Intramolecular carbohydrate-aromatic interaction and intermolecular van der Waals interactions enhance the molecular recognition ability of GM1 glycomimetics for cholera toxin

Article abstract of DOI:10.1002/chem.200400084

The design and synthesis of two GM1 glycomimetics, 6 and 7, and analysis of their conformation in the free state and when complexed to cholera toxin is described. These compounds, which include an (R)-cyclohexyllactic acid and an (R)-phenyllactic acid fra

Full text of DOI:10.1002/chem.200400084

A. Bernardi, J. JimØnez-Barbero et al.
Chem. Eur. J. 2004, 10, 4395 – 4406
DOI: 10.1002/chem.200400084
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4395

FULL PAPER
Intramolecular Carbohydrate–Aromatic Interactions and Intermolecular
van der Waals Interactions Enhance the Molecular Recognition Ability of
GM1 Glycomimetics for Cholera Toxin
Anna Bernardi,*[a] Daniela Arosio,[a] Donatella Potenza,[a]
Inmaculada Sµnchez-Medina,[a] Silvia Mari,[b] F. Javier Caæada,[b] and
Jesffls JimØnez-Barbero*[b]
Dedictated to the memory of Dr. Juan Carlos del Amo, postdoctoral fellow, killed in Madrid, in the terrorist attack of March 11th
Abstract: The design and synthesis of
two GM1 glycomimetics, 6 and 7, and
analysis of their conformation in the
free state and when complexed to chol-
era toxin is described. These com-
pounds, which include an (R)-cyclohex-
yllactic acid and an (R)-phenyllactic
acid fragment, respectively, display sig-
nificant affinity for cholera toxin. A de-
tailed NMR spectroscopy study of the
toxin/glycomimetic complexes, assisted
by molecular modeling techniques, has
allowed their interactions with the
toxin to be explained at the atomic
level. It is shown that intramolecular
van der Waals and CH–p carbohy-
drate–aromatic interactions define the
conformational properties of 7, which
adopts a three-dimensional structure
significantly preorganized for proper
interaction with the toxin. The exploi-
tation of this kind of sugar–aromatic
interaction, which is very well descri-
bed in the context of carbohydrate/pro-
tein complexes, may open new avenues
for the rational design of sugar mimics.
Keywords: carbohydrate–aromatic
interactions · carbohydrate–protein
interactions
·
cholera toxin
·
glycomimetics · oligosaccharides
Introduction
bacterial enterotoxins, the cholera toxin (CT) and the heat-
labile toxin of Escherichia coli (LT). Bothtoxins are hex-
americ AB5 proteins and use the GM1 headgroup pentasac-
charide (o-GM1, 1b, Scheme 1) as their molecular target to
attack and penetrate the host cells. The recognition pair
composed of GM1 and these two bacterial enterotoxins has
been particularly well studied, bothfrom a biochemical and
a structural point of view.[2,3] This information served as the
basis for the rational design of the pseudooligosaccharide 2
(Scheme 1),[4,5] which was found to be as active as GM1 in
binding to CT.[4] More recently, we have reported on a
group of second-generation mimics, 3–7 (Scheme 1),[6,7] ob-
tained by replacing the sialic acid (NeuAc) moiety of 2 with
simple a-hydroxy acids. All of these compounds show mod-
erate to good affinity for CT (Table 1); this affinity depends
critically on the configuration of the hydroxy acid stereocen-
ter and on the nature of the substituent R (Table 1).
Many biologically significant processes are known to be con-
trolled in their early stages by the interaction of proteins
with oligosaccharides, often in the form of glyconjugates.
Thus, oligosaccharide mimics that can antagonize oligosac-
charides at the protein receptor level are receiving much at-
tention, bothas tools to modulate or alter signal transmis-
sion and to be developed into drugs.[1] Our groups have
been exploring this field by using the GM1 ganglioside 1a
as a model system (Scheme 1). GM1 is a membrane glycoli-
pid which functions as the cellular receptor of two related
[a] Prof. Dr. A. Bernardi, Dr. D. Arosio, Dr. D. Potenza,
I. Sµnchez-Medina
Università di Milano—Dipartimento di Chimica Organica
e Industriale e Centro di Eccellenza CISI
via Venezian 21, 20133 Milano (Italy)
Fax : (+39)02-50314072
The design process of this series of mimics was supported
by extensive NMR spectroscopy studies.[8] After the first
group of three ligands (3–5, Scheme 1) was synthesized and
tested, their conformation was investigated first in solution
and then upon binding to the cholera toxin by using trans-
ferred nuclear Overhauser effect (TR-NOE)[9,10] measure-
ments. It was found that CT selects a conformation similar
to the global minimum of the free pseudosaccharides from
E-mail: anna.bernardi@unimi.it
[b] S. Mari, Dr. F. J. Caæada, Prof. Dr. J. JimØnez-Barbero
Dept. de Estructura y Función de Proteínas
Centro de Investigaciones Biológicas, C.S.I.C.
Ramiro de Maeztu 9, 28040 Madrid (Spain)
Fax : (+34)91-5360432
E-mail: jjbarbero@cib.csic.es
Supporting information for this article is available on the WWW
under http://www.chemeurj.org/ or from the author.
4396
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
DOI: 10.1002/chem.200400084
Chem. Eur. J. 2004, 10, 4395 – 4406

4395 – 4406
tional studies in the free state
and in the complex with CT are
discussed. A preliminary com-
munication has been reported.[7]
Definitions and abbreviations:
Residues of the pseudosugars
5–7 are defined as indicated in
Scheme 2. The CHD residue is
numbered as depicted in
Scheme 2, to help comparison
withteh brancihng galactose
unit of GM1. Glycosidic angles
are
defined
as
follows:
Galb(1!3)GalNAc:
F=
GalH1-GalC1-O1-GalNAcC3,
Y=GalC1-O1-GalNAcC3-Gal-
NAcH3; Galb(1!4)CHD: F=
GalH1-GalC1-O1-CHDC4, Y=
GalC1-O1-CHDC4-CHDH4;
hydroxy acid–CHD: F=C(O)-
Ca-O-CHDC3,
Y=Ca-O-
CHDC3-CHDH3. The improp-
Scheme 1. Ganglioside GM1 and its mimics.
Table 1. CT affinity constants of the second-generation GM1 mimics 3–7
as determined by fluorescence titration experiments.[a]
Compound
R (configuration)
Kd [mm]
Ref.
[6]
3
4
5
6
7
H
Me (S)
Me (R)
CH2C6H11 (R)
CH2Ph( R)
750
1000
190
45
[6]
[6]
[7]
[7]
10
[a] Kd values were obtained from fluorescence intensity titrations of
0.5 mm CT solutions (in tris(hydroxymethyl)aminomethane (Tris) buffer,
pH 7, room temperature) and nonlinear fitting (SigmaPlot) of data ob-
tained in duplicate runs.
Scheme 2. GM1 mimics nomenclature, abbreviations, and definitions.
Gal=galactose, GalNAc=N-acetyl galactosamine, CHD=cyclohexane-
diol, HL =H-L in the text and other figures.
the ensemble of the presented conformations, and no evi-
dence of major conformational distortions was obtained. In
the free state the three molecules were shown to be rather
flexible in the hydroxy acid region. In the bound state[8] the
protein appeared to select for binding one or two side-chain
conformations that could reproduce the orientation of the
NeuAc carboxy group in GM1. The NMR data were inter-
preted with the aid of molecular-modeling techniques, which
allowed workable models for the ligand:toxin complexes to
be derived. These models suggested that the higher affinity
of the (R)-lactic acid derivative 5 relative to 3 and 4
(Table 1) could result from van der Waals interactions estab-
lished between its side-chain methyl group and a hydropho-
bic area in the toxin binding site near the sialic acid side-
chain binding region of the CT/GM1 complex. This informa-
tion in turn suggested that the affinity of the pseudo-GM1
binders could be improved by adding appropriate hydropho-
bic fragments to the framework of the (R)-lactic acid GM1
mimic 5.[8] We now report on the pseudo-GM1 ligands 6 and
7, which include a cyclohexyl group and a phenyl group, re-
spectively. Their synthesis is described (see the Supporting
Information) and detailed NMR spectroscopy and computa-
er dihedral angle c (see Figure 3) describes the relative ori-
entation of the carboxy group and the CHD ring and is de-
fined as follows: c=C(O)-Ca-CHDC3-CHDH3. The dihe-
dral angle q (see Figure 4) describes the orientation of the
cyclohexyl group in the hydroxy acid side chain and is de-
fined as q=C(O)-Ca-CH2-CCy.
Results
Synthesis of the ligands: The synthetic pathway followed for
the preparation of ligands 6 and 7 is shown in Scheme 3.
The full synthetic sequence and product characterizations
are reported in the Supporting Information for this paper.
Binding-affinity determination: Dissociation constants (Kd)
were obtained from fluorescence intensity titrations of
0.5 mm CT solutions (in Tris buffer, pH 7, room temperature)
Chem. Eur. J. 2004, 10, 4395 – 4406
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4397

A. Bernardi, J. JimØnez-Barbero et al.
FULL PAPER
Scheme 3. Synthesis of 6 and 7. a) Bu2SnO, benzene, reflux; then CsF, DME, and 9 or 10; b) 13 (0.5 equiv) and TfOH (0.05 equiv) in CH2Cl2, RT!
reflux; c) H2/Pd-C, MeOH; then cat. MeONa in MeOH. Bn=benzyl, DME=1,2-dimethoxyethane, TCA=trichloroacetic acid, Tf=triflate=trifluorome-
thanesulfonyl.
as previously described.[7] In particular, the Kd values shown
in Table 1 were obtained by nonlinear fitting (SigmaPlot) of
data obtained in duplicate runs. Remarkably, the simple
The intraresidue NOE cross-peaks also support the theory
that all the six-membered rings of the molecules are in chair
conformations (see below and Tables 2 and 3).
(R)-2-hydroxy-3-phenyl
pro-
pionic acid (phenyllactic acid)
derivative 7 showed a dissocia-
tion constant of 10 mm, which is
only one order of magnitude
less potent than the monovalent
association of the natural ligand
o-GM1 against the cholera
toxin.[11] The cyclohexyl deriva-
Table 2. Principal NOE contacts for compound 6 in the free state and bound to cholera toxin. The distances r
(ꢀ10%) are estimated according to a full-matrix relaxation approach.[32] The intraresidue Gal and GalNAc
H-1/H-3 and H-1/H-5 contacts were taken as internal references.[a]
Proton pair
Observed intensity
free state
Deduced r
[Š]
Observed intensity
bound state
Deduced r
[Š]
H-1 GalNAc
H-4 CHD
H-L
strong
n.o.[a]
strong
medium
strong
n.o.[a]
2.4
>3.5
2.4
strong
2.4
>3.5
2.4
n.o.[a]
H-1 Gal
H-3 CHD
H-3 GalNAc
H-L
strong
strong
strong
very weak
medium-strong
strong
very weak
very weak
strong
2.7
2.5
[b]
tive
6 showed a somewhat
CH2
H-2ax CHD
H-2eq CHD
H-3 CHD
H-4 CHD
H-2ax CHD
H-1 GalNAc
H-L
lower affinity, with Kd =45 mm;
this is still better than the value
obtained for the simple (R)-
H-L
weak
2.9
2.6
3.1
>3.5
2.5
>3.5
2.7
2.5
3.2
3.3
2.5
3.2
medium
very weak
n.o.[a]
strong
n.o.
lactic
derivative
5
(Kd =
190 mm). This experimental fact
enforces the hypothesis drawn
H-4 CHD
weak
from the previous modeling [a] ax=axial, eq=equatorial, n.o.=no observable NOE contact. [b] CH2 and H-5ax CHD are isochronous.
studies[8] and emphasizes the
importance of setting a nonpo-
lar group bulkier than a methyl group at the proper posi-
tion. The fine details of the differences between 6 and 7 will
be given below.
Vicinal J5,6 couplings for the hydroxymethyl groups could
be measured for 6 and 7. The coupling values were between
7.5–9 Hz for boththe Gal and GalNAc residues, in agree-
ment witha gt:tg equilibrium of the w torsion angle, as is
usually the case for these Gal-type sugars.[13]
NMR studies of the free ligands 6 and 7: NMR spectroscopy
experiments were carried out at 400 and 500 MHz at tem-
peratures of 293–300 K. A complete assignment of the H
The conformation of oligosaccharides is defined by the F
and Y torsion angles around the glycosidic linkages
1
and 13C NMR signals of 6 and 7 was achieved on the basis
of COSY, TOCSY, HSQC, and NOESY experiments. Chem-
ical shifts and coupling constants are reported in the Sup-
porting Information. No changes in the NOESY spectra
were noticed upon addition of calcium ions (up to 5 mm). It
has been reported[12] that GM1 and other gangliosides might
form complexes withcalcium ions. Probably, the lack of a
lipid chain, which might also affect local concentrations,
may somehow influence this type of interaction in this case.
For both compounds, the analysis of the vicinal proton–pro-
ton coupling constants for the six-membered rings indicates
(Scheme 2). For 6 and 7, the rotation around the ether link-
age in the hydroxy acid side chain must also be described.
F,Y definitions can be extrapolated from the glycosidic
bond convention and defined as described in Scheme 2. Ex-
perimental information on these geometrical parameters can
be gathered by using NOE measurements.[14] Specifically, se-
lective 1D NOESY and 2D T-ROESY[14] experiments were
carried out that provided experimental interproton distan-
ces.
These, in turn, allowed the conformation equilibria pres-
ent in solution to be profiled and average molecular confor-
mations to be derived (Tables 2 and 3).
4
that the Gal and GalNAc chairs are in the usual C1 confor-
mation. The diol moiety also adopts a chair conformation
with the ester groups in the equatorial orientations.
The relationship between NOE signals and proton–proton
distances is well established[15,16] and can be worked out at
4398
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 4395 – 4406

Molecular Recognition of Glycomimetics for Cholera Toxin
4395 – 4406
Table 3. Principal NOE contacts for compound 7 in the free state and when bound to cholera toxin. The dis-
tances r (ꢀ15%) are estimated according to a full-matrix relaxation approach.[32] The intraresidue Gal and
GalNAc H-1/H-3 and H-1/H-5 contacts were taken as internal references.
be correlated withone major
orientation
around
the
Galb(1!3)GalNAc F/Y tor-
sion angles. This region corre-
sponds to the global minimum
syn conformation, characterized
by F/Y values of around 608/
ꢁ208.
Proton pair
Observed intensity
free state
Deduced r
[Š]
Observed intensity
bound state
Deduced r
[Š]
H-1 GalNAc H-4 CHD
strong
n.o.
strong
2.8
medium
medium
weak
medium
medium
strong
strong
weak
2.4
>3.5
2.4
medium
2.7
2.8
3.1
2.8
2.8
2.3
2.4
3.0
2.5
>3.5
2.5
>3.5
medium-strong
n.o.
strong
2.8
medium
medium
medium
medium
medium
strong
strong
weak
2.5
>4.0
2.4
H-L
H-1 Gal
PhH-1 GalNAc
H-3 GalNAc
medium
H-L
2.7
2.8
2.8
2.8
2.8
2.3
2.4
3.2
2.5
3.2
2.5
3.2
A similar situation occurs for
H-4 CHD
H-6 GalNAc
H-3 GalNAc[a]
H-5 GalNAc
H-3 CHD[a]
H-2eq CHD
H-1 CHD
H-1 GalNAc
H-L
the
GalNAcb(1!4)CHD
moiety, witha similar NOE pat-
tern.[8,18] In fact, only the very
H-L
strong
interresidue
H-1
GalNAc/H-4 CHD cross-peak
is observed; this cross-peak cor-
responds to a H-1 GalNAc/H-4
CHD distance of 2.5ꢀ0.2 Š.
The average conformation in
solution for this linkage corre-
sponds to the global minimum
syn conformer, characterized by
H-4 CHD
strong
strong
strong
n.o.
strong
weak
strong
weak
H-1 GalNAc
H-L
[a] H-3 GalNAc and H-3 CHD are isochronous.
least semiquantitatively by using a relaxation matrix.[16] The
NOE intensities reflect the conformer populations, and
therefore information on the population distributions in free
solution can be obtained by focusing on the key, mutually
exclusive NOE interactions that characterize the different
possible conformations.[17]
At 400 MHz all the cross-peaks observed in the NOE
spectra of 6 and 7 were very weak. The wtc value is close to
1.1, which provides an almost zero longitudinal NOE con-
tact (tC is the overall correlation time). Although some ini-
tial information could be derived from T-ROESY spectra,
many crucial proton signals (for example, H-L and H-2
GalNAc) were basically isochronous; this prompted us to a
500 MHz analysis. At 293 K, the corresponding NOESY
spectra (Figures 1 and 2) were
F/Y values of around 508/208. Conformers in the anti-F or
anti-Y regions of bothglycosidic linkages withpopulations
above 5% would give rise to exclusive H-1 Gal/H-2, H-4
GalNAc and/or H-1 GalNAc/H-5ax, H-5ax CHD NOE con-
tacts that are not observed in these cases.
Regarding the orientation of the acetamide moiety[18] of
both 6 and 7, the cross-peak of the methyl group with the
corresponding H-2 GalNAc proton is a very weak, and no
cross-peaks to either H-1 or H-3 GalNAc are observed.
Therefore, these observations are in agreement with a major
orientation in solution with the methyl group pointing out
of the pyranoid ring with an anti-like relationship with the
C-2 atom.
good enoughto extract the ob-
served cross-peaks, which are
reported in Tables 2 and 3.
The orientation around the
Galb(1!3)GalNAc linkage can
be defined[8,18] by the NOE con-
tact observed between the
anomeric proton of galactose,
H-1 Gal, and the protons on
the GalNAc moiety, especially
H-3, H-4, and H-2. The H-1
Gal/H-3 GalNAc cross-peak is
very strong for both 6 and 7
(see Tables 2 and 3), withan in-
teraction similar to or even
stronger than the corresponding
intraresidue H-1 Gal/H-3 Gal
and H-1 Gal/H-5 Gal cross-
peaks. As we have already de-
scribed for 3–5,[8] the H-1 Gal/
H-3 GalNAc distance obtained
from the NMR spectroscopy
data (approximately 2.4 Š) can Figure 1. 500 MHz NOESY spectrum of 6 in D2O at 293 K.
www.chemeurj.org
Chem. Eur. J. 2004, 10, 4395 – 4406
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4399

A. Bernardi, J. JimØnez-Barbero et al.
FULL PAPER
determine similar through-
space proton–proton interac-
tions for the reporter H-L.
Thus, the experimental inter-
proton distances generated by
NMR spectroscopy analysis can
be analyzed more conveniently
by using the improper dihedral
angle descriptor c(C(O)-Ca-
CHDC3-CHDH3), which is de-
fined in Figure 3, and univocal-
ly describes the orientation of
H-L (and of the hydroxy acid
carboxy group) relative to the
CHD ring. In principle, three
idealized staggered orientations
can be drawn across the ether
connector (Figure 3); these can
be identified as anti-c, (ꢁg)-c,
and (+g)-c and they give rise to
specific interactions between H-
L and the protons on the C-3,
Figure 2. Fragments of the 500 MHz NOESY spectra of 7 in D2O at 300 K, from which the orientation of the
C-2 and C-4 carbon atoms of
the CHD ring. The major orien-
tation(s) of the hydroxy acid
relative to the CHD ring can
side chain is deduced. The NOE contacts for H-L (top) and for the aromatic protons (bottom) are indicated.
Thus, all data so far concur to describe the “upper” por-
tion of both 6 and 7 as a pseudotrisaccharide with relatively
little flexibility, which can be described by oscillations
around one single major conformation. This is the same sit-
uation observed for all the GM1 mimics synthesized so
far,[8] and for the Galb(1!3)GalNAcb(1!4)Gal fragment
of o-GM1 itself.[19]
The additional torsional degrees of freedom available for
6 and 7, correspond to the hydroxy acid side-chain bonds.
Due to the absence of anomeric effects, the ether linkage
that connects the CHD ring to the hydroxy acid moiety is
more flexible than the interglycosidic linkages described so
far. The NOE contacts observed in this region can arise
form multiple combinations of F,Y values that happen to
thus be defined by focusing on the NOE cross-peaks for the
H-L proton to H-3, H-4, H-2eq, and H-2ax of CHD
(Tables 2 and 3). Further information on the mobility of the
side chain and on its conformation(s) can be gathered from
the NOE contacts observed for the R side-chain substituent.
For the cyclohexyl derivative 6, medium and weak NOE
contacts are observed between H-L and the key CHD pro-
tons (see Table 2). Nevertheless, the NOE intensity of the
H-L/H-3 CHD cross-peak is higher than that of the H-L/H-
2eq CHD connectivity and also stronger than the H-L/H-4
CHD one. The H-L/H-1 GalNAc and H-L/H-2ax NOE
cross-peaks are also negligible. In turn, the cyclohexyl
moiety of the side chain only displays weak cross-peaks to
its vicinal H-3 CHD proton. No cross-peaks are observed
Figure 3. Newman projections along the ether linkage between the CHD ring and the hydroxy acid moiety. The exclusive NOE contacts expected for
H-L in each rotamer are shown. Each of these projections may correspond to multiple F,Y combinations. c=the improper dihedral angle. See text for
further details.
4400
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 4395 – 4406

Molecular Recognition of Glycomimetics for Cholera Toxin
4395 – 4406
between the cyclohexyl side chain and H-2eq or H-4 CHD.
The absence of cross-peaks between H-L and H-1 GalNAc
and the very weak intensity of those between H-L and H-4
CHD and between H-L and H-2 CHD are consistent witha
very flexible side chain and indicate that probably all three
staggered rotamers depicted in Figure 3 are contributing to
the equilibrium.
of 7 in aqueous solution. Importantly, the origin of this con-
formation lock appears to be intramolecular van der Waals
and CH–p interactions between the aromatic ring and the
GalNAc residue; this may be described as hydrophobic
packing. These interactions are not so favorable in 6 where
the cyclohexyl ring obviously differs from the phenyl moiety
in its electronic and geometrical features. The importance of
carbohydrate–aromatic interactions for the molecular recog-
nition of oligosaccharides by the binding sites of proteins
has been well documented.[20–24] Here, we present clear evi-
dence of the importance of this type of interaction for favor-
ing a given type of three-dimensional structure.
Computational models of 6 and 7 were generated by
using previously established protocols (MC/EM conforma-
tion searches and MC/SD molecular-dynamics simulations,
AMBER* force field augmented by the Kolb parameters[25]
for hydroxy acids, GB/SA water solvation) and compared to
the experimental results. The calculations showed that the
Gal–GalNAc–CHD fragment of bothmolecules is populat-
ing the syn conformation for bothglycosidic linkages.
Minima were located at f/y values of 508/08 for the Gal–
GalNAc linkage and 258/308 for the GalNAc–CHD linkage.
By contrast, and in agreement withthe experimental data
discussed above, the computational description of the hy-
droxy acid side chain differs significantly for the two mole-
cules. For 6, MC/EM calculations located three conformers
within 1 kcalmolꢁ1 of the global minimum, at c=150, 65,
and 398 (Figure 4a and Table 4). MC/SD dynamics showed
that the molecule is highly flexible along the variable c
angle and continuously populates an ample region with c
varying from 40–1608 plus a secondary low-energy region at
c=ꢁ608 (Figure 4b). The presence of this extensive confor-
mation equilibrium agrees withthe experimental observa-
tion.
For the conformation analysis of 7, the key NOE contacts
that allow the conformation equilibrium present in solution
to be described were obtained from NOESY and T-ROESY
experiments carried out at 500 MHz and 300 K. Several in-
terresidual cross-peaks were observed. As described above
for 6, the major arrangement of the hydroxy acid side chain
can be defined by focusing on the NOE contacts between
H-L and the CHD protons. In this case, important informa-
tion is also obtained from the through-space interactions be-
tween the aromatic protons and the CHD and GalNAc resi-
dues (Figure 2, Table 3). The H-L proton shows NOE cross-
peaks of strong intensity withH-2eq and H-3 CHD (with
corresponding distances of 2.4 and 2.3ꢀ0.1 Š, respectively),
while no cross-peaks are observed between H-L and H-4
CHD (Figure 2a, Table 3). This pattern, which is markedly
different from what was measured for 6, is in agreement
withone major anti-c orientation (Figure 3) between the
carboxyl group and the H-3 CHD proton. In addition, the
aromatic protons also show clear cross-peaks to a variety of
protons (Figure 2b, Table 3) located not only on the CHD
moiety (H-4), but also on the a face of the GalNAc residue
(H-1, H-3, and H-5, Figure 2b). Thus, it appears that in the
major solution conformation of 7 the phenyl moiety stacks
below the GalNAc residue.
Collectively, these experimental data allow us to conclude
that the phenyllactic derivative 7 is significantly less flexible
in the hydroxy acid fragment than the cyclohexyllactic ana-
logue 6 and suggest that a single conformation of the hy-
droxy acid chain can describe the conformational properties
MC/EM calculations also showed that the side chain of 7
is significantly less flexible than that of 6, although computa-
Figure 4. Conformation analysis of 6. a) Low-energy conformations within 1.2 kcalmolꢁ1 of the global minimum. b) MC/SD dynamics simulations (5 ns).
The improper dihedral angle, c, is plotted against the dihedral angle, qC(O)-Ca-CH2-CCy.
Chem. Eur. J. 2004, 10, 4395 – 4406
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4401

A. Bernardi, J. JimØnez-Barbero et al.
FULL PAPER
Table 4. Three different conformation (conf.) families were found by an MC/EM[a] conformation search, each
witha different value of the improper angle, c (F=C(OO)-Ca-O-CD3; c=C(OO)-Ca-CD3-HD3). The key interatom-
ic proton–proton distances are outlined.
tions of 6 and 7 to the CT
B pentamer. As previously
shown, for ligands which are
not bound tightly and exchange
between the free and bound
states at a reasonably fast rate,
the TR-NOESY experiment
provides an adequate means for
determining the conformation
of the bound ligand.[9,10] As
mentioned above (Table 1), the
present systems bind in the mi-
cromolar range to CT and thus
can properly be studied by this
technique. The addition of chol-
era toxin to a D2O solution of 6
and 7 induced broadening of
Distances [Š]
conf. 1
(DE=0.0 kcalmolꢁ1
c=146.68)
conf. 2
(DE=0.2 kcalmolꢁ1
c=658)
conf. 3
(DE=1.1 kcalmolꢁ1
c=398)
,
,
,
H-L/H-3 CHD
2.51
2.25
3.39
4.39
4.36
2.69
2.81
3.92
4.49
3.22
3.71
2.24
2.42
4.24
4.47
2.47
2.81
2.31
H-L/H-2eq CHD
H-L/H2ax CHD
H-L/H-4 CHD
H-L/H-1 GalNAc
H-1 GalNAc/H-4 CHD
[a] 10000 steps of MC/EM withthe H 2O GB/SA solvent model and the Amber* force field were employed.
tionally the restriction of mobility is mostly confined to the
orientation of the phenyl ring, which is always calculated to
stack either with the Gal or the GalNAc ring (Figure 5).
Two energetically equivalent orientations of the carboxyl
group were calculated, with cꢂ1508 (anti-c) and cꢂ608
((+)g-c; Figure 5a and b). However, in this case, the NMR
experimental data (see above) allows us to conclude that
only one major orientation of the side chain, corresponding
to the anti-c conformation, is in fact present. The best fit
withthe NOE data is represented by a conformation found
near the global minimum, as shown in Figure 5c. Based on
these data, ligand 7 appears to be preorganized for toxin
binding, in contrast with the conformational behavior of its
aliphatic analogue 6, which is fairly flexible.
It is interesting to note that the stacking effect between
the aromatic ring and the sugars is somewhat “understood”
by the molecular mechanics computations, which most likely
“see” it as a solvation effect (the hydrophobicity of the ring)
combined witha positive van der Waals interactions be-
tween the phenyl group and the a face of the sugars. This
latter interaction appears to be precluded to the cyclohexyl
substituent for steric reasons.
the resonance signals in the 1H NMR spectrum, a result indi-
cating that binding occurs (Figures S1 and S2 in the Support-
ing Information). TR-NOESY experiments were performed
on the ligand/CT samples at different mixing times and with
different ligand to toxin molar ratios (25:1 to 50:1). Nega-
tive cross-peaks were clearly observed at 300 K, as expected
for ligand binding.
For the phenyllactic acid based ligand 7, the cross-peak
pattern is similar to the one described above for free 7 in
aqueous solution. Indeed, the same set of interresidue cross-
peaks withbasically identical intensities (relative to the in-
traresidual contacts) were observed (Table 3). The orienta-
tion around the Galb(1!3)GalNAc and GalNAcb(1!
4)CHD fragments is unchanged upon binding, as deduced
by the corresponding H-1 Gal/H-3 GalNAc and H-1
GalNAc/H-4 CHD cross-peaks. The orientation of the phe-
nyllactic acid moiety relative to the GalNAc and CHD resi-
dues is also almost identical to that observed in the free
state, as shown by the H-L/H-2eq CHD, H-L/H3 CHD, and
aromatic/H-1, H-3, H-5 GalNAc short contacts (Figure 6,
Table 3). TR-ROESY experiments allowed spin-diffusion ef-
fects to be excluded for these key cross-peaks.[29]
For 6, TR-NOESY experiments were also performed at a
32:1 ligand/receptor molar ratio. Negative cross-peaks were
clearly observed at 300 K, bothat 400 and 500 MHz, a result
indicating ligand binding (Table 2). In this case, the analysis
NMR studies of the CT complexes: NMR experiments, in-
cluding TR-NOESY[9,10] and saturation transfer difference
(STD),[26–28] were performed to deduce the bound conforma-
Figure 5. Two major conformation families, a and b, were found for the side chain of 7 by the MC/EM conformation search. They are characterized by a
different value of the improper torsional angle, c. According to the experimental results, only the anti-c conformation (cꢂ1508) is found in solution. The
best fit with the NOE data is obtained with conformer c, which belongs to the a family.
4402
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 4395 – 4406

Molecular Recognition of Glycomimetics for Cholera Toxin
4395 – 4406
trum showed that the aromatic
protons, together with H-4 Gal
and H-6 Gal, give rise to the
most prominent STD signals
(Figure 8). Thus, it can be infer-
red that this region (aromatic
ring and Gal moiety) of the gly-
comimetic is in more intimate
contact withthe toxin binding
site. Additionally, the resonance
signals of H-3 and H-1 Gal, H-1
GalNAc, and H-4 CHD also
appeared in the difference spec-
trum but withsmaller intensi-
ties. The weakest STD signals
were observed for the CHD
moiety protons, except for H-4,
a result indicating that this part
of the molecule is not involved
in the molecular recognition
process (Figure 8) to a signifi-
Figure 6. TR-NOESY spectrum of 7 bound to CT (solvent D2O, 293 K).
cant extent.
The STD spectrum of the
of the cross-peaks shows that, out of the conformation equi-
librium observed in the free state, one main conformation,
which features an anti-type relationship of the carboxyl
group to the H-3 CHD proton (anti-c in Figure 3), is select-
ed for binding. In fact, cross-peaks between H-L/H-2eq
CHD and H-L/H-3 CHD were observed (Figure 7) that
agreed with the conformation selection of the anti-c rotamer
(cꢂ1508). A minor proportion of a gauche-c rotamer
cannot be discarded, as indicated by the presence of a very
weak H-L/H-4 CHD TR-NOESY cross-peak.
CT/6 mixture (Figure 9) also allowed the binding epitope to
be deduced. Resonance signals belonging to the Gal moiety
(H-1, H-2, H-4, H-5, H-6) were evident in the difference
spectrum. Additional peaks for H-1, H-4, and H-5 GalNAc
were also clearly observed. However, those belonging to the
CHD ring and to the pendant cyclohexyl moiety were rather
weak, or basically nonexistent, including the H-3 CHD and
H-4 CHD protons (see Figure 9). This indicates that these
moieties establishonly marginal interactions withthe pro-
tein, in contrast to the aromatic ring of the phenyllactic de-
rivative 7.
Finally, in order to map the binding epitope of the glyco-
mimetics bound to the cholera toxin, 1D STD NMR experi-
ments were also performed.[26–28] For 7, the difference spec-
As a final step, a three-dimensional model of the complex
structures was obtained by performing MC/EM calculations
within the binding site of the
LT toxin, according to the pro-
cedure already described for
the lactic acid analogues.[8] The
corresponding views are depict-
ed in Figure 10. It clearly ap-
pears that the phenyl ring in the
LT/7 complex (Figure 10a) is
calculated to be in close prox-
imity to the protein binding site.
On the contrary, and in accord-
ance withexperimental obser-
vations, the cyclohexyl ring in
LT:6 projects away from the
protein surface (Figure 10b).
Dynamics simulations (MC/
SD) also showed that the
phenyl ring of 7 is locked be-
tween the protein and the
GalNAc ring (see Figure S3
(animated gif file) in the Sup-
porting Information). A clear
Figure 7. TR-NOESY spectrum of 6 bound to CT (solvent D2O, 300 K).
intramolecular carbohydrate–ar-
Chem. Eur. J. 2004, 10, 4395 – 4406
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4403

A. Bernardi, J. JimØnez-Barbero et al.
FULL PAPER
omatic stacking interaction
takes place and locates the
phenyl ring in the proper posi-
tion to establishadditional con-
tacts with the hydrophobic
patchin the protein. The typical
Gal–toxin interaction, which
defines the structure of cholera
toxin complexes, is also evident.
In contrast, in the LT/6 com-
plex the cyclohexyl ring appears
to display more flexibility even
in the bound state, with the Gal
moiety establishing the major
contacts withteh toxin. No
stacking of the GalNAc moiety
with the aliphatic cyclohexyl
Figure 8. a) Reference 1D 1H NMR spectrum of 7 withCTB5 (ratio 25:1) in D 2O at 300 K. b) STD spectrum
with 1 s presaturation of the protein envelope protons. Ligand proton signals are evident in the aromatic
region. Protons are those belonging to the Gal moiety (H-4 and H-6). H-1 GalNAc, H-1 Gal, H-4 CHD, and
H-3 GalNAc are also clearly visible, while the other protons of the CHD unit do not appear in the STD spec-
trum.
side chain takes place. The cy-
clohexyl ring appears to move
in and out of the toxin binding
side and to sample at least two
different orientations relative
to the protein cavity (see Fig-
ure S4 (animated gif file) in the
Supporting Information).
Discussion
All the known structural data
on the complexes formed be-
tween CT and GM1 or the
pseudo-GM1 mimics 2–5 show
that the principal interactions
between the protein and the
(pseudo)carbohydrate ligands
are established through the
nonreducing end galactose unit
and the carboxy group of the
NeuAc residue (in 1 and 2) or
of the surrogate hydroxy acid
(in 3–5). NMR and computa-
tional studies performed on the
complexation of 3–5 by CT
Figure 9. a) Reference 1D 1H NMR spectrum of 6 withCTB5 (ratio 32:1) in D 2O at 298 K. b) STD spectrum
with 2 s presaturation of the protein envelope protons. Protons belonging to the GalNAc and Gal moiety (H-1
GalNAc, H-1 Gal, H-4 GalNAc, H-2 GalNAc, H-4 Gal, H-6 GalNAc, H-6 Gal, H-5 GalNAc, H-5 Gal, and
H-2 Gal) are clearly visible.
have suggested that additional van der Waals interactions
can be gained by lipophilic substituents on the (R)-hydroxy
acid side chain,[8] which, in the bound conformation of the
ligand, can extend towards a hydrophobic area of the toxin
binding site close to the region that accommodates the sialic
acid side chain in the CT/GM1 complex.[3] The present work
shows that the pseudo-GM1 ligands 6 and 7, which include
an (R)-cyclohexyllactic acid fragment and an (R)-phenyllac-
tic acid fragment, respectively, do indeed display stronger af-
finity for CT than 5.
Although the two ligands described in this paper are simi-
lar in nature and activity, a detailed analysis of their behav-
ior in solution and in the binding site of the cholera toxin re-
veals striking differences, bothin terms of conformational
flexibility and of binding mode. Intra- and intermolecular in-
Figure 10. a) Molecular model (MC/EM, Amber*, H2O, GB/SA, best fit
withexperimental data for 7). b) Molecular model (MC/EM, Amber*,
H2O, GB/SA, lowest energy conformer for 6).
4404
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 4395 – 4406

Molecular Recognition of Glycomimetics for Cholera Toxin
4395 – 4406
spectroscopy tube to give a final concentration of approximately 0.1–
0.2 mm. TR-NOESY experiments were performed withmixing times of
100, 200, and 300 ms, for molar ratios of ligand/protein between 15:1 and
50:1. No purging spin-lock period to remove the protein background sig-
nals was employed. First, in all cases, line broadening of the sugar pro-
tons was monitored after the addition of the ligand. STD experiments
were carried out by using the method proposed by Meyer, Peters, and
their respective co-workers.[26–28] No saturation of the residual HDO
signal was employed and, again, no spin-lock pulse was employed
remove the protein background signals. In our hands, the use of a spin-
lock period induced artifacts in the difference spectrum. The theoretical
analysis of the TR-NOE contacts of the sugar protons was performed ac-
teractions among the different residues strongly modulate
the conformational features of these molecules in solution
and when bound to the toxin. The major differences be-
tween 6 and 7 may be ascribed to the presence of an intra-
molecular aromatic–carbohydrate interaction in the phenyl-
lactic acid derivative 7 that strongly biases its conformation-
al behavior by severely restricting its conformational free-
dom. This conformational constraint is lacking in the cyclo-
hexyllactic acid derivative 6, which behaves similarly to the
lactic acid derivative 5 in terms of flexibility and preferred
orientations of the hydroxy acid side chain. As a result of
the conformation lock, the side chain of 7 is preorganized in
an anti-c conformation that allows optimal interaction of the
carboxy group in the carboxylate binding region of CT. The
same conformation appears to be attained by 6 in its bound
state, but it has to be selected from a pool of different ro-
tamers that are simultaneously present in solution. NMR
analysis of the bound state of the ligands also reveals that
the phenyl ring of 7 is in close contact withthe protein, as
revealed by the intense signal of the aromatic protons in the
STD spectrum of the CT:7 complex. In contrast, the cyclo-
hexyl group of 6 appears to make a muchloser contact with
the toxin. Thus, the preorganization effect and a more effi-
cient van der Waals interaction between the side-chain sub-
stituent and the protein appear to concur in determining the
high affinity of 7 for CT.
cording to the protocol employed by London, with a relaxation matrix
[8]
withexchange as described.
Different exchange-rate constants, k, de-
fined as pf·k=Kꢁ1 (where pf is the fraction of the free ligand), and leak-
age relaxation times were employed to obtain the optimal match between
the experimental and theoretical results for the intraresidue H-1/H-3 and
H-1/H-5 cross-peaks of the Gal and GalNAc moieties for the given pro-
tein/ligand ratio. Normalized intensity values were used since they allow
correcting for spin-relaxation effects. The overall correlation time, tc, for
the free state was always set to 0.15 ns and the tc for the bound state was
estimated as 15 ns according to the molecular weight of the toxin (tc =
10–12”MW). To fit the experimental TR-NOE intensities, exchange-rate
constants, k, between 100–1000 sꢁ1 and external relaxation times, 1*, for
the bound state of 0.5, 1, and 2 s were tested. Optimal agreement was
achieved when k=150 sꢁ1 and 1*=1 s.
TR-ROESY experiments were also carried out to exclude spin-diffusion
effects. A continuous-wave spin-lock pulse was used during the 250 ms
mixing time. Key NOE contacts were shown to be direct cross-peaks,
since they showed different signs to diagonal peaks. In some cases, they
allowed the detection of intra-Gal and intra-GalNAc H-1/H-4 and H-1/
H-6 cross-peaks that were due to spin diffusion.
The phenyl ring–GalNAc stacking interaction that we
have observed in 7 has been described here for the first
time as an element of a conformation lock in the structure
of a sugar mimic. This kind of sugar–aromatic interaction,
which is very well described in the context of carbohydrate/
protein complexes[20–24] may have a broader application,
which we intend to explore, as an element in conformation-
based design of sugar mimics. Additionally, it has been re-
ported that the presence of the hydrophobic ceramide
moiety may influence the head-group presentation of the
glycan[30] and may even influence the binding to cholera
toxin.[31] Thus, although this report has focused on the im-
portance of the oligosaccharidic part, further studies in our
groups are paying attention to the importance of the lipid
chain for the interaction under different experimental condi-
tions.
Computational methods
Conformation search and dynamics of isolated 6 and 7: The calculations
were performed by using the MacroModel/Batchmin[33] package (ver-
sion 7.0) and the AMBER* force field. Kolbꢁs parameters were used for
the hydroxy acid moiety.[25] Bulk water solvation was simulated by using
MacroModelꢁs generalized Born GB/SA continuum solvent model.[34]
The conformation searches were carried out by using 20,000 steps of the
usage-directed MC/EM procedure according to previously established
protocols.[35, 19] Extended nonbonded cut-off distances (a van der Waals
cut-off of 8.0 Š and an electrostatic cut-off of 20.0 Š) were used.
For the MC/SD[36] dynamic simulations, van der Waals and electrostatic
cut-offs of 25 Š, together with a hydrogen-bond cut-off of 15 Š, were
used. The dynamic simulations were run with the AMBER* all-atom
force field. Charges were taken from the force field (all-atom option).
The same degrees of freedom of the MC/EM searches were used in the
MC/SD runs. All simulations were performed at 300 K, witha dynamic
time step of 1 fs and a frictional coefficient of 0.1 psꢁ1. Typically, 2 runs
of 5 ns eachwere performed, with2 conformations of the substrates as
the starting point; these conformations differed at the hydroxy acid link-
age and were selected from the MC/EM outputs. The Monte Carlo ac-
ceptance ratio was about 4%, and eachaccepted MC step was followed
by an SD step. Structures were sampled every 1 ps and saved for later
evaluation. Convergence was checked by monitoring both energetic and
geometrical parameters.
Experimental Section
NMR spectroscopy experiments: NMR spectra were recorded at 25–
308C in D20 on Varian Unity 500 MHz and Bruker Avance 400 and
500 MHz spectrometers. For the experiments with the free ligand, the
compound was dissolved in D2O and the solution was degassed witha
stream of argon. COSY, TOCSY, and HSQC experiments were per-
formed by using the standard sequences. 2D T-ROESY experiments were
performed with mixing times of 300 and 500 ms. The strength of the
180 pulses during the spin-lock period was attenuated 4 times with re-
spect to that of the 90 hard pulses (between 7.2 and 7.5 ms). In order to
deduce the interproton distances, relaxation matrix calculations were per-
formed by using software written inhouse which is available from the au-
thors upon request.[32] For the bound ligands, STD and TR-NOE experi-
ments were performed as previously described.[8] First, the B chain of
cholera toxin was subjected to 2 cycles of freeze–drying with D2O to
remove traces of H2O. It was then transferred in solution to the NMR
Conformation search (MC/EM) of the LT complexes: The conformation
searches were carried out by using the usage-directed MC/EM procedure,
with slight variations of the protocol used in the study of the LT/psGM1
complex.[4] In brief, the starting structure for LT/7 was obtained by super-
imposing the conformation of 7 that exhibits the best fit with the NOE
data (conformer c in Figure 5) on the psGM1 ligand in our model of the
LT:psGM1 complex by using the galactose coordinates. This “docking”
step was followed by substructure energy minimization. The starting stru-
ture of the LT/6 complex was obtained from the LT/7 complex by graphi-
cally converting the phenyl ring into a cyclohexyl group.
The explicit torsional variables were the same as those used for the free-
state calculations, that is, the Gal–GalNAc and Gal–CHD anomeric link-
ages, the C-5–C-6 bonds of Gal and GalNAc, and the five bonds of the
hydroxy acid moiety. Furthermore, the ligand was allowed to rotate (max
Chem. Eur. J. 2004, 10, 4395 – 4406
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4405

A. Bernardi, J. JimØnez-Barbero et al.
FULL PAPER
1808) and translate (max. 1 Š) within the binding site (MOLS command
of Batchmin). 10000 MC/EM steps were performed. Bulk water solvation
was simulated by using the GB/SA model. Five crystallographic water
molecules were retained, as previously described.[4,35] All calculations
were carried out on a B2 (B+B(+1)) dimer. Only the ligand and a shell
of residues surrounding the binding site of LT were subjected to energy
minimization. All the residues within 6 Š of the sugars were completely
included in the shell. An all-atom treatment was used for the ligand and
for the aromatic residues of the protein. The rest of the toxin was treated
with a united-atom model. The ligand and all of the binding site polar hy-
droxy and amino hydrogen atoms were unconstrained during energy min-
imization. All other atoms that belonged to the substructure being mini-
mized were constrained to their crystallographic coordinates by parabolic
restraining potentials that increased with the distance from the sugar sub-
[7] D. Arosio, S. Baretti, S. Cattaldo, S. Potenza, A. Bernardi, Bioorg.
Med. Chem. Lett. 2003, 13, 3831–3834.
[8] A. Bernardi, D. Potenza, A. M. Capelli, A. García-Herrero, F. J.
Caæada, J. JimØnez-Barbero, Chem. Eur. J. 2002, 8, 4597–4612.
[9] A. A. Bothner-By, R. Gassend, Ann. N.Y. Acad. Sci. 1973, 222, 668–
676.
[10] P. L. Jackson, H. N. Moseley, N. R. Krishna, J. Magn. Reson. Ser. B
1995, 107, 289–292; V. L. Bevilacqua, D. S. Thomson, J. H. Preste-
gard, Biochemistry 1990, 29, 5529–5537; V. L. Bevilacqua, Y. Kim,
J. H. Prestegard, Biochemistry 1992, 31, 9339–9349; H. Kogelberg,
D. Solís, J. JimØnez-Barbero, Curr. Opin. Struct. Biol. 2003, 13, 646–
653;
[11] A. Schon, E. Freire, Biochemistry 1989, 28, 5019–5024.
[12] M. B. Khalil, M. Kates, D. Carrier, Biochemistry 2000, 39, 2980–
2988.
[13] K. Bock, J. O. Duus, J. Carbohydr. Chem. 1994, 13, 513–543.
[14] D. Neuhaus, M. P. Williamson, The NOE Effect in Structural and
Conformational Analysis, VCH, New York, 1989.
[15] T. L. Hwang, A. J. Shaka, J. Am. Chem. Soc. 1992, 114, 3157–3158.
[16] NMR Spectroscopy of Glycoconjugates (Eds.: J. JimØnez-Barbero, T.
Peters), VCH, Weinheim, 2002.
[17] J. Dabrowski, T. Kozar, H. Grosskurth, N. E. Nifantꢁev, J. Am.
Chem. Soc. 1995, 117, 5534–5539.
[18] A. Bernardi, D. Arosio, L. Manzoni, D. Monti, H. Posteri, D. Poten-
za, S. Mari, J. J. Barbero, Org. Biomol. Chem. 2003, 1, 785–792.
[19] P. Brocca, A. Bernardi, L. Raimondi, S. Sonnino, Glycoconjugate J.
2000, 17, 283–299.
[20] N. K. Vyas, Curr. Opin. Struct. Biol. 1991, 1, 732–740.
[21] F. A. Quiocho, Biochem. Soc. Trans. 1993, 21, 442–448.
[22] W. I. Weis, K. Drickamer, Annu. Rev. Biochem. 1996, 65, 441–473.
[23] S. Elgavish, B. Shaanan, J. Mol. Biol. 1998, 277, 917–932.
[24] J. Jimenez-Barbero, J. L. Asensio, F. J. Caæada, A. Poveda, Curr.
Opin. Struct. Biol. 1999, 9, 549–555.
[25] H. C. Kolb, B. Ernst, Chem. Eur. J. 1997, 3, 1571–1578.
[26] M. Meyer, B. Meyer, Angew. Chem. 1999, 111, 1902–1906; Angew.
Chem. Int. Ed. 1999, 38, 1784–1788.
[27] J. Klein, R. Meinecke, M. Meyer, B. Meyer, J. Am. Chem. Soc. 1999,
121, 5336–5337.
[28] M. Vogtherr, T. Peters, J. Am. Chem. Soc. 2000, 122, 6093–6099.
[29] J. L. Asensio, F. J. Caæada, J. Jimenez-Barbero, Eur. J. Biochem.
1995, 233, 618–630.
ꢁ2
strate. The following force constants were used: 100 kJŠ for atoms
ꢁ2
within 0–3 Š of any atom of the ligand; 200 kJŠ for atoms within 3–
ꢁ2
4 Š; 400 kJŠ for atoms within 4–5 Š. The periphery of the restrained
structure was checked with the EdgeD command of MacroModel, and
isolated atoms were included to avoid incomplete functional groups. All
other atoms were ignored.
MC/SD dynamics of the LT:ligand complexes: The simulations were car-
ried out by using the same substructure and explicit MC variables descri-
bed above, but with the lowest energy conformation from the MC/EM
searchas the starting point. Extended nonbonded cut-offs were employed
(van der Waals and electrostatic cut-off of 25 Š, hydrogen-bond cut-off
of 15 Š). The simulations were performed for 1 ns, at 300 K, with a dy-
namic time step of 1 fs and a frictional coefficient of 0.1 psꢁ1. Structures
were sampled every 2 ps and saved for later evaluation.
Acknowledgments
The project was supported by the Azioni Integrate program between
Italy and Spain, and the European programs COST-D13 and “Improving
Human Potential” under contract HPRNCT-2002-00173 (Glycidic Scaf-
folds Network). I.S.-M. acknowledges a Marie Curie Fellowship from the
European Community program ꢁImproving Human Potentialꢁ under con-
tract HPMT-CT-2001-00293. The Madrid group also acknowledges the
Ministery of Science and Technology of Spain for funding (Grant
BQU2003–03550-C01).
[30] D. Roy, C. Mukhopadhyay, J. Biomol. Struct. Dyn. 2002, 19, 1121–
1132.
[31] B. Lanne, B. Schierbeck, J. Angstrom, J. Biochem. 1999, 126, 226–
234.
[32] A. Poveda, J. L. Asensio, M. Martín-Pastor, J. Jimenez-Barbero, J.
Biomol. NMR 1997, 10, 29–43.
[33] F. Mohamadi, N. G. J. Richards, W. C. Guida, R. Liskamp, M.
Lipton, C. Caufield, G. Chang, T. Hendrickson, W. C. Still, J.
Comput. Chem. 1990, 11, 440–467.
[34] W. C. Still, A. Tempzyk, R. Hawley, T. Hendrickson, J. Am. Chem.
Soc. 1990, 112, 6127–6129.
[1] P. Sears, C. H. Wong, Angew. Chem. 1999, 111, 2446–2471; Angew.
Chem. Int. Ed. 1999, 38, 2300–2324, and references therein.
[2] C.-L. Schengrund, N. J. Ringler, J. Biol. Chem. 1989, 264, 13233–
13237.
[3] E. A. Merritt, S. Sarfaty, F. van den Akker, C. LꢁHoir, J. A. Martial,
W. G. J. Hol, Protein Sci. 1994, 3, 166–175; E. A. Merrit, T. K.
Sixma, K. H. Kalk, B. A. M. Van Zanten, W. G. J. Hol, Mol. Micro-
biol. 1994, 13, 745–753; E. A. Merrit, S. Sarfaty, M. G. Jobling, T.
Chang, R. K. Holmes, W. G. J. Hol, Protein Sci. 1997, 6, 1516–1528.
[4] A. Bernardi, A. Checchia, P. Brocca, S. Sonnino, F. Zuccotto, J. Am.
Chem. Soc. 1999, 121, 2032–2036.
[35] A. Bernardi, L. Raimondi, F. Zuccotto, J. Med. Chem. 1997, 40,
1855–1862.
[36] F. Guarnieri, W. C. Still, J. Comput. Chem. 1994, 15, 1302–1310.
[5] A. Bernardi, G. Boschin, A. Checchia, M. Lattanzio, L. Manzoni, D.
Potenza, C. Scolastico, Eur. J. Org. Chem. 1999, 1311–1317.
[6] A. Bernardi, L. Carrettoni, A. Grosso Ciponte, D. Monti, S. Sonni-
no, Bioorg. Med. Chem. Lett. 2000, 10, 2197–2200.
Received: January 26, 2004
Revised: May 3, 2004
Published online: July 27, 2004
4406
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 4395 – 4406