First round of a focused library of cholera toxin inhibitors

Article abstract of DOI:10.1016/j.carres.2007.06.006

C-Galactosides have been used as scaffolds to design a library of non-hydrolysable inhibitors of cholera toxin (CT). Test elements from the library were synthesized and found to inhibit CT binding to an asialofetuin-coated SPR chip with micromolar affinity. Preliminary results are reported.

Full text of DOI:10.1016/j.carres.2007.06.006

Carbohydrate Research 342 (2007) 1651–1660
First round of a focused library of cholera toxin inhibitors
a,
Crtomir Podlipnik, Ingrid Velter,b Barbara La Ferla,b Gilles Marcou,a,à Laura Belvisi,a
ˇ
Francesco Nicotrab,* and Anna Bernardia,*
aDipartimento di Chimica Organica e Industriale, Universita’ degli Studi di Milano, Via Venezian 21, I-20133 Milano, Italy
bDipartimento di Biotechnologie e Bioscienze, Universita’ degli Studi di Milano-Bicocca, Piazza della Scienza 2, I-20126 Milano, Italy
Received 23 February 2007; received in revised form 29 May 2007; accepted 2 June 2007
Available online 9 June 2007
Abstract—C-Galactosides have been used as scaffolds to design a library of non-hydrolysable inhibitors of cholera toxin (CT). Test
elements from the library were synthesized and found to inhibit CT binding to an asialofetuin-coated SPR chip with micromolar
affinity. Preliminary results are reported.
ꢀ 2007 Elsevier Ltd. All rights reserved.
Keywords: Carbohydrate mimics; Cholera toxin; Glycomimetics; C-Glycosides; GM1 ganglioside
1. Introduction
massive loss of fluids, which in turn can lead to dehydra-
tion and shock.4,5
Vibrio cholerae cholera toxin (CT) is the causative agent
of cholera, a disease that is responsible for the death of
hundreds of thousands of people each year in develop-
ing countries.1 CT is a member of the AB5 holotoxin
family. It consists of a single catalytically active compo-
nent, A, and a non-toxic receptor-binding component, a
pentamer of B subunits. The B pentamer (CTB5) is
responsible for binding to the cell surface and this func-
tion is retained even in the absence of the A subunit.2,3
The mechanism of CT action is initiated by binding of
the B subunits to the oligosaccharide head-group of the
GM1 ganglioside, b-Galp-(1!3)-b-GalpNAc-(1!4)-(a-
NeupAc-(2!3)-b-Galp-(1!4)-b-Glc–OH, 1 (o-GM1,
Fig. 1), on intestinal epithelial cell membranes. The A
subunit is then translocated into the host cell, where it
mono-ADP-ribosylates Gsa, a guanine nucleotide-bind-
ing regulatory protein, resulting in persistent activation
of adenylyl cyclase.3 The rising level of cAMP leads to
Small molecules acting as decoys for the toxin’s GM1
binding site could work as toxin inhibitors by competing
with GM1, possibly saturating the toxin, which would
prevent it from binding to the epithelial cells and causing
the symptoms of cholera. Natural and synthetic antago-
nists of CT are known, and they are under investigation
as potential agents for the treatment of cholera and
other toxin-caused enteropathies.6
The interactions of o-GM1 with CTB5, as seen in the
X-ray structure of the complex, are highlighted in Fig-
ure 1.7,8 The terminal galactose stacks on top of Trp88
and is linked through hydrogen bonds to Asn90,
Lys91, Glu51, and Gln61 (Fig. 1). The terminal sialic
acid (NeuAc) is exposed to solvent on the a-face and
it is in van der Waals contact with Tyr12 on the b-face.
Hydrogen bonds are formed between the sialic acid
polar groups with crystallographic water molecules
and with the backbone of Glu11 and Tyr12. The carb-
oxyl group of NeuAc establishes a key interaction with
Trp88, mediated by a key water molecule. The outer
part of the binding pocket, with Glu11 and Arg35, con-
stitutes a dipole that orients the N-acetyl fragment of the
sialic acid (Fig. 1).
*
Corresponding authors. Tel.: +39 02 64483457 (F.N.); tel.: +39 02
50314092; fax: +39 02 50314072 (A.B.); e-mail addresses: francesco.
nicotra@unimib.it; anna.bernardi@unimi.it
Present address: Faculty of Chemistry and Chemical Technology,
Asˇkercˇeva 6, 1000 Ljubljana, Slovenia.
à Present address: Institut de Chimie de Strasbourg, UMR7177,1, rue
Blaise Pascal, 67000 Strasbourg, France.
The terminal galactose and sialic acid residues con-
tribute the most to GM1 binding. Although the two
0008-6215/$ - see front matter ꢀ 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.carres.2007.06.006

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1652
b
a
OH
HO
HO
HO
O
OH
O
O
OH
O
O
OH
OH
HO
O
O
AcHN
HO2C
OH
O
OH
OH
O
1
HO
OH
AcHN
OH
HO
Figure 1. (a) The GM1 oligosaccharide o-GM1, and (b) its key interactions with CT binding site.
isolated monosaccharides show very low affinity for the
protein (Gal 40 mM; NeuAc > 200 mM),9 o-GM1 binds
to CT with a dissociation constant KD = 43 nM.9 This
appears to depend on the conformational preorganiza-
tion of o-GM1,9,10 which has been shown by NMR to
adopt one major conformation11 closely resembling the
bound conformation observed in the CT complex.7,12
According to this interpretation, the pseudo-oligosac-
charide 2 (Fig. 2), which reproduces the 3D structure of
o-GM1, was found to display a similar affinity for
CTB5.10 A second group of pseudo-oligosaccharides
rationally designed to replace the sialic acid moiety
while retaining CT affinity have led to micromolar inhi-
bitors of toxin (Fig. 2).12–18 Similar results have been
achieved by using galactose as a CT-binding anchor
and developing several generations of galactoside librar-
ies3,6a,19–22 (Fig. 2). Some of the molecules discussed
above display good affinity for CT and are structurally
much simpler than the natural ligand GM1. However,
they all are O-glycosides, and are unlikely to be meta-
bolically stable to any significant extent. Furthermore,
the synthetic methods used to connect the pharmaco-
phoric sugar moieties are those of traditional carbo-
hydrate chemistry, which are often laborious and not
high yielding procedures.
To address these shortcomings, we are currently
working toward the development of CT ligands starting
from simple C-galactosides. C-Glycosides are well
known metabolically stable analogs of O-glycosides.23–29
Furthermore, some of us have recently shown that
the functionalized C-galactosides 7 and 8 (Fig. 3) can
be easily synthesized in a few steps directly from galact-
HO
HO
OH
HO
HO
OH
HO
HO
OH
O
OH
O
O
O
O
O
HO
OH
O
O
AcHN
HO2C
AcHN
CO2tBu
CO2tBu
CO2tBu
CO2tBu
O
O
2
HO2C
R'
R
O
3 R= H, R'= Me
4 R= Me, R'= H
5 R= H, R'= C6H11
6 R= H, R'= Ph
HO
OH
OH
HO
AcHN
O
OH
O
OH
OH
N
N
H
O
-galactoside inhibitor
O
O
HO
from ref 21
NO2
Figure 2. Some currently known CT inhibitors.

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1653
Table 1. Computed (Autodock/Glide) and measured activity of CT
ligands
OH
O
OH
O
HO
HO
HO
HO
Entry
Compound
Glide
G-scorea
pIC50
measured
pIC50
computed
HO
HO
7
8
COOH
1
2
3
4
5
6
7
8
9
1
2
ꢀ12.38
ꢀ11.79
ꢀ10.27
ꢀ9.65
5.80b
5.89b
3.70b
3.00b
4.43c
5.00c
2.69d,e
2.45e
3.90e
3.63e
3.51e
3.22e
5.65
5.18
3.97
3.48
4.92
5.08
2.20
3.32
3.82
4.09
3.27
5.01
NH2
3
Figure 3. The C-galactoside anchors used for the development of
metabolically stable CT-ligands.
4
5
ꢀ11.47
ꢀ11.66
ꢀ8.03
6
ose avoiding the need for protecting groups.30 The
approach we have adopted to identify CT inhibitors
involves the following steps: (a) the development and
validation of a docking/scoring protocol based on a
set of known pseudo-GM1 ligands; (b) the design of a
focused library of C-galactosides starting from 7 and
8; (c) the synthesis and affinity evaluation (by SPR) of
selected elements of the library. We report here on our
initial results, which have led to the identification of
simple C-galactosides with micromolar affinity for CT.
9
10
11
12
13
14
ꢀ9.45
ꢀ10.08
ꢀ10.42
ꢀ9.38
10
11
12
ꢀ11.58
a G-score obtained starting from Autodock poses following the pro-
cedure described in Section 3 (Docking experiments).
b Fluorescence measurement, from Ref. 14.
c Fluorescence measurement, from Ref. 15.
d Fluorescence measurement, from Ref. 46.
e This paper (SPR measurement).
2. Results and discussion
we obtained very good results using Autodock when
generating poses for o-GM1 and its derivatives, but
the ranking based on Autodock score was poor. Better
results were obtained by clustering the Autodock poses,
minimizing the cluster leaders in the toxin’s cavity and
ranking the poses by in situ scoring with Glide.33,34
The Glide score (G-Score) obtained ranged from
ꢀ12.38 for the best ligand 1, to ꢀ8.03 for methyl a-D-
galactopyranoside 9, a millimolar CT ligand (Table 1).
The seven test compounds were ranked correctly relative
to their affinity (Table 1, Fig. 4). Visual inspection of the
best ranking pose for each ligand allowed us to recog-
nize the features expected based on the available experi-
mental data.
2.1. Validation of the docking protocol
The validation of the docking protocol was attempted
using a set of seven known CT ligands, which included
o-GM1 1 (Fig. 1), the pseudo-GM1 2–6 (Fig. 2) and
methyl a-D-galactopyranoside 9 (Fig. 4). The CT com-
plex of all these compounds was at least partially defined
either from crystallographic7,19 or NMR studies13,16 and
IC50 from fluorescence titrations were available14,15 and
covered a good range of different affinities (Table 1,
entries 1–7).
Autodock31 was previously reported to reproduce the
bound pose of CT-ligands with good fidelity.32 Indeed,
OH
OH
OH
HO
HO
HO
HO
HO
OMe
OH
O
OH
O
OH
O
O
OMe
OMe
O
NO2
HO
H
N
H
N
H
N
O
10 IC50= 3500 μM
11 IC50= 125 μM
12 IC50= 233 μM
OH
OH OH
HO
HO
HO
OH
OH
OH
HO
HO
O
H
N
O
HO
HO
O
O
HO
H
OH
O
OH
N
O
OMe
O
O
COOH
13 IC50= 305 μM
14 IC50= 595 μM
9 IC50= 2000 μM
15 IC50= 3500 μM
Figure 4. Structure and CT inhibition data (SPR) of ligands 9–15.

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1654
Direct docking (and scoring) with Glide yielded poses
The compounds were synthesized starting from
known C-allyl galactose 15 (Scheme 1).35,36 Compound
15 was acetylated and subjected to Lemieux oxidation
followed by reduction to afford alcohol 17. Tosylation
and treatment with NaN3 afforded the azido derivative
18, which was reduced and deprotected providing amine
8 in good overall yields. Coupling of 8 with diphenyl-
maleic anhydride or freshly prepared cinnamoyl chlo-
rides provided 10–14 in 50–80% yield.
of lower quality and some of the compounds, including
the natural ligand, were especially problematic, possibly
because they are more flexible and larger than ideal for
virtual screening software. However, roughly the same
ranking of the test compounds was obtained, indicating
that direct screening with Glide, which is orders of mag-
nitude faster than the protocol described above, could
be used for the initial analysis of larger libraries.
The inhibition properties of compounds 10–14 were
analyzed in a surface plasmon resonance (SPR) compe-
tition assay, using the cholera toxin B pentamer CTB5
as the target and a SPR chip coated with asialofetuin
(ASF), as previously described.37 A series of measure-
ments with increasing CTB5 concentrations yielded a
binding isotherm and a Kchip of 70 lM.38 Inhibition
studies were then performed using ligands 9–15 and
CTB5 at 70 lM. First, the methyl a-D-galactopyran-
oside 9 (Fig. 4) was measured and an IC50 of 2 mM
was determined, consistent with the known CTB5 affin-
ity of galactose.9,19 No affinity improvement was shown
by the cinnamic acid derivative 10 (IC50 = 3500 lM)
over 9 or the a-C-allyl galactoside 15. On the contrary,
the inhibition potency of the ligands increased by one
order of magnitude for compounds 11–14, which dis-
played additional functional groups. Amide 11, bearing
three methoxy groups on the aromatic ring, is the most
active of the set (IC50 = 125 lM). Inspection of the
Autodock poses (Fig. 5) does not suggest specific H-
bonding interactions for the three methoxy groups,
rather they appear to provide additional hydrophobic
interactions at the bottom of the binding site. Further-
more, the three oxygen atoms of 11, which are largely
exposed to the aqueous solvent, may improve the solva-
tion of the complex relative to the unsubstituted analog
10.
2.2. Design, synthesis and testing of the test library
Based on these results, a virtual screening of the C-
galactoside library was attempted. A library of 309 com-
pounds was built by conjugating scaffolds 7 and 8 with a
group of selected amines or acids, respectively. The com-
pounds were designed to contain either an acid function,
which could substitute for the carboxyl group of sialic
acid, or lipophilic groups, which would be able to
exploit the shallow lipophilic surface surrounding the
CT galactose binding site, or both.
The virtual library was built by in-house software as
described in Section 3.3 and screened using Glide 4.0.
Analysis of the G-score showed that none of the
screened compounds were ranked above the natural
ligand o-GM1, but that several were ranked in a G-score
range associated with ligands of micromolar affinity in
the validation set.
From the sublibrary of cinnamic acid derivatives
obtained starting from scaffold 8, the five amides 10–
14 (Fig. 4) were selected for synthesis using five commer-
cially available cinnamic acids. These five compounds
were re-docked in CT binding site using the Auto-
dock/Glide protocol described above, and the G-score
obtained is reported in Table 1 (entries 8–12). All these
compounds are a-C-glycosides, as opposed to native
GM1, which contains a b-galactosyl residue. According
to the calculated poses, this has no influence on the
binding mode of the galactose anchor, which retains
all the known interactions in the binding site (see
Fig. 1b).
Furthermore, we could use the G-scores and the
experimental pIC50 for the seven test compounds to gen-
erate a simple model for predicting IC50 for CT inhibi-
tors. The model is expressed by Eq. 1, which was then
used to predict activity values for the set of the newly
OAc
OH
HO
OAc
AcO
AcO
AcO
OAc
O
AcO
AcO
c
b
a
O
O
O
HO
AcO
OH
AcO
AcO
AcO
17
18
8
15
16
d
OH
O
N3
OH
HO
e or f
10-14
HO
HO
NH2
Scheme 1. Synthesis of ligands 10–14. Reagents and conditions: (a) Ac2O, pyridine; (b) OsO4, NaIO4, 0 ꢁC!rt, then NaBH4; (c) TsCl, pyridine,
CH2Cl2, 0 ꢁC!rt then NaN3, DMF; (d) H2 Pd–C, MeOH then MeONa, IRA-120H+; (e) ArCOCl, toluene, reflux, 12 h; (f) diphenylmaleic
anhydride, MeOH, K2CO3, 2 h.

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1655
3. Conclusions
In conclusion, we have shown that simple C-galactosides
can be used as anchors for non-hydrolysable ligands
capable of CT inhibition. The small test library that
we have described here is by no means optimized.
Nevertheless, four of the five non-hydrolysable com-
pounds prepared displayed an inhibition potency that
compares well to previously described artificial ligands
that are more complex and/or hydrolysable, such as
those depicted in Figure 4.14,15,21 The library design
was supported by in silico docking protocols which were
validated using a training set of known CT ligands. It is
well known that binding mode prediction and accurate
prediction of binding affinity are different tasks, which
are accomplished with very different efficiency by cur-
rently available docking programs.41 In the case at hand,
we have tested two different programs and found that
one (Autodock) produced good ligand poses (as judged
by comparison with the known X-ray or NMR struc-
tures) and poor affinity ranking and the second one
(Glide) gave a better reproduction of the binding affinity
and poor ligand poses. Eventually we have validated a
method that relies on the initial rapid library screening
with Glide followed by redocking with Autodock and
in situ scoring with Glide. Based on this procedure we
have built a predictive model to estimate a ligands’ inhi-
bition potency from virtual screening scores. With these
tools in hand, we will continue our studies toward the
design and synthesis of more potent analogs.
Figure 5. Poses obtained with Autodock/Glide procedure for com-
pound 11.
4. Experimental
4.1. Synthesis
Figure 6. Correlation between Glide score and affinity for CT
binders. Stars represent entries 1–7 in Table 1, which are included in
the model described by Eq. 1. The numbers represent compounds
10–14.
4.1.1. General methods. All solvents were dried with
molecular sieves for at least 24 h prior to use. Thin layer
chromatography (TLC) was performed on silica gel
60 F254 plates (Merck) with detection using UV light
when possible, or by charring with a solution of
concd H2SO4–EtOH–H2O (5:45:45) or a solution of
(NH4)6Mo7O24 (21 g), Ce(SO4)2 (1 g), concd H2SO4
(31 mL) in water (500 mL). Flash column chromatogra-
phy was performed on silica gel 230–400 mesh (Merck).
1H and 13C NMR spectra were recorded at 25 ꢁC with a
Varian Mercury 400 MHz instrument using CDCl3 as
the solvent unless otherwise stated. Chemical shift
assignments, reported in ppm, are referenced to the cor-
responding solvent peaks. Mass spectra were recorded
with a MALDI2 Kompakt Kratos instrument, using
a-cyano-4-hydroxycinnamic acid (CHCA) or gentisic
acid (DHB) as the matrix; HRMS were recorded on a
QSTAR elite LC/MS/MS system with a nanospray ion
source. Optical rotations were measured at rt using a
synthesized amides 10–14. The predicted values are col-
lected in Table 1 (column 5) and compared to the exper-
imental values derived from the SPR measurements
(column 4). A reasonable correlation between experi-
mental pIC50 and computed Glide scores for the CT
inhibitors can be observed in Figure 6. The largest devi-
ation observed is for the predicted IC50 value of com-
pounds 14, which is two orders of magnitude higher
than that measured by SPR experiment.
pIC50 ¼ ꢀ4:18ðꢁ1:48Þ ꢀ 0:79ðꢁ0:14Þ ꢂ G Score
ðn ¼ 7; r2 ¼ 0:87; F ¼ 34:03; s2 ¼ 0:26;
q2 ¼ 0:73Þ
ð1Þ

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1656
Kruss P3002 electronic polarimeter and are reported in
(1.28 g, 2.41 mmol); [a]D +104.3 (c 1, CHCl3); 1H
NMR (400 MHz, CDCl3): d 5.41 (t, J = 3.2 Hz, 1H,
H-4), 5.22–5.16 (m, 2H, H-2, 3), 4.53 (br dd, J = 11.8,
9.2 Hz, 1H, H-6a), 4.43 (dt, J = 11.5, 3.1 Hz, 1H, H-
1), 4.20–4.16 (m, 1H, H-5), 4.02 (dd, J = 11.8, 3.9 Hz,
1H, H-6b), 3.74–3.73 (m, 2H, CH2CH2OH), 2.11 (s,
3H, CH3C@O), 2.10 (s, 3H, CH3C@O), 2.09 (s, 3H,
CH3C@O), 2.06 (s, 3H, CH3C@O), 1.95–1.86 (m, 1H,
H-10a), 1.69–1.61 (m, 1H, H-10b); 13C NMR
(100 MHz, CDCl3): d 171.0 (C), 170.0 (C), 169.9 (C),
169.8 (C), 69.99 (CH), 69.84 (CH), 69.00 (CH), 67.94
(CH), 67.32 (CH), 61.37 (CH2), 60.06 (CH2), 21.14–
21.03 (CH3); MALDI-TOFMS (CHCA) 377 [M+1],
398 [M+23].
¨
units of 10ꢀ1 deg cm2 gꢀ1
.
4.1.2. 3-(2,3,4,6-Tetra-O-acetyl-a-D-galactopyranosyl)-
propene (16). To a solution of the C-allyl-galactoside
15 (2.50 g, 12.25 mmol) in pyridine (18.0 mL), acetic
anhydride (9.0 mL, 98.04 mmol) was added at 0 ꢁC.
The reaction was stirred for 4 h at rt, diluted with ethyl
acetate (50 mL), and then washed with water
(1 · 30 mL),
10%
CuSO4
(3 · 30 mL),
water
(1 · 30 mL), 5% HCl (1 · 30 mL), water (1 · 30 mL),
and brine (1 · 30 mL). The organic layer was dried
(Na2SO4), filtered and the filtrate concentrated under
vacuum. The resulting residue was purified by flash col-
umn chromatography over silica gel (petroleum ether–
EtOAc, 2.5:1) to afford 16 (4.35 g, 95%); yellow oil,
Rf = 0.6 (petroleum ether–EtOAc, 1.5:1); [a]D +87 (c
4.1.4. 1-Azido-2-(2,3,4,6-tetra-O-acetyl-a-D-galactopyr-
anosyl)ethane (18). To a stirred solution of alcohol 17
(1 g, 2.67 mmol), in pyridine (5 mL), at 0 ꢁC, solid tosyl
chloride (5.60 mg, 2.92 mol) was added. The reaction
mixture was warmed to rt and stirred overnight,
quenched by the addition of H2O (1 mL) and stirred
for another 30 min. The reaction was then diluted with
CHCl3 and washed with H2O, then brine (1 · 50 mL),
dried (Na2SO4), filtered and concentrated under reduced
pressure. The residue was submitted to the next step
without further purification.
1
1, CHCl3); H NMR (400 MHz, CDCl3): d 5.80–5.69
(m, 1H, CH@CH2), 5.42–5.40 (m, 1H, H-4), 5.28–5.25
(dd, J = 4.8, 9.3 Hz, 1H, H-2), 5.22–5.21 (dd, J = 3.2,
9.3 Hz, 1H, H-3), 5.15–5.08 (m, 2H, CH@CH2), 4.31–
4.28 (q, J = 4.8, 10.0 Hz, 1H, H-1), 4.22–4.17 (dd,
J = 8.8, 12.8 Hz, 1H, H-6), 4.12–4.06 (m, 2H, H-5, 6),
2.52–2.41 (m, 1H, H-10), 2.32–2.22 (m, 1H, H-10), 2.12
(s, 3H, CH3), 2.07 (s, 3H, CH3), 2.04 (s, 3H, CH3),
2.03 (s, 3H, CH3); 13C NMR (100 MHz, CDCl3): d
168.1 (C), 167.6 (C), 167.5 (C), 167.3 (C), 130.8 (CH),
115.3 (CH2), 69.1 (CH), 65.9 (CH), 65.6 (CH), 65.3
(CH), 59.2 (CH2), 28.7 (CH2), 18.7 (CH3), 18.6 (CH3),
18.5 (CH3).
To a stirred solution of the crude tosylate (1.25 g) in
DMF (5 mL), sodium azide (NaN3) (229 mg,
3.53 mmol) was added and the resulting mixture stirred
at rt for 12 h and was then concentrated under reduced
pressure. The residue was purified using flash column
chromatography over silica gel (petroleum ether–
EtOAc, 3:1) to provide 18 (852 mg, 80% over the two
4.1.3. 2-(2,3,4,6-Tetra-O-acetyl-a-D-galactopyranosyl)-
ethanol (17). A solution of 16 (2.08 g, 5.6 mmol) in
H2O–acetone–t-BuOH 2:1:1 (40 mL) was cooled to
0 ꢁC and OsO4 (0.01 equiv) was added and the reaction
mixture was stirred for 15 min prior to the addition of
solid NaIO4 (3.64 g, 17.02 mmol). After 3 h, another
portion of solid NaIO4 (2.40 g, 11.22 mmol) was added
and the reaction mixture was stirred overnight at rt.
The suspension was then filtered through a Celite pad
and the filtrate was concentrated. The residue was dis-
solved in EtOAc, filtered again through a Celite pad,
and the filtrate was dried (Na2SO4), filtered and concen-
trated. The resulting aldehyde (1.73 g) was submitted to
the next step without further purification.
steps); 1H NMR (400 MHz, CDCl3):
d 5.41 (t,
J = 3.2 Hz, 1H, H-4), 5.24 (dd, J = 8.6, J = 4.7 Hz,
1H, H-2), 5.17 (dd, J = 8.6, J = 3.22 Hz, 1H, H-3),
4.35–4.29 (m, 2H, H-1, 6a), 4.10–4.06 (m, 2H, H-5,
6b), 3.45–3.38 (m, 2H, CH2N3), 2.11 (s, 3H, CH3C@O),
2.09 (s, 3H, CH3C@O), 2.07 (s, 3H, CH3C@O), 2.05 (s,
3H, CH3C@O), 1.96–1.89 (m, 1H, CHCH2N3), 1.74–
1.66 (m, 1H, CHCH2N3); 13C NMR (100 MHz, CDCl3):
d 170.6 (C), 169.9 (C), 169.8 (C), 169.7 (C), 69.16 (CH),
69.05 (CH), 68.44 (CH), 68.04 (CH), 67.40 (CH), 61.39
(CH2), 47.96 (CH2), 26.25 (CH2), 21.14-21.03 (CH3);
Maldi-Tof MS (CHCA) 374 [M+1ꢀ28].
To a stirred solution of the crude aldehyde (1.73 g,
4.63 mmol) in MeOH (20 mL), at 0 ꢁC, was added
sodium borohydride (70 mg, 1.85 mmol) in three por-
tions. After 30 min, the pH was adjusted to 3 by the
addition of AcOH and the mixture was evaporated.
The residue was then diluted with CH2Cl2 (15 mL),
washed with aqueous HCl, with brine (2 · 10 mL), dried
(Na2SO4) filtered and concentrated under reduced pres-
sure. The residue was purified by flash column chroma-
tography over silica gel (petroleum ether–EtOAc, 6:4) to
provide alcohol 17 in 61% yield over the two steps
4.1.5. 2-(a-D-Galactopyranosyl)ethylamine (8). To a
solution of azide 18 (490 mg, 1.22 mmol) in MeOH
(3 mL), at rt, was added Pd(OH)2–C (100 mg, 10% by
weight). The reaction mixture was then placed under
an atmosphere of H2 and stirred for 16 h. The mixture
was then filtered through Celite and the filtrate evapo-
rated under reduced pressure. The residue contained a
partially deacylated product, so it was dissolved in dry
MeOH (3 mL) and NaOMe (1.2 mL, 0.1 M) was added.
After 15 min the reaction was neutralized with resin IR-

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1657
120 H+, filtered and the filtrate evaporated under re-
duced pressure to afford 8 (237 mg, 94%); amorphous
white solid; 1H NMR (400 MHz, CD3OD): d 4.04–
3.98 (m, 1H, H-1), 3.85–3.80 (m, 2H, H-2, 4), 3.68–
3.52 (m, 4H, H-3, 5, 6a, 6b), 3.05–2.95 (m, 2H,
CH2N), 2.01–1.85 (m, 2H, CH2CH2N); 13C NMR
(100 MHz, CD3OD): d 76.14 (CH), 74.99 (CH), 72.31
(CH), 71.53 (CH), 70.46 (CH), 66.36 (CH2), 63.86
(CH2), 33.9 (CH2).
(CH2), 28.61 (CH2); HRMS calcd for [M+H]+:
405.1274, [M+Na]+: 360.1423. Found: 383.1613,
405.1408.
Compound 13. Yield 72%; 1H NMR (400 MHz,
D2O): d 7.42 (d, J = 15.7 Hz, 1H, CH@), 7.17 (d,
J = 1.7 Hz, 1H, HAr), 7.11 (dd, J = 1.7, J = 8.0 Hz,
1H, HAr), 6.92 (d, J = 8.0 Hz, 1H, HAr), 6.45 (d,
J = 15.7 Hz, 1H, CH@), 6.02 (s, 2H, OCH2O), 4.38–
4.21 (m, 1H, H-1), 4.06–3.92 (m, 2H, H-2, 4), 3.84–
3.70 (m, 4H, H-3, 5, 6a, 6b), 3.43 (br t, 2H, CH2N),
2.04–1.90 (m, 1H, CHCH2N), 1.84–1.65 (m, 1H,
CHCH2N); 13C NMR (100 MHz, CD3OD): d 178.5
(C), 150.4 (2C), 143.4 (CH), 131.9 (C), 126.4 (CH),
124.8 (CH), 111.3 (CH), 109.1 (CH), 104.2 (CH2), 76.8
(CH), 74.7 (CH), 72.4 (CH), 71.6 (CH), 70.6 (CH),
64.0 (CH2), 39.2 (CH2), 26.2 (CH2); HRMS calcd for
[M+Na]+: 404.1321. Found 404.1257.
4.1.6. General procedure for amides 10–13. To a solu-
tion of the corresponding cinnamic acid derivative
(0.5 mmol) in THF (1 mL), at rt, oxalyl chloride
(0.2 mL/mmol) was added. The reaction was stirred
for 2.5 h, then the volatiles were evaporated under re-
duced pressure. The residue was dissolved in toluene
and added to a mixture of amine and Na2CO3 and
heated at reflux for 2.5 h. The solvent was evaporated
and the residue purified by flash column chromato-
graphy on silica gel, using a gradient EtOAc–MeOH,
10:0!8:2.
Compound 14. A solution of galactose amine 8
(16 mg, 0.078 mmol), diphenyl maleic anhydride
(24 mg, 0.097 mmol), Na2CO3 (33 mg, 0.31 mmol) and
H2O (100 lL) in MeOH and THF (1 mL, 1:1, vol:vol)
was stirred at rt for 1.5 h. The mixture was filtered,
the solvents were evaporated under reduced pressure
and the residue was purified using a solvent system gra-
dient EtOAc–MeOH, 10:0!6:4 to yield 10 mg of com-
1
Compound 10. Yield 86%; [a]D +0.183 (c 0.5); H
NMR (400 MHz, D2O): d 7.59–7.52 (m, 2H, HAr),
7.42–7.35 (m, 4H, HAr, CH@), 6.55 (d, J = 15.4 Hz,
1H, CH@), 4.15–4.08 (m, 1H, H-1), 3.91–3.87 (m, 2H,
H-2, 4), 3.70–3.60 (m, 4H, H-3, 5, 6a, 6b), 3.33–3.26
(m, 2H, CH2N), 1.91–1.85 (m, 1H, CHCH2N), 1.81–
1.72 (m, 1H, CHCH2N); 13C NMR (100 MHz, D2O):
d 171.4 (C), 143.7 (CH), 137.1 (C), 132.9 (CH), 131.7
(CH), 130.6 (CH), 122.8 (CH), 76.35 (CH), 74.63
(CH), 72.34 (CH), 71.70 (CH), 70.69 (CH), 63.87
(CH2), 39.26 (CH2), 25.93 (CH2); HRMS calcd for
[M+H]+: 338.1604, [M+Na]+: 360.1423. Found:
338.1765, 360.1302.
1
pound 14 (28%); H NMR (400 MHz, D2O): d 7.30–
7.21 (m, 6H, HAr), 7.18–7.04 (m, 4H, HAr), 4.18–4.05
(m, 1H, H-1), 4.00–3.92 (m, 2H, H-2, 4), 3.82–3.61 (m,
4H, H-3, 5, 6a, 6b), 3.44–3.31 (m, 2H, CH2N), 2.02–
1.90 (m, 1H, CHCH2N), 1.89–1.80 (m, 1H, CHCH2N);
13C NMR (100 MHz, CD3OD): d 174.7 (C), 146.8 (C),
139.4 (C), 138.6 (C), 136.5 (C), 133.0 (CH), 132.6
(CH), 132.3 (CH), 131.9 (CH), 131.6 (CH), 131.1
(CH), 130.9 (CH), 130.7 (CH), 130.6 (CH), 130.5
(CH), 74.8 (CH), 72.4 (2 · CH), 71.7 (CH), 70.9 (CH),
63.9 (CH2), 39.4 (CH2), 26.1 (CH2); HRMS calcd for
[M+NaꢀH2O]+: 462.1529. Found: 462.1719.
Compound 11. Yield 70%; 1H NMR (400 MHz,
D2O): d 7.31 (d, J = 15.7 Hz, 1 H, CH@), 6.82 (s, 2H,
HAr), 6.45 (d, J = 15.7 Hz, 1H, CH@), 4.13–4.07 (m,
1H, H-1), 3.98–3.90 (m, 2H, H-2, 4), 3.85–3.65 (m,
13H, H-3, 5, 6a, 6b, 3OCH3), 3.38 (br t, 2H, CH2N),
1.98–1.83 (m, 2H, CH2CH2N); 13C NMR (100 MHz,
D2O): d 180.4 (C), 155.2 (C), 143.3 (CH), 133.6 (C),
122.6 (CH), 108.0 (CH), 74.63 (CH), 72.33 (CH), 71.66
(CH), 70.67 (CH), 63.84 (CH2), 63.55 (CH), 58.63
(CH3), 39.20 (CH2), 23.63 (CH2); HRMS calcd for
[M+H]+: 428.1921. Found: 428.1967.
4.2. SPR measurements
The SPR measurements were performed on a double
channel BIOCORE instrument at 25 ꢁC, in HBS buffer.
The glycoprotein asialofetuin (ASF) was covalently
attached to a dextran functionalized gold SPR chip via
NHS active ester intermediacy following a protocol pre-
viously established.37 A series of measurements with
increasing CTB5 concentrations yielded a binding iso-
therm and a Kchip of 70 lM.38 After each binding exper-
iment the surface was regenerated with 10 mM NaOH
for 2.5 min. In a typical competition experiment,
CTB5 (70 lM in HBS buffer) was mixed with variable
concentrations of the ligands and allowed to reach equi-
librium for 5 min at rt. The equilibrated samples were
then introduced to the chip-immobilized protein. These
measurements generated a new series of binding iso-
therms, from which equilibrium constants (Req) were
Compound 12. Yield 75%; 1H NMR (400 MHz,
CD3OD): d 8.29–8.20 (m, 2H, HAr), 7.80–7.75 (m,
2H, HAr), 7.59 (d, J = 15.7 Hz, 1H, CH@), 6.81 (d,
J = 15.7 Hz, 1H, CH@), 4.12–4.04 (m, 1H, H-1), 3.96–
3.82 (m, 3H, H-2, 4, 6a), 3.80–3.76 (m, 1H, H-5),
3.72–3.65 (m, 2H, H-3, 6b), 3.59–3.50 (m, 1H, CHN),
3.48–3.362 (m, 1H, CHN), 1.97–1.80 (m, 2H,
CH2CH2N); 13C NMR (100 MHz, CD3OD): d 170.6
(C), 152.3 (C), 145.5 (C), 141.9 (CH), 132.6 (CH),
129.0 (CH), 127.9 (CH), 77.22 (CH), 76.91 (CH), 74.65
(CH), 72.93 (CH), 72.93 (CH), 65.07 (CH2), 40.92

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1658
100
script Pprep was run on protein in the presence of bound
ligand. This script neutralizes all residues of the protein
except those that are close to the ligand or those that
form salt bridges. No H-bonding conflict was found.
We then proceeded to add the hydrogens to the protein
and to the water molecules. One water molecule was
reoriented so as to form hydrogen bond with the car-
boxyl group of sialic acid, as it is suggested by analysis
of the X-ray.8 We then achieved structure preparation
by using the Schro¨dinger, Inc script premin. This script
performs a minimisation in three steps in order to allow
the complex obtained so far to relax without losing the
original structure. The minimization is limited in num-
ber of steps and is not intended to attain convergence.
Instead, the process is stopped as soon as the RMS rel-
80
60
40
20
0
0
2
4
6
8
10
12
14
[ ], mM
˚
ative to the original structure exceeds 0.4 A.
Figure 7. SPR experiments: hyperbolic fitting for compound 10.
For the full minimization of the complexes, we used
the Amber force field with the Still-Senderowitz pyra-
*
nose parameters.42 The Polak-Ribiere Conjugate Gradi-
ent (PRCG) algorithm was used with a limit of 5000
steps, unless the energy gradient converged below
0.05 kJ molꢀ1 before the limit was reached. The minimi-
sation used several shells so as to allow the ligand to re-
lax in its new environment and the protein to be flexible
enough to allow this relaxation. The first shell consisted
obtained using SPR instrument software. A hyperbolic
fitting (y = Rmax IC50/(IC50 + x)) was then used in or-
der to extrapolate the ligand’s affinity to the protein.
In Figure 7, the hyperbolic fitting of 10 is shown.
4.3. Modeling section
˚
4.3.1. Model preparation. We started from the X-ray
CT–GM1 complex (PDB code:3CHB).8 It is a confor-
mation of the B5 subunit of the Cholera Toxin with five
pentasaccharides GM1 bound to the five binding sites.
The binding sites are formed by pairs of B units that
are in contact with one another. So two domains on
the five that constitute the B5 subunit of the Cholera
Toxin are necessary to model the binding site. We may
have chosen any of those pairs, but in the sites involving
the chains D and F lattice interactions in the crystal
form have prevented the position of the terminal glucose
residue to be determined. Therefore they are missing in
the PDB file. So we have chosen the chains G and H
as models. A sulfur atom of unknown origin and a
molecule of 2-(N-morpholino)-ethane-sulfonic acid
coming from the well solution used for crystallization
were removed.
The protein was prepared for the virtual screening
experiments following the protocol recommended by
Schro¨dinger.39 The first phase was to suppress all water
molecules except those that are relevant for the binding.
Two water molecules in the binding site (W7302 and
W7303 in the PDB file) have been identified as playing
an important role for CT binding events40 and were
retained. Hydrogens were added to the ligand automat-
ically via the Maestro interface, leaving no lone pair and
using an explicit all-atom model. Multiple bonds and
formal charges were checked and corrected for the car-
boxyl group of the sialic acid in GM1. After separation
of the complex in separate files for the protein, the
ligand, and the water molecules, the Schro¨dinger, Inc
of the ligand and the polar hydrogens within 5 A of the
ligand. The second shell consisted of all the atoms within
˚
5 A of the ligand that did not belong to the first shell. A
constraint of 200 kJ molꢀ1
A
ꢀ2 was applied to maintain
˚
these atoms in place. The third shell consisted of all
˚
atoms within 15 A of the ligand that were not already in-
cluded. Those atoms were completely frozen, though
they participated in the force field. All other atoms were
ignored.
4.3.2. Docking experiments. The minimized complex
CTB5:GM1 was used for generation of Autodock’s
and Glide’s grids. The grids were needed for flexible
docking with Autodock and subsequent scoring in situ
with Glide. All single bonds in ligands are treated as
freely rotatable in the Autodock procedure, except the
C3–OH, C4–OH, C5–C6 and C6–OH bonds of the gal-
actose anchor.
4.3.2.1. Autodock setup. Autodock 3.0 was used,
starting from the complex structure prepared as de-
scribed in the Modeling section and generating the grid
from the GM1 ligand using the standard Autodock pro-
cedure (from Autodock tutorial). The grid spacing was
standard, and the size is set to include all the CT binding
sites. We used standard setup of the software, which
makes use of a Lamarckian genetic algorithm. Each
docking experiment was repeated 100 times. The popu-
lation size was set to 50, the number of energy evalua-
tions per run was set to 106, and the maximum
number of generations to 54,000. All other parameters

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1659
OH
HO
OH
OH
HO
HO
OH
O
HO
+
O
H
COOH
N
COOH
%90
NH
O
%90
O
OC[C@H]1O[C@H](CC%90(=O)[C@H](O)[C@@H](O)[C@H]1O OC(=O)C[C@H](N%90)C1=CC=CC=C1
.
SMILES TO 3D
CORINA
Figure 8. Virtual chemical reaction using a combination of SMILES strings.
used default values. In particular, the rms tolerance for
clustering was set to 1 A. The poses obtained by this
A virtual combinatorial library was prepared by com-
bining SMILES strings.43 Attaching points on both
reactants have been assigned by prefacing the same
two-digit number with percent sign (%). The new com-
pounds are created by combining SMILES string of
two reactants into the product using dot (.) operator.
Example of such virtual chemical reaction is shown in
Figure 8.
˚
procedure were minimized using the same protocol de-
scribed above for the CT–GM1 complex and scored
in situ with Glide. The result of a docking experiment
following this procedure is reproducible, that is, even
if Autodock uses a genetic algorithm, the final set of
docked poses is always the same as that of the
GlideScore.
The final structure represented by SMILES string was
then converted to 3D structures using CORINA.44,45
4.3.2.2. Glide setup. Glide uses a number of grids to
explore the conformational space of ligands in the bind-
ing site and scores each of them. We have used the com-
plex formed by the GH chains of the B5 subunit of the
Cholera Toxin and GM1 to help Glide define the ligand
site. All the grids are calculated in an enclosing box. The
size of this box has been fitted to GM1 as it was the larg-
est ligand we wanted to dock. Note that the grids were
defined for a given conformation of the binding site.
This conformation changed slightly each time we mini-
mized a ligand in the binding pocket because the orien-
tation of the polar hydrogen was unconstrained. Those
differences were not sufficient nonetheless to justify
recalculations of the grid for each minimized complex.
Acknowledgments
We thank the European Union for financial support
(under contract HPRN-CT-2002-00173, Glycidic Scaf-
folds Network) and the Interdepartmental Center for
Large Instrumentation (CIGA) at the University of
Milan for use of the SPR instrumentation.
References
1. Cholera 1999, WHO Weekly Epidemiolog. Rev. Vol. 75,
August 4, 2000; pp 249–256.
2. Lencer, W. I.; Slaslwsky, D. Biochim. Biophys. Acta 2005,
1746, 314–321.
4.4. Focused virtual library
3. Fan, E.; O’Neil, C. J.; Mithchell, D. D.; Robien, M. A.;
Zhang, Z.; Pickens, J. C.; Tan, X.-J.; Korotkov, K.;
Roach, C.; Krumm, B.; Verlinde, C. L. M. J.; Merritt, E.
A. Int. J. Med. Micr. 2004, 294, 217–223.
4. Moss, J.; Vaughan, M. J. Biol. Chem. 1977, 252, 2455–
2457.
The virtual library was generated using an in-house soft-
ware by systematically combining a set of fragments on
a set of scaffolds. The user defines attachment points for
addition of the fragments to the scaffolds. Thus, we were
always sure to generate libraries of compounds that
were synthetically accessible.
5. Cassel, D.; Pfeuffer, T. Proc. Natl. Acad. Sci. U.S.A. 1978,
75, 2669–2673.

ˇ
C. Podlipnik et al. / Carbohydrate Research 342 (2007) 1651–1660
1660
6. (a) Fan, E.; Merritt, E. A.; Verlinde, C. L. M. J.; Hol, W.
26. Nicotra, F. Modified Carbohydrates and Carbohydrate
Analogs; Chapman & Hall, Blackie Academic & Profes-
sional: London, 1997, pp 384–429.
G. J. Curr. Opin. Chem. Biol. 2000, 10, 680–686; (b) Velter,
ˇ
I. A.; Politi, M.; Podlipnik, C.; Nicotra, F. Minirev. Med.
Chem. 2007, 7, 159–170.
27. Du, Y.; Linhardt, J. R. Tetrahedron 1998, 54, 9913–
9959.
28. Bililign, T. G. B. R.; Thorson, J. S. Nat. Prod. Rep. 2005,
22, 742–760.
29. Yuan, Y.; Linhardt, J. R. Curr. Top. Med. Chem. 2005, 5,
1393–1430.
7. Merritt, E. A.; Kuhn, P.; Sarfaty, S.; Erbe, J. L.; Holmes,
R. K.; Hol, W. G. J. J. Mol. Biol. 1998, 282, 1043–1059.
8. Merritt, E. A.; Sarfaty, S.; van den Akker, F.; L’Hoir, C.;
Martial, J. A.; Hol, W. G. J. Protein Sci. 1994, 3, 166–175.
9. Turnbull, W. B.; Precious, B. L.; Homans, S. W. J. Am.
Chem. Soc. 2004, 126, 1047–1054.
30. La Ferla, B.; Cardona, F.; Perdigao, I.; Nicotra, F. Synlett
˜
10. Bernardi, A.; Checchia, A.; Brocca, P.; Sonnino, S.;
Zuccotto, F. J. Am. Chem. Soc. 1999, 121, 2032–2036.
11. Acquotti, D.; Poppe, L.; Dabrowski, J.; van der Lieth,
C.-W.; Sonnino, S.; Tettamanti, G. J. Am. Chem. Soc.
1990, 112, 7772–7778.
2005, 17, 2641–2642.
31. Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.;
Hart, W. E.; Belew, R. K.; Olson, A. J. J. Comput. Chem.
1998, 19, 1639–1662.
32. Minke, W. E.; Diller, D. J.; Hol, W. G. J.; Verlinde, C. L.
J. Med. Chem. 1999, 42, 1778–1788.
33. Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T.
A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E.
H.; Shaw, D. E.; Shelley, M.; Perry, J. K.; Francis, P.;
Shenkin, P. S. J. Med. Chem. 2004, 47, 1739–1749.
34. Halgren, T. A.; Murphy, R. B.; Friesner, R. A.; Beard, H.
S.; Frye, L. L.; Pollard, W. T.; Banks, J. L. J. Med. Chem.
2004, 47, 1750–1759.
ˇ
´
12. Bernardi, A.; Podlipnik, C.; Jimenez-Barbero, J. GM1
Glycomimetics and Bacterial Enterotoxins. In Protein–
Carbohydrate Interactions in Infectious Diseases; RSC:
Bewley, CA, 2006; pp 73–91.
13. Bernardi, A.; Arosio, D.; Potenza, D.; Sanchez-Medina,
´
I.; Mari, S.; Canada, F. J.; Jimenez-Barbero, J. Chem. Eur.
˜
J. 2004, 10, 4395–4406.
14. Arosio, D.; Baretti, S.; Cattaldo, S.; Potenza, D.; Ber-
nardi, A. Bioorg. Med. Chem. Lett. 2003, 3831–3834.
15. Bernardi, A.; Carrettoni, V.; Grosso Ciponte, V.; Monti,
D.; Sonnino, S. Bioorg. Med. Chem. Lett. 2000, 13, 2197–
2200.
35. Bennek, J. A.; Gray, G. A. J. Org. Chem. 1987, 52, 892–
897.
36. Broxterman, H. J. G.; Kooreman, P. A.; van der Elst, H.;
Roelen, H. C. P. F.; van der Marel, G. A.; van Boom, J.
H. Recl. Trav. Chim. Pays-Bas 1990, 109, 583.
37. Arosio, D.; Vrasidas, I.; Valentini, P.; Liskamp, R. M. J.;
Pieters, R. J.; Bernardi, A. Org. Biomol. Chem. 2004, 2,
2113–2124.
`
16. Bernardi, A.; Potenza, D.; Capelli, A. M.; Garcıa-Herrero,
´
A.; Canada, F. J.; Jimenez-Barbero, J. Chem. Eur. J. 2002,
˜
8, 4598–4612.
17. Bernardi, A.; Arosio, D.; Sonnino, S. Neurochem. Res.
2002, 27, 539–545.
18. Bernardi, A.; Arosio, D.; Manzoni, L.; Monti, D.; Posteri,
38. de Mol, N. J.; Plomp, E.; Fischer, M. J. E.; Ruijtenbeek,
R. Anal. Biochem. 2000, 279, 61–70.
´
H.; Potenza, D.; Mari, S.; Jimenez-Barbero, J. Org.
Biomol. Chem. 2003, 1, 1–9.
39. Schrodinger. FirstDiscovery 2.5, User Manual, March
2003.
19. Minke, W. E.; Roach, C.; Hol, W. G.; Verlinde, C. L.
Biochem. 1999, 38, 5684–5692.
40. Bernardi, A.; Raimondi, L.; Zuccotto, F. J. Med. Chem.
1997, 40, 1855–1862.
20. Minke, W. E.; Hong, F.; Verlinde, C. L.; Hol, W. G.; Fan,
E. J. Biol. Chem. 1999, 274, 33469–33473.
41. Warren, G. L.; Andrews, C. W.; Capelli, A.-M.; Clarke,
B.; LaLonde, J.; Lambert, M. H.; Lindvall, M.; Nevins,
N.; Semus, S. F.; Senger, S.; Tedesco, G.; Wall, I. D.;
Woolven, J. M.; Peishoff, C. E.; Head, M. S. J. Med.
Chem. 2006, 49, 5912–5931.
42. Senderowitz, H.; Parish, C.; Still, W. C. J. Am. Chem. Soc.
1996, 118, 2078–2086.
43. Weininger, D. J. Chem. Inf. Comput. Sci. 1988, 28, 31–36.
44. Sadowski, J.; Gasteiger, J. Chem. Rev. 1993, 93, 2567–
2581.
21. Pickens, J. C.; Merritt, E. A.; Ahn, M.; Verlinde, C. L. M.
J.; Hol, W. G. J.; Fan, E. Chem. Biol. 2002, 9, 215–224.
22. Mitchell, D. D.; Pickens, J. C.; Korotkov, K.; Fan, E.;
Hol, W. G. J. Bioorg. Med. Chem. 2004, 12, 907–920.
23. Postema, M. H. D. C-Glycoside Synthesis; CRC Press:
Boca Raton, 1995.
24. Levy, D. E.; Tang, C. The Chemistry of C-Glycosides.
Tetrahedron Organic Chemistry Series; Pergamon: New
York, 1995; Vol. 13.
45. Sadowski, J.; Gasteiger, J.; Klebe, G. J. Chem. Inf.
Comput. Sci. 1994, 34, 1000–1008.
46. Mertz, J. A.; McCann, J. A.; Picking, W. D. Biochem.
Biophys. Res. Commun. 1996, 226, 140–144.
25. Nicotra, F. Synthesis of C-glycosides of Biological Inter-
est. In Topics in Current Chemistry; Springer: Berlin
Heidelberg, 1997; Vol. 187, pp 56–83.