A simple model system for the study of carbohydrate-aromatic interactions

Article abstract of DOI:10.1021/ja066633g

A molecular scaffold was identified which enables the establishment of intramolecular interactions between a monosaccharide and a nearby phenyl ring. A group of molecules containing four different monosaccharides (glucose, galactose, N-acetyl-glucosamine,

Full text of DOI:10.1021/ja066633g

Published on Web 02/20/2007
A Simple Model System for the Study of
Carbohydrate-Aromatic Interactions
Giancarlo Terraneo,† Donatella Potenza,† Angeles Canales,‡
Jesus Jime´nez-Barbero,*,‡ Kim K. Baldridge,*,§ and Anna Bernardi*,†
Contribution from the UniVersita’ di Milano, Dipartimento di Chimica Organica e Industriale e
Centro di Eccellenza CISI, Via Venezian 21, 20133 Milano, Italy, Centro de InVestigaciones
Biolo´gicas, Consejo Superior de InVestigaciones Cient´ıficas, Ramiro de Maeztu 9,
28040 Madrid, Spain, and Institute of Organic Chemistry, UniVersity of Zu¨rich,
Winterthurerstrasse 190, CH-8057 Zu¨rich, Switzerland
Received September 14, 2006; E-mail: anna.bernardi@unimi.it; jjbarbero@cib.csic.es; kimb@oci.unizh.ch
Abstract: A molecular scaffold was identified which enables the establishment of intramolecular interactions
between a monosaccharide and a nearby phenyl ring. A group of molecules containing four different
monosaccharides (glucose, galactose, N-acetyl-glucosamine, and N-acetyl-galactosamine) was synthesized
and used to investigate the extent and nature of this carbohydrate-arene interaction, as well as the effect
on the overall 3D structure of the molecules involved. The sugar-aromatic distance was evaluated by
rigorous NMR studies supported by molecular modeling and found to be constant throughout the series,
independent of the nature of the sugar and of the conformational behavior of the fragment connecting the
two elements. Ab initio calculations at the B3LYP/DZV(2d,p) level of theory enable the analysis of the
electronic nature of the interaction. The study shows that, given the opportunity, persistent intramolecular
aromatic-sugar interactions can be established and can significantly influence overall molecular shape
and energetics. These results have important implications in the design of structural mimics of oligosac-
charides.
is the driving force for intermolecular interactions.5 NMR studies
have shown the presence of interactions between hydrogens
belonging to the apolar faces of sugar rings and aromatic
residues of protein side chains6 and have linked the strength of
the interaction to the size and the electron richness of the arene.7
In addition, studies concerning the role of N-linked oligosac-
charides during protein folding have revealed8 that the incidence
of aromatic amino acids in proximity to glycans during the
folding process is higher than the standard levels occurring on
the surface or within the protein core. Together, these data
provide significant implications for molecular recognition of
carbohydrates in water solution, and indeed, artificial carbohy-
drate receptors that exploit sugar/aromatic interactions have been
described.9 Carbohydrate-aromatic complexes have also been
investigated using computational methods, suggesting the
Introduction.
Sugar-protein interactions play an important role in a wide
range of biological processes, from regulatory processes such
as fertilization to pathologies such as tumor spread. These
interactions can be responsible for mediating diverse cellular
activities, such as cell recognition, growth, and apoptosis.1,2
A
large variety of proteins, with very different functions and
topologies, are involved in carbohydrate recognition, including
enzymes, periplasmic receptors, antibodies, and lectins.3 Because
of the amphiphilic character of oligosaccharides, a variety of
forces mediate the recognition process, with many types of
interactions revealed in the available X-ray structures of sugar-
protein complexes. Polar groups of the sugars are bound by
H-bond donors and acceptors in the protein backbone and polar
side chains. Apolar regions (formed primarily by the pyranose
CH groups) are complemented by nonpolar surfaces of the
proteins and, in particular, tend to pack against aromatic residues
in the receptor side chains.4 Convincing arguments have been
proposed that suggest that desolvation of such apolar patches
(5) Lemieux, R. U. Acc. Chem. Res. 1996, 29, 373-380.
(6) (a) Jime´nez-Barbero, J.; Peters, T. NMR Spectroscopy of Glycoconjugates;
Wiley-VCH: Weinheim, Germany, 2003. (b) Wormald, M. R.; Petrescu,
A. J.; Pao, Y. L.; Glithero, A.; Elliott, T.; Dwek, R. A. Chem. ReV. 2002,
102, 371-386. (c) Jime´nez-Barbero, J.; Asensio, J. L.; Can˜ada, F. J.;
Poveda, A. Curr. Opin. Struct. Biol. 1999, 9, 549-555.
(7) Chavez, M. I.; Andreu, C.; Vidal, P.; Aboitiz, N.; Freire, F.; Groves, P.;
Asensio, J. L.; Asensio, G.; Muraki, M.; Can˜ada, F. J.; Jime´nez-Barbero,
J. Chem.sEur. J. 2005, 11, 7060-7074.
(8) Petrescu, A. J.; Milac, A. L.; Petrescu, S. M.; Dwek, R. A.; Wormald,
M. R. Glycobiology 2004, 14, 103-114.
(9) (a) Davis, A. P.; Wareham, R. S. Angew. Chem., Int. Ed. 1999, 38, 2978-
2996 and references therein. (b) Mazik, M.; Cavga, H.; Jones, P. G. J. Am.
Chem. Soc. 2005, 127, 9045-9052. (c) Klein, E.; Crump, M. P.; Davis,
A. P. Angew. Chem., Int. Ed. 2005, 44, 298-302. (d) Vacca, A.; Nativi,
C.; Cacciarini, M.; Pergoli, R.; Roelens, S. J. Am. Chem. Soc. 2004, 126,
16456-16465.
† Universita’ di Milano.
‡ Consejo Superior de Investigaciones Cient´ıficas.
§ University of Zu¨rich.
(1) Lis, H.; Sharon, N. Chem. ReV. 1998, 98, 637-674.
(2) Simanek, E. E.; McGarvey, G. J.; Jablonowski, J. A.; Wong, C. H. Chem.
ReV. 1998, 98, 833-862.
(3) (a) Wies, W. L.; Drickamer, K. Ann. ReV. 1996, 65, 441-473. (b) Vyas,
N. K. Curr. Opin. Struct. Biol. 1991, 1, 732-740.
(4) (a) Quiocho, F. A.; Biochem. Soc. Trans. 1993, 21, 442-448. (b) Elgavish,
S.; Shaanan, B. J. Mol. Biol. 1998, 277, 917-932.
9
2890
J. AM. CHEM. SOC. 2007, 129, 2890-2900
10.1021/ja066633g CCC: $37.00 © 2007 American Chemical Society

Study of Carbohydrate−Aromatic Interactions
A R T I C L E S
Figure 2. Compounds used in this study and relative numbering convention.
compound 2 (Figure 2) was chosen as a suitable model to
investigate intramolecular interactions between various monosac-
charide fragments (2c-f, R ) monosaccharide) and the phenyl
ring appended to the ether side chain.
A series of molecules 2a-f with different substituents at the
axial hydroxy group of the dicarboxy cyclohexanediol15 (R )
H, MOM, Glcâ1-, Galâ1-, GlcNAcâ1-, GalNAcâ1-), were
synthesized. Compounds 2a-f can be exploited as models for
deducing chemical features and structural requirements of the
C-H vector responsible for optimal interaction with the
aromatic rings. An NMR-based approach was used to study
these features. The chemical nature of the compounds enabled
the sugar/aromatic interaction to be studied in D2O by analyzing
the nuclear Overhauser effect (NOE) observed between diag-
nostic protons belonging to the two fragments. Moreover, the
mobility of the ether chain could be studied within the series
by NOE-based methods. Spectral variations taking place in the
different glycan fragments were analyzed.
Figure 1. Aromatic-sugar interactions stabilize the bioactive conformation
of glycomimetic 1a (see ref 14). (a) Structure of 1. (b) Preferred
conformation of 1a in water. The arrows show schematic representations
of the closed contact between aromatic and sugar residues as inferred from
NOE measurements, which were observed in the NMR spectrum of 1a in
D2O.
interaction to be primarily intermolecular van der Waals contacts
and/or CH-π interaction.10,11,12
Carbohydrate-arene interactions have also been shown to
direct conformational equilibrium of glycophanes13 and oli-
gosaccharide mimics14 in water solution. We have recently
suggested that aromatic/sugar interactions may find application
in the design of glycomimetics, since intramolecular sugar/
aromatic interactions can help to stabilize bioactive conforma-
tions of small-molecule oligosaccharide mimics. In fact in the
course of our studies directed toward the synthesis of glyco-
mimetic ligands of the cholera toxin (CT),14 we found clear
NMR evidence of a close spatial proximity between the phenyl
ring and the N-acetylgalactosamine residue in compound 1a
(Figure 1), which binds to CT with micromolar affinity.
Comparison with a series of analogues bearing different alkyl
groups on the ether side chain suggested that such an interaction
biases the conformational behavior of 1a by restricting the
conformational freedom of the chain. As a result, the side chain
of 1a is “preorganized” in a suitable spatial orientation that
enables optimal interaction of the binding determinants to the
binding region of CT (Figure 1).14
Results and Discussion.
Synthesis. The synthesis of compounds 2 followed procedures
that have been previously described for the synthesis of 1a.14
Starting from diol 3,15 the monoether 5 was obtained by
regioselective alkylation with the triflate 414 (45% yield; Scheme
1). The alcohol 2a was obtained from 5a by deprotection of
the benzyl ester (H2/Pd). The MOM derivative 2b was obtained
by protection of 5a (MOMCl/TBAI/DIPEA in CH2Cl2), fol-
lowed by removal of the benzyl ester (Scheme 1). Compounds
2c-f were synthesized following a common glycosylation
procedure using the sugar trichloroacetamidate as the donor and
the monoether 5b as the acceptor. The glycosylation reaction
was carried out at -30 °C and catalyzed by trimethylsilyl triflate
for glucose and galactose (2c and 2d). For the two 2-acetamido
sugars (GlcNAc and GalNAc in 2e and 2f, respectively), the
glycosylation reaction was catalyzed by triflic acid and stirred
at room temperature for 3 h before refluxing overnight in
CH2Cl2. For the entire series, the benzyl ester was removed
using H2/Pd and the acetyl groups were removed under standard
Zemplen’s condition (MeONa/MeOH). Experimental details are
given as Supporting Information.
Sugar-aromatic interactions may have a broader application
as a key element in conformational-based design of sugar
mimics and therefore warrant further investigation. To this end,
NMR Studies of Compounds 2a-f in D2O. NMR spectra
of compounds 2a-f were recorded in D2O at 400 MHz and
298-300 K. Chemical shifts and coupling constants are reported
in the Supporting Information. NOESY experiments were
carried out using a delay of 800 ms that was selected by applying
(10) Fernandez-Alonso, M.; Can˜ada, F. J.; Jime´nez-Barbero, J.; Cuevas, G.
J. Am. Chem. Soc. 2005, 127, 7379-7386.
(11) (a) Sujatha, M. S.; Sasidhar, Y. U.; Balaji, P. V. Biochemistry 2005, 44,
8554-8562. (b) Sujatha, M. S.; Sasidhar, Y. U.; Balaji, P. V. Protein Sci.
2004, 13, 2502-2514.
(12) Spiwok, V.; Lipovova, P.; Skalova, T.; Vondrackova, E.; Dohnalek, J.;
Hasek, J.; Kralova, B. J. Comput.-Aided Mol. Des. 2005, 19, 887-901.
(13) Morales, J. C.; Penade´s, S. Angew. Chem., Int. Ed. 1998, 37, 654-657.
(14) Bernardi, A.; Arosio, A.; Potenza, D.; Sanchez-Medina, I.; Mari, S.; Can˜ada,
F. J.; Jime´nez-Barbero, J. Chem.sEur. J. 2004, 10, 4395-4406.
(15) Bernardi, A.; Arosio, D.; Manzoni, L.; Micheli, F.; Pasquarello, S.; Seneci,
P. J. Org. Chem. 2001, 66, 6209-6216.
9
J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007 2891

A R T I C L E S
Terraneo et al.
Figure 3. Idealized staggered conformations describing the orientation of HL relative to the diol moiety (CHD). The improper dihedral angle ø (C(O)-
CR-C4CHD-H4CHD) is used to define the conformations.
Scheme 1. Synthesis of 2a-fa
a (i) Bu2SnO reflux, then 4 at rt, 5a: 33%, 5b: 45%; (ii) H2/Pd, 1 atm, rt, 2a: 95%; (iii) 1. MOMCl/TBAI/DIPEA in CH2Cl2 at rt, 2. (ii) 2b: 60% over
two steps; (iv) 1. DCM/cat; 2. (ii); 3. MeONa in MeOH at rt, 2c: 61%, 2d: 69%, 2e: 64%, 2f: 70% over three steps.
the inversion recovery experiments (t1ir, Bruker library) from
50 ms to 3 s. The 800 ms delay was found to be the most
appropriate mixing time to evaluate aromatic/sugar interaction
in molecules 2c-f. NOE contact tables are collected in the
Supporting Information (Tables SI-1 to SI-6). All compounds
gave satisfactory NOESY spectra.
proton in position 5 of the diol (Figure 2).18 Presence of the
alternative anti conformation would give rise to exclusive G1/
H6axCHD NOE contact, which is never observed in any of the
molecules 2c-f.
Additional degrees of freedom available to compounds 2a-f
correspond to the hydroxy acid side-chain bonds and to the
relative orientation between the monosaccharides and the
aromatic ring in compounds 2c-f. 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 glycosidic
linkage. The NOE contacts observed in this region can arise
from multiple æ,ψ combinations that provide similar through-
space proton-proton interactions for the reporter HL (Figure
2). Thus, the experimental interproton distances obtained by
NMR spectroscopy can be analyzed more conveniently using
the improper dihedral angle descriptor, ø(C(O)-CR-CHDC4-
CHDH4), defined using the Newman-type projections shown
in Figure 3. This descriptor defines the orientation of HL (and
of the hydroxy acid carboxy group) relative to the CHD ring.
The relation between NOE signals and proton-proton
distances is well established16 and can be worked out in a
semiquantitative manner using an interproton relaxation matrix.
The NOE intensities reflect the conformer population, and
therefore information concerning the population distribution in
solution can be obtained by focusing on the diagnostic, mutually
exclusive, NOE interactions that characterize the different
possible conformations.17 A comparative analysis of compounds
2c-f reveals a series of common features throughout the series.
For all compounds, analysis of vicinal proton-proton coupling
constants for the sugar fragment indicates that all the pyranoses
are in the usual 4C1 conformation. The cyclohexanediol moiety
(CHD) also adopts a chair conformation, with the ester groups
in equatorial position as shown by the exclusive NOE contact
between the H2 and H4 protons on the CHD ring (Figure 2).
The orientation around the sugar-diol linkage corresponds to
the syn:syn global minimum conformation for all the monosac-
charides examined, as defined by the G1/H5CHD NOE contact
observed between the anomeric protons of the sugars and the
In principle, although rotations around the O-C bond may
also produce quasi-eclipsed orientations, three idealized stag-
gered orientations are possible involving the ether connector,
identified in Figure 3 as Conf. A (ø ) 180°), Conf. B (ø )
+60°), and Conf. C (ø ) -60°), each associated with specific
(18) (a) Bernardi, A.; Potenza, D.; Capelli, A. M.; Garc`ıa-Herrero, A.; Can˜ada,
F. J.; Jime´nez-Barbero, J. Chem.sEur. J. 2002, 8, 4597-4612. (b) Bernardi,
A.; Arosio, D.; Manzoni, L.; Monti, D.; Posteri, H.; Potenza, D.; Mari, S.;
Jime´nez-Barbero, J. Org. Biomol. Chem. 2003, 1, 785-792.
(16) Hwang, T. L.; Shaka, A. J. J. Am. Chem. Soc. 1992, 114, 3157-3158.
(17) Dabrowski, J.; Kozar, T.; Grosskurth, H.; Nifant’ev, N. E. J. Am. Chem.
Soc. 1995, 117, 5534-5539.
9
2892 J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007

Study of Carbohydrate−Aromatic Interactions
A R T I C L E S
Figure 4. NOESY spectra of 2a-f in D2O: NOE contacts of HL.
Table 1. Diagnostic NOESY Cross-peaks for the HL Proton and
A quantitative evaluation of the NMR results was attempted
with the assistance of molecular modeling techniques. Both
molecular mechanics (MC/EM19) and dynamics (MC/SD20)
calculations were used. The determination of the conformational
distribution involving the ether chain of each compound,
however, involves small differences in the conformational
behavior of a flexible fragment, which is quite difficult to
describe well by computational methods. Our previous experi-
ence with similar compounds18a indicated that molecular
mechanics and dynamics calculations can provide a qualitative
reproduction of the side-chain behavior but cannot quantitatively
reproduce the experimental data. In an effort to obtain the most
accurate description of the conformational distribution of this
elusive fragment, conformer populations were generated using
three different methods and subsequently evaluated by compar-
ing predicted NOE intensities with experimentally obtained
values.21 Three sets of populations were obtained, as described
in detail in the Experimental Section, from molecular dynamics
ensemble averaging (MD set), Boltzmann distributions obtained
from molecular mechanics conformational searches (MM set),
and NOE-fitting procedures (Fitting set). The experimental NOE
intensity was measured as the ratio between the volume of the
cross-peak and the volume of the corresponding diagonal peak.
The expected NOE intensity was calculated for each set of
conformations with the Noeprom22 software, which uses the full
interproton relaxation matrix approach.23 The accuracy of the
calculations was evaluated by measuring the percentage of NOE
intensity |exptl - theor| errors. The set of conformers that
produced the most accurate Noeprom results was then taken as
the best description of the conformational behavior of the side
chain. The results obtained for all compounds across all methods
are reported in the Supporting Information. The optimal
description obtained for each molecule is reported in Table 1.
Conformational Distribution of the Side Chaina
calculated NOE intensityb (%)
(experimental NOE intensity)c
proton pairs
2a
2b
2c
2d
2e
2f
HL-H4
6.4
(6.6)
7.9
(5.6)
1.8
6.5
(6.9)
6.5
(6.7)
0.2
8.2
(7.3)
6.1
(6.2)
0.1
7.7
(6.7)
5.2
(4.8)
0.6
7.7
(2.2)
3.8
(0.4)
1.6
6.5
(6.4)
2.3
(2.0)
1.6
HL-H5
HL-H3eq
(1.6)
(0.0)
(0.0)
(0.6)
(0.2)
(1.8)
A/B/C ratiod 36:64:0 6:94:0 0:100:0 12:88:0 27:73:0 28:72:0
(set)e
(fitting) (MM) (fitting) (fitting) (MM) (fitting)
2.8 0.9 1.7 1.4 10.0 0.5
av error %f
a Spectra recorded in D2O at 400 MHz and 300 K. b Calculated using
Noeprom. c The NOE % is expressed as the ratio between the cross-peak
volume and the volume of the corresponding diagonal peak. d Relative
population of the limit conformations of Figure 3 which allowed the best
reproduction of the NOE cross-peak intensity. e Set of conformations that
produced the lowest error. MD, molecular dynamics; MM, multiminimi-
zation; fitting, populations from NOE data fitting with MM geometries.
f Total error obtained for the three diagnostic contacts of Table 1.
interactions between HL and the protons on the C4, C3, and
C5 carbon atoms of the CHD ring. The primary orientations of
the hydroxy acid relative to the ring can thus be defined by
focusing on the NOE cross-peaks of the HL proton to H4, H5,
H3eq, and H3ax of the cyclohexanediol: conformer B is
associated with the exclusive HL/H5 NOE contact; conformer
A, with the HL/H3eq contact; and conformer C, with the HL/
H3ax cross-peak (Figure 3). Finally, the orientation of the
aromatic ring relative to the pyranose moiety in 2c-f can be
investigated by examining the NOE contacts between the
hydrogens on the R face of the sugar and the aromatic protons.
A first examination of the NOESY spectra shows, for most
compounds, the simultaneous presence of NOE contacts between
the HL/H5 and HL/H3eq proton pairs (Figure 4 and Table 1).
This fact suggests a fair amount of conformational freedom
within the side chain, with a general prevalence of conformer
B. These observations contrast with the spectrum of 1a, which
shows no NOE contacts between HL and H5 and a strong NOE
cross-peak between HL and H3eq.14
(19) Chang, G.; Guida, W. C.; Still, W. C. J. Am. Chem. Soc. 1989, 111, 4379-
4386.
(20) Guarnieri, F.; Still, W. C. J. Comput. Chem. 1994, 15, 1302-1310.
(21) Mari, S.; Can˜ada, J. F.; Jime´nez-Barbero, J., Bernardi, A.; Marcou, G.;
Motto, I.; Velter, I., Nicotra, F.; La Ferla, B. Eur. J. Org. Chem. 2006,
2925-2933.
(22) Martin-Pastor, M. http://desoft03.usc.es/mmartin/software.html
(23) Cumming, A. C.; Carver, J. P. Biochemistry 1987, 26, 6664-6676.
9
J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007 2893

A R T I C L E S
Terraneo et al.
Table 2. Correlation Time (τc) and Interproton Distances (r)
Obtained by 1D NOE and 1D ROE Experiments for the Pairs
HL/H5, HL/H4, and HL/H3eq
2c
2d
2e
2f
proton pair
τca (ps)
rb (Å)
τ
c (ps)
r (Å)
τc (ps)
r (Å)
τ
c (ps)
r (Å)
HL/H4
HL/H5
HL/H3eq
210
213
187
2.63
2.87
3.53
164
221
162
2.61
2.74
3.53
231
233
296
2.71
3.18
3.07
168
168
207
2.64
3.08
2.98
average τc
203
182
253
181
a Correlation time in ps. b Distance in Å.
Table 3. Conformational Distribution of the Side Chain
Conformers Estimated from 1D NOE and 1D ROE Experiments
ratio
compounds
Conf. A/Conf. B
2c
2d
2e
2f
18:82
22:78
36:64
32:68
The errors obtained, defined as the difference between the
experimental and calculated NOE cross-peak intensities, are
consistently lower than 3% for compounds 2b, 2c, 2d, and 2f
and 10% for compound 2e, for which the 2D-NOESY spectrum
is noisier than the others.
In order to further address the description of the ether chain
behavior in the sugar series (2c-2f), we also performed 1D
NOE and 1D ROE experiments in D2O. The buildup curves
for the NOE and ROE effect were measured at different mixing
times, and effective correlation times (τc) were estimated for
each of the proton couples involved (see Experimental Section).
It is known24 that the ratio between the longitudinal and
transversal relaxation rates for a given proton pair and the related
NOE gives a mathematical function that is independent of the
distance and only depends on the effective correlation time (τc)
for that particular proton pair. Using this function, it is possible
to estimate the spectral density J(nω0) from which the inter-
proton distances can be rigorously derived (see Methods
section). This approach allows one to account for the existence
of distinct motions in different parts of the molecule and to
estimate the correlated distances for any given proton pair
without using an internal reference.
Figure 5. Low energy conformations of 2d showing the two main side-
chain conformations A (ø ) 180°) and B (ø ) +60°). The interaction
between the phenyl ring and the sugar is similar in both conformers.
40:60 (2e and 2f) depending on the sugar (Figure 5). Stabiliza-
tion of the A cluster (ø ) 180°) in the GlcNHAc- and GalNHAc-
substituted compounds 2e and 2f is likely related to the
establishment of an electrostatic interaction between the hex-
osamine NH group and the carboxy group on the ether side
chain, which in the A conformation is directed toward the sugar
ring.
The distance and orientation of the aromatic ring relative to
the monosaccharides can also be directly determined by
examining the NOE contacts between the two fragments. The
cross-peaks observed in D2O are shown in Figure 6, and their
relative intensities are reported in Table 4. The reported NOE
intensities are measured relative to the intensity of the cross-
peak between the geminal protons of the benzyl group, which
is assigned a 100% value. The protons of the phenyl ring appear
as a broad multiplet, and in the aromatic line for all compounds,
many NOE contacts are detected. Cross-peaks between the
aromatic ring, the side chain protons (HL), and the cyclohexane
ring protons (H5) are clearly observed for all compounds
examined. It is worth noting that the HAr/H5 cross-peak is
consistently more intense in the spectra of 2e and 2f, containing
GlcNAc and GalNAc, respectively. Various contacts are also
detected with the protons belonging to the R face of the sugars.
For all compounds 2c-f, HAr/G1, HAr/G3, and HAr/G5 cross-
peaks are present, all with the same intensity (ca. 10%) and
independent of the nature of the sugar. In addition, the NOESY
spectra of 2d and 2f, which are substituted with galactose and
galactosamine moieties, respectively, also display an NOE
contact between the aromatic protons and the R-proton on the
sugar 4 position (HAr/G4).
According to this analysis, all average effective correlation
times for the vectors that interconnect proton pairs of the ether
side chain with the CHD moiety are in the region 217 ( 32 ps,
which indicates that the analysis shown in Table 1 can be safely
used to address the conformational distribution. No major
differences exist among the four studied derivatives. Using the
average interproton distances for the MM clusters A and B and
the experimental interproton distances of Table 2, the confor-
mational distributions shown in Table 3 could be estimated.
In summary, for all compounds and experimental approaches
used, the R-hydroxyacid chain is found to be flexible, with
conformer C (ø ) -60°) represented very little, if at all. The
majority of the conformations belong to the B cluster (ø )
+60°), and the A/B ratio varies between 20:80 (2c and 2d) and
(24) (a) Poveda, A.; Santamaria, M.; Bernabe, M.; Rivera, A.; Corzo, J.; Jime´nez-
Barbero, J. Carbohydr. Res. 1997, 304, 219-228. (b) Poveda, A.; Asensio,
J. L.; Martin-Pastor, M.; Jime´nez-Barbero, J. J. Biomol. NMR 1997, 10,
29-43. (c) Poveda, A.; Asensio, J. L.; Martin-Pastor, M.; Jime´nez-Barbero,
J. Carbohydr. Res. 1997, 300, 3-10.
9
2894 J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007

Study of Carbohydrate−Aromatic Interactions
A R T I C L E S
Figure 6. NOESY spectra of 2c-f (400 MHz): the line of the aromatic protons (HAr) in D2O.
Table 4. Intensity of the NOESY Cross-peaks between the
Aromatic Protons and Those on the Monosaccharide Ring in
Compounds 2c-f (D2O, 400 MHz)a
Table 5. Correlation Time (τc) and Distances (r) between
Aromatic Protons (HAr) and Sugar Protons Obtained by 1D NOE
and 1D ROE Experiments
2c
2d
2e
2f
τca
rb
τca
rb
τca
rb
τca
rb
cross-peak
(ps)
(Å)
(ps)
(Å)
(ps)
(Å)
(ps)
(Å)
G1-HAr
G3-HAr
G5-HAr
124
129
123
3.19
3.62
3.26
ovc
ovc
158
3.13
122
3.05
229
3.59
n.e.d
n.e.d
ovc
ovc
a Correlation time (τc) in ps. b Distance in Å. c Not detected because
signal overlaps. d Not estimated.
estimated. These correlation times are somewhat shorter than
those measured between the side chain and the CHD moiety,
although not to a significant degree. These values provide an
indication that the different parts of these molecules are moving
in solution at different frequencies. Thus, all the structural and
dynamical data seem to point to the aromatic/sugar interaction
as a predominant feature in these model compounds in water
solution, occurring independent of the conformational motions
of the appended ether chain.
a NOE % relative to the intensity of the cross-peak between the geminal
protons of the benzyl group, which is assigned 100% intensity. b G1 and
H5 overlap. c Not detected. d G3, G5, and one of the benzyl proton overlaps.
e G3 and G5 overlap.
Computational Studies of Compound 2c. To further analyze
the nature of the carbohydrate-aromatic interaction, ab initio
calculations were performed. The optimized structure of the
glucose-substituted compound 2c, chosen as the model com-
pound, was determined at the B3LYP/DZV(2d,p) level of theory,
starting from the lowest energy B conformation located by
molecular mechanics and validated by NMR for 2c. The
hydrogen bond network resulting from the hydroxyl groups in
position G2, G3, and G4 was adopted in a counterclockwise
arrangement in order to preserve the “cascade” donor-acceptor
between the groups. Because the â-glucose structure adopts an
all-equatorial hydroxyl position, the hydroxyl in G4 acts as a
donor for the acceptor G3, which in turns acts as a donor to the
hydroxyl in position G2, which then enables further hydrogen
bonding with the ether oxygen. In the glucose moiety the C5-
C6 bond, characterized by the torsion angle ω (O6-C6-C5-
O5), has three possible staggered conformations: gauche-trans
(gt), trans-gauche (tg), and gauche-gauche (gg) (Figure 7). In
the gt and gg conformations, the G6 hydroxyl acts as a donor
to the intraring oxygen; in the tg conformation, the G6 hydroxyl
acts as a donor to the G4 hydroxyl group. To better understand
the implications of this conformational freedom on the sugar-
aromatic interaction, we investigated the gt, tg, and gg conform-
ers of 2c. Their energy differences are shown in Table 6,
together with the distance between the aromatic ring and the
The nature and the intensity of the contacts suggest that the
conformational freedom of the ether chain does not affect the
position of the aromatic ring to a measurable degree. This in
turn suggests that either the differences are too small to be
identified or the phenyl-sugar interaction is insensitive to the
conformational flexibility of the side chain, since the aromatic
ring can be accommodated below the sugar in both the A and
B conformations assumed by the R-hydroxyacid chain. This
latter hypothesis is supported by molecular mechanics calcula-
tions, which reveal a low-energy minimum featuring a stacked
conformation of the two rings in both the A and B clusters.
Structures found for 2d are shown in Figure 5 for illustration.
In both conformation A and B, the aromatic ring is at an average
distance of 4.2-4.4 Å from the sugar (distance between the
rings’ centroids) consistently in all compounds. The interproton
distances have be evaluated quantitatively using the 1D experi-
ments described above (see Methods) and are reported in
Table 5.
The data in Table 5 clearly show that the G1-HAr distance
is indeed similar in all compounds of the series. The G3-HAr
and G5-HAr distances are also similar across all cases where
estimation could be made. The corresponding effective correla-
tion times are 170 ( 50 ps for all cases in which they were
9
J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007 2895

A R T I C L E S
Terraneo et al.
Figure 7. Three staggered (ω-angle) calculated conformations of 2c.
on â-methyl glucoside.27 The distance between the aromatic ring
and the sugar ring does not appear to be affected by the C5-
C6 rotamer.
Single-point MP2/DZV(2d,p)//B3LYP/DZV(2d,p) calcula-
tions provide more accurate estimates of the relative energies
for such interactions and are reported in Table 6. The rotamer
distribution obtained at this level of theory (gt/tg/gg, 46:13:48
at 300 K) is close to the experimental data25,26 and consistent
with similar population of the gg and gt rotamers and low
contribution of the tg conformer.
Table 6. B3LYP/DZV(2d,p) and MP2/DZV(2d,p)//B3LYP/
DZV(2d,p) Calculated Relative Energies for the Three Staggered
Conformations (ω-Angle) of 2c
∆
E c
∆
E a
distance
MP2/DZV(2d,p)//
B3LYP/DZV(2d,p)
ω-angle
B3YLP/DZV(2d,p)
CGlc
−
Carb (Å)
gt
tg
gg
0.0 (0.0)
2.60 (0.62)
1.13 (0.21)
4.43
4.42
4.42
0.12 (0.03)
3.24 (0.77)
0.0 (0.0)
a kJ/mol; values in parentheses are kcal/mol at B3LYP/DZV(2d,p).
b Distance between the glucose centroid and centroid of aromatic ring at
optimized geometry at B3YLP/DZV(2d,p). c kJ/mol; values in parentheses
are kcal/mol at MP2/DZV(2d,p)//B3LYP/DZV(2d,p) and correspond to a
46:13:48 population distribution at 300 K.
An estimate of the interaction energy (Eint) between the Glc-
fragment and the CHD-fragment in compound 2c was ap-
proximated from the calculated values as
Eint ) Emol - Emon
Here, Emol is the calculated energy of compound 2c, and Emon
is the calculated energy of the Glc-fragment (EGlc) with the
CHD-fragment (ECHD). Full geometry optimizations were
performed for all components, and the proper isodesmic
relationship was established. To achieve this relationship, the
two monomers were obtained by fragmenting at the glycosylic
bond (C1Glc-O-C5CDH), followed by saturating the remaining
structures with hydrogens, as appropriate (Figure 8). For glucose,
we have kept the terminal bond as R-OH, and for the CHD-
fragment, the fragmentation is fixed as CHD-OH. The isodes-
mic relatonship is balanced with one molecule of water (Ewater).
The resulting isodesmic relationship for 2c together with the
two monomers (C21H31O10) provides the energy of interaction
as
Eint ) [(Emol + Ewater) - (EGlc + ECHD)]
Figure 8. Molecular components exemplified as used for calculation of
interaction energies, Eint. Elements in red indicate placement of bonds that
are made and broken.
The components of Eint have been fully optimized at the
B3LYP/DZV(2d,p) level of theory for each conformation of
2c. All the energies and geometrical parameters are available
in the Supporting Information.
The calculated interaction energies (Eint) for each conforma-
tional variant of 2c are shown in Table 7. The three interaction
energies are very similar, all around 5 kcal/mol, favoring the
complex Emon. These results would suggest that loss of the
glucose moiety from a position above the aromatic ring involves
glucose moiety. The calculated energy differences for the three
conformers of 2c (gt, tg, gg, see Table 6) are in qualitative
agreement with the experimental rotamer distribution in the solid
state,25 in solution,26 and with ab initio calculations performed
(25) (a) Jeffrey, G. A.; McMullan, R. K.; Takagi, S. Acta Crystallogr., Sect. B
1977, 33, 728-737. (b) Marchessault, R. H.; Perez, S. Biopolymers 1979,
18, 2369-2374.
(26) Ohrui, H.; Nishida, Y.; Watanabe, M.; Hori, H.; Meguro, H. Tetrahedron
Lett. 1985, 26, 3251-3254.
(27) Kirschner K. K.; Woods R. J. Proc. Natl. Acad. Sci. U.S.A. 2001, 98,
10521-10545.
9
2896 J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007

Study of Carbohydrate−Aromatic Interactions
A R T I C L E S
Figure 9. (a) CPK surface. (b) Calculated molecular orbital, HOMO-6, displayed with isosurface ( 0.01 contours.
Figure 10. 2c-gt conformation: views a and b. Electrostatic potential mapped on electron density isosurfaces of 2c in the a and b view: c and d (isosurface
0.002 au), e and f (isosurface 0.005 au), and g and h (isosurface 0.007 au), respectively.
Table 7. Interaction Energy (Eint) for 2c (gt, tg, and gg) Using Full
Geometry Optimization at the B3LYP/DZV(2d,p) Level
Table 8. Relative Interaction Energy (Eint) for DFT and MP2
Methods
a
b
c
2c
Eint
2c
Eint
∆Eint
∆Eint
ω-angle
(kcal/mol)
ω-angle
(kcal/mol)
(kcal/mol)
(kcal/mol)
gt
tg
gg
-5.42
-4.72
-5.18
gt
tg
gg
-5.42
-4.72
-5.18
0.0
0.70
0.23
0.0
0.74
0.21
a B3LYP/DZV(2d,p) interaction energy (Eint) at B3LYP/DZV(2d,p)
optimized geometries. b B3LYP/DZV(2d,p) relative interaction energy (Eint)
at B3LYP/DZV(2d,p) optimized geometries. c MP2/DZV(2d,p) relative
interaction energy (Eint) at the B3LYP/DZV(2d,p) optimized geometries.
a decrease in stabilization for the entire system. This behavior
reinforces our notion that sugar-aromatic interactions have an
electrostatic character that provides energy stabilization on the
order of roughly 5 kcal/mol, for these types of systems. Such
interaction magnitudes constitute weak interactions of the wdV
type and are consistent with previously reported values.10-12
As well, one can deduce that 5 kcal/mol roughly accounts for
the sum of the contributions between the hydrogens on the
R-face of the sugar (G1, G3, and G5) and the π-system of the
aromatic ring.
E
int as described above and reported in Table 8. In fact, for this
particular analysis, the MP2 calculations are in agreement with
the B3LYP values, with the same trend of relative energy
interactions predicted.
The analysis of weak interactions is notoriously controver-
sial.28 In particular, verification of interaction and identification
of the stabilizing atom-atom interactions between these two
fragments is quite difficult; therefore we have considered several
possibilities. For these investigations, we have concentrated on
We have also evaluated this interaction energy using more
conventional MP2 methods, as they have been shown to be a
more accurate assessment of weak interactions in many exam-
ples. As such, we report MP2/DZV(2d,p)//B3LYP/DZV(2d,p)
(28) (a) Dunitz, J. D.; Gavezzotti, A. Angew. Chem., Int. Ed. 2005, 44, 1766-
1787. (b) Gatti, C. Z. Kristallogr. 2005, 220, 399-457.
9
J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007 2897

A R T I C L E S
Terraneo et al.
the structure with the lowest complexation energy, Eint, con-
former gt. In Figure 9a, the CPK29 surface of 2c-gt is displayed,
illustrating the G3 hydrogen to be within van der Waals contact
of the aromatic ring and the G1 and G5 hydrogens to be
extremely close to the van der Waals surface of the ring.
A further representation of the sugar/aromatic interaction for
this model is provided by the analysis and visualization of the
molecular orbital structure, displayed in Figure 9b. In fact,
molecular orbital HOMO-6, where HOMO is the highest
occupied molecular orbital, shows a bonding interaction between
the sugar moiety and the aromatic ring, i.e., the G1, G3, and
G5 hydrogens and the aromatic ring. The fact that this interaction
is placed well below the HOMO (EHOMO-6 ) -4.580 eV)
suggests a strong contribution toward the stability of the system.
The intramolecular distances between the aromatic centroid
(CAr) and the sugar protons for the compound 2c are in the
range of NOE contacts (CAr-G1 ) 3.50 Å, CAr-G3 ) 3.03
Å, and CAr-G5 ) 3.87 Å).
Further assessment of the weak complex interaction can be
evaluated using electrostatic potential contour maps. The
electrostatic potential (ESP) of 2c was mapped onto the total
electron density as calculated from the ab initio methods and
various contours evaluated in succession for illustrative pur-
poses. The electron density isosurfaces evaluated begin from
isosurface 0.002 au, representing >98% of the electronic charge,
to isosurface 0.007 au, (i.e., closer to the nucleus), representing
an increase of 4.39 kcal/mol for the electronic charge. All
isosurfaces are mapped in the same range of ESP value ((0.223
au), and the contour colors represent the range including the
most electron-rich areas (red) to the most electron-poor areas
(blue). In this case, since there is a negative charge on the
carboxyl group, the entire molecule shows negative character.
Figure 10a and b show two different views of the complex
involving 2c, with indicated levels of isosurface displayed for
each in Figure 10c-h. Figure 10c and d show the isosurfaces
0.002 au and are analogous to the contour surface view offered
by the CPK method. The relative ESP surface around the G1
and G3 hydrogens clearly indicates contact with the aromatic
ring, suggesting and confirming the intermolecular bonding-
type CH-π interaction, between the aromatic ring and the sugar
moiety. Further illustration of this phenomena is observed by
displaying the other surfaces, Figure 10e and f, at isosurface
0.005 au, encompassing a slightly smaller magnitude of electron
density, and showing the G3/aromatic ring as still intact, again
suggesting strong interaction between the two moieties at this
hydrogen position. Moving yet closer to the nuclei, Figure 10g
and 10h, at isocontour 0.007 au, now show no contact between
any of the hydrogens on the R-face of the sugar with the
aromatic ring. Together, this analysis enables an estimation of
the extent of the weak interaction, placed between 0.002 au and
0.007 au or between 1.2 and 4.4 kcal/mol.
conclusions concerning the nature of the sugar/aromatic type
interaction and the effect on the 3D structure of the molecules
involved.
As revealed by NOE measurements, the phenyl ring and the
monosaccharides appear to adopt a well-defined orientation,
which is consistent across the series examined. NOESY spectra
of 2c did not depend on the solution pH in the range 3.6-7,
and the intensity and position of the observed cross-peaks were
identical for the corresponding methyl ester,30 thus ruling out a
major contribution of the carboxy group to the establishment
of the molecular conformation. The sugar-aromatic distance
was evaluated by rigorous NMR studies supported by molecular
modeling and found to be constant throughout the series and
independent of the nature of the sugar. Since all the aromatic
proton signals strongly overlap in the spectra examined, the
analysis of the NOE contacts may have missed some fine details
of the possible differences in sugar-aromatic orientation.
However, it is worth noting that the above conclusion is
supported both by the molecular modeling results and by DFT
calculations (B3LYP/DZV(2d,p)) not described here.30 Thus it
appears that, given the opportunity, persistent intramolecular
aromatic sugar interactions are established in small molecules
and can significantly influence the oVerall molecular shape. This
is not to say that all the remaining internal degrees of freedom
can be frozen or fully controlled by exploiting the formation of
a sugar/aromatic intramolecular complex: indeed the work
reported here shows that the mobility of the ether side chain in
compounds 2c-f does display some variation as a function of
the carbohydrate involved. Furthermore, the dynamic behavior
of the side chain in the unglycosylated compounds 2a-b does
not differ significantly from that observed in the glycosylated
set (see Table 1), whereas our previous studies show that it is
significantly influenced by the configuration of the included
stereocenter.18a Rather it seems that the soft nature and low
directionality of the sugar/aromatic interaction enables the
structure to be resiliently conserved despite other molecular
changes. This, in turn, allows the molecule to maintain its global
shape. Such behavior can have important implications in the
design of structural mimics of oligosaccharides, as we have
found serendipitously in the course of previous studies.14 This
type of mechanism could also be at work in the formation of
sugar/lectin complexes and fits well the current understanding
of the sugar code as an exquisitely tunable one.31
Ab initio calculations suggest an energy in the range of 5
kcal/mol for the sugar/aromatic type interaction. This estimate
is in agreement with previously reported values.10-12 Aromatic-
sugar contacts have been previously suggested as driven by CH/
π,10-12 van der Waals,32 or hydrophobic33 interactions. More
generally, the nature and origin of weak interactions in
molecules and crystals are the subject of intense and heated
debate.28 In our previous studies14 we observed that 1b (Figure
1), the cyclohexyl analogue of 1a, had a dynamic behavior very
different from that of 1a and did not display any NOE contacts
between the cyclohexyl ring and the carbohydrate portion. Visual
inspection of the combined ab initio and density functional
Conclusions
Compound 2 was chosen as a model for investigating the
nature of intramolecular interactions between four different
monosaccharides (glucose, galactose, N-acetyl-glucosamine, and
N-acetyl-galactosamine in 2c-f, respectively) and the phenyl
ring appended to the ether side chain. NMR studies supported
by molecular modeling and ab initio calculations led to several
(30) Terraneo, G. Ph.D. thesis, Universita’ di Milano, Milano, Italy, 2006.
(31) Gabius, H-J.; Siebert, H.-C.; Andre´, S.; Jime´nez-Barbero, J.; Ru¨diger, H.
ChemBioChem 2004, 5, 740-764.
(32) Toone, E. J. Curr. Opin. Struct. Biol. 1994, 4, 719-728 and references
therein.
(29) Gavezzotti, A. J. Am. Chem. Soc. 1983, 105, 5220-5225.
(33) Elgavish, S.; Shaanan, B. Trends Biochem. Sci. 1997, 22, 462-467.
9
2898 J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007

Study of Carbohydrate−Aromatic Interactions
A R T I C L E S
The first conformer set (MD set) was obtained by running three MC/
SD runs of 5 ns each, starting from three conformers corresponding to
the three idealized geometries of Figure 3 (Conf. A, Conf. B, Conf. C,
15 ns total simulation) and saving 5000 structures for each run, using
an electrostatic cutoff of 20 Å, a van der Waals cutoff of 8.0 Å, and a
hydrogen bond cutoff of 4 Å. The simulation was performed at 300 K,
with a dynamic time step of 1.5 fs; the Monte Carlo acceptance ratio
was less 4%, and each accepted MC step was followed by an SD step.
Structures were sampled every 1 ps and saved for later evaluation,
monitoring both energetic and geometrical parameters. As expected,
the MC/SD protocol allowed for fast interconversion of the starting
conformers, so that the populations obtained from the individual runs
appeared to be converged within 5% and the interproton distances
differed by no more than 0.1 Å.
model of 2c in Figure 9 shows that the apolar hydrogen atoms
on the R-face of the sugar are in proximity with the van der
Waals surface of the aromatic residue, and the molecular orbital
representation of the HOMO-6 orbital (Figure 9b) indicates a
stabilizing bonding interaction between the sugar CH vectors
and the aromatic ring. The electrostatic potential of the aromatic
ring (Figure 10) appears slightly more positive on the face which
is in contact with the sugar, suggesting that the sugar CHs are
acting as weak acids on the π-electron cloud of the phenyl ring.
Preliminary results obtained from NOESY spectra of 2c in
CD3OD solution30 appear to indicate a weakening of the NOE
contacts and possibly suggest a role for hydrophobic packing
in defining the molecule conformation. In light of the above
observations, the interaction between the aromatic ring and the
carbohydrate fragment in 2c-f could be best described as a
relatively extended contact between the two hydrophobic
surfaces involving significant electron overlap.
The second set (MM set) was obtained by Multiple Minimization
of the MD set (i.e., of the total 15 000 snapshots saved during the
molecular dynamics simulations) using the same force field and cutoff
applied in the MD set. The third set of conformations (Fitting set) was
obtained by fitting the NOE data, using a procedure that we have
recently described.21 Briefly, the MM set of conformations was clustered
in the three clusters A, B, and C of Figure 3 using the ø(C(O)-CR-
CHDC4-CHDH4) descriptor as detailed above. Interproton distances
and NOE intensities (%) were derived for each cluster using a full
matrix relaxation calculation extended to all the members of each
cluster. An initial value was assigned to the weight of each cluster,
and the conformer distribution was estimated by fitting the Noeprom-
based % overall NOE intensities to the experimental ones.
Aromatic amino acid residues are often present in the binding
sites of carbohydrate-binding proteins. The resulting carbohy-
drate-protein complexes are characterized by a placement of
the sugar in a roughly parallel orientation relative to the plane
of the aromatic ring, similar to the one observed in compounds
2c-f. These molecules, therefore, can serve as simple and
tunable models for study of the nature of carbohydrate-aromatic
interactions and the implications in the molecular recognition
of carbohydrates.
B. Ab initio Calculations. Full geometry optimizations, including
structural, orbital, and various electrostatic property analysis, were
carried out using the GAMESS software package.38 Structural computa-
tions were performed using hybrid density functional theory (HDFT).
The HDFT method employed Becke’s three-parameter functional in
combination with a nonlocal correlation provided by the Lee-Yang-
Parr expression with both local and nonlocal terms, B3LYP.39 The
DZV(2d,p) basis set40 was used. Full geometry optimizations were
performed and uniquely characterized by calculating and diagonalizing
the matrix of energy second derivatives (Hessian) to determine the
number of imaginary frequencies (0 ) minima; 1 ) transition state).
All energetics and Cartesian coordinates for the optimized geometries
are available in the Supporting Information.
Experimental Section
Computational Methods. A. Molecular Mechanics Calculations
and Generation of the Three Population Sets. All calculations were
performed using the MacroModel/Batchmin 8.534 package (Maestro-
version 6.0) and the AMBER* force field. Kolb’s parameters were used
for the hydroxyacid moiety.35 Bulk water solvation was simulated by
using MacroModel’s generalized Born GB/SA continuum solvent
model,36 which treats the solvent as an analytical continuum starting
near the van der Waals surface of the solute, and uses a dielectric
constant (ꢀ) of 78 for the bulk water and 1 for the molecule.
From the fully optimized structures, single-point energy calculations
were performed using the MP2 dynamic correlation treatment for further
analysis of energetics and properties. The optimal method was
determined using several levels of theory to establish self-consistency
in terms of basis sets as well as effects of dynamic correlation.
Additionally, these methods have been previously shown to be reliable
for the types of compounds considered here.41 Molecular orbital contour
plots and electrostatic potential plots, used as an aid in the discussion
of the results, were generated using the program 3D-PLTORB42 and
GAMESS, respectively, and depicted using QMView43 and MOLE-
KEL,44 respectively.
An initial conformational search was carried out using 10 000 steps
of the usage-directed MC/EM procedure following previously estab-
lished protocols.37 Extended nonbonded cutoff distances (a van der
Waals cutoff of 8.0 Å and an electrostatic cutoff of 20.0 Å) were used.
The output structures were clustered based on the improper dihedral
angle descriptor ø(C(O)-CR-CHDC4-CHDH4), which describes the
side-chain conformation. Three clusters were obtained, corresponding
to Conf. A, Conf. B, and Conf. C of Figure 3. Conformations with
values of the ø improper dihedral angle in the range 180° ( 60° were
clustered in the A family, those with ø in the range 60° ( 60° were
clustered in the B family, and those with ø in the range of -60° ( 60°
were clustered in the C family. In this way all the space was investigated
in all directions, without loosing information and no conformers were
outside of the defined staggered conformations (A, B, or C). The above
criteria were used also to define the A, B, and C populations of the
following sets.
NMR Methods. NMR spectra were recorded in a 5 mM D2O
solution using Bruker AVANCE 400, at 300 K. 1D, as well as 2D,
COSY, HSQC, and NOESY experiments were recorded using the
(38) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon,
M. S.; Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S.;
Windus, T. L.; Dupuis, M.; Montgomery, J. A. J. Comput. Chem. 1993,
14, 1347-1363 (www.msg.ameslab.gov/GAMESS/GAMESS.html).
(39) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652.
The cluster leaders were used as starting structures for the dynamics
simulations.
(40) Dunning, T. H. J. Chem. Phys. 1970, 53, 2823-2833.
(34) Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Liskamp, R.; Lipton, M.;
Caufield, C.; Chang, G.; Hendrickson, T.; Still, W. C. J. Comput. Chem.
1990, 11, 440-467.
(41) (a) Milet, A.; Korona, T.; Moszynski, R.; Kochanski, E. J. Chem. Phys.
1999, 111, 7727-7735. (b) Perez-Jorda, J. M.; San-Fabian, E.; Perez-
Jimenez, A. J. J. Chem. Phys. 1999, 110, 1916-192. (c) Guerra, C. F.;
Bickelhaupt, F. M. J. Chem. Phys. 2003, 119, 4262-4273. (d) Zhang, Y.;
Pan, W.; Yang, W. J. Chem. Phys. 1997, 107, 7921-7925.
(35) Kolb, H. C.; Ernst, B. Chem.sEur. J. 1997, 3, 1571-1578.
(36) Still, W. C.; Tempzyk, A.; Hawley, R.; Hendrickson, T. J. Am. Chem. Soc.
1990, 112, 6127-6129.
(42) 3D-PLTORB; San Diego, 3D version, 1997.
(37) (a) Brocca, P.; Bernardi, A.; Raimondi, L.; Sonnino, S. Glycoconjugate J.
2000, 17, 283-299. (b) Bernardi, A.; Raimondi, L.; Zuccotto, F. J. Med.
Chem. 1997, 40, 1855-1862.
(43) Baldridge, K. K.; Greenberg, J. P. J. Mol. Graphics 1995, 13, 63-68.
(44) Flu¨kiger, P.; Lu¨thi, H. P.; Portmann, S.; Weber, J. MOLEKEL 4.0; Swiss
Center for Scientific Computing: Manno, Switzerland, 2000.
9
J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007 2899

A R T I C L E S
Terraneo et al.
standard pulse sequences. NOESY experiments were carried out using
a mixing time of 800 ms. Exclusive NOE contacts for both compounds
were identified and integrated from the NOESY spectrum relative to
the diagonal peak of the corresponding rows.
0.75 s. The longitudinal relaxation rate (σnoe) and the transversals
relaxation rate (σroe) for each proton pair were calculated as the slope
of the line for mixing time vs NOE intesity. The ratio between σnoe
and σroe gives a mathematical function (eq 1) that allows determining
the τc for a proton pair without knowing the distance (ω0 is the
frequency of the magnet in rad/s).
Cross-peaks from NOESY experiments were integrated using the
MestReC software45 obtaining the related NOE intensities (%) as the
ratio between the diagonal peak and the cross-peak. Their intensities
were compared with those estimated by Noeprom,22 as described below.
The theoretical NOE intensities were generated from the computational
models using the Noeprom software22 based on a full matrix relaxation
approach46 and using a rigid isotropic model.47 The NOE volumes
reported in Table SI-7 (Supporting Information) correspond to a 400
MHz spectrometer, using a 800 ms mixing time and a correlation time
(τc) of 200 ps. The MD predictions were obtained as averages of all
the conformations collected during the 15 ns dynamic runs The MM
predictions were obtained as Boltzmann-weighted averages of all the
conformations obtained by Multiple Minimization of the dynamics
output, within 3 kcal/mol from the global minimum. The NOE Fitting
prediction was obtained by using the Simplex algorithm.48 An initial
value was assigned to the weight of a given cluster (Conf. A, Conf. B,
and Conf. C) and an estimation of the average % NOE intensities
calculated; then, the deviation between the calculated and experimental
values was iteratively minimized (up to 10 000 steps), and thus, the
best-fitting population weights were obtained.
2
4
4
2
4
4
σ
noe/σroe ) (5 + ω0 τc2 - 4ω0 τc )/(5 + 22ω0 τc2 + 8ω0 τc ) (1)
After knowing the τc, σnoe, and σroe for a proton pair, the interproton
distance for this proton couple could be calculated. The density spectral
function (J(nω0)) correlates the molecular motion and the distance, and
so, by using the eqs 2-4 the distance could be deduced.
J(nω0) ) τc /1 + (nω0τc)2
σnoe ) C[6J(2ω0) - J(0)]
σroe ) C[2J(0) + 3J(ω0)]
(2)
(3)
(4)
µ0 2 p2γ4
H
C )
r-3 2
(4π)
10
Acknowledgment. This project was supported in part by the
European Union under Contract HPRN-CT-2002-00173 (Gly-
cidic scaffolds Network). K.K.B. acknowledges the Swiss
National Science Foundation (SNF-200021-107979) for support
of this work. We thank Dr. Carlo Gatti for the many helpful
comments and discussions.
The Noeprom results were analyzed to estimate the % NOE intensity
errors, which are collected in Table SI-7 of the Supporting Information,
together with the associated population distributions and distances (Å).
1D Experiments. 1D NOE and 1D ROE NMR spectra were recorded
in a 5 mM solution of D2O using Bruker AVANCE 500, at 300 K
using the standard selnogp and selrogp sequence from the Bruker
library. The NOE buildup was measured at different mixing times, for
1D NOE 0.2, 0.4, 0.6, 0.8, 1 s and for 1D ROE 0.2, 0.3, 0.4, 0.5, 0.6,
Supporting Information Available: Experimental procedures,
1
compound characterization, H NMR and 13C NMR of com-
pounds 2a-f, tables of NOE contacts and conformational
analysis results for compounds 2a-f, Cartesian coordinates and
absolute energies from QM calculations of compound 2c (gt,
tg, and gg). This material is available free of charge via the
Internet at http://pubs.acs.org.
(45) Cobas, J. C.; Sardina, F. J. Concept. Magn. Reson. 2003, 19, 80-96
(www.mestrec.com).
(46) Borgias, B. A.; James, T. L. Methods Enzymol. 1989, 176, 169-183.
(47) Lipari, G.; Szabo, A. J. Am. Chem. Soc. 1982, 104, 4546-4559.
(48) Dantzig, G. B. Linear Programming and Extension; Princeton University
Press: 1998; Chapter 5: The Simplex Method.
JA066633G
9
2900 J. AM. CHEM. SOC. VOL. 129, NO. 10, 2007