Structural and thermodynamic studies on cation-II interactions in lectin-ligand complexes: High-affinity galectin-3 inhibitors through fine-tuning of an arginine-arene interaction

Article abstract of DOI:10.1021/ja043475p

The high-resolution X-ray crystal structures of the carbohydrate recognition domain of human galectin-3 were solved in complex with N-acetyllactosamine (LacNAc) and the high-affinity inhibitor, methyl 2-acetamido-2-deoxy-4-O-(3-deoxy-3-[4-methoxy-2,3,5,6-tetrafluorobenzamido] -β-D-galactopyranose)-β-D-glucopyranoside, to gain insight into the basis for the affinity-enhancing effect of the 4-methoxy-2,3,5,6- tetrafluorobenzamido moiety. The structures show that the side chain of Arg144 stacks against the aromatic moiety of the inhibitor, an interaction made possible by a reorientation of the side chain relative to that seen in the LacNAc complex. Based on these structures, synthesis of second generation LacNAc derivatives carrying aromatic amides at 3′-C, followed by screening with a novel fluorescence polarization assay, has led to the identification of inhibitors with further enhanced affinity for galectin-3 (K_d ≥ 320 nM). The thermodynamic parameters describing the binding of the galectin-3 C-terminal to selected inhibitors were determined by isothermal titration calorimetry and showed that the affinity enhancements were due to favorable enthalpic contributions. These enhancements could be rationalized by the combined effects of the inhibitor aromatic structure on a cation-Π interaction and of direct interactions between the aromatic substituents and the protein. The results demonstrate that protein-ligand interactions can be significantly enhanced by the fine-tuning of arginine-arene interactions.

Full text of DOI:10.1021/ja043475p

Published on Web 01/21/2005
Structural and Thermodynamic Studies on Cation-Π
Interactions in Lectin-Ligand Complexes: High-Affinity
Galectin-3 Inhibitors through Fine-Tuning of an
Arginine-Arene Interaction
Pernilla So¨rme,†,‡ Pascal Arnoux,§,| Barbro Kahl-Knutsson,‡ Hakon Leffler,*,‡
James M. Rini,*,§ and Ulf J. Nilsson*,†
Contribution from Organic and Bioorganic Chemistry, Lund UniVersity, P. O. Box 124,
SE-221 00 Lund, Sweden, Section MIG, Department of Laboratory Medicine, UniVersity of Lund,
So¨lVegatan 23, SE-223 62 Lund, Sweden, and Departments of Medical Genetics and
Microbiology and Biochemistry, UniVersity of Toronto, Toronto, Ontario, M5S 1A8, Canada
Received October 28, 2004; E-mail: ulf.nilsson@bioorganic.lth.se
Abstract: The high-resolution X-ray crystal structures of the carbohydrate recognition domain of human
galectin-3 were solved in complex with N-acetyllactosamine (LacNAc) and the high-affinity inhibitor, methyl
2-acetamido-2-deoxy-4-O-(3-deoxy-3-[4-methoxy-2,3,5,6-tetrafluorobenzamido]-â-D-galactopyranose)-â-D-
glucopyranoside, to gain insight into the basis for the affinity-enhancing effect of the 4-methoxy-2,3,5,6-
tetrafluorobenzamido moiety. The structures show that the side chain of Arg144 stacks against the aromatic
moiety of the inhibitor, an interaction made possible by a reorientation of the side chain relative to that
seen in the LacNAc complex. Based on these structures, synthesis of second generation LacNAc derivatives
carrying aromatic amides at 3′-C, followed by screening with a novel fluorescence polarization assay, has
led to the identification of inhibitors with further enhanced affinity for galectin-3 (Kd g 320 nM). The
thermodynamic parameters describing the binding of the galectin-3 C-terminal to selected inhibitors were
determined by isothermal titration calorimetry and showed that the affinity enhancements were due to
favorable enthalpic contributions. These enhancements could be rationalized by the combined effects of
the inhibitor aromatic structure on a cation-Π interaction and of direct interactions between the aromatic
substituents and the protein. The results demonstrate that protein-ligand interactions can be significantly
enhanced by the fine-tuning of arginine-arene interactions.
Introduction
papers in ref 11). Nevertheless, the galectins have been clearly
implicated in physiological and pathological mechanisms in
The galectins are a family of proteins defined by a carbo-
hydrate recognition domain (CRD) of about 135 amino acids
with certain conserved sequence motifs and affinity for â-
galactosides.1,2 They appear to be involved in a wide variety of
both extracellular (cell adhesion, induction of cell signaling
including apoptosis) and intracellular activities (regulation of
growth and apoptosis and RNA-splicing), although in many
cases their functions remain unclear (see refs 3-10 and additional
inflammation, immunity5,12-14 and cancer.6,7,10,15 In addition, the
modulation of galectin expression or activity has been linked
to disease outcome;5,7,16 most recently galectin-2 17 was found
to be associated with increased risk for myocardial infarction
and galectin-4 18 with inflammatory bowel disease. For these
(8) Patterson, R. J.; Wang, W.; Wang, J. L. Glycoconjugate J. 2004, 19, 499-
506.
(9) Hsu, D. K.; Liu, F. T. Glycoconjugate J. 2004, 19, 507-515.
(10) Lahm, H.; Andre, S.; Hoeflich, A.; Kaltner, H.; Siebert, H. C.; Sordat, B.;
von der Lieth, C.-W.; Wolf, E.; Gabius, H.-J. Glycoconjugate J. 2004, 20,
227-238.
(11) Leffler, H., Ed. Glycoconjugate J. 2004, 19, 433-468.
(12) Almkvist, J.; Karlsson, A. Glycoconjugate J. 2004, 19, 575-581.
(13) Hirashima, M.; Kashio, Y.; Nishi, N.; Yamauchi, A.; Imaizumi, T. A.;
Kageshita, T.; Saita, N.; Nakamura, T. Glycoconjugate J. 2004, 19, 593-
600.
† Lund University.
‡ University of Lund.
§ University of Toronto.
| Current address: CEA/Cadarache, DSV, DEVM, Laboratoire de
Bioe´nerge´tique Cellulaire, 13108 Saint Paul lez Durance, France.
(1) Houzelstein, D.; Goncalves, I. R.; Fadden, A. J.; Sidhu, S. S.; Cooper, D.
N.; Drickamer, K.; Leffler, H.; Poirier, F. Mol. Biol. EVol. 2004, 21, 1177-
1187.
(14) Sato, S.; Nieminen, J. Glycoconjugate J. 2004, 19, 583-591.
(15) Partridge, E. A.; Le Roy, C.; Di Guglielmo, G. M.; Pawling, J.; Cheung,
P.; Granovsky, M.; Nabi, I. R.; Wrana, J. L.; Dennis, J. W. Science 2004,
306, 120-124.
(2) Leffler, H.; Carlsson, S.; Hedlund, M.; Qian, Y. Glycoconjugate J. 2004,
19, 433-440.
(3) Zick, Y.; Eisenstein, M.; Goren, R. A.; Hadari, Y. R.; Levy, Y.; Ronen, D.
Glycoconjugate J. 2004, 19, 517-526.
(16) John, C. M.; Leffler, H.; Kahl-Knutsson, B.; Svensson, I.; Jarvis, G. A.
Clin. Cancer Res. 2003, 9, 2374-2383.
(4) Ochieng, J.; Furtak, V.; Lukyanov, P. Glycoconjugate J. 2004, 19, 527-
535.
(17) Ozaki, K.; Inoue, K.; Sato, H.; Iida, A.; Ohnishi, Y.; Sekine, A.; Odashiro,
K.; Nobuyoshi, M.; Hori, M.; Nakamura, Y.; Tanaka, T. Nature 2004, 429,
72-75.
(5) Rabinovich, G. A.; Toscano, M. A.; Ilarregui, J. M.; Rubinstein, N.
Glycoconjugate J. 2004, 19, 565-573.
(6) Huflejt, M.; Leffler, H. Glycoconjugate J. 2004, 20, 247-255.
(7) Takenaka, Y.; Fukumori, T.; Raz, A. Glycoconjugate J. 2004, 19, 543-
549.
(18) Hokama, A.; Mizoguchi, E.; Sugimoto, K.; Shimomura, Y.; Tanaka, Y.;
Yoshida, M.; Rietdijk, S. T.; de Jong, Y. P.; Snapper, S. B.; Terhorst, C.;
Blumberg, R. S.; Mizoguchi, A. Immunity 2004, 20, 681-693.
9
10.1021/ja043475p CCC: $30.25 © 2005 American Chemical Society
J. AM. CHEM. SOC. 2005, 127, 1737-1743
1737

A R T I C L E S
So¨rme et al.
Herein we report the 1.6 Å resolution crystal structure of the
human galectin-3C in complex with inhibitor 3, the first structure
of a galectin-3 complex with a synthetic ligand. Moreover, we
have now extended the resolution of the LacNAc complex to
1.4 Å resolution. The crystal structure of the galectin-3C-3
complex points to the importance of an arginine-arene interac-
tion and encouraged us to synthesize a further 56 inhibitors (4-
59) consisting of LacNAc derivatives carrying aromatic amides
at 3′-C. Evaluation of the galectin-3 inhibitory activity of these
inhibitors (4-59) with a novel fluorescence polarization assay26
revealed significantly improved inhibitors with Kd values as low
as 320 nM. The thermodynamic parameters describing the
binding of these improved inhibitors, determined by isothermal
titration microcalorimetry (ITC), showed that complex formation
was enthalpy-driven.
Figure 1. Structures of the parent methyl glycoside of N-acetyllactosamine
1 and the corresponding 3′-benzamido- and 3′-(4-methoxy-2,3,5,6-tetra-
fluorobenzamido) high-affinity inhibitors 2 and 3, respectively.
Results and Discussion
To understand the structural basis for the high affinity of
compound 3 for galectin-3 and hence to obtain information to
guide us in the design and synthesis of even better inhibitors,
the crystal structures of galectin-3C in complex with LacNAc
and 3 were solved and refined at 1.4 and 1.6 Å resolution,
respectively. Although these crystals are essentially isomorphous
with those reported previously,25 the LacNAc structure was
solved by the MAD phasing method using the sodium bromide
quick-soak approach as a test of the ability of this method to
provide high-quality experimentally determined phase informa-
tion. Experimental phases were calculated to 2.0 Å resolution,
resulting in very high quality electron density maps. The starting
point for refinement of the complex with compound 3 was that
of the galectin-3C coordinates obtained from the LacNAc
complex. (See Supplementary Information for phasing and
refinement statistics.)
Structure of Galectin-3 in Complex with LacNAc. Overall,
the LacNAc complex refined here at 1.4 Å resolution using
synchrotron data collected at 100 K agrees well with the
structure refined previously at 2.1 Å resolution using data
collected on a rotating anode at room temperature.25 Neverthe-
less, the structure has been improved in a number of ways. First,
one more residue (Pro 113) is clearly visible in the electron
density at the N-terminus of the molecule. The conformation
of the molecule in this region is of interest as it forms the
connection to the N-terminal domain unique to galectin-3, a
conformation that is not compatible with the subunit interactions
found to mediate dimer formation in the canonical dimeric
galectins-1 and -2.27-29 Second, the orientations of a number
of side chains have been better defined. Among those close to
the binding site, the side chain of Arg144 is slightly different
in that the guanidino group is rotated about the Cδ Nꢀ bond
and all three nitrogen atoms are now found to form hydrogen
bonds with surrounding water molecules. In addition, the side
chain Arg129 is oriented in such a way that it makes a hydrogen
bond with the carboxyl terminus of the molecule through its
Nꢀ and NH2. Third, the number of ordered water molecules is
reasons, galectin inhibitors would be valuable not only as tools
for probing galectin function but also as lead compounds in
the future development of therapeutics.
Lectin-monosaccharide complexes typically involve two to
five hydrogen bonds complemented by hydrophobic and van
der Waals interactions and are in general weak (Kd’s in the
millimolar range). Consequently, the design of small high-
affinity lectin inhibitors has proven to be a considerable
challenge. Since many lectins are oligomeric in nature, the most
successful lectin inhibitors, to date, have been multivalent
compounds designed to interact with more than one of the
lectin’s subunits.19-21 Galectin-3, however, is mainly monomeric
in solution, albeit in equilibrium with a lower percentage of
oligomers,22 and multivalent compounds did not seem to give
higher inhibitory potency.23 Therefore, we set out to design high-
affinity inhibitors capable of tightly binding monomeric galectin-
3.24 The approach was based on the fact that the three-
dimensional structure of the galectin-3 CRD (galectin-3C), in
complex with LacNAc,25 had shown that the 3-OH of the
galactoside residue points toward an extension of the lactose/
LacNAc binding site thought to be responsible for conferring
the selectivity shown by galectin-3 for longer oligosaccharides.
Chemical modifications at this position led us to the discovery24
that benzamide substituents at LacNAc 3′-C substantially
increased the affinity of galectin-3 for these compounds. One
inhibitor (3; Figure 1), carrying a 4-methoxy-2,3,5,6-tetra-
fluorobenzamido group at 3′-C was shown to inhibit intact
galectin-3 50 times better than the parent methyl 2-acetamido-
2-deoxy-4-O-â-D-galactopyranose-â- D-glucopyranoside (1). The
benzamide substitution pattern was concluded to be of critical
importance, as the unsubstituted inhibitor 2 was only 15 times
better than compound 1.
(19) Blixt, O.; Collins, B. E.; van den Nieuwenhof, I. M.; Crocker, P. R.; Paulson,
J. C. J. Biol. Chem. 2003, 278, 31007-31019.
(20) Gestwicki, J. E.; Cairo, C. W.; Strong, L. E.; Oetjen, K. A.; Kiessling, L.
L. J. Am. Chem. Soc. 2002, 124, 14922-14933.
(21) Kitov, P. I.; Shimizu, H.; Homans, S. W.; Bundle, D. R. J. Am. Chem.
Soc. 2003, 125, 3284-3294.
(22) Ahmad, N.; Gabius, H.-J.; Andre´, S.; Kaltner, H.; Sabesan, S.; Roy, R.;
Liu, B.; Macaluso, F.; Brewer, C. F. J. Biol. Chem. 2004, 279, 10841-
10847.
(26) So¨rme, P.; Kahl-Knutsson, B.; Wellmar, U.; Nilsson, U. J.; Leffler, H.
Methods Enzymol. 2003, 362, 504-512.
(23) Ahmad, N.; Gabius, H.-J.; Sabesan, S.; Oscarson, S.; Brewer, C. F.
Glycobiology 2004, 14, 817-825.
(27) Lobsanov, Y. D.; Gitt, M. A.; Leffler, H.; Barondes, S. H.; Rini, J. M. J.
Biol. Chem. 1993, 268, 27034-27038.
(24) So¨rme, P.; Qian, Y.; Nyholm, P.-G.; Leffler, H.; Nilsson, U. J. ChemBio-
Chem 2002, 3, 183-189.
(28) Liao, D. I.; Kapadia, G.; Ahmed, H.; Vasta, G. R.; Herzberg, O. Proc.
Natl. Acad. Sci. U.S.A. 1994, 91, 1428-1432.
(25) Seetharaman, J.; Kaningsberg, A.; Slaaby, R.; Leffler, H.; Barondes, S.
H.; Rini, J. M. J. Biol. Chem. 1998, 273, 13047-13052.
(29) Bourne, Y.; Bolgiano, B.; Liao, D. I.; Strecker, G.; Cantau, P.; Herzberg,
O.; Feizi, T.; Cambillau, C. Nat. Struct. Biol. 1994, 1, 863-870.
9
1738 J. AM. CHEM. SOC. VOL. 127, NO. 6, 2005

Cation−Π Interactions in Lectin−Ligand Complexes
A R T I C L E S
an attractive target for the design of further optimized inhibitors
as described below.
The complex with inhibitor 3 shows that the side chain of
Arg144 makes what appears to be a cation-Π interaction31 with
the benzyl substituent of the inhibitor. The 4.3 Å distance seen
between Nꢀ of the arginine side chain and the centroid of the
benzamido ring is in the range of what has previously been
observed for arene-guanidino distances within proteins struc-
tures.32 Cation-Π interactions have now been shown to be
important in protein stability, ligand binding, and enzyme
catalysis and typically, Arg side chains, Lys side chains, metal
ions, or ligand cations are found to interact with the side chains
of Trp, Tyr, or Phe.33-35 Experiments designed to measure the
strength of cation-Π interactions in biological systems have
shown that they range from -0.4 to -2.4 kcal/mol.35 Given
that fluorinated arenes have electron-poor Π-systems (in fact,
the quadrupole moment of 4-methoxy-2,3,5,6-tetrafluorobenz-
amide is small according to ab initio calculations) and thus form
weak cation-Π interactions,36,37 it may seem surprising that
the affinity-enhancing effect of the 4-methoxy-2,3,5,6-tetra-
fluorobenzamido-guanidino interaction is so large. However,
it has been shown that the electrostatic attraction between the
cation and the permanent quadrupole moment of the Π system
does not account entirely for the observed energetics in systems
involving cation-Π interactions31,35 and that van der Waals
interations,38 Π-Π interactions, charge-transfer effects, and
polarization effects must also be considered. In this case, the
methoxy group of 3 is positioned close to the guanidino group
of Arg144, which may allow for additional electrostatic interac-
tions that further increase the affinity.39 Moreover, fluorinated
hydrocarbons are poorly solvated in water and desolvation of
the 4-methoxy-2,3,5,6-tetrafluorobenzamido moiety can be
expected to contribute substantially to the affinity of 3 for
galectin-3.
Figure 2. 2F0-Fc electron density map contoured at 1σ showing the
electron density associated with the inhibitor 3. The protein surface
accessible is shown and colored according to the charge (blue is positive,
and red is negative).
slightly increased in the present high-resolution structure to 180,
as compared to the 120 seen at lower resolution. Finally, we
observed six residues in alternate conformation, namely, Ile134,
Asn179, Asn222, Ser232, Ser244, and Met249. None of these
residues is in contact with the LacNAc disaccharide. Excluding
the atoms in alternate conformations, the root mean square
deviation (RMSD) of all atoms between the low- and high-
resolution structures is 0.26 Å. The high-resolution LacNAc
complex structure now includes 137 residues, 180 water
molecules, six residues in alternate conformations, one chloride
ion, and the LacNAc disaccharide.
Structure of Galectin-3 in Complex with Inhibitor 3.
Inhibitor 3 is found to bind in the galectin-3C LacNAc binding
site with the benzyl substituent in the extended site as expected
(Figure 2). The benzamide nitrogen at 3′-C of the galactose
residue is hydrogen-bonded to a water molecule, which is in
turn hydrogen-bonded to Oδ1 of Asn160 and NH1 of Arg144
(Figure 3a,b). The benzamide carbonyl oxygen is hydrogen-
bonded with two water molecules, one of which is further
hydrogen-bonded to the Nꢀ1 of Trp 181. These interactions with
the amido functionality may explain why the corresponding
phenyl sulfonamides or ureas at 3′-C are less efficient inhibi-
tors.24
Upon binding of inhibitor 3 the only significant change in
the galectin-3C structure is the conformational change observed
in the side chain of Arg144. The Arg144 guanidino group, which
in the complex with LacNAc is involved in a hydrogen bond
with water molecules at the galectin-3C surface, is moved 3.5
Å to interact with the benzamido ring of 3 (Figure 3c,d). From
being aligned along the galectin-3C surface in the complex with
LacNAc, the side chain of Arg144 adopts an all-staggered
conformation and points out toward the solvent in the complex
with 3. This type of ligand-induced change in conformation is
an example of one of the complications associated with the
structure-based approached to ligand design as recently re-
viewed.30 The central carbon (Cú) of the Arg144 guanidino
group is stacked on one side of the benzamido ring at a distance
of 3.5-4.2 Å from the carbon atoms of the aromatic ring. On
the other side of the benzamido ring, two water molecules are
found within 4.5 Å of its centroid. Additionally, one edge of
the benzamido ring sits in an extended nonpolar pocket formed
by the side chains of Arg144(Câ to Cδ), Ala146(Câ), and
Asn160(Câ). This extended nonpolar pocket thus emerged as
Synthesis of Second Generation Inhibitors. Earlier studies
in natural systems37,40 and in minimalized artificial systems in
water41 and organic solvents39,42 have concluded that electron-
withdrawing substituents (i.e. fluorine or nitro) on the aromatic
component weaken while electron-donating substituents (i.e.
dimethylamino) strengthen cation-Π interactions, observations
that suggest that the Π-electron density of the aromatic
component is of primary importance. To investigate the influ-
ence of the structure of the aromatic component of an arginine-
arene cation-Π interaction in a natural system and to discover
further improved inhibitors of galectin-3, we prepared 42
(31) Ma, J. C.; Dougherty, D. A. Chem. ReV. 1997, 97, 1303-1324.
(32) Flocco, M. M.; Mowbray, S. L. J. Mol. Biol. 1994, 235, 709-717.
(33) Gallivan, J. P.; Dougherty, D. A. Proc. Natl. Acad. Sci. U.S.A. 1999, 96,
9459-9464.
(34) Zacharias, N.; Dougherty, D. A. Trends Pharm. Sci. 2002, 23, 281-287.
(35) Meyer, E. A.; Castellano, R. K.; Diedrich, F. Angew. Chem., Int. Ed. 2003,
42, 1210-1250.
(36) Mecozzi, S.; West, A. P., Jr.; Dougherty, D. A. J. Am. Chem. Soc. 1996,
118, 2307-2308.
(37) Zhong, W.; Gallivan, J. P.; Zhang, Y.; L., L.; Lester, H. A.; Dougherty, D.
A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 12088-12093.
(38) Mao, L.; Wang, Y.; Liu, Y.; Hu, X. J. Am. Chem. Soc. 2003, 125, 14216-
14217.
(39) Rensing, S.; Arendt, M.; Springer, A.; Grawe, T.; Schrader, T. J. Org.
Chem. 2001, 66, 5814-5821.
(40) Mu, T.-W.; Lester, H. A.; Dougherty, D. A. J. Am. Chem. Soc. 2003, 125,
6850-6851.
(41) Ngola, S. M.; Dougherty, D. A. J. Org. Chem. 1998, 63, 4566-4567.
(42) Hunter, C. A.; Low, C. M. R.; Rotger, C.; Vinter, J. G.; Zonta, C. Proc.
Natl. Acad. Sci. U.S.A. 2002, 99, 4873-4876.
(30) Davis, A. M.; Teague, S. J.; Kleywegt, G. J. Angew. Chem., Int. Ed. 2003,
42, 2718-2736.
9
J. AM. CHEM. SOC. VOL. 127, NO. 6, 2005 1739

A R T I C L E S
So¨rme et al.
Figure 3. Interaction of the inhibitor 3 with galectin-3C. (a) Schematic of the inhibitor-protein contacts. The colored circular lines around the inhibitor
indicate solvent accessibility with dark for buried and light for accessible. (b) View of selected residues and water molecules in the vicinity of the binding
site. The Arg144 guanidino group (red) in the galectin-3C-LacNAc complex (c) adopts a different conformation to form a cation-Π interaction in the
galectin-3C-3 complex (d). The extended hydrophobic pocket formed by Arg144 Câ to Cδ is indicated by a blue arrow in d.
potential LacNAc-based inhibitors all having aromatic amides
at 3′-C (4-46, Figure 4) following our recently described
procedure.24
substitution is detrimental to the interaction with galectin-3.
Fluorination (4-7) of the aromatic ring results in inhibitors
whose affinities are worse than the parent unsubstituted benz-
amide 2.24 This is probably due to a decrease in the Π-electron
density in the fluorinated benzamides and thus a decrease in
the cation-Π interaction strength. As a consequence, the
p-methoxy group of 3 can be concluded to be essential for high
affinity.
Galectin-3 Inhibition in a Competitive Fluorescence
Polarization Assay. A novel high-throughput galectin-3 solution
phase assay based on fluorescence polarization26,43 was em-
ployed to evaluate compounds 4-46. Indeed, the chemical
nature of the benzamido moiety strongly affects the interaction,
as the inhibitory powers of 4-46 varied significantly. All
compounds 4-46 were better than the unsubstituted LacNAc
methyl glycoside 1 (Kd ) 67 µM), and 35 was better than the
unsubstituted parent benzamido derivative 2 (Kd ) 6.7 µM).
Although several (15, 24, 32, 34, 37; Kd ) 0.95-1.3 µM) were
similar to 3 in affinity for galectin-3, only two compounds (41
and 44; Kd ) 0.48 and 0.32 µM, respectively) were better than
3 (Kd ) 0.88 µM).
If it is assumed that the aromatic moieties of the good
inhibitors stack against the Arg144 guanidino group in a manner
similar to that of 3, conclusions can be drawn concerning the
effects of the arene structure and substitution pattern on the
cation-Π interaction. In general, we see that para- and meta-
substitution of the benzamide is beneficial, while ortho-
Methoxy-substitution at the meta-position is beneficial; the
m-methoxy group of 9 can be expected to form favorable direct
(oxygen) electrostatic interactions with the guanidino group of
Arg144, as well as to fill the cleft formed by Arg144 Câ to
Cδ. Presumably p-methoxy (10) is somewhat less efficient than
m-methoxy (9) at making favorable direct interactions due to
the increased distance from the guanidino group and the cleft
formed by Arg144 Câ to Cδ. o-Methoxy-substitution (8) gives
inhibitory power worse than no substitution (2). The observa-
tion that two o-methoxy substituents (12) are detrimental to
galectin-3 binding is not unexpected, since, upon stacking of
the Arg144 guanidino group onto the aromatic ring, one or the
other of the two methoxy groups would interfere with the protein
surface. The observation that the introduction of two m-methoxy
groups (15) leads to an activity surpassing that of 3 seemed
surprising at first, because if 15 forms a complex with galectin-3
(43) So¨rme, P.; Kahl-Knutsson, B.; Huflejt, M.; Nilsson, U. J.; Leffler, H. Anal.
Biochem. 2004, 334, 36-47.
9
1740 J. AM. CHEM. SOC. VOL. 127, NO. 6, 2005

Cation−Π Interactions in Lectin−Ligand Complexes
A R T I C L E S
Figure 4. Synthetic LacNAc derivatives (4-59) carrying aromatic amides at 3′-C. Kd values (µM) as determined by a competitive fluorescence polarization
assay were determined at 8 µM inhibitor concentration and are shown in brackets. *Kd for 44 had to be determined at 1.6 µM concentration of 44 because
of its high affinity, and the Kd for 4, 12, and 55 had to be determined at 40 µM concentration because of their low affinities for galectin-3.
structurally similar to that of 3, then one of the methoxy groups
would point more or less into bulk solution. However, the
presence of two symmetrically placed methoxy groups would,
of course, be entropically favorable, because two degenerate
conformers of 15 would interact with galectin-3 with the same
enthalpic contribution. Compound 17 also carries two m-
methoxy groups, in addition to a p-methoxy group, but it is
significantly less active than 15. However, in 1,2,3-trisubstituted
benzene rings, the methoxy groups are typically arranged
alternatively up and down relative to the benzene ring, a
geometry that would result in a steric clash in a galectin-3-17
complex structurally similar to that of galectin-3-3. This
hypothesis is further supported by the fact that 21 is a slightly
better inhibitor than 14. In 21 the methylene bridge between
the m- and p-oxygens can adopt an orientation similar to the
methyl group of 9.
Taken together, simply introducing an electron-donating
methoxy functionality on an aromatic partner in an arene-
guanidino interaction does not automatically lead to enhanced
interaction. The direct electrostatic and steric interactions of the
introduced substituent with the protein surface can overshadow
electron-donating effects and have to be taken into consideration,
as do the effects of the substituent on ligand conformational
preferences. In the small arginine receptors studied by Schrader,
for example, it has been suggested that the direct electrostatic
interactions between the aromatic methoxy group and the argi-
nine guanidino group were more important than the cation-Π
interactions in stabilizing the complexes.39 Their hypothesis was
supported by electrostatic potential calculations, which gave
similar results with benzene and anisole.
The electron-donating dimethylamino-substituent in the para-
position (26) only marginally improves the inhibitor affinity.
9
J. AM. CHEM. SOC. VOL. 127, NO. 6, 2005 1741

A R T I C L E S
So¨rme et al.
Table 1. Thermodynamic Parameters for the Interactions between
Galectin-3C and Selected Inhibitors
This result may be explained by the fact that, compared to the
p-methoxy compound (10) discussed above, it is sterically less
complementary to the hydrophobic cleft formed by Arg144 Câ
to Cδ. m-Dimethylamino substition as in compound 25 does
not affect the inhibitory power, presumably because such a
substituent would have to be positioned away from the Arg144
side chain for steric reasons. Furthermore, the higher degree of
protonation of the dimethylamino substituent at near-neutral pH,
as compared to the methoxy, negatively influences both a direct
electrostatic interaction with the guanidino group and the indirect
interaction via electron donation to the benzene ring. Unpro-
tonated dimethylamino-substitution has been shown to enhance
cation-Π interactions in an artificial model system in chloro-
form.42
1
15
24
32
∆G (kcal/mol)
∆H (kcal/mol)
T∆S ((kcal/mol)/K) -4.0
-5.6
-7.2
-7.2
-7.2
-9.6 ( 0.6 -11.2 ( 0.1 -11.7 ( 0.1 -12.0 ( 0.1
-4.0
-4.5
-4.8
Kd (µM) 84.0 ( 10.0 5.0 ( 0.3
4.8 ( 0.3
5.7 ( 0.4
interactions have been described in the literature,23,44-49 and the
present study is the first to analyze small synthetic inhibitors
with galectin-3C. Notably, the best monovalent ligand described,
to date, is the blood group A tetrasaccharide with Kd ) 72 µM
(as compared to 191 µM for LacNAc).47 The isothermal titration
microcalorimetry experiments were performed with galectin-
3C to avoid the possibility that the N-terminal domain of intact
galectin-3 might mediate higher order oligomers.22,50 In addition,
galectin-3C is considerably more soluble than intact galectin-
3. The methyl glycoside of N-acetyllactosamine 1 was included
in the experiments as a reference ligand; it was found to have
Kd ) 82 µM, a value lower than that obtained with reducing
N-acetyllactosamine and intact galectin-3 47 (Kd ) 191 µM).
It should also be noted that the Kd values obtained for the
various compounds by ITC are approximately 5-fold higher
than those obtained by fluorescence polarization (the relative
affinities are essentially identical). Although the basis for this
difference is not known, the fluorescence polarization experi-
ments were performed on intact galectin-3 where effects arising
from the N-terminal domainssuch as oligomerizationsmay be
important.
The optimized inhibitors 15, 24, and 32 exhibited Kd values
of 5.0, 4.8, and 5.7 µM, respectively, in the ITC experiments,
an approximate 15-fold increase in affinity compared to the
parent methyl glycoside of N-acetyllactosamine 1 (84 µM).
These increases correspond to a change in binding free energy
(∆∆G) of approximately 1.7 kcal/mol, a value that agrees well
with that observed for cation-Π interactions in other biological
systems.35,37 As shown in Table 1, these changes are a direct
reflection of changes in the enthalpy of binding (∆∆H between
1.6 and 2.4 kcal/mol). In two cases, the favorable enthalpy
changes are offset somewhat by unfavorable changes in
entropy51-53 relative to compound 1 (i.e. 24 and 32 with ∆∆S
) 0.5 and 0.8 kcal/mol, respectively). Interestingly, the m,m-
dimethoxy-substituted benzamido derivative 15 displays no
entropy penalty relative to the parent LacNAc 1.
Among the other substituents investigated, only p-nitro (24),
p-methyl (32), p-tert-butyl (34), and p-acetyl (37) give rise to
high-affinity inhibitors. The nitro-substituent of 24 extends the
Π-system of the benzamido group closer to the Arg144 residue,
thus leading to an improved interaction with the Arg144
guanidino group. However, the possibility that the nitro group
forms direct bidentate interactions with the Arg144 guanidino
group in a structurally different fashion cannot be excluded.
Surprisingly, p-carboxy substitution (39) results in a rather small
affinity enhancement, although it can be regarded as an isostere
of 24.
Among the larger aromatic systems, 2-naphthamides proved
to be exceptionally good at enhancing the affinity for galectin-
3. The 2-naphthamido derivatives 41 and 44 are well-suited to
forming a cation-Π interaction with the Arg144 guanidino
group and to fill the hydrophobic cleft lined by Arg144 Câ-
Cδ. Large and more polarizable Π-systems typically form strong
cation-Π interactions.35,39 The o-carboxy substituent of 44 has
only a small affinity-enhancing effect, but it may be useful as
a means of increasing the solubility of the inhibitors. The
heteroaromatic amides, exemplified by the indol 29 and pyrroles
27 and 28, are surprisingly poor even though indol (i.e. Trp) is
common in protein cation-Π interactions and pyrrole is an
electron-rich Π-system. This suggests that the benzamide-
Arg144 interaction is indeed specific and very sensitive to the
exact positioning of the Π-system, a conclusion supported by
the observation that phenylacetamide 11, 1-naphthamide 42, and
naphthylacetamide 43 are less efficient inhibitors.
The Kd of 320 nM for 44 is more than 2 orders of magnitude
better than that of the parent LacNAc 1 (Kd ) 67 µM) and
represents an unusually high affinity for a monovalent ligand-
lectin interaction. The excellent inhibitory power of the
dimethoxy-substituted 15 suggested the synthesis of even further
optimized inhibitors by means of replacing the methoxy with
other alkoxy groups. The synthesis of 13 derivatives (47-59)
did indeed show that 3,5-dialkoxy-substitution of the benzamide
was beneficial for complex formation as 11 of them were better
inhibitors than the unsubstituted reference compound 2. How-
ever, none of them was better than the 3,5-methoxy-substituted
benzamide 15.
Conclusions
Analysis of high-resolution crystallographic data, together
with synthesis and evaluation of a large panel of benzamido
derivatives and ITC measurements, provides important informa-
tion about inhibitor-galectin-3 interactions, which is essential
(44) Gupta, D.; Moonjae, C.; Cummings, R. D.; Brewer, C. F. Biochemistry
1996, 35, 15236-15243.
(45) Schwarz, F. P.; Ahmed, H.; Bianchet, M. A.; Amzel, L. M.; Vasta, G. R.
Biochemistry 1998, 37, 5867-5877.
(46) Bharadwaj, S.; Kaltner, H.; Korchagina, E. Y.; Bovin, N. V.; Gabius, H.-
J.; Surolia, A. Biochim. Biophys. Acta 1999, 1472, 191-196.
(47) Bachhawat-Sikder, K.; Thomas, C. J.; Surolia, A. FEBS Lett. 2001, 500,
75-79.
Thermodynamic Studies on Galectin-3: Inhibitor Com-
plexes. To further characterize the underlying basis for the
affinity-enhancing effect of the 3′-benzamido-substitution of
LacNAc, three of the best inhibitors (15, 24, and 32), with
suitably high water-solubility, were selected for ITC experi-
ments. Only a few thermodynamic studies on galectin-ligand
(48) Ahmad, N.; Gabius, H.-J.; Kaltner, H.; Andre´, S.; Kuwabara, I.; Liu, F.-T.
Can. J. Chem. 2002, 80, 1096-1104.
(49) Brewer, C. F. Glycoconjugate J. 2004, 19, 459-465.
(50) Brewer, C. F. Biochim. Biophys. Acta 2002, 1572, 255-262.
(51) Toone, E. J. Curr. Opin. Struct. Biol. 1994, 4, 719-728.
(52) Lemieux, R. U. Acc. Chem. Res. 1996, 29, 373-380.
(53) Gohlke, H.; Klebe, G. Angew. Chem., Int. Ed. 2002, 41, 2644-2676.
9
1742 J. AM. CHEM. SOC. VOL. 127, NO. 6, 2005

Cation−Π Interactions in Lectin−Ligand Complexes
A R T I C L E S
for the development of inhibitors with higher affinity and
specificity. In particular, an Arg144-arene stacking that
involves a cation-Π interaction appears to be of importance
for high affinity, which suggests that exploitation and fine-tuning
of cation-Π interactions may very well emerge as a more
general strategy toward the enhancement of protein-ligand
interactions. However, the present work also reenforces the fact
that direct interactions of the arene substituents with the protein
surface are also very important in the inhibitor design process.
In addition, the synthesis of potent specific monovalent inhibi-
tors of galectin-3, as described herein, provides valuable research
tools as well as an advance toward the goal of developing
galectin-based therapeutic drugs.
(France), as well as grants from the program “Glycoconjugates
in Biological Systems” sponsored by the Swedish Strategic
Research Foundation, the Swedish Research Council, the
Swedish Medical Research Council (Grant No. 12165), and the
Medical Research Council (Canada)/Canadian Institutes of
Health Research.
Supporting Information Available: Data collection and
refinement statistics for galectin-3-ligand complexes, experi-
mental procedures, NMR and HRMS-FAB data for compounds
4-59, synthesis and NMR data for 3,5-dialkoxybenzoic acids,
and 1H NMR spectra of 4-59 (pdf). This material is available
free of charge via the Internet at http://pubs.acs.org.
Acknowledgment. This project was partly supported by a
grant from the Association pour la Recherche sur le Cancer
JA043475P
9
J. AM. CHEM. SOC. VOL. 127, NO. 6, 2005 1743