the eye.
The median effective concentration (EC50) values for aggre-
gation reveal the effect of multivalency; the concentration of 2
required to give a 50% turbidity response was 3.3 lM (0.13 mM
histidine residues) whilst the concentration of 3 that gave a
50% response was only 0.25 lM (0.055 mM histidine residues).
Furthermore, even when compared on a valence-corrected basis
the longer poly-L-histidine 3 is over twice as effective at aggregating
Cu(1)/DSPC vesicles compared to 2.
2500 | Org. Biomol. Chem., 2007, 5, 2498–2505
This journal is
The Royal Society of Chemistry 2007
©

View Article Online
The ITC analysis conditions were identical to the conditions
used for the turbidimetric analyses. Thus solutions of ligands
2, 3 and 4 were made up in MOPS buffer (20 mM) at pH 7.4
with each solution 2 mM in histidine residues. Titration of these
solutions into vesicle-free buffer showed the respective heats of
dilution were small (∼0 kJ mol−1 for 2, < −0.4 kJ mol−1 for 3,
and < −0.9 kJ mol−1 for 4) and fitting of the ITC data showed
no measurable binding. To check for non-specific binding to
the surface of DSPC vesicles, these three ligand solutions were
also titrated isothermally into undoped DSPC vesicles under the
same conditions. In each case little heat release was observed
(∼0 kJ mol−1 for 2, < −0.5 kJ mol−1 for 3, and < −0.8 kJ mol−1 for
4) and there was no measurable binding. In conjunction with the
absence of turbidity changes, these data suggested that none of 2,
3 or 4 associated with undoped DSPC vesicles. These solutions
of ligands 2, 3 and 4 were then titrated into suspensions of
800 nm diameter DSPC vesicles doped with 5% mol/mol Cu(1)
(0.1 mM bulk concentration of vesicle-embedded Cu(1)). In each
case significant heat release was observed upon addition of the
histidine-containing ligands to the vesicle suspensions, with the
lowest enthalpy change observed during the titration of 4 into
Cu(1)/DSPC vesicles.
turbidimetric data: N-acetyl-L-histidine, which does not aggregate
vesicles, bound to Cu(1) only slightly more weakly than the
multivalent ligands, whilst short poly-L-histidine 2 bound with
the same affinity (within error) as the more strongly aggregating
and higher valency poly-L-histidine 3.
Analysis and binding model
Several different theoretical frameworks have been developed to
explain the binding of multivalent ligands to membrane-bound
receptors. However, these frameworks have regarded multivalent
ligands either as “stickers” that crosslink vesicles2b or chelating
ligands that bind solely to a single vesicle surface.27 We propose
here a model that accommodates both inter- and intramembrane
binding of multivalent ligands, and can be used to predict and
control the extent of vesicle crosslinking by multivalent ligands.
The binding of multivalent ligands to membrane-bound recep-
tors is often considered to be “cooperative” because the binding
constant, calculated using the bulk concentrations of the binding
species, is higher than that calculated for binding to analogous
receptors that are not membrane-bound. This apparent binding
enhancement actually largely stems from the concentrating effect
of embedding receptors within a vesicle membrane.28,29 The effec-
tive molarity of a membrane-bound receptor correlates with the
percentage loading of receptor in the membrane, v, and it has been
demonstrated that as v increases intramembrane binding becomes
increasingly favourable.28,29 Whether or not this is regarded as true
cooperativity depends upon the definition used; some definitions
of cooperativity include concentrating effects caused by linking
binding units together or embedding them in a membrane,30 whilst
others exclude such concentrating effects.31 The lower polarity of
the interface can also enhance binding to membrane-embedded
receptors, but our monovalent control 4 has a value of K(4) (3 ×
Analyzing the ITC data on a per ligand basis for these highly
multivalent poly-L-histidine ligands was impractical, so instead the
binding isotherms were analysed on a per binding unit (valence-
corrected) basis.22,23 This afforded the average binding constants
Kav , each of which is the geometric mean of the individual
microscopic binding constants.24,25 Since Cu(1) should have two
free sites for coordinating histidine, the refined ITC data with the
respective heats of ligand dilution subtracted was fitted using both
R
the one-site and two-site binding models in the Originꢀ curve-
fitting package accompanying the Microcal VP-ITC. Both binding
models were found to fit the data, but the second binding constant
in the two-site model was found to be redundant. These binding
constants were then checked by fitting the cumulative heat release
to Cu(IDA) + (His) and Cu(IDA) + 2(His) binding isotherms.26
The 1 + 1 model gave the best fit to all the data sets and gave
103
M
−1) that lies within the range observed for the binding of
histidine ligands to Cu(IDA) complexes in solution (1 × 103 M−1
to 6.3 × 103
M
−1).32
The per unit binding constant for (His)39 2 to Cu(1) in DSPC
R
similar values of Kav as those obtained with Originꢀ. If the data
bilayers, Kav (2) = 6 × 103
M
−1, is only slightly greater than that of
from the control titrations of undoped DSPC vesicles with ligands
2, 3 or 4 were subtracted from the Cu(1)/DSPC vesicle titration
data, the calculated binding constants were the same within error.25
For ligands 2, 3 or 4 complexing to Cu(1) we found the measured
affinities per histidine residue lay in a small range between 3 and
N-acetyl-L-histidine. Similarly, with the longer analogue (His)226
3, the per unit binding constant for membrane-embedded Cu(1),
Kav (3) = 6 × 103 M−1 is, within error, the same as that observed
for 2. The similarity between these three values suggests that
following immobilization of poly-L-histidine on the vesicle surface,
the strength of which can be estimated from K(4), subsequent
binding of histidine residues to Cu(1) in the membrane is only
slightly more favourable than the initial binding interaction. It
would appear that despite the very high effective concentration
of Cu(1) in the vesicle membrane (5% mol/mol corresponds
to a local concentration of 0.06 M), intramembrane binding is
actually quite weak. However, such relatively weak intramembrane
binding is essential for vesicle aggregation to occur in our system;
it allows intermembrane binding (crosslinking) to compete with
intramembrane binding (chelation) of the poly-L-histidine.
7 × 103
M
−1. In all cases, binding of histidine-containing ligands
2, 3 or 4 to membrane-embedded Cu(1) was exothermic (−7 to
−14 kJ mol−1) and entropically favourable (25 to 45 J mol−1 K−1).25
Table 1 shows the striking similarity between the valence-corrected
affinities of the different ligands for Cu(1) embedded in DSPC
vesicles, quite at odds with the wide variation observed in the
turbidimetric data.25 Indeed, there is little correlation with the
Table 1 Binding constants obtained at 298 K from the ITC titrations
of Cu(1)/DSPC vesicles (0.1 mM Cu(1) and 2 mM in total lipid) with
histidine-containing ligands 2, 3 or 4
It may be that molecular constraints in our system inhibit
intramembrane binding and promote vesicle aggregation; for
example, poly-L-histidine may adopt conformations that hinder
coordination of all histidine residues to embedded Cu(1) and/or
the gel-phase DSPC membrane may prevent intramembrane
clustering of Cu(1) in response to binding.33 However, multivalent
Ligand
Kav /M−1
2 (His)39
3 (His)226
4 AcHis
(5.6 1.6) × 103
(5.8 1.5) × 103
(3.2 0.5) × 103
This journal is
The Royal Society of Chemistry 2007
Org. Biomol. Chem., 2007, 5, 2498–2505 | 2501
©

View Article Online
ligand-mediated vesicle aggregation has been observed in other
systems that use highly flexible ligands and/or receptors embedded
in fluid-phase membranes, two factors that should promote
intramembrane coordination over vesicle aggregation.7–11,34 This
suggests there must be other factors that affect the balance between
inter- and intramembrane binding.
To understand this interplay between inter- and intramembrane
binding, we have expanded upon a published binding model28
to develop a simple new model for the binding of multivalent
ligands to vesicle-bound receptors that also accounts for vesicle
crosslinking. Fig. 4 shows the first binding event, represented by
K1, and the two subsequent binding possibilities; intermembrane
binding (K2inter) and intramembrane binding (K2intra).
where [R]T = the total concentration of the receptor in the
solution (in mol L−1), and Vm = molar volume of the phospholipid
(in L mol−1).
A new binding constant, K2memb, is used in eqn (1); this binding
constant is calculated from the concentrations of binding partners
within the volume of the membrane only and is independent of
membrane loading. Experimentally it can be useful to keep the
bulk receptor concentration ([R]T) constant because the effect of
varying v on ligand binding is immediately apparent from direct
comparison of the titration curves.28
The observed binding constant for the second binding event
is composed of two microscopic binding constants that represent
inter- and intramembrane binding:25
K2obs = K2inter + K2inter
(2)
or
K2memb
100 [R]TVm
K2obs = K2inter + j2v where j2 =
(3)
Thus by extension each ith binding event of a multivalent ligand
can be represented by:
Kiobs = Kiinter + jiv
(4)
We can use this model to analyse the behaviour of our system.
At v = 5, the geometric mean of K1, Kiinter and Kiintra (2 ≥
i ≥ 39) for 2 (Kav (2)) is only slightly greater than the initial
binding event K1 determined using ligand 4. This suggests that
intramembrane binding of Cu(1) is intrinsically unfavourable,
given that at a membrane loading of 5% mol/mol the effective
local concentration of Cu(1) available for intramembrane binding
is 0.06 M. Furthermore, the valence-corrected binding constant
does not increase significantly when the valency of the poly-L-
histidine is increased, i.e. Kav (2) ∼ Kav (3), though more extensive
vesicle aggregation is observed for 3 compared to 2. This suggests
each of Ko2bs to Kobs could be relatively similar; indeed, if we
n
assume all Kobs (2 ≥ i ≥ n) have a similar value (i.e. ∼ Ko2bs) then
i
the averaging effect of valence-correcting binding constants for
highly multivalent ligands means that Kav will approximate K2obs
for large values of n, and increasing ligand valency further will not
change Kav significantly.25 However, due to the concomitant six-
fold increase in ligand length decreasing the unfavourable effect
of vesicle-vesicle repulsion, the extent of crosslinking increases
(though this increase in Kiinter for ligand 3 seems not to be sufficient
to increase Kav (3) significantly). Our results and binding model
(Fig. 4b) show the crucial role played by membrane loading of
receptor; in effect it can be more important than ligand valency in
determining the net binding strength Kav .
Fig. 4 (a) Schematic representation of the initial binding events between
an n-valent ligand and vesicle-embedded receptors. (b) Graph depicting
the predicted dependence of the intermembrane binding constant (K2inter
)
and the intramembrane binding constant (Ki2ntra) on the membrane
concentration of receptor (v).
In our simple model, following immobilization of the n-valent
ligand on a vesicle surface, each ensuing ith binding event has
these two possible binding modes (Kiinter and Kiintra, 2 ≥ i ≥ n).
For example, the binding constants K1 and Ki2nter are analogous
and independent of the membrane loading of receptor, yet K2inter
should be weaker than K1 due to the repulsive interactions between
vesicles. In contrast, the intramembrane binding constant Ki2ntra is
proportional to the percentage loading, v, of the receptor in the
membrane (eqn (1));28 experimentally this linear dependence of
Ki2ntra on v has been borne out by studies of bivalent IgG antibodies
binding to membrane-embedded synthetic antigens.35
Conversely, researchers sometimes desire strong multivalent
binding to a single vesicle surface without vesicle aggregation.
Coating vesicles with PEG-capped lipids has often been used
to prevent unwanted vesicle aggregation, as it increases steric
21,36
hindrance between vesicles and diminishes Kiinter
.
Our model
suggests an alternative method to decrease crosslinking; building
structural features into the receptor that give high intrinsic
membrane binding constants (Kmi emb), thus leading to high values
of Kiintra at low membrane loadings and promoting intramembrane
over intermembrane binding. On this point, it is interesting to
note that multivalent ligand-mediated vesicle aggregation has
been observed in many synthetic systems, often despite high
ꢀ
ꢁ
K2memb
100 [R]TVm
K2intra
=
v
(1)
2502 | Org. Biomol. Chem., 2007, 5, 2498–2505
This journal is
The Royal Society of Chemistry 2007
©

View Article Online
receptor membrane concentrations (up to 9% mol/mol).9–11,34
This suggests that in synthetic systems intramembrane binding
is often intrinsically weak, and specific features to encourage
intramembrane binding over crosslinking need to be incorporated
into synthetic receptors, for example perfluoroalkyl groups to
facilitate intramembrane clustering of receptors.37
70.9, 71.1, 72.9, 83.7 (all glycol CH2), 128.4 (tosyl CH), 130.3 (tosyl
CH), 131.9 (tosyl CSO3), 143.6 (tosyl CCH3); m/z (ES-HRMS)
371.1135 (M + H+. C15H24NaO7S+ requires 371.1140). Elemental
analysis: Found: C, 50.27; H, 7.17; S, 8.71; Calc. for C15H25O7.5S
(C15H24O7S·0.5H2O) C, 50.41; H, 7.05; S, 8.97%.
11-Bis(tert-butoxycarbonylmethyl)amino-3,6,
9-trioxaundecanol (7)
Conclusion
11-Hydroxy-3,6,9-trioxaundecyl para-toluenesulfonate 6 (174 mg,
0.5 mmol, 1 eq.) was dissolved in a stirred solution of di(tert-butyl)
iminodiacetate (173 mg, 0.5 mmol, 1 eq.) and sodium carbonate
(106 mg, 1 mmol, 2 eq.) in acetonitrile (10 mL). The resulting
solution was allowed to stir at reflux for 5 days under dry air.
The solvent was removed under reduced pressure and the crude
product purified by column chromatography (98 : 2 chloroform–
methanol, silica gel) to give the product (187 mg, 0.42 mmol, 84%)
as a colourless oil. Rf = 0.15 (98 : 2 chloroform–methanol, silica
gel). vmax(Nujol mull)/cm−1 1149, 1223, 1735, 3419; dH(300 MHz,
We have used vesicle aggregation by poly-L-histidine to explore
how increasing ligand valency affects the balance between intra-
and intermembrane binding of multivalent ligands. In our model
system, poly-L-histidine ligands of different valencies were added
to DSPC vesicles containing the synthetic copper(II) iminodiac-
etate receptor Cu(1) embedded in their membranes. Turbidity and
ITC studies revealed that at a 5% mol/mol membrane concen-
tration of Cu(1), increasing the valency of the poly-L-histidine
increased vesicle aggregation but did not lead to significantly
stronger binding when analysed on a per binding unit basis.
To explain our observations we have proposed a binding
model that accommodates intra- and intermembrane binding.
This model suggests the ability of poly-L-histidine to form multiple
links to receptors on the same vesicle surface is intrinsically poor
but should improve at high membrane loadings of receptor Cu(1).
Indeed, the difficulty of forming multiple bonds to receptors
embedded within the same vesicle surface must be a common
effect in synthetic systems, as many other researchers also observe
extensive vesicle crosslinking by multivalent ligands. Our binding
model also suggests that at low membrane loadings of receptor
the multivalent ligand will bind more weakly than the separated
monovalent binding units.25 In this case vesicle crosslinking will
predominate over intramembrane chelation.
◦
CDCl3, 25 C) 1.48 (18H, s, 2 × C(CH3)3), 2.97 (t, J = 6.0 Hz,
t
2H, CH2N), 3.52 (s, 4H, 2 × CH2CO2 Bu), 3.63–3.70 (m, 12H,
6 × glycol CH2), 3.75 (br t, J ∼ 5 Hz, 2H, CH2OH); dC(75 MHz,
◦
t
CDCl3, 25 C) 28.6 (C(CH3)3), 53.8 (CH2N), 57.0 (CH2CO2 Bu),
62.1, 70.6, 70.7, 70.9, 71.0, 72.9, 81.2 (all glycol CH2), 171.1
(CO2 Bu); m/z (ES-HRMS) 422.2748 (M + H+. C20H40NO8
t
+
requires 422.2748).
11-Bis(tert-butoxycarbonylmethyl)amino-3,6,9-trioxaundecyl
4-(pyren-1-yl)butanoate (8)
To a solution of compound 7 (170 mg, 0.4 mmol, 1 eq.) in dry
dichloromethane (2 mL) was added 1,3-dicyclohexylcarbodiimide
(90 mg, 0.44 mmol, 1.1 eq.), 1-pyrenebutyric acid (126 mg,
0.44 mmol, 1.1 eq.) and N,N-dimethylaminopyridine (4.80 mg,
0.04 mmol, 0.1 eq.). The reaction mixture was stirred under
argon at room temperature for 2 days. After evaporation of
the solvent under reduced pressure, the residue was dissolved
in ethyl acetate (5 mL), and stirred at 0 ◦C for 1 hour, then
filtered and the precipitate washed with cold ethyl acetate. After
evaporation of the filtrate, the residue was purified by column
chromatography (1 : 1 : 1 : 0.03 ethyl acetate–cyclohexane–
dichloromethane–triethylamine, silica gel) to give the desired ester
(121 mg, 0.175 mmol, 44%) as a yellow oil. Rf = 0.24 (1 : 1 : 1 :
0.03 ethyl acetate–cyclohexane–dichloromethane–triethylamine,
silica gel). vmax(Nujol mull)/cm−1 1149, 1218, 1249, 1605, 1732;
dH(300 MHz, CDCl3, 25 ◦C) d = 1.45 (s, 18H, 2 × C(CH3)3),
2.21 (tt, J1 = 7.7 Hz, J2 = 7.2 Hz, 2H, ArCH2CH2CH2), 2.51
(t, J = 7.2 Hz, 2H, ArCH2CH2CH2), 2.94 (t, J = 5.8 Hz, 2H,
CH2N), 3.41 (t, J = 7.7 Hz, 2H, ArCH2CH2CH2), 3.50 (s, 4H,
We are now endeavouring to create synthetic receptors and
multivalent ligands with structural features that will favour either
intramembrane binding or intermembrane crosslinking. We hope
this will give some insight into how cells control the balance
between receptor clustering and agglutination, both of which
involve cells binding to multivalent ligands.
Experimental
11-Hydroxy-3,6,9-trioxaundecyl para-toluenesulfonate (6)
To a stirred solution of tetra(ethylene glycol) 5 (344 mg, 1.7 mmol,
1 eq.) in dichloromethane (20 mL) was added fresh Ag2O
(720 mg, 3.1 mmol, 1.8 eq.), para-toluenesulfonyl chloride (398 mg,
2.1 mmol, 1.2 eq.), and KI (66 ◦mg, 0.4 mmol, 0.23 eq.). The
reaction mixture was stirred at 0 C for 15 minutes, then filtered
through a small pad of silica gel and washed with ethyl acetate.
The solvent was removed from the filtrate under reduced pressure
and the residue purified by column chromatography (ethyl acetate,
silica gel) to give the desired monotosylated product (465 mg,
1.3 mmol, 79%) as a colourless oil. Rf = 0.17 (ethyl acetate, silica
gel); vmax(Nujol mull)/cm−1 664, 768, 815, 920, 1009, 1095, 1126,
1177, 1596, 1724, 3421; dH(300 MHz, CDCl3, 25 ◦C) 2.48 (s, 3H,
tosyl CH3), 3.6–3.8 (m, 14H, 7 × glycol chain CH2), 4.19 (t,
J = 4.8 Hz, 2H, CH2OTs), 7.37 (d, J = 8.5 Hz, 2H, 2 × meta
aromatic CH), 7.83 (d, J = 8.5 Hz, 2H, 2 × ortho aromatic CH);
dC(100 MHz, CDCl3, 25 ◦C) 22.1 (tosyl CH3), 62.2, 69.1, 69.7, 70.7,
t
2 × CH2CO2 Bu), 3.58–3.64 (m, 10H, 5 × glycol CH2), 3.70 (t,
J = 4.9 Hz, 2H, glycol CH2), 4.26 (t, J = 4.7, 2H, CO2CH2),
7.87 (d, J = 7.5 Hz, 1H, Ar CH), 7.97–8.20 (m, 7H, 7 × Ar CH),
8.32 (d, J = 9.4 Hz, 1H, Ar CH); dC(75 MHz, CDCl3, 25 ◦C)
27.2, (ArCH2CH2CH2), 28.6 (C(CH3)3), 33.1, ArCH2CH2CH2),
t
34.2, ArCH2CH2CH2), 53.8 (CH2N), 57.0 (CH2CO2 Bu), 63.9,
69.5, 70.7, 70.8, 70.9 (all glycol CH2), 81.2 (C(CH3)), 123.7, 125.1,
125.2, 125.3, 126.2, 127.1, 127.8, 127.9 (all Ar CH), 129.1, 130.3,
131.4, 131.8, 136.1 (all Ar C), 171.1 (CO2R), 173.8 (CO2R); m/z
+
(ES-HRMS) 692.3793 (M + H+. C40H54NO9 requires 692.3793).
This journal is
The Royal Society of Chemistry 2007
Org. Biomol. Chem., 2007, 5, 2498–2505 | 2503
©

View Article Online
◦
Elemental analysis: Found: C, 67.75; H, 7.53; N, 1.90 Calc. for
C40H55NO10 (C40H53NO9·H2O): C, 67.68; H, 7.81; N, 1.97%.
to equilibrate for 5 minutes at 25 C before addition of external
reagents. After each addition of an aliquot of ligand solution
(containing ligand 2, 3 or 4, each solution 2 mM in histidine
residues in 20 mM MOPS, 100 mM NaCl, pH 7.4) the absorbance
spectra of the suspensions were measured from 200 to 900 nm. All
turbidity measurements were repeated several times.
11-Bis(carboxymethyl)amino-3,6,9-trioxaundecyl
4-(pyren-1-yl)butanoate (H21)
Compound 8 (120 mg, 0.2 mmol, 1 eq.) was dissolved in
trifluoroacetic acid (1.65 mL). The solution became brown in
colour after a few minutes stirring. After stirring for 5 hours
at room temperature, further trifluoroacetic acid (0.75 mL) was
added, and the mixture stirred for another 5 hours. The solvent
was removed under reduced pressure, then the resultant sticky
grey-green oil triturated with dry ether (3 mL). The residue was
redissolved in dichloromethane and the solution evaporated under
reduced pressure, which gave the product as a yellow sticky tar.
The crude product was then purified by reversed-phase column
chromatography (60 : 40 acetone–water, octadecyl-functionalised
silica gel) to give the desired product as a colourless oil (59 mg,
0.10 mmol, 51%). vmax(Nujol mull)/cm−1 714, 850, 1138, 1192,
1676, 1735; dH(300 MHz, CDCl3, 25 ◦C) 2.13–2.20 (m, 2H,
ArCH2CH2CH2), 2.46 (t, J = 5.8 Hz, 2H, ArCH2CH2CH2), 3.35
(t, J = 7.4 Hz, 2H, ArCH2CH2CH2), 3.44–3.66 (m, 16H, 8 ×
glycol CH2), 4.23 (s, 4H, 2 × CH2CO2H), 7.83 (d, J = 7.5 Hz,
1H, Ar CH), 7.94–8.15 (m, 7H, 7 × Ar CH), 8.27 (d, J = 9.0 Hz,
1H, Ar CH); dC(75 MHz, CDCl3, 25 ◦C) 27.1 (ArCH2CH2CH2),
33.1 (ArCH2CH2CH2), 34.1 (ArCH2CH2CH2), 56.0 (CH2N), 56.4
(CH2CO2H), 63.7, 65.6, 69.4, 70.4, 70.7 (all glycol CH2), 123.7,
125.2, 125.3, 126.3, 127.1, 127.8, 127.9 (all Ar CH), 129.0,
130.2, 131.2, 131.7, 136.0 (all Ar C), 168.2 (CO2CH2), 174.2
(CO2H); m/z (ES+) 580.4 ([M + H]+), 602.4 ([M + Na]+), 618.4
Fluorescence microscopy
Vesicle aggregates were observed on glass slides after dilution of
the parent vesicle solutions 1 in 10 with buffer solution to give
solutions that were 2 mM in lipid.
ITC measurements
Ligand binding was monitored after dilution of the parent vesicle
solution 1 in 10 to give a solution 2 mM in lipid. Heat flow to and
from the sample was measured after each addition of an aliquot
of ligand solution (containing ligand 2, 3 or 4, each solution 2 mM
in histidine residues in 20 mM MOPS, 100 mM NaCl, pH 7.4). All
calorimetric measurements were repeated several times.
Fluorescence
The parent vesicle solutions were diluted 1 in 1000 with buffer
solution, then added to a cuvette and left to equilibrate for 5
minutes at 25 ◦C. Pyrene fluorescence was measured by scanning
the emission spectrum from 360 to 560 nm (excitation 346 nm).
Acknowledgements
([M + K]+); m/z (ES-HRMS) 580.2524 (M + H+. C32H38NO9
+
This work was carried out with the support of the BBSRC
(Grant B19722). We would like to thank Prof. P. Bailey for use
of an isothermal calorimeter, Dr D. Jackson for fluorescence
microscopy images and Dr J. Micklefield for the use of a UV-
visible spectrophotometer.
requires 580.2541). Elemental analysis prior to reverse phase
chromatography: Found: C, 52.80; H, 5.08; N, 1.75; Calc. for
C35H42Cl2F3NO12 (C34H38F3NO11.CH2Cl2.H2O): C, 52.77; H, 5.31;
N, 1.76%. Elemental analysis after reverse phase chromatography:
Found: C, 60.80; H, 6.01; N, 2.23; Calc. for C32H39NNaO10.5
(C32H36NNaO9·1.5H2O) C, 61.14; H, 6.25; N, 2.23%.
References and notes
1 R. Coico, G. Sunshine and E. Benjamini, Immunology: A Short Course,
John Wiley and Sons, New Jersey, USA, 5th edn, 2003, ch. 4 and 5.
2 (a) L. K. Tamm and I. Bartoldus, Biochemistry, 1988, 27, 7453–7458;
(b) N. J. Lynch, P. K. Kilpatrick and R. G. Carbonell, Biotechnol.
Bioeng., 1996, 50, 169–183; (c) K. D. Lee, A. B. Kantor, S. Nir and J. C.
Owicki, Biophys. J., 1993, 64, 905–918.
3 (a) L. L. Kiessling, J. E. Gestwicki and L. E. Strong, Angew. Chem., Int.
Ed., 2006, 45, 2348–2368; (b) J. E. Gestwicki, L. E. Strong, C. W. Cairo,
F. J. Boehm and L. L. Kiessling, Chem. Biol., 2002, 9, 163–169; (c) F. M.
Menger, J. Bian, E. Sizova, D. E. Martinson and V. A. Seredyuk, Org.
Lett., 2004, 6, 261–264; (d) W. Meier, Langmuir, 2000, 16, 1457–1459.
4 (a) P. Barton, C. A. Hunter, T. J. Potter, S. J. Webb and N. H. Williams,
Angew. Chem., Int. Ed., 2002, 41, 3878–3881; (b) J. Kikuchi, K. Ariga,
T. Miyazaki and K. Ikeda, Chem. Lett., 1999, 253–254.
5 (a) F. M. Menger, J. S. Keiper and S. J. Lee, Langmuir, 1997, 13, 4614–
4620; (b) V. Marchi-Artzner, L. Jullien, T. Gulik-Krzywicki and J.-
M. Lehn, Chem. Commun., 1997, 117–118; (c) V. Marchi-Artzner, L.
Jullien, L. Belloni, D. Raison, L. Lacombe and J.-M. Lehn, J. Phys.
Chem., 1996, 100, 13844–13856.
6 (a) F. M. Menger, V. A. Seredyuk and A. A Yaroslavov, Angew. Chem.,
Int. Ed., 2002, 41, 1350–1352; (b) F. M. Menger and H. Zhang, J. Am.
Chem. Soc., 2006, 128, 1414–1415; (c) C. M. Paleos and D. Tsiourvas,
Top. Curr. Chem., 2003, 227, 1–29; (d) C. M. Paleos, Z. Sideratou and
D. Tsiourvas, ChemBioChem, 2001, 2, 305–310.
Synthesis of vesicles
Unilamellar vesicles were prepared by dissolving the appropriate
phosphatidylcholine (16 mg, 20 lmol) and the required amount
(1 lmol, 5 mol% of synthetic lipid relative to phospholipid) of
synthetic lipid in spectroscopic-grade ethanol-free chloroform
(5 mL), followed by removal of the solvent to give a thin film
of phospholipid on the interior of a round-bottomed flask. The
appropriate buffer (20 mM MOPS, 100 mM NaCl, pH 7.4 at 25 ◦C,
1 mL) was added to the flask, and the thin film detached by vortex
mixing to give a suspension of multilamellar vesicles. These were
extruded through a single 800 nm polycarbonate membrane at
∼60 ◦C, a temperature above the Tm of DSPC (54 ◦C) using
an Avestin Liposofast extrusion apparatus to give unilamellar
vesicles. The final concentration of phospholipid in these parent
vesicle suspensions was 20 mM, and that of the synthetic lipid
1 mM.
Turbidimetric measurements
7 (a) S. Chiruvolu, S. Walker, J. Israelachvili, F. J. Schmitt, D. Leckband
and J. A Zasadzinski, Science, 1994, 264, 1753–1756; (b) P. Vermette,
S. Taylor, D. Dunstan and L. Meagher, Langmuir, 2002, 18, 505–511.
The parent vesicle suspensions were diluted 1 in 10 to give
suspensions 2 mM in lipid, which were added to a cuvette and left
2504 | Org. Biomol. Chem., 2007, 5, 2498–2505
This journal is
The Royal Society of Chemistry 2007
©

View Article Online
8 E. T. Kisak, M. T. Kennedy, D. Trommeshauser and J. A. Zasadzinski,
Langmuir, 2000, 16, 2825–2831.
9 E. C. Constable, S. Mundwiler, W. Meier and C. Nardin, Chem.
Commun., 1999, 1483–1484.
10 I. Tsogas, D. Tsiourvas, G. Nounesis and C. M. Paleos, Langmuir, 2005,
21, 5997–6001.
11 Z. Sideratou, J. Foundis, D. Tsiourvas, I. P. Nezis, G. Papadimas and
C. M. Paleos, Langmuir, 2002, 18, 5036–5039.
23 S. M. Dimick, S. C. Powell, S. A. McMahon, D. N. Moothoo, J. H.
Naismith and E. J. Toone, J. Am. Chem. Soc., 1999, 121, 10286–10296.
√
n
24 Kav
=
K1 × K2 ... Kn where Ki is the microscopic binding constant for
the ith binding interaction. See N. C. Price, R. A. Dwek, R. C. Ratcliffe
and M. R. Wormald, Principles and Problems in Physical Chemistry for
Biochemists, Oxford University Press, New York, 3rd edn, 2001, ch. 4.
25 See the Supporting Information.
26 P. Kuzmic, Anal. Biochem., 1996, 237, 260–273.
12 S. J. Webb, L. Trembleau, R. J. Mart and X. Wang, Org. Biomol. Chem.,
2005, 3, 3615–3617.
27 J. Huskens, A. Mulder, T. Auletta, C. A. Nijhuis, M. J. W. Ludden and
D. N. Reinhoudt, J. Am. Chem. Soc., 2004, 126, 6784–6797.
28 E. L. Doyle, C. A. Hunter, H. C. Phillips, S. J. Webb and N. H. Williams,
J. Am. Chem. Soc., 2003, 125, 4593–4599.
29 (a) A. Mulder, T. Auletta, A. Sartori, S. Del Ciotto, A. Casnati, R.
Ungaro, J. Huskens and D. N. Reinhoudt, J. Am. Chem. Soc., 2004,
126, 6627–6636; (b) C. W. Lim, B. J. Ravoo and D. N. Reinhoudt,
Chem. Commun., 2005, 2627–2629; (c) H. Jiang and B. D. Smith, Chem.
Commun., 2006, 1407–1409.
30 M. Mammen, S.-K. Choi and G. M. Whitesides, Angew. Chem., Int.
Ed., 1998, 37, 2754–2794.
31 A. Mulder, J. Huskens and D. N. Reinhoudt, Org. Biomol. Chem., 2004,
2, 3409–3424.
32 (a) F. H. Arnold, Bio/Technology, 1991, 9, 151–156; (b) S. Sun, M. A.
Fazal, B. C. Roy, B. Chandra and S. Mallik, Inorg. Chem., 2002, 41,
1581–1590; (c) S.-S. Suh and F. H. Arnold, Biotechnol. Bioeng., 1990,
35, 682–690.
33 The latter appears to be a significant yet relatively small effect, as pre-
liminary studies in fluid-phase DSPC/cholesterol vesicles containing
5% mol/mol Cu(1) have given Kav to multivalent ligands 2 and 3 that
are enhanced 2–3-fold relative to DSPC.
13 G. S. Jacob, C. Kirmaier, S. Z. Abbas, S. C. Howard, C. N. Steininger,
J. K. Welply and P. Scudder, Biochemistry, 1995, 34, 1210–1217.
14 The turbidity of aggregated vesicles does not necessarily vary linearly
with size, so binding strengths can only be inferred. See: E. V. Pozharski,
L. McWilliams and R. C. MacDonald, Anal. Biochem., 2001, 291, 158–
162.
15 (a) B. J. Ravoo, J. Kevelam, W. D. Weringa and J. B. F. N. Engberts,
J. Phys. Chem. B, 1998, 102, 11001–11006; (b) H. J. Heerklotz, J. Phys.:
Condens. Matter, 2004, 16, R441–R467; (c) A. Pantos, D. Tsiourvas, Z.
Sideratou and C. M. Paleos, Langmuir, 2004, 20, 6165–6172.
16 D. W. Pack and F. H. Arnold, Chem. Phys. Lipids, 1997, 86, 135–152.
17 (a) A. Bouzide and G. Sauve, Org. Lett., 2002, 4, 2329–2332; (b) The
long reaction time required may be due to the b-alkoxy-substituent;
these decrease the rate of SN2 reactions approximately 10-fold. See:
F. B. Tutwiler and R. L. McKee, J. Am. Chem. Soc., 1954, 76, 6342–
6344.
18 T. A. Waggoner, J. A. Last, P. G. Kotula and D. Y. Sasaki, J. Am. Chem.
Soc., 2001, 123, 496–497.
19 Previous work using vesicles containing encapsulated 5/6-
carboxyfluorescein has shown that the vesicles remain intact and non-
leaky upon aggregation with 2. See reference 12.
34 S. Iwamoto, M. Otsuki, Y. Sasaki, A. Ikeda and J. Kikuchi, Tetrahedron,
2004, 60, 9841–9847.
20 J. Israelachvili and H. Wennerstro¨m, Nature, 1996, 379, 219–225.
21 (a) A. Pantos, Z. Sideratou and C. M. Paleos, J. Colloid Interface
Sci., 2002, 253, 435–442; (b) D. A. Noppl-Simpson and D. Needham,
Biophys. J., 1996, 70, 1391–1401.
22 (a) M. Ambrosi, N. R. Cameron, B. G. Davis and S. Stolnik, Org.
Biomol. Chem., 2005, 3, 1476–1480; (b) T. K. Dam, H.-J. Gabius, S.
Andre´, H. Katner, M. Lensch and C. F. Brewer, Biochemistry, 2005,
44, 12564–12571.
35 (a) T. Yang, O. K. Baryshnikova, H. Mao, M. A. Holden and
P. S. Cremer, J. Am. Chem. Soc., 2003, 125, 4779–4784; (b) O. D.
Hendrickson, A. V. Zherdev, A. P. Kaplun and B. B. Dzantiev, Appl.
Biochem. Microbiol., 2003, 39, 75–81.
36 M. Kasbauer, D. D. Lasic and M. Winterhalter, Chem. Phys. Lipids,
1997, 86, 153–159.
37 R. J. Mart, K. P. Liem, X. Wang and S. J. Webb, J. Am. Chem. Soc.,
2006, 128, 14462–14463.
This journal is The Royal Society of Chemistry 2007
Org. Biomol. Chem., 2007, 5, 2498–2505 | 2505
©